The construction of a hybrid ring combiner: Using the same duplexer, feedline and antenna for two repeaters/links/transmitters

Two transmitters, one antenna

There are occasions when it is desirable to combine two transmitters - and receivers - onto a single antenna - perhaps at a busy repeater site where more than one link or repeater is required.

The most obvious answer to this would be to have a separate antenna - and duplexer - for each of the links - but there are potential problems with this:

• Antennas are expensive
• Feedline is expensive
  
• Duplexers are even more expensive
• If you are on a shared site, you may have difficulty installing - or even getting permission to install - another antenna - and if you are renting tower space, this "other" antenna will be an ongoing expense.
  

If you have the luxury of planning ahead and you have a bit if extra link margin ^{Footnote 1} (e.g. you have "excess" signal and can afford a bit of loss), by placing two repeaters and/or link radios on adjacent - or nearly adjacent - frequencies, you can probably use a single antenna, feedline and duplexer - potentially saving money and hassle in the long run.  Let's take as an example as a hypothetical pair of full-duplex links at a single site in the U.S. on the 70cm amateur band:

• Transmit frequencies:  421.000 and 421.100 MHz
• Receiver frequencies:   434.500 and 434.600 MHz

As these pairs of frequencies are only 100 kHz apart from each other, they will handily fall into each others' notches in their duplexers meaning that neither transmitter is likely to bother either receiver.  What this means is that you could use a single duplexer along with a bit of extra gear (more on that later) to put both radios on the same antenna.  This same technique could also be applied if, for some reason, you had two repeaters at the same site (say an analog and a digital) as will be mentioned a bit later.

Note:  While this article describes usage on the 70cm band, there's no reason why it could not be applied to other bands - higher and lower - as well.

First, let's briefly cover using a single antenna to feed more than one receiver.

Receive:  Splitting the receive signal path

Receiving is pretty easy:  Just use a splitter.  If you are using a 2-way splitter this will result in 3.0-3.5 dB loss on receive, and there is - in most cases - likely to be enough link margin so that this won't be a problem, but if not, placing a preamplifier of modest gain (say, 10dB) in front of the splitter will overcome these losses:  Avoid the temptation to use a higher-gain amplifier than this to minimize the probability that it might make the problem even worse due to overloading (desensing) and producing intermodulation products that will cause other interference issues.

In a pinch, a 2-way "Cable TV" type of splitter - with appropriate adapters - will work fine for 2 meter and 70cm even though they are designed for 75 ohm system:  They will have reasonably low loss (less than 4dB for a good-quality unit) and a port-to-port isolation of 20dB or better.  For more information see the article on this blog:  Using TV (F-connector) 75 ohm splitters and taps in 50 ohm systems on the amateur HF, VHF and UHF bands (link).

Methods of combining closely-spaced verses more-distantly spaced transmitters   When combining transmitters, there are two general situations that arise:  Those that are very closely-spaced in terms of frequency, and those that are spaced farther away from each other. For two transmitters whose frequencies are separated by a reasonable distance (say, 5-10% of the frequency) it is usually practical to use a combination of notch cavities and phasing lines to separate them from each other - not unlike the way a duplexer allows both transmit and receive at the same time on a given antenna:  There's no reason why one couldn't have two transmitters on a duplexer.  If done carefully, such combination can incur additional losses of only a couple of dB. The problem changes if the two transmitters are quite close in frequency:  Eventually, the separation is so small that it is not possible to use resonant cavities to separate them from each other:  For 2 meters, this spacing is about 400kHz or narrower while on 70cm a spacing of less than 1.5-2 MHz starts to become more difficult.  As the spacing gets narrower and narrower, losses go up and in the case of combining just two transmitters, if the losses exceed about 3 dB, it may be better to simply use a hybrid combiner like those described here rather than potentially large, expensive notch cavities.  If there are more than just two transmitters to be combined, everything gets even more complicated and careful system design and frequency planning are of paramount importance to minimize losses.

You can also find on the surplus market (e.g. EvilBay) splitters - typically made by Mini-Circuits - that are native 50 ohm.  It's worth noting that Mini-Circuits tends to rate their products very conservatively and even a splitter/combiner rated for "only" 200 or 250 MHz at the top end will work just fine at 450 MHz.  One device frequently found on the surplus market is the Mini-Circuits ZFSC-2-1(+) which is rated for 5-500 MHz and is suitable for all repeater bands 10 meters through 70cm.

Transmit:  Combining multiple transmitters onto a single feedline

This is a bit trickier.  

If your frequencies are very near each other (within 1-2 MHz at 70cm, closer than 500 kHz at 2 meters) it's not likely to be practical to use complicated band-pass/notch schemes to isolate the two transmitters with less than about 3 dB of loss.  Again, we are presuming that your link budget will tolerate an additional 3dB of loss - and in most cases in the "real" world, this is likely to be true.

The "easiest" way to combine two transmitters - whether the frequencies are close to each other or not - is with a hybrid combiner.  If the frequencies are quite close together - as in our example above - then the same duplexer, feedline and antenna can be used.

Note:  On 2 meters, frequencies closer than 100 kHz can likely be combined and mutually filtered with a single duplexer/band-pass cavity while on 70cm you may be able to get away with up to 500 kHz of separation - and both may require a bit of careful tuning of the duplexers/band-pass cavities and - at the wider spacings noted above, the possible need to tolerate a bit of extra loss or reduction of isolation.  Clearly, the closer the better!

Wilkinson Power Combiner

There are two common types of combiners that would be suitable for this, both of which having the property of being useful only "near-ish" their design frequency - say, less than +/- 10% or so.  The simplest of these is the Wilkinson type, depicted in Figure 1.  This circuit is frequently used in power amplifiers when there are multiple gain stages in parallel that are phased identically, being fed from the same source and combined to get more output.^{See Footnote 6}  However, if you are combining different transmitters at different frequencies you will theoretically lose half the power - but there's really no way around this.  Again, it's quite likely that a loss of half of the transmit power (3 dB) at the antenna will have little actual effect on the coverage of the transmitter/link in a real-world situation.

Figure 1:
  Wilkinson combiner diagram (from Wikipedia)
In a 50 ohm system, this uses 72 ohm line and a
100 ohm resistor between P2 and P3.

How it works

Ports P2 and P3 represent the connections to the two transmitters and P1 is the combined output.  As can be seen, there is a total of 1/2λ (half  wavelength) between P2 and P3:  Any signal from P2 will arrive at P3 180 degrees out of phase and vice-versa, the ultimate result being that the signal from P3 is cancelled out at P2 and the signal from P2 is cancelled out at P3, effectively isolating the two transmitters from each other - which is important, as we'll later see.

Note that there a resistor between P2 and P3 which has a resistance of twice the system impedance - or in the case of a 50 ohm system, it should be 100 ohms.  Additionally, half of the TOTAL transmit power will be dissipated by this resistor if the signal sources are dissimilar (e.g. NOT being used to combine two parallel stages in an RF amplifier).  Finding a 100 ohm resistor suitable for 70cm that is also capable of handling 10s of watts of RF with a good return loss is tricky - and then placing it across two terminals that are "hot" with RF - makes the Wilkinson less attractive in this application.

Note that this device requires transmission line that is √2 * Zo - or in the case of a 50 ohm system, it's 50 * 1.414 = 70.7 (e.g. 71 ohms).  For our purposes, "75 ohm" cable like RG-59 or RG-11 would be fine if we were to construct one but most implementations at UHF use stripline on circuit boards.

Note that the Wilkinson may also be used to divide (split) power equally to two loads should that be required.  This is most commonly done in a power amplifier where two, separate amplifier stages need to be fed with the same signal after which another combiner would be used to sum the outputs again. 

Hybrid ring combiner

Another type combiner that will fit the bill is the so-called "Hybrid Ring" (a.k.a. "Rat Race") combiner, depicted in Figure 2.

Figure 2: 
Hybrid Ring Combiner (from Wikipedia)
a.k.a. "Rat Race" combiner.  Like the Wilkinson,
this uses 72 ohm feedline in a 50 ohm system, but
the terminator (usually at P4) is 50 ohms.

Referring to Figure 2, our two transmitters are connected to ports P1 and P3 while output P2 contains the sum of the P1 and P3 signals and P4 contains the difference of the P1 and P3 signals.  All ports must be sourced/terminated at 50 ohms and we would typically connect 50 ohm transmitters at P1 and P3 and 50 ohm loads (antennas or dummy loads) at P2 and P4 as appropriate.  

If we look carefully, we can see that between P1 and P3 (via P2) there is 1/2λ of feedline between the two if we go clockwise around the circle - but if we go the other way around the circle we have 1/4λ plus 3/4λ - or a total of 1λ - which which means that there is 1/2λ difference in feedline length between them - thus a 180 degree phase difference - any RF being input to P3 are cancelled out at P1 and vice-versa.  With P2 being halfway between P1 and P3 - and going the other way we have 1.5λ of feedline between the same points and no cancellation we get the sum of the two transmitters.  Like the Wilkinson, you could also use the Hybrid ring to split power between two loads:  To do this the transmitter (source) would be applied to P2 and the two loads would be connected at P1 and P3.

If we were combining two in-phase RF amplifiers from the same transmitter - as we might if we had two identically-phased amplifiers that we wanted to combine for more power, we would get zero power at P4 and the combination of the two amplifiers would appear at P2:  A dummy load - connected at P4 - would see little or no power if the two two amplifiers were operating with equal power and phase.

In our case, we are combining two different transmitters - on different frequencies - and this means that half of our power will end up each on P2 and P4, so we would typically connect our dummy load at P4 and our output to our antenna (via the duplexer) at P2.

Figure 3:
Andrews FSJ1-75 75 ohm Heliax - unfortnately
no longer being made, but other types of 75 Ohm
cable may be used as well as described in the text.
Click on the image for a larger version.

Like the Wilkinson power combiner, the ring combiner requires transmission line that is √2 * Zo (e.g. 71 ohms) - and again, 75 ohm cable is fine for most purposes in a 50 ohm system.  A distinct advantage of the ring combiner is that unlike the Wilkinson, the dummy load in the ring combiner is ground-referenced rather than having a resistor across two "hot" RF terminals - simplifying design.  Additionally, because the load that we would connect to P4 is the same as our system impedance - 50 ohms - finding a suitable device on the new or surplus market is quite easy.

Either the hybrid-ring or the Wilkinson combiners should yield well over 25dB of isolation between ports when properly constructed over a wide frequency range (8-10% or more) - and values of over 50dB are attainable on the workbench with known-good loads and sources across narrower frequency ranges.  Having good isolation between ports is vital when combining transmitters:  Not only does good isolation imply lower loss  (e.g. closer to the theoretical 3dB) but in the case of two different transmitters, minimizing the amount of energy that transmitter "A" gets from transmitter "B" reduces "wasted" power due to this cross-coupling, but more importantly it reduces the likelihood of the generation of IMD (Intermodulation Distortion) products that could cause receive performance degradation and interference - more on ways to prevent this will be discussed later.

Again, the reader is reminded that both the Wilkinson and Hybrid Ring combiners - unlike their counterparts that use transformers - are not inherently broadband:  They are typically used within less than +/-10% or so of their design frequency for best performance as this is the limit of efficacy of the quarter-wave phasing sections.  Later in this article we'll analyze the constructed hybrid ring combiner and demonstrate its frequency range.

Choice of transmission line

As noted, the impedance of 75 ohm feedline is "good enough" for these two types of combiners on a 50 ohm system in most cases and this sort of coaxial cable is readily available:  RG-59 and RG-11 types are suitable - as are some RG-6 cables such as Belden 1694A that have tinned-copper shields and solid copper center conductors to which we can easily solder - but note that most RG-6 cables use aluminum shields and copper-covered steel center conductors:  In theory these would work, but making a good connection to the shield is a complication.

NOTE:  There are PTFE (Teflon) 75 ohm coaxial cables available - see Footnote 2 at the bottom of this article for more information.

Figure 4: 
A hybrid-ring combiner, constructed as a "ring" using
FSJ1-75 Heliax and "N" type connectors on pigtails with bits
of brass tubing at the joints.
Click on the image for a larger version.

While I had all three of the typical 75 ohm cable types available to me, I found - while rummaging around - some Andrews FSJ1-75 1/4" Heliax ^tm coax.  Unfortunately, this cable hasn't been made since 2011, but it may still be found on the surplus market occasionally - and the data sheet for it is still online (link).  This cable is lower loss than any of the other options mentioned - but this is probably not important as such short lengths of it are used:  Reactance losses due to the construction itself are likely to be greater than the cable losses, anyway - and I would have gotten very similar results with the other choices.  The use of the smaller cable also implies a smaller bending radius - which will be important as we'll soon see.

In theory, one could construct a hybrid ring like this using RG-11-type coaxial cable, N-type connectors and known-good ^{See Footnote 3} N-type "tee" connectors - but you will have to carefully calculate the added lengths of the connectors and adapters when doing so.

Form factor

There are several ways that this ring combiner could be built.  The most obvious is in the form of an actual ring as depicted in Figure 4 - a combiner that I built over 25 years ago, also using the same Andrews FSJ1-75.  This uses short pieces of brass tubing to connect the segments together and holes in the sides of the tubing allow solder connections to be made between the segments and to the short "pigtails" with "N" connectors on them:  It measures better than 35 dB TX port isolation at its intended frequency.   This particular combiner had been used to combine a two UHF repeaters - an FM voice repeater and a 9600 baud packet repeater, on frequencies just 25 kHz apart - onto the same duplexer/antenna for several years with excellent results using the methods described below. 

Figure 5: 
For the ring combiner on this page, it needed to be a bit more
compact and rugged, housed in a Hammond 1590D box.
Click on the image for a larger version.

For the combiner that I recently built there was the need for something more "compact":  The "ring" described above (in Figure 4) is about 10" (25cm) in diameter and the connectors are awkwardly spaced, on flexible leads - and the entire thing is a bit fragile.

 

For 70cm, the feedline lengths are short enough that the sections can be put into a large-ish die-cast box - particularly if the cable used has a fairly tight allowed bending radius, which would rule out RG-11.

 

In this case I had on hand a Hammond 1590D which is 7.4" (18.8cm) x 4.7" (11.95cm) x 2.2" (5.6cm).  I also happened (as one does) a number of chassis mount "N" connectors with short lengths of UT-141 (hardline) already connected to them which would allow (literally!) some flexibility as to how the internal phasing sections of feedline could be oriented and connected.  It's worth nothing that if RG-179 were used (see Footnote #2) a smaller-still metal box could be used - even for a 2 meter combiner!

 

Running the numbers

First, calculate the wavelength at the desired center frequency.  In our case, we need 421.0 MHz:

300 / F (MHz) = Wavelength

300 / 421.0 = 0.713 Meters - The wavelength at 421.0 MHz.

Figure 6:
Preparing and measuring the cable pieces.  For the ring
combiner there is one 3/4λ and three 1/4λ pieces.  They are
intentionally cut slightly long to allow stripping and soldering.
Click on the image for a larger version.

Now, we need to find the velocity factor of our particular cable, a number which indicates the tendency for electricity to move slower than light when it is conveyed through conductors in the presence of a dielectric.  For best results, find the manufacturer of the specific cable that you are using as this varies - particularly between solid and foam dielectric cables and the precise type of dielectric.  For our FSJ1-75, the stated velocity factor is 0.78 (78%).  Knowing this, we can calculate the length of one electrical wavelength of our cable:

Vf * Length = Electrical length  So we plug in the numbers:

0.713 * 0.78 = 0.556 Meters

Figure 7: 
The stripped/prepped end of a cable segment as described in
the text.  The calculated lengths are measured between the
ends of the dielectric, where the center conductor protrudes.
Click on the image for a larger version.

As we can see from the diagram of Figure 2, we need four pieces of cable:  One that is 3/4λ (0.75λ) and three more that are 1/4λ (0.25λ), so we multiply the above to get those:

0.556 * 0.75 = 0.417 Meters for 3/4λ section  - approx. 16-3/8"

0.556 * 0.25 = 0.139 Meters for the 1/4λ sections  - approx. 5-1/2"

Prepping the cable

As we need a bit of extra cable to expose the center conductor to which we solder we need to make the sections slightly longer - about 1/2" (13mm) at each end - or about 1" (25mm) longer overall for each piece.  After cutting the pieces of cable, put a mark at the center of each and measure about 1/2" (1cm) less than half the length that we calculated above to the end strip back the outer jacket:  This will leave a section of bare shield to which we can later solder.  Now remove the shield at a point about 1/4" (0.5cm) less than half the length that we calculated above and remove the dielectric, but leaving a couple of mm (about 1/16") out from the end of the shield:  And example may be seen in Figure 7.

Note that when we calculate the length of the cable - 0.139 meters for a 1/4λ section in our example - we are measuring at the points where the center conductor protrudes from dielectric.  As can be seen in Figure 8, the center conductors are then bent over level with the top of the dielectric to be soldered to to the other cable segments and the distances are measured from the points where they are bent over.

Figure 8:
The junction of the three cables, with the center conductors
bent over and soldered - and with the shields firmly soldered
together as well, using 26 AWG bare wire during assembly.
Click on the image for a larger version.

At this point it would be a good idea to consult the specifications of the cable that you are using and determine the minimum bending radius.  For FSJ1-75, this is specified as being 1.25" (31.75mm) and for Belden 1694A this is about 2.75" (70mm):  Try to bend it less than this if you can:  Cut a piece of cardboard or paper with a circle of this radius as a comparison.

To get an idea as to how everything should be routed inside the box, the connectors+cables were first installed and tightened and the four pieces were laid out and moved about to determine what made sense.  As can be seen from Figures 8-11, each of the cables were bent at a gentle right angle so that the center conductors all came together at one point and the shields in parallel with each other - all without bending the cables at too-tight a radius.

Once the orientation of a cable end was determined, the center conductors were bent over - leaving a gap of about 1/16" (2mm) between it and the shield and the connections tacked together.  After this, some tinned 26 AWG wire was wound tightly around the shield and then soldered, making a both a very short-length electrical connection and providing mechanical rigidity as can be seen in Figure 8.

Figure 9: 
A better view of the cables being assembled in the box,
showing how the 26 AWG bare wire first used to tie the
springy cables together and then flooded with solder.
Click on the image for a larger version.

A very hot soldering iron is a must here as it allows connections to be made very quickly and prevent melting of the dielectric.

 

If you have it, use tin-lead solder (e.g. 60/40 or 63/37) as it melts at a lower temperature than "lead free" solder and is less likely to melt and damage the cable's dielectric. Additionally, a few drops of either "no-clean" or rosin flux will help make good, clean joints with minimal heat.

As it turns out, the FSJ1-75 is relatively forgiving in terms of heat if you use a hot iron and work quickly - but if you use something like 1694A or other flexible coaxial cable, you may want to tin the braid prior to assembly - starting with a few bits of "practice" coaxial cable to avoid melting pieces that you have already cut to length and prepared.

Figure 10:
The cable segments were originally fitted in the box and
tacked together as seen in Figure 9, but were carefully removed
so that the connections could be fully soldered on both sides.
Click on the image for a larger version.

Testing

A NanoVNA is a good tool to test the combiner - but you will also need TWO known-good 50 ohm terminators (dummy loads).  For this, you probably won't want to use anything but good-quality units - which are available surplus - and you will want to use an ohmmeter to verify that they are in the range of 50-52 ohms - and don't forget to take into account the resistance of your ohmmeter leads!

Before proceeding, be sure to do an "SOLT" (Short-Open-Load-Through) calibration of your VNA for the frequency range of interest.  In our case, it's 400-500 MHz. 

As you will be measuring "through" loss, I suggest that for measuring this device that you use the "receive" (second) port of your NanoVNA instead of the termination when doing the SOLT calibration as it will likely not be quite as good as the load that came with the NanoVNA and would otherwise make VSWR/return loss measurements "appear" to be worse than they are:  If you are interested in single-port (S11) measurements only, use your SOLT load for calibration and either it or a known-good load when testing..

Referring again to Figure 2, note which of the four connectors correlate with P1, P2, P3 and P4 (mark them with a pen or label) and connect the dummy loads to P2 and P4.  Then, connect the "In" and "Out" leads of the NanoVNA (or similar) to P1 and P3 - it doesn't matter which way.

Figure 11: 
Initial testing of the combiner, showing the NanoVNA
connected to ports P1 and P3 and the dummy loads on ports
 P2 and P4.
Click on the image for a larger version.

If all goes well, the VSWR should be below 1.25:1 and the "through loss" (which is the isolation between the two transmitters to which P1 and P3 will be connected) should be well above 25 dB.  At some frequency - probably a bit above the intended design frequency you may see the through loss dramatically increase (a "dip" in amplitude) of port-to-port isolation between P1 and P3.  If you do not see significantly more than 25 dB of isolation between P1 and P3, re-check your connections:  If you see way under 20dB you have probably misidentified your connections and should check again - but if you think that you have properly identified everything, re-check your math from when you calculated the cable lengths - and don't forget to include the velocity factor.

 

As we'll see later, a device like this should yield in excess of 30dB transmit-transmit port isolation over a very wide frequency range - even without any "adjustments".  In Figure 11 you can see the initial testing of the ring combiner and in the background on the NanoVNA - on the blue trace - the "dip" showing the frequency of the best port-to-port isolation - which, before tuning, was near 435 MHz.

Having verified that the isolation between P1 and P3 is good, connect one of the NanoVNA leads to P2 and move the dummy load onto the to which the NanoVNA had been connected:  If the unit is working properly, the insertion loss will be about 3dB, which is exactly correct.

Figure 12: 
A "tuning strip" used to make slight adjustments.  The
"grounded" strip is moved closer to the exposed center
conductors to increase capacitance.
Click on the image for a larger version.

Making adjustments

While the "ring" depicted in Figure 4 is elegantly simple, it can't be adjusted:  Scrunching it into the box as shown in Figures 10 and 11 allow a bit of tweaking to optimize both isolation and matching by virtue of the exposed center conductors.

When I built the combiner pictured I noted that the best insertion loss between P1 and P3 appeared as a "dip" around 435 MHz with the lowest VSWR occurring around 440 MHz.  I observed that very lightly touching my finger at the junction of P4 caused this "dip" to shift down in frequency, indicating that a very small amount of capacitance might better things at our 421.0 MHz design frequency.

Rather than connect a small variable capacitor - which would need to be of very low capacitance values (possibly less than 1pF) I did something simpler:  I cut a small strip of brass sheet and soldered one end to the shield of the coax at the junction of the cables as depicted in Figure 12.  This strip was then bent around the exposed junction - but kept at 1/8-1/16" (2-3mm) away from the center conductors to avoid shorting:  This added a bit of capacitance at that point and shifted the "dip" down in frequency, and by adjusting this brass strip closer/farther away from the center conductors of the cables, I was able to "dial" it in at 421 MHz.  If you can't get the strip close enough to the center conductor to bring the frequency down, solder a smaller strip of metal to the center conductor to form a larger capacitor plate, allowing more capacitance with greater distance.

Figure 13:
Inside the combiner - tuning strips installed on P1 and P4
which are used to slightly tweak the performance at the
intended operating frequency.
Click on the image for a larger version.

Similarly, I noted that lightly touching the connection at P1 reduced the VSWR slightly (which had started at around 1.2:1) and a similar brass strip was installed there.  The two interacted slightly with each other and I was able to get both a lower VSWR (under 1.1:1) and very good isolation (more than 50dB) at 421 MHz.  Note that when connected to the "real world" (e.g. actual transmitters and an antenna) which will probably have some reflected power and will NOT likely have precisely 50 ohms of impedance, the isolation will likely be less than what we measure on the workbench.  The lack of a really low VSWR (e.g. higher return loss) anywhere across the frequency range can likely be attributed to the fact that we used nominally 75 ohm feedline for the construction of the combiner rather than "70.7 Ohm" feedline - which is difficult to find in the form of coaxial cable!

With a bit of finessing I was able to get the port-to-port isolation and VSWR shown in the plot below:

Figure 14:
Port-to-port isolation (blue line) and VSWR (green) with a Smith chart in the background
and a legend in the upper-right corner showing the readings at the marked frequencies on the plots.
There is well over 50dB of isolation between P1 and P3 at the intended frequency.
Click on the image for a larger version.

As we can see from Figure 14, the port-to-port isolation at our target frequency is quite good - in excess of 50 dB - while the VSWR is around 1.1:1 as noted in the upper-left corner.  It's worth noting that the port-to-port isolation is better than 30dB between 400 and 445 MHz - a span of about 10% - and it's better than 20 dB from somewhere below 400 MHz to 491 MHz - and no-where on this plot does the VSWR exceed 1.4:1.  What this means is that if your goal is 30dB isolation, the design is quite forgiving.  As can be seen from the Smith chart in the center, the matching is pretty well-behaved.

What about insertion loss?  This plot gives us the answer:

Figure 15: 
TX port to antenna insertion loss (blue) and VSWR (green) - again with a Smith chart in the
background and the legend in the upper-left corner.
This plot shows just 0.05dB above the theoretical at the 421 MHz design frequency and a
good VSWR as well.
Click on the image for a larger version.

This graph shows the measured insertion loss and at our target frequency of 421 MHz, it is around 3.05 dB - very close to the theoretical 3dB loss and getting very near the uncertainty of our measurement.  Over the same "30dB port-to-port isolation" frequency range that we measured above (400-445 MHz) we see that the insertion loss is lower than 3.2 dB and that the VSWR over this range never exceeds 1.2:1.  Even up at 491 MHz where we had only about 20dB port-to-port isolation the insertion loss is a bit over 4 dB with a VSWR of 1.4:1 - not great, but still usable in a pinch.

Putting it into practice

Figure 16:
A two-stage UHF isolator with loads - a Ma-COM 7R011,
owned by the author, designed to work at UHF ^{Footnote 5}.
Although the two loads are only rated for 12 watts, the
 unit itself can handle 125 watts if the load on the left
(nearest the "output") were sized accordingly.
Click on the image for a larger version.

Being able to bash two transmitters on "nearby" frequencies together is a good thing - but we do need more than just the combiner to do so effectively and "cleanly".  (Note:  If the transmitters' frequencies were spaced farther apart, a different scheme would be required - see Footnote #4). 

 

The entire point of having a ring combiner with good port-to-port isolation is to minimize losses - here, we limit them to about 3dB - but it also prevents the RF output power of one transmitter from getting into another where mixing can occur, producing low-level IMD (intermodulation distortion) products that could cause interference to your own gear and - more importantly - other users on site.  As the port-to-port isolation is likely to diminish with "imperfect" sources and loads, we need to do more to prevent these mixing products from occurring than just the combiner.

To minimize the probability of this occurring it is also a very good idea to install a ferrite isolator on the output of each transmitter, prior to the ring combiner as depicted in Figure 18.  These devices - one of which is pictured in Figure 16 - can be through of as a sort of "diode" for RF:  They will let transmit power go through them, toward the antenna, but any power that comes the other way - reflected due to VSWR or RF energy from another transmitter - will end up in its dummy load(s) - effectively preventing RF from getting back into either of the final amplifiers:  Figure 17 shows what a properly-functioning isolator does to block RF coming back from the "load" port - in this case it reduces that energy by around 60dB at the frequency to which it is adjusted.

Figure 17:
The reverse (isolation) plot of the 2-stage isolator of Figure
16.   At the frequency to which it is tuned (450 MHz)
it has over 60 dB of reverse isolation. In the other direction
(not shown in this plot) its loss is only about 1dB.
Note that +/-25 MHz from center, the isolation drops below
30dB which is why a pass cavity is suggested!
Click on the image for a larger version.

These isolators (which are "circulators" with included "dump" loads) generally come in two flavors:  Single stage and double stage - the latter having two devices in the same package like the one in Figure 16.  Typically yielding 20-30 dB of reverse isolation per stage, a double stage device typically has between 40 and 60 db reverse isolation as shown in Figure 17.  This plot also shows something else:  While the reverse isolation is very good at its tuned frequency, it decreases as you move away:  If you have other users on site - and you are not using a pass cavity between the isolator and antenna - these "off-frequency" signals will not be as well-attenuated by the isolator as at the design frequency and its efficacy will be reduced while the probability of IMD being generated in the transmitter's output will increase.

 

While isolators can be quite expensive when new, they are frequently found on the surplus market.  Most isolators are tunable, and this will need to be done to optimize performance on your operating frequency - but this can easily be accomplished with a NanoVNA and instructions as to how to do this may be found on the Repeater Builder web site (link) as well as on videos on YouTube.  Even the extra expense of isolators and a band-pass cavity may well be less expensive than and preferable to the installation of two separate antennas, feedline and duplexers - particularly if another antenna were to incur recurring costs such as maintenance and tower rent!

 

If your combiner had a rather "average" 30dB of port-to-port isolation and you were using a two-stage isolator on each of your transmitters this would mean the RF energy from one transmitter getting into the other would be down between 70 and 90dB - and this nearly guarantees that IMD products will be extremely low and likely undetectable.

To further reduce the probably of harmful IMD product escaping your system, a band-pass cavity should be inserted between the output of the combiner and the "transmit" port of the duplexer as depicted in Figure 18:  As noted in another article on this blog (See the article:  When Band-Pass/Band-Reject (Bp/Br) duplexers really aren't band-pass - link) almost all duplexers used in amateur repeater service will NOT offer much filtering once you move a few MHz away from their tuned frequency.

Figure 18: 
A typical application using a ring combiner, along with isolators and a band-pass cavity in the TX leg.
The isolator on TX #1 - and the ring combiner - minimize the amount of energy that it "sees" from
TX #2 and vice-versa while the band-pass cavity limits off-frequency energy that can enter the system.
Click on the image for a larger version.

 

By including a band-pass cavity in the "transmit" leg of the duplexer - prior to the duplexer - energy that is away from the operating frequency will be attenuated significantly.  This is important as many isolators have a rather limited frequency range over which they are most effective as seen in the figure above.  Additionally, if there are very low-level IMD products produced by mixing within your two transmitters even with isolators and the port-to-port isolation of the combiner, these will be significantly quashed by the band-pass cavity, practically eliminating even the possibility of self-generated interference.

 

Figure 19:
The completed UHF ring combiner.
The labeling on the sides of the box
identifying the ports is not visible in this photo.
Click on the image for a larger version.

Conclusion:

With a bit of planning and foresight is is possible to combine multiple transmitters - or even full duplex link radios or repeaters - onto a single antenna.  The techniques described here are most useful if you are able to do frequency planning by placing the transmit/receive frequency pairs quite close to each other, permitting the use of a common duplexer and single band-pass cavity in the TX and another in the RX leg.  Even if the frequencies are separated a bit and a common duplexer is not possible, these techniques can still be adapted.

 

In most cases, the extra 3dB of loss on receive and transmit can be ignored as such a reduction in sensitivity radiated power is likely to be unnoticed - and it well may be worth the trade-off in terms of minimizing infrastructure and even cost.

 

* * * * *

Footnotes:

1. Link margin refers to the amount of signal between a transmitter and receiver - and specifically, the degree to which it could degrade and still produce acceptable results.  Consider the following:  Let us suppose that your goal is to have a link with 20dB SINAD or better and your receiver was capable of producing a signal of this quality with -110dBm at its input terminals.  If the strength of your receive signal from the other end of the link was -90dBm, it could be reduced - via fading or other path degradation - by 20dB before it would be considered to be "faded".  If - in the process of putting two radio systems on the same antenna you were to reduce either the transmitter at the far end or the receiver at the near end by about 3dB, in our example we would still be left with about 17dB of "fade margin".
   
2. RG-179 and RG-302 - both being 75 ohm available with PTFE (Teflon) and similarly-resistant jacket - would be excellent choices in the construction of a ring combiner owing to their flexibility and heat resistance.  RG-179 is available from  both Mouser Electronics - link -and Digi-Key Electronics link at the time of writing and while rather expensive, it doesn't take a lot of cable to construct a ring combiner.  While the PTFE is preferred, it's also available with other types of dielectric/jacket which will work, but care would be required to avoid melting it during soldering.  The PTFE cables' velocity factors are typically around 69% while the polyethylene versions are around 79% - but always check the manufacturer's data sheet!
   
3. If you ever use "Tee" type adapters, be extremely careful what you get - particularly if you are using type "N".  Many "foreign-made" N-type connectors are very poorly built - both mechanically and in terms of RF - and some of them having been found to use springs to make contact (highly inductive - VERY BAD!) and/or compression-type connections that tend to oxidize.  If you can afford it, such "Tee" connectors from Pasternak (link) will likely be good - as are genuine old mil-spec devices from reputable manufacturers (Amphenol, Kings, etc.).  If you don't know their quality, be prepared to measure them carefully using a VNA - and better yet, buy an extra so that you can cut it apart and see for yourself if it resembles anything like a "constant impedance" device with solid, reliable construction.
4. Note that since in our example the two transmitters' frequencies are very close together (only a few hundred kHz at 70cm) there would be little point in putting a band-pass cavity between the output of the isolator and ring combiner, but if the transmitters were several MHz apart - a spacing at which the cavities would offer reasonable rejection - it might make sense to do so.  If one did have several MHz spacing, it's less likely that the band-pass cavity in Figure 18 would be appropriate and that one could get away with using a single duplexer, either. 
5. The two-stage UHF isolator in Figure 16 was bought by the author "as is" - with the pair of 12 watt loads - at a swap meet for about $20 - the low price being due to the fact that it had clearly been submerged in water for a while and was showing some corrosion.  It was very carefully being disassembled and thoroughly cleaned - which included washing the trimmer capacitors with denatured alcohol to remove any contaminants - and resoldering the internal connections as it was reassembled, it once again worked, more than meeting the manufacturer's specifications for both reverse isolation and forward insertion loss.
6. The Wilkinson divider/combiner is often seen in high-power amplifiers - both to split drive power to identical amplifier stages, and to combine them again after amplification - and in this case, the same signal in terms of frequency, phase and power is being combined/split.  In cases like this where the inputs and outputs are operating at identical power levels (e.g. combining two, signals or putting two identical signals together) the 100 ohm load across P2 and P3 in Figure 1 sees NO (or very little) voltage difference and dissipates little/no power.  If, however, the two signal paths become different (e.g. one of the two amplifiers fails) this resistor will then dissipate a significant percentage of the remaining signal power.  Anyone who has had to repair a power amplifier that uses such a combiner/splitter arrangement knows that in addition to needing to repair the amplifier stages - and make sure that their outputs are phased equally (there's often a capacitor to do this) that equalizing resistor will have been "smoked" (destroyed) almost instantly when one of the amplifier sections failed as they typically use a resistor that is rated for a fraction of the power output:  If everything is fine, the resistor survives - but if not...
   

* * * * *

This page stolen from ka7oei.blogspot.com

[END]

An RV "Generator Start Battery" regulator/controller for use with a LiFePO4 power system

I was recently retrofitting my brother's RV's electrical system with LiFePO₄ batteries (ReLi³on RB-100's).  This retrofit was done to allow much greater "run time" at higher power loads and to increase the amount of energy storage for the solar electric system while not adding much weight, not needing to vent corrosive fumes.  (These types of batteries, LiFePO₄,  are very safe - e.g. they don't burst into flame if damaged or abused.)

While I was doing this, I began to wonder what to do about the generator "start" battery.

Charging LiFePO4 batteries in an RV The voltage requirements for "12 volt" Lead-Acid batteries are a bit different from those needed by LiFePO4 "12 volt" batteries: Lead acid batteries need to be kept at 13.2-13.6 volts as much as possible to prolong their life (e.g. maintained at "full charge" to prevent sulfation). LiFePO4  batteries may be floated anywhere between 12.0 and their "full charge" voltage of around 14.6 volts - but they will not be fully recharged unless they are held near the high end of this range. Routinely discharging lead-acid batteries below 50% can impact their longevity - and they must be recharged immediately to prevent long-term damage. LiFePO4  batteries may be discharged to at least 90% routinely - and they may be left there, provided their voltage is not allowed to go too low. Lead acid batteries may be used without any management hardware:  Maintaining a proper voltage is enough to ensure a reasonable lifetime. LiFePO4 batteries must have some sort of battery management hardware to protect against overcharge and over-discharge as well as to assure proper cell equalization.  Many modern LiFePO4 batteries (such as the "Rel3ion" devices used here) have such devices built in. Conventional RV power "converters" are designed to apply the proper voltage to maintain lead-acid batteries (e.g. maintain at 13.6 volts.) Because LiFePO4 batteries require as much as 14.6 volts to attain 100% charge (a reasonable charge may be obtained at "only" 14.2 volts) connecting them directly to an existing RV with this lower voltage means that they may never be fully-charged!  Modern, programmable chargers (e.g. inverter-chargers, solar charge controllers) have either "LiFePO4 " modes or "custom" settings that may be configured to accommodate  the needs of LiFePO4 batteries.  While the lower voltage (nominal 13.6 volts) will not hurt the LiFePO4 batteries, they likely cannot be charged to more than 40-75% of their rated capacity at that voltage.  (approx. 13.6-13.7 volts is the lowest voltage were one can "mostly" charge a LiFePO4 battery.) Because of Peukert's law, one can only expect 25-50% of the capacity of a lead-acid battery to be available at high amperage (e.g. 0.5C or higher) loads. With LiFePO4 batteries, more than 80% of the battery's capacity can be expected to be available at similar, high-amperage.  What this means is that at such high loads, a LiFePO4 battery can supply about twice the overall power when compared with a lead-acid battery of the same amp-hour rating.  At low-current loads the two types of batteries are more similar in their available capacity. In short:  Unless an existing charging system can be "tweaked" for different voltages and charging conditions, one designed for lead-acid batteries may not work well for LiFePO4 batteries.  In some cases it may be possible to set certain "equalize" and "absorption" charge cycle parameters to make them useful with LiFePO4s, but doing this is beyond the scope of this article.

Originally the RV had been equipped with two "Group 24" deep-cycle/start 12 volt batteries in parallel (a maximum of, perhaps, 100 amp-hours, total, when brand new, for the pair of "no-name" batteries supplied) to run things like lights, and the pump motors for the water system, jacks and slide-outs and as the "start" battery for the generator.  Ultimately we decided to wire everything but the generator starter to the main LiFePO₄ battery bank.

Why?

Suppose that one is boondocking (e.g. "camping" away from any source of commercial power) and the LiFePO₄ battery bank is inadvertently run down. As they are designed to do, LiFePO₄ battery systems will unceremoniously disconnect themselves from the load when their charge is depleted to prevent permanent damage, automatically resetting once charging begins.
 
If that were to happen - and the generator's starter was connected to the LiFePO₄ system - how would one start the generator?

Aside from backing up the towing vehicle (if available), connecting its umbilical and using it to charge the system just enough to be able to get the generator started, one would be "stuck", unable to recharge the battery.  What's worse is that even if solar power is available, many charge controllers will go offline if they "see" that the battery is at zero volts (e.g. when they are in that "disconnected" state) - even if the sun is shining, preventing charging from even starting in the first place!

What we needed was a device that would allow the starting battery be be charged from the main battery, but prevent it from back-feeding and being discharged.


Note:

It is common in many RVs for the generator to not charge its own starting battery directly, via an alternator.  The reason for this is that it is assumed by the makers of the generators and RVs that the starting battery will be charged by the towing vehicle and/or via the RV's electrical system via its AC-powered "voltage converter", powered from "shore" power or via the generator's AC output.

But first, a few weasel words:

• Attempt to construct/wire any of the circuits only if you are thoroughly familiar with electronics and construction techniques.
• While the voltages involved are low, there is still some risk of dangerous electric shock.
• With battery-based systems extremely high currents can present themselves - perhaps hundreds or even thousands of amps - should a fault occur.  It is up to the would-be builder/installer of the circuits described on this page - or anyone doing any RV/vehicle wiring - to properly size conductors for the expected currents and provide appropriate fusing/current limiting wherever and whenever needed.  If you are not familiar with such things, please seek the help of someone who is familiar before doing any wiring/modifications/connections!
• This information is presented in good faith and I do not claim to be an expert on the subject of RV power systems, solar power systems, battery charging or anything else.
• You must do due diligence to determine if the information presented here is appropriate for your situation and purpose.
• YOU are solely responsible for any action, damage, loss or injury that might occur.  You have been warned! 

Why a "battery isolator" can't be used:

If you are familiar with such things you might already be saying "A device like this already exists - it's called a 'battery isolator'" - and you'd be mostly right - but we can't really use one of these devices because LiFePO₄ batteries operate at a full-charge voltage of between 14.2 and 14.6 volts, and the battery isolator would pass this voltage through, unchanged.  If you apply 14+ volts to a "12 volt" lead-acid battery for more than a few days, you will likely boil the away electrolyte and ruin it!

What is needed is a device that will:

• Charge the generator start battery from the main (LiFePO₄ ) battery system
• Isolate it from the main battery, and 
• Regulate the voltage down to something that the lead-acid chemistry can take - say, somewhere around 13.2-13.6 volts.

In this case the main LiFePO₄ battery bank will be maintained via the AC-powered (generator or shore) charging system and/or the solar power converters at its normal float voltage, so it makes sense to use it to keep the start battery fully-charged.

The solution:

After perusing the GoogleWeb I determined that there was no ready-made, off-the-shelf device that would do the trick, so I considered some alternatives that I could construct myself.

Note:  The described solutions are appropriate only where the main LiFePO₄ bank's voltage is just a bit higher (a few volts) than the lead-acid starting battery:  They are NOT appropriate for cases where a main battery bank of a much higher voltage (e.g. 24, 48 volts, etc.) is being used to charge a "12 volt" starting battery.

Simplest:  "Dropper diodes":

Because we need to get from the nominal 14.2-14.6 volts of the LiFePO₄ system down to 13.2-13.7 volts it is possible to use just two silicon diodes in series, each contributing around 0.6 volts drop (for a total drop of "about" 1.2 volts) to charge the starting battery, as depicted in Figure 1, below.  By virtue of the diodes' allowing current flow in just one direction, this circuit would also offer isolation, preventing the generator's battery from being discharged by back-feeding into the main battery.

To avoid needing to use some very large (50-100 amp) diodes and heavy wire to handle the current flow that would occur when the starter motor was active - or if the start battery was charging heavily - one simply inserts some series resistance to limit the current to a few amps.  Even though this would slow the charging rate somewhat, the starting battery would be fully recharged within a few hours or days at most - not a problem considering the rather intermittent use of the starting battery - more about that later.

Figure 1.
This circuit uses a conventional tungsten-filament "1157" tail/turn signal bulb (NOT an LED replacement!) with both filaments tied together, providing more versatile current limiting.  Please read notes in the text concerning mounting of the light bulb.
The diodes (D1 and D2) should be "normal" silicon diodes rather than "Shottky" types as it is the 0.6 volt voltage drop per diode that we need to reduce the voltage from the LiFePO₄ stack to something "safe" for lead-acid chemistry.  If one wished to "tweak" the voltage on the starting battery, one could eliminate one diode or even replace just one of them with a Shottky diode to increase the lead-acid voltage by around 0.2-0.3 volts.
The use of a current-limiting device like a tungsten light bulb allows lighter-gauge wire to be used to connect the two battery systems together.
Click on the image for a larger version.

In lieu of a large power resistor, the ubiquitous "1157" turn signal/brake bulb is used as depicted in Figure 1.  Both filaments are tied together (the bulb's bayonet base being the common tie point) providing a "cold filament" resistance of 0.25-0.5 ohms or so, increasing to 4-6 ohms if a full 12 volts were placed across it.  The reason for the use of a light bulb will be discussed later.

Although not depicted in Figure 1, common sense dictates that appropriate fusing is required on one or both of the wires, particularly if one or more of the connecting wires is quite long, in which case the fuse would be placed at the "battery" end (either LiFePO₄ or starting battery) of the wire(s) to provide protection should a fault occur between that source and the charge controller:  Fusing at 5-10 amps is fine for the circuit depicted.

This circuit is "good enough" for average use and as long as the LiFePO₄ bank is floated at 14.2 volts with occasional absorption peaks at 14.6 volts, the lead-acid starting battery will have a reasonably long life.

A regulator/limiter circuit:

As I'm wont to do, I decided against the super simple "dropper diode and light bulb" circuit - although it would have worked fine - instead, designing a slightly fancier circuit to do about the same as the above circuit, but have more precise voltage regulation.  While more sophisticated than two diodes and a light bulb, the circuit need not be terribly complicated as seen in Figure 2, below:

Figure 2:
The schematic diagram of the slightly more complicated version that provides tight voltage regulation for the starting battery.  As noted on the diagram, appropriate fusing of the input/output leads should be applied!
This diagram depicts a common ground shared between the main LiFePO₄ battery bank and the starting battery, usually via the chassis or "star ground" connection.  In the as-built prototype, Q2 was an SUP75P03-07 P-channel power MOSFET while D1 was an MR750 5 amp, 50 volt diode. A circuit board is not available at this time.
NOT SHOWN is the fusing of the input and output leads, near-ish their respective batteries/source connections, with 10 amp automotive fuses.
Click on the image for a larger version.

How it works:

U1 is the ubiquitous TL431 "programmable Zener".  If the "reference" terminal (connected to the wiper of R5) of this device goes above 2.5 volts, its cathode voltage gets dragged down toward the anode voltage (e.g. the device turns "on").  Because R4, R5 and R6 form an voltage divider, adjustable using 10-turn trimmer potentiometer R5, the desired battery float voltage may be scaled down to the 2.5 volt threshold required by U1.

If the battery voltage is below the pre-set threshold (e.g. U1 is "seeing" less than 2.5 volts through the R4/R5/R6 voltage divider) U1 will be turned off and its cathode will be pulled up by R2.  When this happens Q1 is biased on, pulling the gate of P-channel FET Q2 toward ground, turning it on, allowing current to flow from the LiFePO₄ system, through diode D1 and light bulb "Bulb1" and into the starting battery.

By placing R1 and R2 on the "source" side of FET Q2, the circuit is guaranteed to have two potential sources of power:  From the main LiFePO₄ system, through D1, and from the starting battery via the "backwards" intrinsic diode inside Q2.  The 15 volt Zener diode (D2) protects the FET's gate from voltage transients that can occur on the electrical system.

Figure 3:
The completed circuit, not including the light bulb, wired on a small
piece of perforated prototype board.
A printed circuit board version is not available at this time.
Click on the image for a larger version.

Once the starting battery has attained and exceeded the desired float voltage set by R5 (typically around 13.5 volts for a "12 volt" lead-acid battery) U1's reference input "sees" more than 2.5 volts and turns on, pulling its cathode to ground.  When this happens the voltage at the base of Q1 drops, turning it off and allowing Q2's gate voltage, pulled up to its source by R1, to go high, turning it off and terminating the charge.

Because the cathode-anode voltage across U1 when it is "on" is between 1 and 2 volts it is necessary to put an additional voltage drop in the emitter lead of Q1, hence the presence of LED1 which offsets it by 1.8-2.1 volts.  Without the constant voltage drop caused by this LED, Q1 would always stay "on" regardless of the state of U1.  Capacitor C1, connected between the "reference" and the cathode pins of U1 prevent instability and oscillation.

In actuality this circuit linearly "regulates" the voltage to the value set by R5 via closed loop feedback rather than simply switching on and off to maintain the voltage.  What this means is that between Q2 and the light bulb, the voltage will remain constant at the setting of R5, provided that the input voltage from the LiFePO₄ system is at least one "diode drop" (approx. 0.6 volts) above that voltage.  For example, if the output voltage is set to 13.50 volts via R5, this output will remain at that voltage, provided that the input voltage is 14.1 volts (e.g. 13.5 volts plus the 0.6 volts drop of diode D1) or higher.

Because Q2, even when off, will have a current path from the starting battery to the main LiFePO₄ bank due it its intrinsic diode, D1 is required to provide isolation between the higher-voltage LiFePO₄ "main" battery bank and the starting battery to prevent a current back-feed.  Were this isolation not included, if the main battery bank were to be over-discharged, current would flow backwards, through FET Q2, from the generator starting battery and discharge it, possibly to the point where the generator could not be started.

Again, D1's 0.6 volt (nominal) drop is inconsequential provided that the LiFePO₄ bank is at least 0.6 volts above that of the starting battery, but this will occur very frequently if the charge on that bank is properly maintained via generator, solar or shore power charging.  A similar (>= 5 amp) Shottky diode could have been used for D1 to provide a lower (0.2-0.4 volt) drop, but a silicon diode was chosen because it was on hand.

Testing the device:

Assuming that it is wired/built correctly, connect a variable power supply to the input lead to simulate the LiFePO₄ battery bank.  Setting the voltage a volt or two higher than the expected float voltage (e.g. 14.5-16 volts) adjust R5 to attain the desired start battery float voltage (13.50-13.7 volts is recommended - I use 13.55 volts) as measured on either side of "Bulb1".  Adjust the power supply voltage up and down a bit (e.g. below 12 volts and up to 17 volts) and if working correctly, the output voltage from the circuit should be rock-steady as long as the input voltage is about 0.6 volts above the set output voltage.

Now short the output leads (e.g. the "positive" output lead should be going through "Bulb1") and the light bulb should illuminate fully - assuming that your variable voltage supply is capable of supplying the 3-ish amps needed for the lamp.  Measuring directly at the circuit board's "ground" (common "battery negative") terminal and at the connection between Q2 and "Bulb1" you should still have the voltage set by R5 within a few hundredths of a volt.

Note:  If you were to measure connect the negative lead of the voltmeter to the power supply or the shorted output leads the measured voltage would be a bit lower owing to voltage drop along the wires.

Shorting the output leads and measuring the voltage as done in the previous step demonstrates two important design points:

• That the voltage at the output of Q2 remains steady from no-load to maximum current conditions.
• That the light bulb is properly acting as a current limiting device.

While doing this "short circuit" test, make sure that the heat from the light bulb rises away from the circuit board itself and that the means of mounting it is capable of withstanding the bulb's heat without burning or melting anything.

Connecting the device:

On the diagram only a single "Battery negative" connection is shown and this connection is to be made only at the starting battery.  Because this circuit is intended specifically to charge the starting battery, both the positive and negative connections should be made directly to it as that is really the only place where we should be measuring its voltage!

Also noted on the diagram is the assumption that both the "main" (LiFePO₄ ) battery and the starting battery share a common ground, typically via a common chassis ("star") ground point which is how the negative side of the starting battery ultimately gets connected to the negative side of the main LiFePO₄ bank:  It would be rare to find an RV with two battery systems of similar voltages where this was not the case!

Finally, it should go without saying that appropriate fusing be included on the input/output leads that are located "close-ish" to the battery/voltage sources themselves in case one of the leads - or the circuit itself - faults to ground:  Standard automotive ATO-type "blade" fuses in the range of 5-10 amps should suffice.  In order to safely handle the fusing current and to minimize voltage drop while charging the connecting wires to this circuit should be in the range of 10 to 16 AWG with 12-14 AWG being ideal.

What's with the light bulb?

Figure 4:
The circuit  board mounted in an aluminum chassis box along with the
light bulb.  Transistor Q2 is heat-sinked to the box via insulating hardware
and the board mounted using 4-40 screws and aluminum stand-offs.  The light
bulb is mounted to a small terminal lug strips using 16 AWG wire soldered
to the bulb's base and the bottom pins:  A large "blob" of silicone (RTV)
was later added around the terminal strip to provide additional support.
Both the bottom of the box (left side) and the top include holes to allow
the movement of air to help dissipate heat.  Holes were drilled in the back
of the box (after the picture was taken) to allow mounting.
This box is, in this picture, laying on its side:  The light bulb would be
mounted UP so that its heat would rise away from the circuitry via
thermal convection.
Click on the image for a larger version.

The main reason for using a light bulb on the output is to limit the current to a reasonable value via its filament.  When cold, the parallel resistance of the two filaments of the 1157 turn-signal bulb is 0.25-0.5 ohms, but when it is "hot" (e.g. lit to full brilliance) it is 4-6 ohms.  Making use of this property is an easy, "low tech" way to provide both current limiting and circuit protection and, when the filament is cold (e.g. charging battery "mostly" charged), increase the amount of charging current that can flow.  Taking advantage of this changing resistance of a light bulb allows higher charging current that would be practical with an ordinary resistor.


In normal operation the light bulb will not glow - even at relatively high charging current:  It is only if the starting battery were to be deeply discharged and/or failed catastrophically (e.g. shorted out) that the bulb would begin to glow at all and actually dissipate heat.  

Limiting the charging current to just a few amps also allows the use of small-ish (e.g. 5 amp) diodes and reduce the heat that could be dissipated during regulation allowing the use of an aluminum box as Q2's heat sink, but more importantly it allows much thinner and easier-to-manage wire (as small as 16 AWG) to be used since the current can never be very high in normal operation.  Limiting the charging current is just fine for the starting battery due to its very occasional use:  It would take only an hour or two with a charge current to top off the battery after having started a generator on a cold day!

As noted on the diagram and in previous text the light bulb must be mounted such that its operating temperature and heat dissipation at full brilliance will not burn or melt any nearby materials as the glass envelope of the bulb can will easily exceed the boiling temperature of water!  With both the "simple" diode version in Figure 1 and the more complex version in Figure 2 it is recommended that the bulb is mounted above the circuitry to take advantage of air convection to keep the components cool as shown in Figure 4.  If a socket is available for the 1157 bulb, by all means use it, but still heed the warnings about possible amount of heat being produced.

In operation:

When this circuit was first installed, the starting battery was around 12.5 volts after having sat for a week or two (during the retrofit work) without a charging source and having started the generator a half-dozen times.  With the LiFePO₄ battery bank varying between 13.0 and 14.6 volts with normal solar-related charge/discharge cycles, it took about 2 days for the start battery to work its way up to 13.2 volts, at which point it was nearly fully charged - and then the voltage quickly shot up to the 13.55 volts as set by R5.  This rather leisurely charge was mostly a result of the LiFePO₄ bank spending only brief periods above 13.8 volts.

Even though this doesn't very quickly charge the battery under normal conditions, as we'll see below, this isn't really important.

How much of the starting battery's capacity is being used?

If one were to assume that the generator was set to run once per day and pull 100 amps (a current likely seen on a very cold day!) from the battery for 5 seconds this would represent (100 amps * ( 5 sec/3600sec )) = about 0.14 amp-hours: - This happens to be about the same amount of energy as is contained in 4 fresh hearing-aid batteries or about 1/10th of the capacity of a single AAA cell!

From this we can see that this "100 amps for 5 seconds" is an average current of just over 5 milliamps (1/200th of an amp!) when spread across 24 hours - a value likely comparable the self-discharge rate of the battery itself.   By these numbers you can see that it does not take much current at all to sustain a healthy battery that is used only for starting!  Because this battery is never used for running things like lights or motors, it really never gets abused by being deeply discharged.

A standard group 24 "deep cycle starting" battery was used since it and its box had come with the RV.  In this particular application, for generator starting only, a much smaller battery - such as one used for starting 4x4s or motorcycles - would have sufficed and saved a bit of weight and space.

The advantage of the group 24 battery is that it, itself, isn't particularly heavy and it is readily available in auto-parts, RV and "big box" stores everywhere.  Because it is used only for starting the generator, it need not have been a "deep cycle" type, but rather a normal "car" battery - although the use of something other than an RV-type battery would have necessitated re-working the battery connections as RV batteries have handy nut/bolt posts to which connections may be easily made.


Final comments:


There are a few things that this simple circuit will not do, including "equalize" the lead acid battery and compensate for temperature - but this isn't terribly important, overall in this application.


Concerning equalization:

Even if the battery is of the type that can be equalized (many sealed batteries, including "AGM" types - those mistakenly called "gel cells" - should never be equalized!) it should be remembered that it is not the lack of equalization that usually kills batteries, but rather neglect:  Allowing them to sit for any significant length of time without keeping them floated to above 2.17 volts/cell (e.g. above 13.0 volts for a "12 volt" battery) or, if they are the sort that need to be "watered" and not keeping their electrolyte levels maintained.  Failure to do either of these will surely result in irreversible damage to the battery over time.

It is also common practice to adapt the float voltage to the ambient temperature, but even this is not necessary as long as a "reasonable" float voltage is maintained - preferably one where water loss is minimized over the entire expected temperature range.  Again, it is more likely to be failure of elementary battery maintenance that will kill a battery prematurely than a minor detail such as this.

Practically speaking, if one "only" maintains a proper float voltage and keeps them "watered" the starting battery will likely last for at least the 3-5 year expected lifetime, particularly since, unlike battery in standard RV service, this starting battery will never be subjected to the deep discharge cycles which can really take a toll on a lead-acid battery.  While an inexpensive, no-name "group 24" battery, when new, may have a capacity of "about" 50 amp-hours, it won't be until the battery has badly degraded - probably to the 5-10 amp-hour range - where one will begin to notice starting difficulties.

Important also is the fact that the starting battery in this RV is connected to part of the main LiFePO₄'s battery monitoring system (in this case, a Bogart Engineering TM-2030-RV).  While this system's main purpose is to keep track of the amount of energy going into and out of the main LiFePO₄ battery, it also has a "Battery #2" input connection where one can check the starting battery's voltage - always a good thing to do at least once every day or two when one is "out and about".

Finally, considering the very modest requirements for a battery that is used only for starting the generator, it would take only a very small (1-5 watt) solar panel (plus regulator!) to maintain it.  While this was considered, it would have required that such a solar panel be mounted, wires run from it to the battery (not always easy to do on an RV!) and everything be waterproofed.  Because the connections to the main battery bank were already nearby, it was pretty easy to use this circuit, instead.

[End]

This page was stolen from "ka7oei.blogspot.com"