Frequency response of the RX-888 SDR at the high and low ends (above 30 MHz and below 1.5 MHz)

Figure 1:
The RX-888 Mk2.

The RX-888 Mk 2 (hereafter referred to as the  RX-888 or '888) is a versatile device, essentially providing a means by which "all of HF" (0-30 MHz - or even 0-60 MHz) may be sampled and presented to a computer for processing via a multi-gigabit USB3 interface.  As it has no onboard signal processing, this device is practically "future proof" in that as all computations are performed on the host computer and there are no frequency or bandwidth limitations regarding the sort of signals - or how many - may be processed, presuming adequate processing capacity.

Comment:

I've seen at least three different "sub-versions" of the RX-888 Mk2 - each one looking slightly different (different circuit board color and other minor differences) - since this device was released.  It's also very likely that components also vary a bit with different manufacturers so the actual frequency response of units of different "builds" may also change.  Unfortunately, I only "have what I have" and haven't been able to compare differences - if any.

The highs and the lows

Like any receiver, it has limits of its frequency response - both at the upper end where the high-pass filter dominates and at the bottom end where the component selection as well as the design itself will limit low-frequency response.

Let's look at the low end first.

The lows

The low end limit to the frequency response (somewhere below 1 MHz) of the '888 has not previously been well defined.  This low frequency response is set by component limitations within the HF signal path, including:

• Coupling capacitors.  DC blocking capacitors in series with the signal path will act as high-pass filters, rolling off the low frequencies.
• The Bias-Tee inductor.  The RX-888 has the ability to supply power via the antenna port to an amplifier.  This inductor has a finite inductance and it, too, will force a high-pass response as well.  This inductor's value was measured as being 10uH (nominal) which presents a reactance of 50 ohms at about 800 kHz.  This is the major contributor to low-frequency roll-off as discussed below.
  
• The coupling transformer.  The RX-888 has a transformer that couples the input of the variable gain amplifier (VGA) from the attenuator.  As with any transformer, this, too, has defined low-frequency response.  This transformer was measured and found to have an inductance of 125uH of its primary (a reactance of 50 ohms at 64 kHz) with the secondary (the side facing the VGA) being about 760 uH.
  

This low-frequency roll-of is not uncommon in broadband receivers:  Most amateur transceivers suffer severe performance degradation at LF and VLF frequencies for the simple reason that the designers presume (correctly!) that very few of the users of that gear would ever be interested in that range - and making this assumption simplifies the design somewhat and reduces cost. 

Using a signal generator with a constant output, the response of the RX-888 (Mk2) was measured, using the signal strength at 1500 kHz as a reference:

Frequency (kHz)	Attenuation (db) Unmodified unit	Attenuation (db) Bias-Tee inductor removed
1500 (Reference)	0	0
1250	0.3	0
1000	0.4	-0.1
750	1.0	-0.2
500	2.6	-0.3
475 (630 meters)	2.9	-0.3
400	4.5	-0.2
300	7.4	-0.2
250	10.9	-0.2
200	19.9	-0.2
150	17.9	0
137 (2200 meters)	14.7	0
100 (Loran C)	9.7	0.3
75  (DCF77 approx.)	7.5	0.7
60  (WWVB, JJY)	6.8	1.1
50	6.7	1.9
40	7.4	5.3
30 (Submarine comms)	11.3	7.4
25 (Submarine comms)	12.8	10.5
20 (Submarine comms)	17.5	15.0
15	22.5	22.7
10	30.7	32.8
7.5	36.4	40.0
5	45.5	49.5
2.5	61	60
1	80	87

Table 1:  Attenuation measurements at 1.5 MHz and below using both an unmodified RX-888 and the same one after the bias-Tee inductor was removed.

Comments about the frequency response of the unmodified unit:

As can be seen from the table above, the "stock" RX-888 is flat within about 2 dB or so across the AM broadcast band (520-1700 kHz) but it falls off precipitously between 100 and 300 kHz with a bit of "rebound" in the 40-150 kHz area, likely due a very low "Q" resonance of inductance and capacitance of the aforementioned components (inductors, transformers) in the signal path.

In the "VLF" range (30 kHz and below) the unmodified receiver may be somewhat usable when using an active antenna to overcome losses, but at 20 kHz and below the response drops off like a rock and, as the chart shows, it's pretty much unusable below 5-10 kHz.

The factors above conspire to prevent a flat frequency response at lower frequencies - say, those below 1.5 MHz.  For the table below, my reference amplitude and frequency is 1.5 MHz as it seemed to be more or less representative of the amplitude response above this, in the HF range - and it seemed to be comfortably above that at which the aforementioned high-pass effects of the components were having a significant effect.

Figure 2:
The red arrow points to the location of the
10uH bias-Tee inductor.  As seen in
Table 1, its removal can significantly improve
MF and LF performance.
Click on the image for a larger version.

Can anything be done to improve LF/VLF response?

YES, it is possible to modify the RX-888 to improve the low and frequency response by removing the Bias-Tee inductor from the HF port and as can be seen from the above data there is a dramatic difference in usable sensitivity at frequencies below 1 MHz - particularly below 400 kHz.

This is the easiest modification as it entails the removal of a single component and Figure 2 shows the location of this inductor.  It may most easily be removed with a hot-air rework tool, but it should be possible to carefully use a solder-wetted iron to heat it and remove with a pair of tweezers (temporarily remove any thermal pad below that portion of the board if it's present) or a very sharp pair of diagonal flush-cut pliers to remove it (perhaps destructively) as well.

There are other ways by which the low frequency response may be improved, including:

• Replacing the coupling transformer.  The transformer used in the RX-888 is likely specified for a low-end frequency response of 1 MHz or so, so it's not surprising that this may be the worst offender (once the bias-tee inductor has been removed) in low-frequency roll-off.  Replacing it with a different unit with larger inductance (a commercial or hand-made unit) would certainly help.  It may also be possible to simply replace the transformer with coupling capacitors (say, 0.1uF) - but this would be at the expense of sensitivity and performance across the entire frequency range, something that might be acceptable if one's primary interest was in the MF/LF/VLF spectrum.  As the inductance of the transformer's primary is known to be about 125uH, we can see that this is likely the main cause of attenuation below 60 kHz.
• Increasing value of coupling capacitors.  The coupling capacitors in series with the signal path are likely not ideal for coupling VLF frequencies.  A value such as 0.1 uF or larger would be suggested.

For VLF use (30 kHz and below) if you have interest in this frequency range you may be better off not trying to use the RX-888 - at least directly.  Some possibilities include:

• Use a VLF up-converter.  Converting the frequencies 0-30 kHz to a higher frequency range will put this spectrum within the useful range of the RX-888 and practically any other modern receiver.  There have been a number of VLF up-converter units for sale in the past, but I don't have a specific recommendation.  If this up-converter is clocked from the same source as the RX-888's clock (e.g. using its onboard 27 MHz oscillator, or both from a common, external clock) then frequency drift could be minimized.
  
• Use a sound card.  A modest computer sound card with a 192 kHz sample rate and a 16 to 24 bit A/D converter is perfectly capable of ingesting frequencies up through at least 80 kHz and down (nearly) to DC.

Having a receiver capable of VLF (3-30 kHz) or ELF (300-3000 Hz) is one thing, but having an antenna system capable of this is a different matter altogether.  There are many available E-field active whips that will work well down into the 10-20 kHz region, but below that frequency you are into the realm of specialized gear - and listening at "audio" radio frequencies in all but the most rural areas devoid of power lines and other forms of civilization can be fraught with frustration and disappointment due to the likely pick-up of mains-related energy and its harmonics.

Here are a few links related to equipment for LF/VLF reception.  Note that I have not necessarily built, bought or used the equipment described below, so your mileage may vary.

• The BBB-4 Natural Radio Receiver.  This is a simple circuit is intended for "Audible" RF frequency ranges of a few hundred Hz to a few 10s of kHz and is suitable for detecting "spherics" like whistlers and chirps.
• Some thoughts on E-Field Whistler Receiver Design.  This article describes design aspects of ELF and VLF receivers
  
• An up-converter for receiving long and very long waves.  This article describes an upconverter that may be built.
  
• AMRAD low-frequency up-converter.  This is another upconverter design for VLF/LF use.  The low-end frequency response is limited by T1 in this design.
  
• Heros VLF-LF-MF up-converter.  This is a commercial LF/VLF upconverter. 
• The AMRAD Active LF Antenna.  This article - from the September, 2001 QST - describes a high-performance antenna capable of reception down to at least 10 kHz.  In lieu of the CP666 transistor, one may also use the Russian KП903 (a.k.a. "KP903") power JFET which may be found on EvilBay and other sites.
  

The effective reception of signals in the LF, VLF and ELF frequency range is highly contingent on having a "quiet" receive site, largely free of local noise sources and also on scrupulous attention to detail when it comes to decoupling the feedline (going to the "noisy" chassis of the receiver) from the antenna to prevent unwanted signals from being conveyed - but that's a topic of its own!

See the article A (semi)-typical suburban E-field whip receive system for the 630 and 2200 meter amateur bands - link. for a few details on how this might be done.

Real-world observations

At the Northern Utah WebSDR - where there are, at the time of writing, full-time WSPR receivers - it so-happens that there are currently some KiwiSDR and RX-888 based receivers sharing the exact, same signal path.  The KiwiSDR - which is capacitively coupled (e.g. you can hear "tinny" audio from the receiver tuned to 0 Hz and you apply the source to the antenna connector) has quite good response well into the VLF range.

Compared to the RX-888, the KiwiSDR performs noticeably better on the 2200 meter amateur band (137 kHz) in decoding WSPR and FST4W signals in which the '888 is about 15dB down.  As the '888 based system can't hear the 2200 meter signals as well, this indicates that signal levels feeding the '888 are a bit too low for it to "hear" the noise floor of the antenna system - but it also indicates that, perhaps, a few dB of boost in the signal path may remedy this:  This RX-888 has NOT had its bias-Tee inductor removed - but that's on the "to do" list:  After the bias-Tee inductor is removed I expect that it will perform comparably to the KiwiSDR at 2200 meters and I'll update this web page after having done so.

As the "LF/VLF" antenna system at the Northern Utah WebSDR is separate from that of the HF signal path - being combined in a special filter/amplifier module - boosting only the LF/VLF path would be the most beneficial as it wouldn't compromise HF reception by potentially overloading the A/D converter as would boosting everything.

The Highs

The RX-888's specifications state that it contains a "60 MHz" low-pass filter - but the precise nature of its response is not noted.

Comment about sample rates and aliasing - and the need for additional low-pass filtering

The use of the 60 MHz low-pass filter implies that the designers intended an A/D converter sample rate of more than twice that frequency - and since the RX-888 will happily sample at more than 130 MHz, this fits the need.  Many users do not operate their RX-888 at 130 MHz, however, as their interest does not extend beyond HF and operate it, instead, at around 65 MHz to reduce CPU and power loading.

A bit of warning here:  With a 65 MHz sample rate, the '888 will happily respond to signals above the Nyquist frequency (half of the sample rate, or 32.5 MHz) and these signals - spectrally "inverted" - will naturally appear at lower and lower frequencies as the original source signal's frequency increases.  Since the '888s low-pass filter is set at around 60 MHz, it will do nothing to prevent this:  The far right column of Table 2, below, shows the aliases of the test frequencies.

What this means is that users of the RX-888 using it at a sample rate lower than 130 MHz should be using an outboard low-pass filter.  With a sample rate of 65 MHz, a good-quality 30 MHz Low-Pass filter is strongly recommended and will suppress aliased signals that would otherwise appear above Nyquist.  Such filters may be found online via the usual retailers, but do not overlook an old 30 MHz transmit-type low-pass filter of the sort used to prevent interference to analog TV by an HF transmitter - often found at amateur radio swap meets or on EvilBay for cheap.

 The amplitude response, relative to 30 MHz, is shown below:

Frequency (MHz)	Attenuation (dB)	Alias frequency (MHz) @130 MHz sample rate	Alias frequency (MHz) @65 MHz sample rate
30 (Reference)	0	-	-
40	0.8	-	25
50	3.8	-	15
54	5.5	-	11
60	8.5	-	5
64	10.5	-	1
70	14.1	60	5  (double alias)
75	17.7	55	10  (double alias)
80	21.7	50	15  (double alias)
85	27.1	45	20  (double alias)
90	32.5	40	25  (double alias)
95	37.8	35	30  (double alias)
100	42.8	30	30  (triple alias)
105	48.0	25	25  (triple alias)
110	52.9	20	20  (triple alias)

 Table 2:  Sensitivity response of the RX-888 relative to 30 MHz

Table 2 shows the amplitude response of the RX-888 (Mk2) relative to 30 MHz.  The third and fourth column show the resulting aliased frequencies at sample rates of 130 and 65 MHz, respectively.

"Could I intentionally use aliases to receive higher frequencies than my sample rate would allow?"   After reading this, you might ask yourself "If I operate at a sample rate of 65 MHz, could I intentionally do this to receive spectrally-inverted 6 meter signals between 15 and 11 MHz?"   The answer is yes, you could - and as the chart above shows, they would be only 3.8-5.5dB down from the "real" signals across that same 15-11 MHz range.  Intentionally allowing aliases to occur is often done to allow the detection of signals well above the sample rate.  The caveat here is that one would want to sharply filter the source of the "above Nyquist" frequencies to limit them to the band of interest as well as prevent noise on the aliased frequency (15-11 MHz in this example) by filtering those frequencies as well.   Doing this works just fine as long as proper filtering is done to keep out the "unwanted" signals (at the higher and lower frequencies) along with appropriate amplification make up for losses.   In the example above, the lower part of 6 meters would appear just above the 20 meter band - but if one adjust the sample rate, the alias could be moved farther away from 20 meters and, with proper filtering, one could receive both 6 and 20 meters on the same receiver hardware.

What the above table above also shows is that the 60 MHz low-pass filter isn't very good:  By the time you get to the bottom of the FM broadcast band (88 MHz) we know that the attenuation is only around 32 dB.  Here in North America it's common for an FM broadcast station to have many 10s of kilowatts of ERP which means that if you live anywhere near such a station - even if you are using an antenna that wasn't designed to receive FM broadcast frequencies - you may experience some interference around the alias frequencies noted in Table 2.

No matter the sample rate at which you operate your RX-888, it's recommended that you carefully check for aliased responses of FM transmitters.  If you find them - and even if you don't - I'd recommend a separate FM broadcast band blocking filter be installed to quash ingress from strong signals:  Without it you'll probably get some leakage of moderate-to-strong signals in the 22-42 MHz range (frequency-inverted) if you are running at a 130 MHz sample rate or in the 23-32 MHz range if you are running at a sample rate of 65 MHz.

Figure 2 also demonstrates why - if you operate the '888 with a sample rate of 65 MHz - you should really be using a good 30 MHz low-pass filter with it:  Any signals above 30 MHz - including noise - will be attenuated only to the extent shown in the table and will interfere with the desired 0-30 MHz signals.

* * * * *

Other RX-888 related posts at this site:

• Measuring signal dynamics of the RX-888 -  This page discusses the gain distribution of the RX-888, its apparent sensitivity and steps that one should take to maximize performance when used for simultaneous "all of HF" reception.
• Improving the thermal management of the RX-888 (Mk2) - The internal power dissipation of the RX-888 exceeds its ability to get rid of the heat that it produces, reducing reliability - particularly in environments with elevated temperature.  This page discusses what to do to remedy this.
• Using and external clock with the RX-888 (Mk2) - Although the RX-888's TCXO is pretty good, you may wish to use an external reference to provide very high frequency accuracy and stability - and this page gives advice and warnings about doing so.
• Repairing a dead RX-888 (no A/D converter clocking) - While external clocking of the RX-888 (Mk 2) is desirable, it must be done with a bit of care to protect the circuitry involved.  If you do manage to damage your '888, this page may be helpful in its repair.
  

 

This page stolen from ka7oei.blogspot.com

[END]