Wednesday, December 20, 2017

Does the Tesla Powerwall 2 produce RFI (Radio Frequency Interference)?

Follow-up article:  There is a follow-up article to this one describing how I mitigated what interference was being produced by the Powerwall 2:  Reducing RFI from the Tesla Powerall 2
Figure 1:
A typical Powerwall 2 installation.
Left to right:  Utility meter/original load center fed from an underground
power feed, the"new" load center to which the household circuits now
connect, the Powerwall "Gateway" (with two 4G antennas on top
 - not used in my installation), AC disconnect for the
Powerwalls, sub-panel for the Powerwalls (containing
a circuit breakers for each unit) and finally, the two Powerwalls.
This type of system is typically installed outside, near the utility's
connection to the house.
Click on the image for a larger version.

Now that I have an installed and operating Tesla Powerwall 2 system I've had the opportunity to answer a question that I've not seen answered elsewhere:

Does the Powerwall 2 cause radio interference?

Why I care:

Being an amateur radio operator that uses a wide range of frequencies across the electromagnetic spectrum (from below 137 kHz to at least 24 GHz) and often "listens" over wider ranges than that I'm always on the look-out for devices that unintentionally produce radio frequency energy which will be manifest as radio interference, reducing my ability to receive signals.

This sort of interference is increasingly commonplace, the incidence having accelerated with the prevalence of "switching" type "wall-warts" (a.k.a. "power cubes") that ubiquitously power nearly anything that is plugged into the wall.  As part of their power conversion, these small devices contain powerful oscillators - typically operating in the 20-100 kHz range - that have the potential to cause radio interference, even at frequencies far removed from their operating range.

What this means is that the inclusion of even more of these devices in my household - including a Tesla Powerwall 2, which is a really big switching power converter - all have the potential of adding to this sea of noise.

What is a Powerwall?

A Powerwall is the Tesla-specific name for what amounts to a "whole house UPS" (Uninterruptible Power System).  There are other manufacturers of similar systems and they have their own nomenclature, generically called an "AC Battery" because they internally perform the AC to DC conversion for charging and DC to AC inverting to provide external AC power.

As the name implies, if the mains power disappears, this system can provide electricity to the entire house (or a portion of it) during the power outage.  As you might expect, very large, high duty-cycle loads such as whole-house air conditioning, electric water heaters, electric clothes dryers and electric furnaces are typically not backed up by a system like this as they would draw down the battery very quickly.

When integrated with a PV (solar electric) system it can be charged from solar energy and if the grid remains unavailable, the house can run indefinitely from the Solar+Powerall, provided that the short-to-medium-term power budget is positive - that is, more solar power is produced than is being used and the battery is not discharged so much between charges (e.g. overnight, on cloudy days) that it reaches the point of cut-off.  My system has two Powerwall units which, working in tandem can provide at least 10kW of power with a storage capacity of a bit more than 26kWh - enough for about a day (without any solar input) with normal usage or several days (without solar input) or little/no sun at all (e.g. the dark of winter with snow-covered panels) if serious power conservation measures are taken.

In areas where there are significant electric rate (tariff) differences between "peak" and "off-peak" hours, this type of system can be used to "zero out" (or reduce) utility usage during peak hours and charge during off-peak hours from the grid and/or with solar.  In my area, this is not relevant as the power rates remain constant throughout the day and it is configured to charge only from solar - which also makes it eligible for the (current as of the time of writing) 30% federal tax credits.

Having one of these systems is a bit like having a back-up generator - except that if the sun is shining, the "gas tank" can be refilled.  Practically speaking this system is unlikely to save me any money in the same way that a back-up generator probably wouldn't, so I would consider it to be a sort of extravagance - like owning an RV, boat or some 4 wheelers - a bit like a somewhat expensive hobby, but more utilitarian.  Being an amateur radio operator I'm also interested in having back-up power in case there is some sort of event that causes the loss of the grid for a period of time, hence the concern about possible radio interference.

How it's connected:

Figure 2, below, shows how a typical "AC Battery" might be wired into a household power system and integrated with a PV inverter.
Figure 2:
A generic block diagram of an "AC Battery" type of back-up power system.
In "Tesla speak", the "Gateway" comprises the functions depicted in the box labeled "Supervisory Control" while the
Powerwall(s) themselves are depicted by the boxes labeled as "Battery-backed inverter/charger system(s)".  If one has large loads that you do not want to be backed up by the Powerwall (e.g. electric water heater, central air conditioner, etc.) these would be connected "outside" the backed-up circuits via a panel connected at the point marked "To Utility Metering/Breaker".
A manual disconnect are required to give first responders an easy way to kill power to the entire house should it be necessary - such as in the event of a fire or other disaster - as simply killing the mains circuit alone would
not do this as the Powerwall would simply restore power automatically!
Click on the image for a larger version.
Comment:  As the time of writing there are some parts of the world - many being in Europe - where, due to regulations, the "whole house back-up" during a grid failure is not available.  The radio interference potentials described below still apply in these cases.

As can be seen, in normal operation the AC battery system is in parallel with the house's power bus and the power grid.  When "charging" from the solar, the system simply monitors the output power of the PV system and adjusts its charge rate to match.  In the "Self-Powered" mode (described below) when there is a grid connection it will charge/discharge at a rate that precisely matches the house's usage, effectively zeroing-out the power going to/from the grid when charging the battery, or export excess power to the grid once the battery has been charged in the same way as a typical "net meter" installation.

If the mains power fails the "Grid Isolation Relay" opens, disconnecting the house from the grid, allowing power to the backed-up loads to be maintained without back-feeding the utility.  The process of detecting a grid failure, disconnection from the grid and full restoration of the power seems to take between 200 and 750 milliseconds but the return to a grid connection - after the mains power has returned and stabilized for a few minutes - is nearly instantaneous.

If there is sufficient power budget (e.g. enough battery storage to last until the next day's sun and there is enough PV capacity to run the house and recharge the battery) it is possible to run "off grid" entirely - theoretically, indefinitely.  The downside of running in an "island" mode is that if you have a "Net Metering" arrangement, the "excess" power that would normally be exported back to the grid would simply be lost and no credit would be available for it:  Since there's really no down-side to having a system that disconnects the power grid during a failure, anyway, why not get credit for "excess" power?

So, does the Tesla Powerwall 2 cause radio frequency interference?

Yes and no.

The "no" part:

On the HF bands I have determined that in my particular case (and prior to mitigation techniques described later) the interference potential on the HF bands to be minimal or negligible.

When the unit is idle (neither charging or discharging) no interference can be detected on any LF, MF or HF band.  No interference has been noted on VHF or UHF bands, either.

When the unit is operating (either charging or discharging) and I am using my normal HF antenna system I cannot detect any interference from it on the HF amateur radio bands of 80 through 10 meters (e.g. 3.5-30 MHz).  Additionally, I cannot detect any interference from the Powerwall 2 system on any VHF or UHF band, either.

If I walk up to the Powerwall 2 system with a portable shortwave radio while it is operating I can hear a bit of noise when I am within a foot or so (less than a meter) that is likely due to short-range magnetic fields, but this noise energy doesn't seem to be being coupled to the connecting wires outside the unit.

The "Yes" parts:

160 meters:

Prior to noise mitigation techniques (mentioned below) on the highest MF band, 160 meters (1.8-2.0 MHz), the story is a bit different:  When the unit was operating, I could just detect a bit of noise from the unit in the far background, just below the local noise floor - but whether or not I could hear this at all depended on which antenna I'm using for receive.  For example, on an active E-field whip I could just hear this noise, but it is not at all audible when using a wire antenna.

Now that I've done a bit of noise mitigation, it is no longer audible on 160 meters with any antenna.

On lower frequencies (below 160 meters):

Prior to noise mitigation techniques (mentioned below) going down lower in frequency - into and below the AM Broadcast band (e.g. below 1.7 MHz) - the RF noise being produced by the Powerwall 2 (again, when it is charging or discharging) gradually increases, was fairly obvious by the time one got to the bottom of the AM broadcast band (e.g. 530 kHz).  Below the AM broadcast band are two more amateur bands - relatively recent additions to amateur radio in the U.S. - and both of these are bands on which I operate:  The 630 meter band (472-479 kHz) and the 2200 meter band (135.7-137.8 kHz).

At these lower frequencies the interference from the Powerwall 2 (when it is operating) ranged from "significant" at 630 meters to "considerable" at 2200 meters - but this is not surprising.  It would appear that the main power converter(s) inside the Powerwall(s) operate at 32 kHz - and the 2200 meter band is at only about 4 times this frequency.  Because the 2200 meter band's frequencies are comparatively close to the operating frequency of the inverter and its 4th harmonic at 128 kHz - and because RF interference filtering works better as frequency is increased while the harmonics of these converters (and their significant mains-frequency modulated sidebands!) also decrease in amplitude - the amount of energy at 2200 and 630 meters will naturally be higher than it would be on the HF bands.

In short:  If you do not plan to operate on the 160, 630 or 2200 meter bands, you will likely not experience any interference at all, even if no mitigation techniques are used.

I can only speak from experience with my system:  Other systems may be better or worse in terms of interference, depending on the situation.

An interference source that can be controlled - "The devil you know":

One of the biggest problems with interference is that the source is often unknown - but if you know what is causing interference, it can often be controlled - by adding some sort of filtering, replacing the device with an equivalent that is less likely to cause interference, or simply not use it during those time during which interference would be a problem.  Following this last point, if the RF interference from the Powerwall 2 were to be of great concern it's worth noting that the user has pretty good control of when this might happen as interference from the Powerwall 2 seems to occur only in two possible states:  When it is charging, or when it is discharging.  What this means is that even if you use the MF (160 or 630 meters) or LF (2200 meters) bands it will not cause interference when it is "idle."

A typical Powerwall 2 owner would operate it in one of two modes, selectable from a phone app:
  • Backup-only.  In this mode the Powerwall 2 operates only as a "whole house UPS" - that is, it is not producing power except when the utility mains is offline (e.g. a power failure or the user has disconnected it from the grid).  In this configuration and in a typical installation, charging of the Powerwall 2's battery is done only with energy from the PV system (solar + inverter) when it needs to do so - and this usually occurs only if the battery has been discharged below 95% or so.
  • Self-powered.  In this mode the Powerwall 2 monitors the net inflow and outflow of power from the house.  In this configuration the Powerwall will either output enough power to "zero out" the usage of the house so that there is, on average, no power going to/from the utility and/or it will take excess power from the PV system to charge its battery which will also "zero out" the power to/from the utility.  If the battery is fully-charged, excess power from the PV system will be fed back into the Grid, just as is done in a normal "Net Metering" situation.
Note:  At the time of this writing there is expected to be a "load leveling" There is now a mode offered in the near future available where the Powerwall may be configured to charge/discharge at specific times to take advantages of time-based tariffs (e.g. lower-cost power during "off" hours).  This does not apply to me and such operation is beyond the scope of this article, but the interference potential could be inferred.

In the "Backup-only" mode the Powerwall 2 system is not usually operating (charging/discharging) and will thus not typically produce any noise on any amateur band - but in the "Self Powered" mode, the only time that interference would not be being produced would be when the Powerwall 2's battery is fully-charged and the excess PV power is being exported to the utility grid.

What this means is that if there is the possibility of interference, one would typically operate in the "Backup-only" mode where it is fairly rare for the unit to operate at all.  In my case, the charging portion of the inverter will operate only for a few hours in the morning as soon as the PV system starts to produce power, one or two days a week when it "tops off" the battery.

If, for some reason one wanted to completely eliminate the possibility of the unit going active - say, during some sort of contest - the Powerwalls could simply be turned off, but this would be done at the risk of losing the power back-up capability in the event of a grid failure, but this is something that you would probably consider only if you were operating on 160 meters or lower.

"My neighbor is putting in a Powerwall system - Will I hear it on the HF bands?"

Probably not.  As noted above, I could hear it only on 160 meters and only on a small active vertical whip which is far more sensitive to such things than a typical wire antenna.  If your antenna isn't very near the Powerwall, you probably won't even hear it at all.

One thing that I have going for me is that my power feed is underground - but this is largely unimportant in my case:  My house is connected directly to the wiring of the Powerwall and if it were to conduct a lot of stray RF onto the AC wiring, I would certainly hear it as it would be conducted directly into my home's wiring and be radiated.  On that basis, I do not believe that even with overhead wiring, the Powerwall would be radiating enough "grunge" from the powerlines alone to be audible.

Another point:  If your neighbor is in the process of installing a Powerwall - or already has a system -  there are several opportunities to determine if QRM is coming from it:
  • Does the QRM go away in the dark?  If so, it's not the Powerwall - more likely some microinverters or "optimizer" modules on each panel.
  • Has the Solar+Powerwall system apparently causing QRM been installed, but not "commissioned" by the utility yet?  If the system in question is not currently operating in a net-metering fashion (e.g. feeding power back to the grid) then the Powerwall will, by design remain offline.  It will not be until the system is fully "on-line" with the solar power system feeding back into the grid and charging the Powerwall that the Powerwall itself is likely to be active and capable of producing any RF interference at all - if it's going to do so.
  • Some microinverters are known to cause interference even at night - if the mains power is applied to them.  If interference is suspected, be sure to check, at night, with the mains power to the microinverter system shut off.
  • The possibility of conincidence should not be ruled out:  Something else could have been installed, causing interference - in the house with the solar power system, a different neighbor's house or even your own!
Follow-up article:

There is a follow-up article to this one describing how I mitigated what interference was being produced by the Powerwall 2:  Reducing RFI from the Tesla Powerall 2

Other articles related to the mitigation of interference from switching power supplies:
Some of the above articles contain additional links to other web pages on related topics.


This page stolen from

Wednesday, December 6, 2017

KA7OEI now QRV on 630 and 2200 meters

Figure 1:
The LF/MF transmit station, configured for 630 meter operation.
At the time this picture was taken I had not yet completed the gear and
put the various pieces in their boxes, hence the mess of clip-leaded-
together modules sitting on my workbench.  Not visible is
the low-pass filter in the power amplifier box or the thermocouple-type
RF ammeter.  The pictured 630 meter variometer has been supplanted
with a "new" one wound with 660/42 Litz wire for lower loss (see Figure 6).
Click on the image for a larger version.
It so-happened that I had a few days off around Thanksgiving and I took this time to throw together a fairly simple transmit converter for the "new" amateur LF and MF bands - notably 2200 and 630 meters.  Having had already obtained my approval from the UTC to operate on both of these bands, I was "good to go".

It took only two evenings to put together the transmit converter and power amplifier as I had the parts on hand - and none of them were particularly exotic.  While the transmit converter will be described in greater detail in a future post, the signal path for the transmitter is approximately thus:

(See the block diagram in Figure 2, below.)
  • A 5 MHz IF is used, allowing a "broadbanded" FT-817 (with TCXO) to serve as the exciter.
  • The 5+ MHz signal (about 5137 kHz for 2200 meters, 5475 for 630 meters) is mixed (using a 74HC4066) with a 5 MHz local oscillator (a 10 MHz OCXO divided-by-two) to yield frequency-stable LF/MF signals.
  • A low-power post-mixer amplifier boosts this signal to a level capable of driving the power amplifier.
  • A single-ended MOSFET-based broadband power amplifier, running on 12-30 volts, provides between 10 and 50 watts of RF at either 630 or 2200 meters.  Because the transmit converter is broadband, it is agnostic to the operating frequency meaning that one needs only use the appropriate low-pass filter to change bands.  (The 630 meter low-pass filter is always in line - another filter is added for 2200 meter operation.)  This power amplifier is designed to be driven by either the transmit converter or another device, such as a QRP Labs Ultimate 3S beacon transmitter configured for these bands.
  • The 50 ohm output of the power amplifier goes to a tapped autotransformer wound on what is probably an FT-240-61 toroidal ferrite core and is used to match the transmitter's output to input resistance of the loading coil.
  • Also in the drawing is a relay the disconnects the loading coil from the autotransformer when not transmitting.  This was necessary to prevent the transmit antenna from "sucking out" some of the receive signal being intercepted by my E-field whip and also to prevent the transmit antenna from conducting "house noise" from the transmitter onto the transmit antenna which gets coupled into the receive antenna, reducing ultimate sensitivity. 
  • The loading coil, placed in series, cancels out the capacitive reactance of the antenna system.  For 630 meters my antenna requires about 230uH while about 2.5mH is needed to resonate the same antenna at 2200 meters.
Figure 2:
Block diagram of the 630 and 2200 meter transmit chain.  The transmit converter is broadband, capable of operating
from below 100 kHz to at least 500 kHz which means that one need only provide appropriate matching and low-pass filters to operate on either band.
Click on the image for a larger version.

When I made my first-ever transmission I had not yet constructed the variometer, but I fished around in my "box-o-inductors" and found several Litz-wound ferrite inductors that were probably rescued from some scrapped TVs or computer monitors and wiring enough of these in series I was able to achieve  resonance with about 750mA of antenna current.  On the very first WSPR transmission I managed to be "heard" by several stations (See Figure 3, below.)
Figure 3:
A screen shot (from of the very first 630 meter WSPR transmission that I made with the badly-kludged loading coil.
Not too bad for a temporary lash-up!
On the next night, after observing a few stations engaging in JT-9 QSOs, I answered a CQ by VE7SL and he replied, giving me a signal report of -22dB while I gave him -19dB.  This was quickly followed by two other QSOs as both W7IUV and NC0B noticed the "new guy" on the band!

Over the next several days I got around to constructing the "new" variometer depicted in Figure 4 and this boosted my antenna current to about 1.25 amps - a theoretical improvement of about 4.4dB with more QSOs to follow - including 2.5 (one "partial") CW contacts on the band.  After operating for a while it became apparent that, for the most part, I could work anyone that I could "hear".

A few days later I constructed yet another variometer for 630 meters - this time using some 660/42 (e.g. 660 strands of 42 AWG) Litz wire which reduced the skin-effect losses by a significant amount and this, along with minor improvements of the ground system, decreased losses and resulted in a further increase of antenna current to a bit over 2 amps - a theoretical ERP improvement of more than 8.5dB as compared to my original configuration. The measured resistance at the input of the 630 meter Litz coil is about 13.5 ohms, implying an overall antenna system efficiency roughly 1% - but still enough to work quite a few stations with a few 10s of watts of RF.

Figure 4:
 The "Mark 1 version of the 630 meter variometer.  This device is wound
on "4 inch" ABS triple-wall sewer pipe using 22 AWG insulated hookup
wire.  Inside is "3/4-inch" ABS waste pipe (actual O.D. about 1-1/8")
that forms the rotatable portion of the variometer.  This unit has an
adjustment range of approximately 175-235 uH.
Click on the image for a larger version.
The "Q" of the antenna system+Litz wire coil is now such that if I QSY from 475.75 kHz for WSPR operations down to about 475.0 kHz for JT-9 I actually see noticeable drop in antenna current until I readjust the variometer, but if I QSY from 475 kHz down to 473 kHz for CW operation the antenna current plummets to a few hundred milliamps and I absolutely must retune!

As is the custom on both the LF and MF bands, my WSPR signal reports not the transmitter power, but rather the estimated EIRP.  I've typically been reporting 0.5 watts (+27dBm) which, assuming about 25 watts of RF power, implies an antenna efficiency of about 2% which, while in the general ballpark, may still be a bit optimistic.  With the recent changes/improvements in my system (mostly improving the grounding, radials and counterpoise network) I will have to re-analyze my estimated system efficiency.

Operation on 2200 meters:
Figure 5:
Antenna and ground system of my LF/MF TX antenna system.  The
yellow line represents the outline of the "Lazy Loop" - a horizontal HF
antenna fed with 450 ohm window line with both conductors of the
feedline being tied together and fed as a tophatted vertical on LF/MF.
The total circumference of this antenna is about 215 feet (65 meters) -
dimensions mostly dictated by the locations of trees at an average
height of roughly 30 feet (9 meters).
The red lines show the extent of my ground/radial system showing
extra wires, including sections of chain-link fences with electrically-
bonded sections and wires buried in the ground, including an
abandoned CATV line.  The roofs of both the house and garage are
metal which are ultimately tied into the ground/radial network.  There
are several ground rods near the feedpoint of the antenna to which
all of the grounds/radials are connected.
Click on the image for a larger version.

I have since wound yet another variometer (visible in Figure 6, below) - also on 4" ABS pipe - for 2200 meters.  This coil, adjustable from about 1.7-2.0mH, uses the same 22 AWG hook-up wire as my original 630 meter loading coil.  As it turned out this coil, by itself, doesn't have quite enough inductance to resonate my antenna at 137 kHz so I place the other two 630 meter coils in series with it.  As compared to the 630 meter loading coils, it is somewhat lossy, but I am able to obtain about 900mA of antenna current:  Not surprisingly, this coil runs slightly warm in operation due to the losses - but these are, no doubt, minor in comparison with the ground losses.

Update - 12 December, 2017:  After improving the ground system my antenna current is now around 1.1 amps on 2200 meters, implying an improvement of at least 1.7dB from current alone.  The actual far-field improvement, based on readings seen from monitoring stations on WSPR, appears to be in the area of 2-3dB.

The measured resistance at the input of this loading coil is about 43 ohms implying an overall antenna system efficiency of well under 0.1%.   Based on estimated antenna efficiency, I've configured WSPR to report my ERP as 50mW, which assuming a transmitter output power of about 25 watts implies an actual antenna efficiency of about 0.2% which is probably very optimistic!

Update - 5 May, 2018:  After some old-fashioned number-crunching, several antenna simulations and comparing my signal to other beacons of "known" EIRP I've revised my estimate of radiated power to be closer to 20 milliwatts, even after taking into account that I've increased my amplifier's output power to between 60 and 80 watts.  The previously-optimistic calculations assumed lower ground (and other losses - such as those due to nearby vegetation) than I originally thought that I had.

* * *

Not surprisingly, operation on 2200 meters - even at this power level - can be a bit hazardous.  With the rather low antenna capacitance the voltages on the feed are quite high - an estimated 5000-8000 peak volts!  What this means is that the feed wire has to be kept well clear of other conductors or else corona will occur, sapping transmit power, filling the room with ozone and becoming a potential fire hazard.  Fortunately, at this modest power level - and with the current-regulated power supply that I'm using - almost any sort of fault will detune the antenna system to the point that the high voltage will all but disappear and/or the power supply will go into current limiting and effectively shut down the transmitter.

Figure 6:
Left to right:  The original 630 meter variometer (seen in figure 4
wound with 22 AWG stranded wire , the new 630 meter wound with
660/42 Litz wire and the 2200 meter variometer, wound with the
same 22 AWG stranded wire and insulated with PET tape to allow it
to withstand the high voltages.  In the lower right corner is the
autotransformer wound on an FT-240 ferrite core.  With my current
(pun intended!) antenna I must put all three of these variometers in
series to resonate the system at 2200 meters.
Click on the image for a larger version.

Despite this simple arrangement I've managed to be "heard" by at least seven other stations in the western U.S. and Canada using WSPR to date, but I've not yet made any 2-way contacts.  The relative scarcity of stations that listen or transmit on 2200 meters - coupled with my rather weak signal - means that a contact will probably have to be arranged and conducted using a weak signal mode like JT-9 or QRSS.


There are plenty of improvements to be made, most notably getting the feed of my antenna a bit higher, laying out a few additional ground wires to further-reduce losses and improving the variometer for 2200 meters - but there are only so many things that I can do on my relatively small city lot.  This entire arrangement has so far been precariously sitting on my workbench meaning that the high RF voltages are also also nearby, just waiting to leap out at me when I reach over to tweak a variometer.

At some point I'll "remote" the matching network outside, but I need to get/build a few other items first, namely some stepper motors, control circuity, more vacuum relays and a means of remotely monitoring the antenna current.

Comment:  Despite having the feedpoint in my shack, I've not had any problems at all with transmit RF getting into computer speakers or other devices in my house.

* * * * * * * *

My recent operation, as of the date of this post, seems to be the only actively transmitting station on either 630 or 2200 meters in Utah.  I have been running WSPR on 2200 meters most of the time, occasionally switching to 630 meters in the local evenings when the activity level on that band is highest.

If you are QRV on 2200 or 630 meters and would like to arrange a CW, JT-9 or QRSS contact with me, or if you are interested in just "hearing" my signal (via your ears or with a computer+sound card) drop me a line using my callsign at arrl dot net.

Other entries on related topics found at this site:
Other web sites that have information on 630 and 2200 meters: 

This list is by no means comprehensive.  Peruse the "links" sections on the sites below for even more information.
  • NJD Technologies - link  - This web page has a wealth of information related to 630 meter operation, propagation and reports of activity, plus lists of known-active operators on both 630 and 2200 meters.  This web site also has many links to others that have credible information on LF and MF band topics.
  • W1TAG's web site - link  - John, W1TAG, has long been an experimenter and operator on the MF and LF bands.  This site has details on equipment both for operating and measuring performance at these frequencies.
  • W1VD's web site - link - Jay, W1VD, has long been an experimenter on the LF/MF bands and this page offers a lot of information on equipment for transmitting and receiving on these bands.
  • Antennas by N6LF - link - The callsign gives  you the clue that this guy likes LF/MF operation.  This page includes detailed information on LF/MF antennas and how to characterize/improve them.


This post stolen from

Friday, December 1, 2017

Containing RF noise from a sine wave UPS

An amateur radio friend of mine (WA7X) has a cabin in the mountains.  It is not a particularly "rustic" cabin as it is festooned with radios, antennas, propagation beacons, computers and cameras and has an internet connection, but because it is in a remote location it occasionally has 3-6 hour power outages and thus it also has a UPS (Uninterruptible Power System) to keep many of these things online in the interim.

This "sine wave" UPS is a 1.5kVa unit that was cast off from by someone for the same reason most UPSs are cast off:  Its internal battery went bad.  Rather than simply replacing the battery, its previous owner simply got another UPS and asked the question "Do you want this?"

Instead of replacing the internal battery, the DC connections for the batteries were brought out and a bank of six 12 volt lead-acid batteries was wired up (two sets of three parallel batteries connected in series amounting to a nominal 200-ish amp-hours at 24 volts) to provide the needed 24 volts and a DC circuit breaker was added for safety, this battery capacity allowing the unit to run for far longer than it could have on the original battery.

While this UPS was more efficient than a previous unit and produces a fairly nice sine wave rather than the typical, ugly "modified sine wave" there was a price to pay:  RF Interference (e.g. RFI) that was present whether the unit was active or on standby.

But first, a few weasel words:
  • This project involves high voltages and/or currents:  Do not attempt to construct a similar device unless you are thoroughly familiar with electrical safety and the wiring of such devices.
  • If you use an external battery bank with a UPS it is imperative that you include some sort of current liming, such as a fuse or circuit breaker rated for both the expected current and battery voltage.  Such devices are available from auto-parts stores.
  • If you use an external battery bank with a UPS you must determine if this battery bank is "mains referenced" internally by the UPS or not.  If it is mains referenced (e.g. connected directly or indirectly to the mains AC voltage) then the low-voltage DC terminals will pose a line voltage safety hazard and care must be taken - the least of which is enclosing the battery and any exposed DC terminals to prevent accidental contact.  This UPS's battery terminals were isolated from the mains and given that the room in which the UPS resides has restricted access, the low-voltage battery connections themselves were deemed to be "adequately safe" left exposed.
  • YOU are responsible for the safety and for any liability if you choose to do something similar to what is described on this page.  You have been warned!
    Figure 1:
    The completed AC/DC filter.  This
    box contains a "brute" force L/C filter
    for the battery (DC) leads as well as
    separate filters for the AC mains in/out
    power connections.
    Click on the image for a larger version.


The radio frequency "grunge" from the UPS manifested itself on the HF ("High Frequency" or shortwave) bands in several ways.  Most obvious was a loud "buzz" every 40-50 kHz on the lower (80 and 40 meter) bands, but there was also more subtle interference that pervaded these and higher bands:  A background "hiss" that might initially escape the notice of the casual listener until one realizes that this noise masked signals that should have still been perfectly audible.  If one switched the receiver to AM it would be observed that this "hiss" was subtly modulated at twice the mains frequency, 120 Hz.

To determine the extent of this sort of interference the typical procedure is to start turning things in the house off one-at-a-time (or, more reliably, the reverse:  Turn everything off at the breaker panel and then turn on one thing at a time) until the culprit is found.  This was done with the UPS and the magnitude of its "radio interfering" nature was determined.  Clearly, this "grunge" was being conducted from the power leads going in and out of the UPS.

Further experimentation revealed the true extent of the noise:  Even with everything disconnected from the UPS - that's to say, with it running on its battery, unplugged from the mains and the load disconnected - there was still detectable noise getting to the antenna and a quick check with a portable shortwave receiver proved that the remainder of this noise seemed to be being radiated by the physically-large battery bank and the wires that connected it.

While the proper application of a "brute force" AC line filter would likely quash the noise conducted in and out of the UPS on the mains power leads, something else would be required to minimize/eliminate the noise emanating via the DC lines.

"Brute force" line filters:
Figure 2:
A typical "brute force" L/C line filter typically found on devices
to minimize conduction of extraneous RF energy on the
AC mains.  For a 1500kVa UPS the filter would need an
appropriate current rating - particularly the fuse!
In some filters, two sections are used, with the components
L1, C1, C2 and C3 repeating.  This same filter
topology is used for the described DC filter.
Click on the image for a larger version.

One of the best ways to eliminate or minimize the amount of RF energy that might be conducted out of a potentially interference-generating device is to apply a combination of inductance and capacitance to that line as depicted schematically in Figure 2.

With the noise coming from the power supply (the UPS in our case) capacitor "C4" effectively "shorts" this RF noise to both sides of the power line, leaving the AC signal (pretty much) unchanged.  Inductor L1 consists of two equal windings on a ferrite core and it is practically invisible to signals that are equal and opposite, but it acts as a series choke for signals that are "common mode" - that is to say, equal on both sides of the power supply's mains leads - such as the RF noise energy.

On the "mains" side of the filter capacitor C3 reinforces the common mode again while capacitors C1 and C2 shunt any remaining RF energy from the power supply - its impedance now made much higher by inductor L1 - to ground - which would be the metal enclosure in which everything was mounted.

As it happens, these filters are available on the surplus market, and we would need three of them:
  • One for the AC mains connection to the UPS.
  • Another for the UPS's AC output
  • A third one for the battery connection to the UPS.
With the UPS being rated for 1.5kVa, some surplus AC mains filters, rated for 16 amps, were obtained while a filter designed specifically to filter DC lines rated for at least 50 amps at 60 volts was found (at The Electronic Goldmine - item G21652 - link) to filter the power connection between the UPS and the battery:  50 amps at 24 volts is not quite 1.5kVa, but there was nowhere near this amount of power being drawn from the UPS and the added circuit breaker/disconnect would provide the required safety - and the filter should be able to handle a brief overload, anyway.

Putting the filters in a box:
Figure 3:
Inside the filter box.  Along the top edge is the main AC input.  On the bottom edge - and just below the outlet on the left side - are the short leads that conduct the "dirty" AC power to/from the UPS, each through its own, separate filter - the two black boxes.
On the far right is the high-current DC filter with stud (bolt) connections being used to make the connections between the battery and the DC input of the UPS.  Barely visible along the bottom, one of the three studs is used to connect a piece of heavy wire or braid to bond this box to the chassis of the UPS to minimize conducted/radiated RF.
In the middle is a circuit board that contains a mains transformer, a high-current mechanical relay and a small solid-state relay.  This board - added later in the design - allows one pair of the outlets to be turned on and off with a simple contact closure of an internet-connected remote power switch.  On the control cable for the relay (the thin white wire) is a ferrite device which minimizes the amount of RF energy that might possibly be conducted in or out of the box on that control lead.
Click on the image for a larger version.

Ideally, one would have put the UPS and the batteries in a large metal box and passed the power leads in and out of this box only via the filters, but this simply wasn't practical, so the next best thing had to be done:  Put the filters in a single, metal box that would be electrically bonded to the UPS chassis and make the connections in and out of the UPS using short leads.  By keeping the leads short and bonding our new filter box to the UPS, we'd do our best to limit the number and length of conductors that carried the RF interference and, most importantly, preventing RF circulating currents from finding their way on external connections.
Figure 4:
The end of the box with the mounted outlets.  As noted,
one pair of these outlets is connected to a relay to allow the
connected devices to be remotely controlled.
Click on the image for a larger version.

To that end a power distribution box was found at a home improvement store (Lowes Depot, I think) for less than $25 and the "guts" removed (the box with "guts" was cheaper than just an empty box by itself!) and the filters mounted inside.

With the short-as-possible conductors between the UPS and the filter, they will have minimal ability to directly radiate RF while the RF conducted on these lines will be filtered by the circuitry in the box itself with the bonding of the two boxes minimizing differential RF currents.  The power "to" the UPS was made with a short length of "SO" cord with a female connector on it while the power "from" the UPS is via a short cable with a male connector, plugged into one of the UPS's outlets - and being the ONLY thing plugged into the "dirty" AC output of the UPS.


Figure 5:
Perhaps a bit cluttered, this is the UPS sitting atop the filter, installed
and working.  In the lower-left corner of the picture, above two
batteries may be seen the DC circuit breaker/disconnect that protects
the DC circuit for short circuits.
Click on the image for a larger version.
Being that this is a remote location, the filter unit was installed a few weeks after it was constructed, having been tested (as best as could be done) on the workbench:  Power flowed through the various filters and the load control relay worked properly.

When installed, the filter box was placed underneath the UPS as it had a slightly larger footprint - and to minimize the length of the "noisy" DC and AC power leads from the UPS, along with the lead used to bond the two cases together.

To connect the DC, the cable coming from the batteries was effectively cut so that there was just enough of it emerging from the UPS to connected to the "output" side of the filter:  On both sides of this cable heavy "ring" lugs were attached and these were connected  to the studs of the DC filter.  To eliminate the probability of accidental shorting, the ground stud on the "input" side of the filter was removed and near both connectors a plastic wire clamp (one may be seen in Figure 6, below) to keep the positive wire in a fixed position and rotating into and shorting to the other stud.

Figure 6:
The back side of the UPS, showing the coiled power cord and the
(green) bonding wire that connects the chassis of the UPS and filter
box firmly together.  At the bottom of the picture can just be seen the
plastic clamp the keeps the positive wire lifted and away from the
the negative wire, to prevent shorting.
The ONLY thing plugged directly into the "dirty" AC output
of the UPS is the cord going to the filter:  Plugging anything else
directly into the UPS would at least partially negate the filtering! 
Click on the image for a larger version.
The AC input and output of the UPS was simply "plugged in" to the cables and the excess cordage was neatly coiled and stowed at the back end of the UPS using plastic "zip" ties:  It was important to do this because these cables are unfiltered and are "noisy" with RF meaning that they should be kept as small and as close to the metal of the cabinet as possible and kept away from any other conductors to minimize cross-coupling which would defeat the purpose of this filter.

To the remaining ground stud of the DC filter was attached a short piece of heavy (8 AWG) green wire with a ring lug on each end, this wire being visible in Figure 6.  The other end of this wire was attached to a marked grounding screw on the UPS chassis to bond the two boxes together and minimize RF circulating currents and to prevent the UPS chassis itself from being a source of radio frequency interference by direct radiation.

The result:

Taking the 20 meter (14 MHz) amateur band as an example, the UPS caused an extra 2 "S" units or so of noise above that of the typical ionospheric noise floor when it was powered up, before the filter was installed - this, being detected on a Carolina Windom antenna lofted between two trees high above the cabin's roof.  After this filter was installed the noise from the UPS was completely undetectable on any HF band, revealing other weaker low-level noises from other devices - the quieting of some of these will be discussed in later installments.

Figure 7:
A general block diagram of how the parts are interconnected.
The external UPS battery bank is protected with a DC-rated circuit breaker/
disconnect switch.  This particular UPS operates from 24 volts, hence the two
series-connected 12 volt batteries.  If the battery is inside the UPS's metal
cabinet, the DC filtering and connections are not needed.
Note that the "output" side of both AC line filters are both connected to
the UPS:  This is done because many filters are designed such that the
"output" side is that best-suited for connecting to RF-noisy circuits.
This diagram does not include the remote relay described.
Click on the image for a larger version.
Before the installation of the filter I'd placed my FT-817 (a small, portable HF transceiver) across the small room from the UPS, receiving with a short antenna and tuned to the 20 meter band.  When the UPS was operating and connected to its loads its noise was clearly audible, causing several S-units of indication on the signal meter.

After the filter was installed and the unit was powered up again with the loads connected, the noise was barely audible in the FT-817 and from across the room, it went away completely when the green wire was connected, bonding the UPS and filter chassis together.  If the radio was moved to within a foot or two of the UPS I could start hearing the "hash", but it seemed to be emanating only from the coils of AC cables "zip"-tied to the back end of the unit.   Because of the short length of the wires - and their being close to the metal case and not near any other wires into which this noise could be coupled it is unlikely that these short conductors will radiate any detectable noise at a distance greater than a few feet.

Doing the same on another UPS:

As part of his "noise abatement" strategy, WA7X did some RF sniffing around his house and discovered that a major contributor of RF noise was a sine wave UPS of a different brand.  While the UPS itself was completely different, the same things needed to be done to it as was
Figure 8:
Line input filter.  This was mounted externally, using very
short wiring, as it was difficult to insert a filter directly
into the AC mains path via the connector that was
directly mounted to the UPS's circuit board.
Click on the image for a larger version.
done to the UPS at his cabin:
  • A "brute force" line filter on the AC input.
  • Another line filter on the AC output.
  • Because the batteries on this one were, like the other one, external there needed to be a filter on the battery DC line as well.
This UPS was a bit smaller - 750VA (about 600 usable watts assuming an average 0.8 power factor) and it, too, had been modified to use a much larger set of external batteries to maximize run time.   Like the other UPS, it produced copious quantities of RFI whether it was "UPSing" (providing back-up power) or idle.

Figure 8 shows the input filter - and its mounting represents a slight complication:  The AC (mains) input went directly to the UPS's circuit board - and no on-board RF filtering was seen.  While we could have probably removed the board, cut some traces and soldered wires to "intercept" the AC flow between the input mains cord and the circuitry, we did something a bit "cheesier" - but still effective:  Attached a filter to the outside.  Going to our local electronics surplus store we found a suitable line filter with an overkill rating of about 10 amps (6 amps would have sufficed for a 750 kVA UPS) with a built-in IEC power connector.  As can be seen from Figure 8 we cut up an IEC power cord and with minimal length, soldered it directly to the terminals of the line filter and insulated them
Figure 9:
Output line filter.  This was mounted in the empty battery
compartment.  The "dirty" AC power from the inverter
is via the black/white wire on the left and the "clean" AC
power to the outlets is on the right.  The "clean" power
was wired using green wires with black/white markings
as that was the only solid wire - used to push into the
outlet's rear connector - that was on-hand.  These two
sets of wires are kept physically separate to prevent RF from
capacitively coupling from one to the other.
Click on the image for a larger version.
with heat shrink tubing and copious quantities of RTV sealant.  Because this filter was intended to mount via screws to the back panel of a piece of equipment, it had no mounting lugs, so we soldered some to the tin-plated case and used these for the ground connection as well.  It looks a bit kludgy, but its "pretty safe"!

For the AC mains output we disconnected the wires from the rear-mounted power socket and spliced short lengths of wires to it, sleeving the splices with several layers of heat-shrinkable tubing, connecting the output of the UPS's inverter to one side of the filter mounted in the battery compartment.  Connected to the AC outlets were short pieces of wire which were then soldered to the other side of the mains filter.  Although not seen in the pictures, there were no chassis holes that permitted the passage of these two sets of wires, so a pair of holes (protected with rubber grommets) were drilled with the unit tilted so that metal shavings (e.g. "swarf") would not end up in the circuitry - one hole for the "dirty" AC power from the inverter and another hole for the "clean" AC power to the socket on the output.  Having these two sets of holes and routing the two sets of wires away from each other prevents the RF grunge from coupling from one set of wires to the other.
Figure 10:
The DC line filter.  Too big to fit in the battery compartment,
this filter was mounted to a metal partition present in the UPS,
holes in that partition and in the chassis below to pass the
metal studs that carried the DC power.  Underneath, through
the battery compartment, was later wired the cabling for the
24 volt, fused DC input.
Click on the image for a larger version.

Finally, there is the DC filter.  Using the same surplus filter obtained from Electronic Goldmine as was used in the other UPS, this filter was mounted in an empty space adjacent to the circuit board.  There, a metal divider provided a handy mounting point for the "dirty" side of the DC connection (e.g. the input to the inverter).  A metal hole punch and a nibbling tool was used to fashion the holes necessary to pass the power studs through them with a similar hole being punched through the case into the battery compartment.  Before the filter was installed we declared which terminal was the positive side and clearly indicated it with a red permanent marker.

In Figure 8 one can see a grommet through the panel to the right of the filter:  This was the original passage of the DC power cables which were not installed at the time the pictures were taken.  When these cables were added later, appropriate DC fusing was included and the wires were carefully routed through the battery compartment, keeping them away from the "dirty" side of the filter to prevent coupling of RFI.

The result:

As with the other power supply, a portable AM/shortwave radio held near the UPS was not able to detect any noise unless it was held within an inch or two (couple of cm) where it was likely that magnetic coupling occurred.  When checked, there was no noise heard on the HF spectrum that was attributable to the inverter - only a few more "new" noises that will need to be chased down now that this UPS has been quieted!

Links to other articles about power supply noise reduction found at this site:


This page stolen from

Tuesday, November 21, 2017

Matching the 2 and 6 meter cycloid dipoles to 50 ohms

The "Cycloid" dipole - a circularly-polarized antenna capable of (more or less) omnidirectional (toward the horizon) radiation was discussed in previous posts:
Figure 1:
The 6 and 2 meter Cycloid dipoles at the WA7X beacon.
The 2 meter cycloid dipole has been in service since 2001 and while the 6 meter
cycloid was (mostly) built at about the same time, it has been in service
only since 2015.
Click on the image for a larger version.
The 2 and 6 meter antennas were installed at the site of the WA7X beacon to impart a circular polarization on the transmitted signals, making them (generally) agnostic to the antenna used by the listener - which is to say that it wouldn't matter whether the receive antenna was vertically or horizontally polarized.  The use of circular polarization also reduces the problem where the ionospheric reflection may rotate the signal to the "other" plane and cause fading at the receive location owing to cross-polarization.

While this antenna is described as being "omnidirectional", that is not true in the proper sense of the word.  Its circularity and most of its radiated power is directed toward-ish the horizon (at some elevation angle) in all directions while relatively little energy is radiated at a high angle upwards or downwards - and what is being radiated in those directions not likely to be very circular.  As with any antenna, the proximity of the coaxial cable, metallic support and feedline will, no doubt, skew the pattern in some way - and this antenna is no exception - but this is unavoidable.

While the dimensions of the "antenna" part of the Cycloid dipole are spelled out in the linked article(s) above, details related to matching of these antennas to 50 ohms is not - with only the suggestion that a "1/2 wave matching network" be used.  While this matching network is very simple, it may be unfamiliar to some, so what follows is a paraphrased response to an email on this very question.

Matching the cycloid dipole to 50 ohms:

While I carefully noted the dimensions of the dipole when I designed them, we never precisely measured the various dimensions of the matching networks of the 2 and 6 meter Cycloid Dipoles as the need for precise replication would render them as mere "starting points" and they are simply stub-tuned 1/2 wave sections - only the dimensions of the actually "antenna" portion are on the web page.

In retrospect, a full 1/2 wave section was probably an overkill as a 1/4 wave may have sufficed - but that "extra" bit of open-wire balanced transmission line (e.g. the portion of the pipe between the coax tap and the "cycloid" part of the antenna) worked out well to provide physical support, rigidity, and counterbalance, would not cause any significant loss, and it all but guaranteed that we would be able to find a good match. in almost any conceivable situation.

The details of the matching network and its tuning are thus:
  • We used about 1/2 wave length of copper tubing protruding from the "back" of the antenna, used also to support the main antenna body.  As can be seen from the pictures, it was folded upon itself, zig-zagging to reduce its overall size.  This extra weight can help to counter-balance the antenna itself.
  • There is a 1/2 wavelength coaxial balun to go from 50 ohms unbalanced to 200 ohms balanced using small 50 ohm coaxial cable.  This type of balun well-described in literature and one of several online calculators may be found here: .  For the 2 meter antenna we actually used some small, 50 ohm hardline (the RG-58-sized equivalent of "UT-141" PTFE coax) that was obtained on the surplus market, but RG-8x or even RG-58 would have been fine.
  • When the antenna was tuned, it was mounted on a nonmetallic support (a fiberglass ladder) placing it several feet/meters above the ground and an MFJ analyzer was connected to the far end of the coax (10-15 feet away) to minimize the effect of having a person too close to the antenna and affecting tuning.  For initial tuning, it should be mounted to the same type of mast as that which will be used for permanent mounting.  For the antennas at the WA7X beacon, plastic pipe has proven to be durable with the 6 meter antenna being mounted using black ABS sewer pipe.  In the case of the 2 meter antenna, it was mounted using some PVC piping that seems to be holding up despite being out in the weather for well over a decade.
  • We prepared two nonmetallic sticks - 5-6 feet long (1x1, wood dowels, bamboo, small plastic pipe, etc.) and one of these had a piece of heavy wire to use as a shorting stub and the other had the balanced (200 ohm) side of the coaxial balun, also connected to 2 wires. The wires/balun were simply taped to the end of the stick to allow contact to be made.
  • Make sure that the copper pipe from which the matching section is made is clean and free of oxide using steel wool or sanding with fine-grit paper to allow a reliable connection while sliding the connections back and forth - both for finding a match and for ease of soldering.
  • At the position farthest from the "antenna" portion, the sliding shorting bar was placed while sliding connection to the balun was placed near it, on the "antenna" side of the shorting bar.
  • With the sticks, the two pieces (coax attachment and short) were slid around to achieve 50 ohm match.  While the two sliding portions are held in place, another person marks their position with a permanent marking pen on the antenna when a match is found.  It is easier to move the connection on the balun back and forth while watching the VSWR while slowly moving the shorting bar back and forth, looking carefully for a match.  Typically, the two connections will be fairly close to each other as seen in Figure 2, below.
  • Once a preliminary match is found, the sliding shorting bar is replaced with a piece of heavy, solid wire (#10-#14 AWG) that is wrapped around the pipes at the marked position. The other sliding bar (on the balun) is then re-checked for a good match, the shorting bar's position tweaked as necessary.
  • Once the position of the shorting bar has been established, the wire on the balun section is wrapped around the pipe at the location of the best match, allowing the wooden stick to be removed.  The positions of the two connections are then tweaked by sliding the wires back and forth for best match.
  • The two connections are soldered in place, and the match re-checked.  If it is OK, the connections are sealed and the match re-checked and adjusted as necessary.
  • In both Figure 1 and Figure 2 one can see small pieces of acetal (e.g. Delrin tm) plastic on the matching network - this material being chosen for its low RF loss characteristics and its durability to UV exposure.  Note that PTFE (a.k.a. Teflon tm) would have also worked well.  Some of the pieces (those at the far left edge of the matching section) are used for mechanical support, but the others are used for fine tuning:  The position of these pieces of dielectric slightly alter the tuning.  After the antenna was fully assembled, these were moved back and forth for the best match and secured in place with blobs of RTV (e.g. Silicone tm) sealant on both sides.  (RTV does not stick to this plastic, but the blobs keep it from moving about.)
As can be seen in Figure 2 on both the 2 and 6 meter antennas, the attachment point of the balun is fairly close to the shorting bar.  The proximity of the coaxial cable balun to the match may affect tuning a bit, so it must be fixed into place before the final tuning is done.

Figure 2:
Annotated image showing the locations of the shorting bars, coax baluns, balun connections and the rain shields on
the 6 and 2 meter cycloid dipoles.  The matching network on the 6 meter antenna is longer than
necessary to allow its far end - which is electrically neutral - to be clamped to the mounting pipe and attached to
a ground wire for static discharge/lightning protection.  The connecting cables were secured with good-quality
electrical tape and black "zip" ties, which were also covered with electrical tape to protect them from UV.
Click on the image for a larger version
Note that for each antenna a "snow/rain shield" was placed over the top of the matching section to minimize the effects of moisture, but the addition of this shield did change the tuning, as did the addition of RTV sealant on some of the connections, so final tuning must be done with such hardware and sealant in place.

The entire procedure is a lot easier if there are 2 or 3 people participating as it is pretty tricky for a single to hold two wires on sticks in place and mark them. If there is only one person available, the shorting bar wire would be wrapped around the (clean!) pipes at a position correlating to about 0.4 wavelength on the pipe and the balun portion slid back and forth to see a "dip" in the VSWR, iteratively adjusting the shorting bar back and forth experimentally while sliding the connection from the balun to get the best match.

As I noted, it is possible that a 1/4 wave section would have been fine, but we just used the 1/2 section as there would be no doubt that it could be matched - and we wanted to minimize the hassle related soldering/unsoldering things as much as possible.  Importantly, this type of match - using the large pipes and "open wire" line - is very low loss compared to many other matching networks (e.g. those using small wound coils and discrete capacitors) and it contributes to the mechanical strength of the antenna itself.


This page stolen from


Tuesday, November 14, 2017

An alternate PA transistor for the QRP Labs Ultimate 3S beacon for more output power

The QRP Labs Ultimate 3S beacon (kit) is a reasonably-priced, compact and self-contained unit.  Despite its apparent simplicity and low price it is capable of transmitting in a variety of modes, such as WSPR, CW, Hellschrieber to name but a few on any amateur band from 2200 meters through the U.S. 222 MHz band.

Sort of.

Actually, it's not really that simple:  For operation on 160 through 10 meters, the construction is rather straightforward but for the higher (VHF) and lower (LF, MF) frequency bands, a few "mods" have to be made - at the very least, some "custom" low-pass filters need to be made and modifications to the power amplifier.  On the bands 160 through 10 meters the design of the circuitry means that at best, only a few hundred milliwatts of RF is possible with the parts supplied, the output power dropping off as one goes up in frequency.  For WSPR, even a few 10s of milliwatts will usually yield the desired results (e.g. detection of band openings) but there are instances where more power

The power amplifier:
Figure 1:
The front panel of my Ultimate 3S beacon, WSPRing away on
20 meters.  You can tell that I live in the U.S. by the position of
the "power" switch!

Click on the image for a larger version.

To minimize the cost, the power amplifier section (Q1-Q3) of the Ultimate 3S beacon uses BS170 N-channel low-power MOSFET transistors.  These devices are capable of dissipating about 1/3-2/3 of a watt each and there is room for three of these devices.  If an efficiency of about 50% can be obtained, it should be possible to safely get between 0.1 and 0.5 watts out of the beacon on the lower bands (e.g. 160-30 meters) - plenty for modes that allow very weak signals to be detected such as WSPR.

But, there is a problem.  The BS170 is not an RF transistor, but designed for low-power switching such as level conversion, turning on LEDs, small motors and relays.  At low frequencies - up to several MHz - it actually works quite well, capable of about a watt if three devices are installed, but by the time one gets to 10 meters it is, in this application, rather challenging to get more than about 100 milliwatts from the Ultimate 3 without a bit of tweaking.

Besides getting the available power amplifier kit for the beacon, one of the options the builder can choose is whether to wire the PA transistors for 5 volts or connect it to a higher-voltage power supply.  In general, using a higher-voltage supply - say, 12-15 volts - will enable somewhat higher RF output power, but this also means that the same amount of bias current at 5 volts will result in higher power dissipation and finding the best value - without blowing up the transistors - is a bit of a delicate dance.

The problems:

The problems with this device at higher bands such 10/12 meters (and up) include:
  • Device capacitance.  There are a number of parasitic reactances involved - including the input and Miller capacitance.  All of these conspire to make it more difficult achieve a wide voltage swing and/or to turn the FET on and off quickly - something that needs to be done to amplify higher frequencies efficiently
  • The drive capability is rather limited.  The power amplifier section of the Ultimate 3 beacon is driven directly by the synthesizer which, for older units (mostly the non-"S" version) could be a DDS board, but the more recent versions use the Si5351 synthesizer chip which has a somewhat lower output level.  Neither of these devices produce enough output to "fully" drive the FET's gate.
The upshot is that while the rather simple amplifier circuit works pretty well at low frequencies, it drops off rapidly as one goes up.

One of the methods to deal with limited drive signals is to bias the transistor slightly.  Because it - like any similar FET - takes a volt or three to start turning on, biasing the transistor toward "on" with a fixed DC voltage means that the limited RF drive signal doesn't have as "far to go" when it comes to driving the device.

Adding this bias works well - but only to a point:  Eventually, the transistor is conducting so much DC current that it is dissipating heat at/near its maximum rating and increasing the bias even more to further increase its effective gain is not an option.  One option is to add heat sinking (by gluing the transistors to a piece of aluminum or copper) to keep them cool, but this is of limited utility.

The Ultimate 3 beacon has the capability of using up to three of these transistors in parallel and while this can improve the power output at lower frequencies (maybe) the limited drive capability of the synthesizer - plus the fact that each transistor has its own capacitance - doesn't necessarily help.  One other factor often overlooked is that FETs are notoriously inconsistent in their DC characteristics:  Unless one goes through pains to sort and match individual FETs - even devices from the same lot - when several are placed in parallel and biased, one is inevitably going to pull more drain current than the others.  This means than when several parallel devices are used, one or two are going to be doing most of the work and under stress while the other two (or one) will be doing comparatively little.

All of this would seem to be an argument to use a single, more capable amplifying transistor to obtain more output power.


The BS170 is quite popular in QRP transmitters because it is cheap, but it can be made to work "less badly" and the best way to do this is to strongly drive its gate with an RF signal.  Often, high-speed CMOS gates are used for this such as a 74AS04 or equivalent with multiple sections wired in parallel.  Doing this "brute force" drive technique can greatly improve the output capability of this otherwise low-frequency device and if done correctly, a DC bias is unneeded, saving a lot of hassle.  Unfortunately, the Ultimate 3 beacon doesn't have a device like this in its signal path, instead connecting the output of the synthesizer (more or less) directly to the gate of the output transistor(s), but one could hack the circuit and wire such a device into the circuit.

Another work-around would be the use of a transistor specifically designed for RF use.  While there are many such devices available, most are quite expensive or hard to find.

One such device is the RD16HHF1 made by Mitsubishi and recommended for the optional 5 watt PA board available from QRP Labs, but this transistor is becoming increasingly difficult to find.  Taking into account the fact that it may difficult to use the original (small!) holes for the BS170, using this device should work well - provided that it is operated within the capabilities afforded by the limited ability to dissipate heat.

The RD16HHF1 is also a favorite for counterfeiters that take an ordinary FET's die, put it in a package and label it as the real thing:  This fakery may work on lower bands, but it falls apart at higher bands for the same reasons that the BS170's efficacy drops off.  Some counterfeiters don't even bother to mount a fake die, instead take an ordinary power FET and label it as an RD16HHF1:  Because the drain and source connections of the RD16HHF1 is "backwards" from "normal" FETs, a device like this will simply short out the power supply!  The only way to be absolutely sure that one has a genuine RD16HHF1 is to put it into a "component tester" - those inexpensive (<$20) devices that will identify practically anything - and see if its pin-out is correct and that its gate capacitance is in the 60-100pF area - and then try it in circuit.

The PD85004:

In perusing the catalogs I determined that a likely candidate device was the PD85004, made by ST Microdevices and available from a number of vendors such as Mouser Electronics.  This device, designed to operate from 13.8 volts, is rated to output several watts at 900 MHz, so it should surely be coaxed to work at HF, right?

This device is a bit more expensive than the original, in single quantities costing about $3.25 each as opposed to about $0.50 each for the BS170 - but the expense isn't very onerous, and it is probably cheaper than a genuine RF16HHF1 - and it is also rated for operation at 13.8 volts.
Figure 2:
The PD85004 mounted to an EvilBay SOT-89-4 carrier.  Short wire
leads go in holes 1, 2 and 3 to effectively turn it into a through-hole
device.  The heat sink was not yet added when this picture was
taken.  As described in the text, I eventually used a 10 ohm
resistor for the gate lead (position 1) to prevent circuit instability.
Click on the image for a slightly larger version.

One complication with the use of this device is that it is available only in a surface-mount package.  Fortunately, I had on hand some SOT-89-4 "carrier" boards (readily available on EvilBay - search for "SOT-89 adapter board") to which I soldered the device, effectively turning it into a leaded device that can be wired into the original FETs' board locations.  These boards cost anywhere from $0.03-$0.20 each, if you buy 10 or more - and that price often includes shipping!

To improve device dissipation a piece of copper flashing was carefully soldered to the tab of this device (which is the grounded source lead) after it was mounted to the carrier (see Figure 3.)  While the rated dissipation of this device is 6 watts, the mechanical layout of the Ultimate 3 beacon significantly limits the size of the heat sink as well as how much heat can be radiated/conducted to its surroundings.

Modifying the U3 for use with the PD85004:

Earlier versions of the U3S specified a simple inductor ("L1") for the drain circuit of the power amplifier while later versions depict either a simple series inductor or a bifilar-wound transformer designated as "T1" - the latter being capable of somewhat more power with the original BS170.  My U3S is of the earlier version with "L1".  I've only tested it with the simple inductor, but this modification should work well with the "T1" configuration as well - but since it is untested, one should be particularly wary of instability.

Initially, I simply wired a PD85004 in place of a BS170 - but owing to the fact that the new device was designed to operate near 1 GHz - and that the layout of the U3 circuit board and interconnects didn't look to be particularly "VHF friendly" - so I expected that there could be some problems.  Before powering it up for the first time I turned the bias potentiometer all of the way down and pre-set my bench supply to current-limit at 500 milliamps - just in case I'd miswired something or I'd managed to turn the bias control all of the way up, instead.

Powering up the beacon and temporarily disabling transmit (easily done in WSPR mode by disconnecting the GPS antenna that is used for timing) I noted the current consumption - about 350 mA, much of that being the LCD's back light - and carefully adjusted the bias to cause a 100mA increase in current consumption.  With the antenna output of the beacon connected to a dummy load via a wattmeter I then reconnected the GPS antenna, readjusted the power supply current limiting and waited for the unit to come online and cycle through the various amateur bands while listening, in turn, to each frequency on a local receiver - using it as an oft-overlooked piece of useful test equipment that most amateur operators already own!

The result, not unexpected, was that I was able to get a power reading on each band, but on some bands - 160 through 40 meters - I heard a loud "hiss" +/- about 20 kHz from the transmit frequency instead of a CW note while the higher bands, 30 through 10 meters, sounded normal.  This just goes to show that at these frequencies this GHz-rated device may need some "taming" to prevent the apparent low-frequency instability and that a wattmeter alone is not necessarily useful for determining if an amplifier is working properly!

To tame the amplifier, I did several things:
  • I installed a 0.1uF between the wiper of the bias adjustment potentiometer R5 and ground.
  • I placed a 220 ohm resistor in parallel with R6 from the bias supply.  This, along with the added capacitor, helped "swamp" the drive signal and provide some lower impedance, low-frequency termination of the device's gate.
  • I replaced the wire lead on the SOT-89 carrier that provided the gate connection with a 10 ohm resistor - pin "1" on the device carrier board in the pictures.  This added resistance helps to break up effects of spurious reactances that can cause the transistor to behave badly in-circuit.
  • Update:  I decided wire in a series-connected 0.1uF capacitor and 1.5k resistor between the gate (pin 1) and drain (pin 3) to quell a suspected VHF/UHF instability under some conditions, manifested by the drain current changing slightly on some bands when I touched the grounded(!) heat sink.
  • Update:  I had to take the U3 apart for another reason and while I was at it I connected a 470 ohm, 1/2 watt resistor across L1 - between V+ and the drain - for good measure.  See notes below about this change.
In testing with the dummy load I found that the amplifier section was now stable.

I then proceeded to carefully adjust the bias and watch the power meter.  Turning up the bias to several hundred milliamps I observed that I could get 2-3 watts of "clean" RF on on every band - but this much power was too great for the heat sink - especially in a "closed up" case - but had an "infinite" heat sink been possible I'm certain that I could have safely operated at this power level.  Monitoring the temperature of the heat sink I found that I could safely get about 1.5 watts out on 160-20 meters, dropping to about a watt on 10 meters, but erring on the side of caution I backed this off a bit to 0.5-0.75 watts on 10 meters which correlated with about 50mA of idle current.
Figure 3:  
The installed PD85004 with heat sink in the Ultimate 3S beacon
for initial testing.  A piece of copper was soldered to the tab of the
transistor to allow it to dissipate a couple watts of heat.  Unfortunately,
there isn't really an easy way to make the heat sink capable of dissipating
significantly more heat without mechanically complicating things.
Not visible in this picture are the added 0.1uF capacitor between the wiper
of R5 (the pot visible to the left of the transistor) and ground
and the 220 ohm resistor in parallel with R6.
  Note that the connections from two of the three original
BS170 positions were used to wire this transistor into the circuit.  Behind
the transistor is the "OCXO" version of the synthesizer and I could have
soldered the heat sink to its case to improve dissipation, but this may risk
reducing the OCXO's own thermal and frequency stability.
This picture was taken before the 10 ohm resistor was put in series
with the transistor's gate to quell instability, replacing the wire lead
at position "1" on the carrier and before the addition of the series
1.5k and 0.1uF capacitor between the gate and drain..
Again, the PA section of my beacon is wired to operate from 13.8 volts.
Click on the image for a larger version.

On the air testing:

My "main" HF antenna is a "lazy loop" of about 225 feet (approx. 70 meters) circumference at an average height of around 30 feet (about 10 meters) feed with 450 ohm window line with a 1:1 balun in the shack - designed to be connected to an antenna tuner.  Because I am not using the tuner with the U3 this means is my antenna it is not resonant (at 50 ohms) on any particular frequency, typically having a VSWR of greater than 5:1 on most bands.  While this may sound bad, the window line itself contributes negligible loss of its own and a reasonably-designed power amplifier should be able to tolerate such a mismatch.  I've run it this way for months with the BS170 finals without a problem and I've gotten reasonable signal reports.

When I connected the modified Ultimate 3S beacon to this antenna, everything worked fine - until I got to 40 meters, at which point I'd hear a loud "click" on the local receiver and the display would go blank.  Apparently, the bad (reactive) termination of the antenna caused the amplifier section to "take off" into some sort of mode of instability and somehow crash the beacon's processor.

Adding a "wee bit" of attenuation:

The work-around was to add an (approximately) 1.5dB resistive pad in series with output antenna connection.  Consisting of two 3.9 ohms resistors and a 220 ohm resistor in a "Tee" arrangement, this prevented the return loss as seen by the beacon from ever exceeding about 3dB, or a VSWR of about 6:1.  This little bit of padding reduced the transmit power only a fraction of an "S" unit, but with its added 3dB of return loss was sufficient to keep the amplifier stable on all of the bands.
Figure 4:
Typical "T" type resistive attenuator that can be useful for preventing
instability and/or damage to the final transistor in the event of a
poor match to 50 ohms.  These resistors were wired/mounted
at the RF output connector, after any low-pass filtering.
For Ra I used the standard values of 3.9 ohms and for Rb, 220 ohms,
resulting in approximately 1.5dB of attenuation.  For the 2.5 dB
attenuator on the WA7X beacon, Ra was 6.8 ohms and Rb was 180
ohms.  Neither of these sets of values are precise 50 ohm matches,
but they are more than "good enough"!  At up to a watt of RF, 1/4 watt
resistors may be used for Ra but a 1 watt resistor (or multiple,
lower-power resistors) should be used for
Rb so that it can tolerate very high mismatch conditions.

The addition of a bit of attenuation on a transmitter like this isn't necessarily a bad thing as it can offer a bit of protection - both in terms of VSWR and things like lightning strikes, offering both a DC discharge path and act as a bit of a "sponge" for induced spikes and excess power.  Even with this bit of attenuation I can safely coax about 1 watt of RF output on 160 through 20 meters, dropping to a bit over 0.5 watts on 10 meters - about 10dB better than I'd managed with a single BS170 on that band.  Because of the mismatch and commensurate losses I've currently set the WSPR beacon to report half this amount of power on some bands, but will increase it again when (?) I get around to putting up a multi-band matched antenna system.

The use of a resistive pad on the output of the transmitter had a precedent.  Upon installing another Ultimate 3S for the WA7X beacon at a remote cabin - this system operating exclusively on 10 meters - we discovered, the hard way, that the optional 5 watt amplifier (using an RD16HHF1)  didn't like it when the 10 meter vertical was temporarily detuned due to snow, causing a mismatch that resulted in the coincident failure of the output transistor.  In that case we added a 2.5 dB resistive pad (5 dB added return loss) to prevent the beacon from ever seeing worse than a 4.5:1 VSWR (even if the antenna connection were accidentally removed) and instead of 5 watts, the beacon is now operating at "2 watts" and is pretty "bullet-proof", reliability being very important for a remotely-controlled beacon at a remote location.

If that hadn't stopped the instability...

If the amplifier hadn't been adequately stabilized by the aforementioned modifications, there would have been two more things that I would have tried:
  • (Now done - see note above) - Add a 220-470 ohm, 1/2 watt resistor across T1 (or L1), the output transformer (coil) between V+ and the output transistor(s) drain(s).  This resistor can help "Q-spoil" a low frequency resonance on the inductance of  L1 that may cause similar oscillations.  This sort of instability is quite common, often due to the fact that RF devices can have tremendous gain at very low frequencies and the interaction with the rather large amount of inductance of the coupling transformer.  My U3 is a bit older and uses "L1" which is simply an inductor between the drain of Q1-Q3 and V+.  If you have chosen a bifilar/trifilar transformer (e.g. "T1") then this resistor would be wired between V+ and the drain.
  • (Already done - see note above) - Add a series 1k resistor and 0.1uF capacitor between the output transistor drain and gate.  This degenerative feedback will also help quell spurious oscillations.
Ultimately, I hope to connect the beacon to an antenna (perhaps a trapped vertical) that is resonant on at least most of the bands on which the beacon operates, but in the mean time, this seems to be working out pretty well.

Final comments:

I started out this blog entry with the mention of bands above 10 meters, which naturally brings up the question:  Will this same modification work on 6 meters and higher?

The answer is yes, probably.

While I can imagine that it should be possible to obtain, perhaps, 0.25-0.5 watts on 6 meters with this same device and using similar techniques, going up much higher in frequency and getting some RF power will probably require a bit of modification as the board layouts and interconnects start to get a bit "iffy" at VHF and higher, requiring special care to avoid excessive harmonic content and other spurious signals.


Stolen from