Friday, August 31, 2012

Problems with Lithium Iron Phosphate (LiFePO4) Batteries

For an update about what turned out to be happening with these batteries - and one possible solution - see the May 18, 2013 post, "Lithium Iron Phosphate batteries revisited - Equalization of cells" - link

In 2010, About 2 5 years ago over the period of several months, I got three 13 volt, 6+ amp-hour Lithium Iron Phosphate (LiFePO4) packs from for about $95 each.  These packs seemed to be a reasonable alternative to my old standby of portable battery power - the ubiquitous 12 volt, 7 amp-hour sealed lead-acid (SLA) battery, often (mistakenly) called "Gel Cells."

Why switch from SLAs?
The three LiFePO4 battery packs in question.

LiFePO4 packs seemed to be attractive for the following reasons:
  • LiFePO4's were lighter than the same-capacity SLAs - roughly 1/2-2/3 as much weight.
  • Claimed 1000-2000 charge durability for LiFePO4's versus 100-200 or so for SLAs.
  • Claimed 10 year lifetime for the LiFePO4's versus 3-5 years for SLAs and conventional Lithium-Ion packs (even the polymer types.)
  • With all of the above, the relatively high initial cost ($95) of the LiFePO4 batteries that would last 10 years seemed to be reasonably comparable to $15-$30 (when new) on a "per-year" basis with typical 7 amp-hour Lead-Acid packs - and the lighter weight was a plus!
As it turns out the LiFePO4 packs aren't quite as "energy dense" as  "normal" Lithium Ion cells - that is, when cylindrical LiFePO4 cells are assembled in a battery pack it takes about the same amount of space as a lead acid battery of the same capacity - but they weigh much less.  It's also worth remembering that conventional LiIon packs will typically last 3-5 years from the date of manufacture and thus didn't have much longevity advantage in that respect over SLAs.

So, over the period of several months, I ordered three of these 6.2 amp-hour LiFePO4 packs that put out about 13-ish volts over their discharge cycle - slightly higher than SLAs, but still well within the realm of what typical "12 volt" gear will accommodate.

As I typically do with newly-acquired batteries I checked the amp-hour capacity of each of the three battery packs shortly after arrival using my West Mountain Radio Computerized Battery Analyzer at 700 milliamps and found that they were reasonably close to the advertised capacity - that is, around 5.8 amp hours:  Typically such batteries are rated at the "20 hour" rate which would have been about 310 milliamps and the higher rate that I used would reduce the measurement by 10-20% so I was pleased with the results.

At about the same time I acquired some 2 year-old 12 volt, 7 amp-hour lead-acid batteries that had been pulled from UPS service on a routine basis and these were found to have about 6.2-6.5 amp-hour capacity at the same 700 milliamp rate.

In the intervening years I used these batteries (both LiFePO4 and SLA) about equally, running radio equipment and the like and earlier this year I suddenly realized that something was amiss:  The LiFePO4 packs were dropping out far earlier than they should have.

A bit of explanation here:

All rechargeable lithium-ion packs (should!) have built-in circuitry to protect against excess over-discharge, the reason being that if you run a lithium battery down too far an irreversible chemical change occurs and they cannot be safely recharged ever again.  For this reason when a lithium pack runs down too far it will suddenly drop off, the internal circuit disconnecting the battery to protect it.

Lead Acid packs, on the other hand, do not do this:  Their voltage slowly drops down and their effective internal resistance goes up and one eventually realizes that the equipment being powered is no longer working correctly.  (Note:  This ignores longer-term permanent damage from sulfation that will occur if a lead-acid cell remains discharged for a long time.)

As it turns out both Lithium-Ion and Lead-Acid packs are charged in similar ways.  One simply connects a power supply of voltage appropriate for the type of battery pack and let it charge.  Both types of batteries, when discharged, will pull more charge current but this will gradually drop off as the battery approaches full charge and for this reason it's typical for these power supplies to be current-limited as well as be fixed voltage.

A major difference between how one treats Lithium-Ion (including LiFePO4) and Lead-Acid (SLA) batteries appears at the point of full charge:
  • For SLAs one obtains the best lifetime by continuously maintaining them at a constant voltage - typically 13.5-13.8 volts for a "12 volt" lead acid battery
  • Lithium types should not be maintained at the "full charge" voltage after full charge has been achieved.

What happens with Lithium-Ion batteries (including LiFePO4) is that if you maintain the "full charge" voltage its internal chemistry degrades much more rapidly than if you were to fully-charge the battery and then immediately disconnect the source, allowing the voltage to sink down a bit on its own.

What this means is that you will get much better longevity out of a Lithium pack if you do not keep a high-level float charge on it.  In fact, the best longevity of Lithium-type rechargeable batteries can be obtained if you store them in a half-discharged state - provided that you check once in a while to verify that their self-discharge hasn't caused their voltage to go so low that they become damaged from that!

* * *

That is how I treated the LiFePO4 battieries:  I would attach the pack to a 1-amp, regulated 14.2 volt 1.5 amp power supply for 12-18 hours and then disconnect it and then place it on the shelf, possibly topping it off briefly just before using it.  The Lead-Acid batteries, on the other hand, are left connected to a 13.6 volt power supply and allowed to sit there all of the time when not being used.

I was, therefore, chagrined when after just two years the now 4 year-old SLAs were outlasting my LiFePO4 packs.

This observation spawned some further testing, so I put the LiFePO4 packs back on my battery tester I was further distressed to note that those that had originally tested out as having 5.8-6 amp hour capacity were now, at the very most, in the 1.5-2 amp-hour range while the much older SLAs were still in the 5.0+ amp-hour range.

 In the time since I did the testing for this entry, the LiFePO4 packs have continued to degrade at about the same, alarming rate while the old Lead-Acid cells are still holding in, degrading much more slowly.


So, what's the deal?  Why are the 4+ year old SLAs still in better shape than the 2 year old LiFePO4 packs?

I really don't know.  I've attempted to correspond with the sellers of the LiFePO4 batteries ( to find out their "take" on this observation, but I've not heard back from them - too bad since I've had reasonable luck with their customer service in the past...

Perhaps they got a batch of "bad" cells - but since the three LiFePO4 packs were actually purchased several months apart it would seem to me that it's more a problem with manufacture/chemistry of the cells themselves. 

What to do?

At the moment I'm sticking with the old, heavy SLAs since I'm now understandably "gun shy" when it comes to LiFePO4s since the former do seem to be fairly predictable in their longevity and performance - at least when treated properly!


For an update about what turned out to be happening with these batteries - and one possible solution - see the May 18, 2013 post, "Lithium Iron Phosphate batteries revisited - Equalization of cells" - link

Update on battery longevity (June, 2016):

I recently re-tested the three batteries depicted above and found that their capacity ranged between 4.8 and 5.4 amps-hours - this for batteries that were at least six years old.  Based on their capacity when they were new, they have lost somewhere around 20% of their original capacity in that time.

While I'm a bit skeptical that they will make it to the 10 or 20 year mark, it is worth noting that practically any lead-acid battery of this same age would have since been relegated to the recycler!


This page stolen from

Tuesday, August 14, 2012

A more practical capacitor-Powered Flashlight

In an earlier post, "A mechanically-powered capacitor flashlight" I wrote about those cheap LED-based "shake-powered" flashlights that were seen on many an annoying commercial several years ago.

You might recall that their promise was that they would never need batteries and one simply shook them back-and-forth to generate all the power that was needed.  In that same post I also noted that many of these same flashlights actually did contain batteries and that while they still worked if those batteries were removed, it took several minutes of shaking to get any usable light and that it was quite an effort to maintain a useful light output!

At the end of this article, I mentioned a few things that might make such a light more practical and useful, including:
  • A better capacitor.  The cheap flashlight had a rather small (0.22 Farad) capacitor for energy storage - not very much energy, really, approximately 6.6 Joules maximum or less than 1/1000th of what a single AA alkaline cell contains!  Being a standard "super cap" its internal resistance was quite high (10's of ohms) which meant that a large percentage of the energy dumped into it during charging and that extracted from it to run the LED was lost as heat - not much heat, but heat just the same.
  • A switching converter to run the LED.  The LED didn't even begin to light until 2.7-3.0 volts or so appeared across the capacitor and it isn't usefully bright until there is 3.6-4.2 volts available which meant that a significant portion of the energy in the capacitor (all of that at voltages of 3-ish volts and below) was unusable.  A simple switching converter would allow both extraction of that additional energy as well as regulate the LED's current so that its brightness was more consistent over the entire charge range and, in theory, could also be adjusted upwards or downwards as necessary.  The efficacy of trying this with a capacitor of high internal resistance would probably be dubious...
One of the conclusions in this earlier article was that the back-and-forth shaking motion wasn't a very efficient means of generating electricity - both in terms of expended muscle energy (since you have to move and stop the entire mass of your arm!) and compared to a conventional crank-type generator - and it would necessarily be larger and heavier in order to be more efficient.  By using a conventional spinning generator and gearing up rotational speed, one can more-efficiently rotate a smaller magnet faster amongst a larger number of poles with a motion that requires less human effort.  What's more, a crank-type generator is quite "scalable" in its input:  You could crank it fairly gently for a long time or do so vigorously for a shorter time and get roughly comparable results in terms of total energy output - within reason, of course.
Figure 1:
The prototype capacitor-based flashlight using a Maxwell Energy
2600 Farad, 2.5 volt "Boostcap".
Click on the image for a larger version.

What is more likely in most situations is that one actually has a source of power somewhere (an already-charged battery, solar panels, a plug-in power supply, etc.) that can be used to charge the flashlight and that it's unnecessary to actually bring along the means of charging the battery with you.

Such devices are already available in the form of batteries, particularly rechargeables, so having a capacitor-powered rechargable flashlight is more of an intellectual exercise rather than one of practicality, but being practical has not always been much of a deterrent to the experimenting nerd!

Some time ago The Electronic Goldmine in Arizona had a large quantity of  Maxwell BCAP0010 BoostCaps tm* available. These were obtained for just $6 each had a rated capacity of 2600 Farads (yes, that's 2.6 kF or kiloFarads!) at 2.5 volts with a "surge voltage" of 2.8 volts - whatever that means...

Comment:  I noted that at other times they had models that were rated at around 3 kiloFarads at 2.7 volts, but these were sold for far more than $6 each.  Alas, as is the nature of surplus, the supply was limited and they sold out fairly quickly.  Sometimes these types of capacitors will show up elsewhere on the surplus market so if you want some, it would pay to look around!

Compared with the 0.22F capacitor in the original flashlight, these units have 10000+ times larger capacity (albeit lower voltage) and very low internal resistance - in the milliohm area - as their intended use was to provide a large burst of current for a short time, say, on an electric vehicle.

To demonstrate, I charged one of these capacitors to 2.5 volts and then I carefully shorted out the terminals with a length of #14 AWG bare copper wire, holding it in pliers.  Within a second or so the current from the capacitor had burned this wire open and in so-doing, it only lost about 0.1-0.15 volts!  For these particular capacitors the maximum rated current is on the order of 600 amps so I have no doubt that I could have repeated the same trick (not recommended!) with larger gauge wire!

What this means is that resistive losses of this type of capacitor (e.g. a "Boostcap") are negligible when it comes to its being charged by a power source and then being discharged by an LED.  As an example, let's assume that we need to draw 100 milliamps to run our hypothetical LED circuit from two different types of capacitors:
  • A standard "supercap" with an internal resistance of 10 ohms - an nice, round value, typical of these types of capacitor.
  • A "boostcap" power system with an internal resistance of 100 milliohms - that value being mostly that of thin wires connecting to the capacitor:  The capacitor itself would likely have an internal resistance a fraction of this!
If we take the formula:  P = I^2 R (that is, power equals the square of the current multiplied by the resistance) with the resistance values above and assuming an LED current of 100 milliamps - and ignoring other losses we get:
  • A loss of 100 milliwatts from a standard super cap.
  • A loss of 1 milliwatt from the "boostcap" and its connecting wires.
Now, if that LED were running from, say, 2 volts at 100 milliamps, the total LED power in each case would be 200 milliwatts - but you can see that the super cap would be losing 100 milliwatts of that in heat while the boost cap would be losing just 1 milliwatt - a considerable difference!  (This assumes that we are somehow ignoring the power loss of the resistance when we are running our LED...)

Clearly, the use of a boostcap offers superior efficiency when discharging, but it also works in reverse:  One could dump many amps into the capacitor (if you used thicker connecting wire) and charge it very quickly and efficiently.

We still have the problem of running the LED, however.  The boostcap capacitors that I obtained were designed to be charged to just 2.5 volts or so and this is too low to run a standard white LED, which needs 3.6-4.2 volts just to light up brightly, so an electronic boost circuit is required and this was accomplished using a variation of the ubiquitous "Joule Thief" circuit:
Figure 2:
Schematic of the flashlight.  This diagram includes a "blocking oscillator" (a.k.a. "Joule Thief") and a current sensing circuit.
See the text for recommendations on transistors to use for Q1.
Click on the image for a larger version.

Important Note:
  • This discussion assumes that one is using an LED with a 3.6 volt threshold as is typical for most white and blue LEDs.  LEDs with lower voltages (e.g. typical red or yellow that operate in the 1.6-2.5 volt region) can't be used with this circuit because their operational voltage would be below that of the full-charge voltage of the capacitor and would be immediately destroyed by the current, from the capacitor, through T1.

While there are more efficient circuits out there, there are almost none that are simpler than the Joule Thief and adaptable to parts that might be found in scrounging around the junk box.

What I came up with is the circuit in the diagram.  At it's heart (Q1, T1, R1, LED1)  it is the Joule Thief circuit comprising a "Blocking Oscillator (link)" that, using inductive "kick" from T1, will produce a voltage higher than that of the power supply (our capacitor), sufficient to light the LED.

While the simplest version of the circuit using the aforementioned components did work, it was very bright at the higher capacitor voltage (above, say, 1.8 volts) but it got noticeably dimmer - but still useful - at lower voltages.  Since the intent was to provide a "useful" amount of light I decided that I didn't need "maximum brightness" at the higher voltages and that I'd be happy with a much dimmer - but consistent - brightness at a much wider range of capacitor voltages.  This also had the obvious and beneficial side-effect of allowing a much longer run-time since, overall, the power consumption was reduced to a fairly steady level over the entire voltage range.

To regulate the LED current a simple circuit was added consisting of T2, D1, R2, R3, C2 and Q2.  The way this circuit works is that the AC current through the LED goes through the primary of T2 and is then integrated by D1, R3 and C2 and if this resulting voltage is too high (correlating with higher average LED current) Q2 would conduct, "pinching" off the drive to Q1.

Originally, a circuit consisting simply of a series resistor along with a transistor like Q2 was tried in which the current through the resistor - if it exceeded the 0.6 volts required to turn on the transistor - would be used to turn off the oscillator and regulate it, but this added resistor required that a bit of the LED's current to be lost as heat through it - plus, it just didn't work very well!

Using a simple transformer arrangement to "transduce" the current into voltage reduced the efficiency losses that occurred with a series resistor while still being fairly simple.  Being simple also meant that there was still a fair amount (say, 25% or so) of LED brightness variation between the target 1.1-2.5 volt range, but that was considered to be acceptable for a simple circuit.  This circuit is also somewhat affected by temperature owing to the fact that not only do the various current gains of the transistors change, but so do the threshold voltage of the transistors and D1.

In this circuit there's really only one critical component and that's Q1, an NPN transistor that was specifically designed for use in photoflash inverters and as such it can switch several amps of current with low collector-emitter drop, this rating being several times that of the more ubiquitous 2N3904 or equivalent.  While a standard NPN like the '3904 will work, it will not work very as well and will be much less efficient.  The KSD5041 may be bought from Mouser Electronics, substituted with a 2SC695, an NTE11 or maybe even found on the flash board of a discarded disposable camera.

An even better alternative for Q1 was suggested by Brooke Clarke (a link to one of his web pages analyzing the Joule Thief may be found here) and that is the Zetex ZTX1048A, available through Mouser and Digi-Key for approximately $1 each in small quantities.  This device - like the KSD5041 and 2SC695 - offer increased efficiency by virtue of its very low collector-emitter saturation voltage - an important consideration when one has conflicting needs of both high current and low voltage in a circuit such as this and according to the specification sheets, the '1048 offers the possibility of even lower saturation voltage than the '5041!

Figure 3:
The capacitor flashlight's circuitry.
Click on the image for a larger version.
The two inductors were toroids salvaged from a defunct computer power supply - and even some of the original wire was salvaged!  In this particular power supply - and several others that I have seen - it's common to see several different-sized toroidal inductors and I happened to choose the larger one for T1.

The circuit itself was built "dead bug" - that is, components were hanging in free space, soldered to each other's leads with the entire assembly being "potted" in thermoset ("hot-melt") glue to stabilize the components to prevent shorting and lead breakage.  As can be seen from the pictures a small piece of PVC pipe was used to not only contain the circuit, but also to shield the positive terminal of the capacitor so that it was not possible to accidentally short it out - something that could conceivably start a fire!

The LED itself is a 3-watt Luxeon III Star that I had kicking around but it's not being run at anywhere near its maximum ratings so about any 1-3 watt white LED that you might find would suffice.  While it's not running a watt of power, the converter probably produces too much output for a single, epoxy-cased white LED, but 3-6 identical units in parallel would probably have be fine with the added benefit that they could be aimed so that their built-in lenses could be used to advantage to shape the resulting beam of light.

Originally, I considered putting a lens on the single LED to concentrate the light but I soon realized that without using a special lens designed specifically for this LED I'd end up with less light overall due to optical inefficiencies.  Even with the LED being "bare" its light output is more than enough to be useful, even walking along a mountain trail in the dark, and its beam is broadly cast so that one isn't as subject to the "spotlight effect" of some LED flashlights where you can see only that which is directly in the beam while the surroundings disappear!

To charge the flashlight I set a variable-voltage bench supply at exactly 2.60 volts and then applied it to the connector (not visible in the pictures) which is wired directly across the capacitor.  From a state of complete discharge (0 volts) it will take several hours for a 1 amp bench supply to fully-charge the capacitor!  Whatever you do, do not allow the capacitor's voltage to exceed its maximum ratings or else it may be damaged:  I have no idea what actually happens if you do this, but I wouldn't recommend trying!

It's worth mentioning at this point that my charging method is extremely inefficient since, when using a linear supply, most of the power input would be lost as heat!  A far more efficient (and somewhat more complex) method would be to use a switching converter to provide the capacitor charge current and have its maximum voltage set to 2.60 volts and this would be much preferred in a power-limited situation where one had only battery or solar as the energy source.

* * *

Update - As of the time of this posting (August, 2013) I've used this flashlight for more than a year, now (since August, 2011) - both around the house and at night while hiking in the mountains and in that time I have only charged it once - and it's still going strong.

Additional Update - As of this update (January, 2017) there is still enough remnant of the original charge on the capacitor to power the light to reasonable brightness.  For most of the years this device has been sitting on a shelf, having been used for a while during an extended power failure to find another flashlight.

* * *

While it may sound like this capacitor can store a reasonable amount of energy storage (and it can!) it's worth noting that the total amount of energy stored in one of these capacitors when it is fully-charged (approximately 8200 joules) is in the same ballpark as the amount of energy contained in a single fresh AA alkaline cell!  Anyone who has actually used a reasonably efficient AA-cell powered LED flashlight knows that it's perfectly capable of providing 10's of hours of useful light, so the duration of the single charge thusfar shouldn't be too surprising.  Just for fun, I dug up some typical numbers:
  • For an AA Alkaline cell, given an average of about 1.25 volts and a usable capacity of 2.2 amp/hours at that voltage, this correlates with a energy storage capacity of 9000-9500 joules, depending on load, temperature, end-of-charge voltage, etc.
  • These calculations ignore the fact that some of the energy being stored in the capacitor or battery at low voltage is not usable as the LED's converter circuit will not operate below approximately 0.9 volts and be able to extract energy.  This is arguably a greater factor with the capacitor because by the time an alkaline battery drops below 0.9 volt, it has almost no residual energy (only a few percent, at most) while 10-12% of the original energy remains in the capacitor.

These numbers are a bit misleading since not all of that energy is usable with equal efficiency in each case over the entire voltage/charge range, but it gives a general idea as to the magnitude.

So, does this flashlight actually work?  Yes, it does!

Is this flashlight really practical?  No, not really.

As it turns out the capacitor itself is not only fairly heavy - about 525g (1 pound) - but it is also quite large - 60mm (2-3/8") diameter and 172mm (6-3/4") long - not including the circuity or bolts:  I have fairly large hands and I find myself moving the flashlight from one to the other as I hike along owing to a bit of muscle fatigue from its diameter and weight.  Again, the capacitor itself was $6 from a surplus seller but that was just a fraction of its original cost (perhaps $150-$200) and one could buy an awful lot of AA cells for its original price!

The main advantage of the capacitor is that unlike a battery, it really doesn't have a fixed number of cycles that it will last before wearing out.  Another advantage is that by knowing its voltage, one can precisely gauge how much useful power remains - a tricky proposition with batteries, especially considering that over time, they lose capacity as they age to a degree that isn't easily determined ahead of time.

I suppose that as time goes on capacitor technology will improve and eventually the power/size/weight will approach (and even surpass!) that of conventional battery technology, but until then a flashlight such as this is a bit of a nerdy novelty!

* "BoostCap" is a trademark of Maxwell Technologies 


This page stolen from

Thursday, August 2, 2012

Two repeaters, one frequency (part 2)

In Part One I'd described why it might be advantageous to place multiple repeaters of a linked system on the same frequency.  In short:
  • A single frequency conserves spectrum.
  • Being on the same frequency over the system's coverage area is more convenient to the user as it eliminates the need to try to figure out which frequency might works best for a given area.
  • The whole system is greater than the sum of its parts because of the probability that brief periods of poor coverage may be augmented by another site.
 In the first part only the implementation of the receive portion of the system was discussed in which multiple receivers were used in a voting scheme - that is, a system in which the signals from the various receivers in the system were analyzed and the best one at that instant was sent on all transmitter.

How, then, does one implement multiple transmitters on the same frequency without their clobbering each other?

This comes again to one of the peculiar aspects of Frequency Modulation (FM) mentioned in the first part of this series:  The Capture Effect.  Briefly, this is the tendency for the stronger of two FM signals to override the weaker - and if they are of sufficiently different signal strength, there may not even be evidence of the weaker signal.

As it turns out, for a number of reasons this effect is more obvious on wideband FM as used in broadcast and you may have even observed a different FM station to suddenly "pop in" in an area where there was overlap.  On the narrowband FM used on amateur radio this effect is somewhat less dramatic and "doubling" (two stations inadvertently transmitting at once) is typically detectable by there being a rather obvious squeal and distorted speech behind the stronger station transmitting or, in cases where the signals are almost exactly of equal strength, neither party wins as the two obliterate each other in an unintelligible mess of noise.

What is worth noting in the above example is that the two transmitters involved are:
Figure 1:
Frequency control/crossband repeater unit at Scott's.  There is
an identical unit at the other site - at Farnsworth Peak.
The 10 MHz oven-controlled oscillator is in the upper-left corner
while the standard GE "EC" channel element is in the upper-
right corner.  This unit - like its twin - is hand-wired on glass-
epoxy prototyping board.
Click on the image for a larger version.
  • On different frequencies.  It's likely that the two transmitters that operated at the same time were on slightly different frequencies - even several hundred Hertz apart.  This frequency difference resulted in a heterodyne (squeal) that decreased intelligibility.
  • The two transmitters were definitely not carrying the same audio.
As it turns out if there are two transmitters that are both held to very tight frequency standards (within a few 10's of Hz at most) and they carry exactly the same audio, they tend not to clobber each other to nearly the same degree if they are of similar signal strength. What's more is that these "similar" transmissions seem to bother each other less as the difference in their respective signal strengths become greater.

Again, the system is laid out thusly:
  •  Farnsworth Peak is the "hub" and the audio for all transmitters in the system originates from there.  The audio to the auxiliary sites (such as Scott's) is conveyed via a UHF link and retransmitted on VHF.
  • All audio from all receivers ends up at Farnsworth and the "best" audio is what is transmitted to all sites.
  • The auxiliary sites (such as Scott's) are essentially crossband repeaters:  2 meter audio is received and relayed to Farnworth on UHF where it is voted upon and this audio is transmitted from Farnsworth  on UHF where it is repeated on VHF at the auxiliary sites.
What this means is that at Scott's, there's a box called the "Disciplined Oscillator" that contains a precision, oven-controlled 10 MHz oscillator that is capable of holding the VHF transmit frequency to within 1-2 Hz of where it is intended to be.  As it so-happens, this same box also contains the intelligence to function as a controller for a pair of crossband repeaters that goes from VHF to UHF for signals that are received and then again from UHF to VHF as the master audio from Farnsworth is transmitted.  This box also provides a few other basic functions such as timeout timer (in the event a link gets "stuck") as well as providing a Morse ID on the UHF link from Scott's to Farnsworth - just to keep things legal.  This same box also has an RS-485 serial interface to allow it to be connected on a bus with other devices so that it may be remotely controlled, configured and polled as needed.

 When we originally designed the system we anticipated that we may need to adjust a few parameters in order to successfully have two transmitters operating on the same frequency without their causing objectionable mutual interference.  The first - and most obvious - of these was frequency control.

Because we use independent oven-controlled crystal oscillators, we couldn't nail the frequencies of the transmitters down precisely to match each other as would be possible were we to have used a GPS or Rubidium-based reference, but we could count on their being within 1-2 Hz of where we had parked them.  Once the system was put on the air we solicited the help of someone who happened to live in an area where the strength of the two transmitters was precisely equal and then tweaked the frequency offsets and then made a subjective analysis as to what was "least annoying."

As it turned out, there were two ranges that seemed to be reasonable in terms of frequency offset:
  • 3-6 Hz offset.  This caused a bit of a "whooshing" sound if the two signal strengths were fairly close and fairly weak.  If the signals were exactly the same strength then the periodic nulls could cause it to drop out briefly and make the signal unintelligible, but even a slight reposition of the receive antenna could mitigate this, however.
  • 40-60 Hz offset.  This caused a buzzing somewhat akin to the sound of a subaudible tone as heard on a signal with severe multipath distortion.
Ultimately, we settled for the 3-6 Hz offset as it was deemed to be the most "user friendly" overall - especially when one considered that one was by far more likely to be traveling mobile through the overlap areas than stationary and that the Doppler shift of a moving vehicle might not only exceed the amount of frequency offset anyway (if it was only 3-6 Hz, at least) but that the "dwell" time in a precise null where the signals of multiple transmitters canceled each other out was going to be extremely short.

Another factor often considered in multiple-transmitter systems is that of audio delay to match the time-of-arrival of the different distances between transmitters - plus additional delay in the audio links used to tie the disparate systems together.  Before we were to go through any hassle of adding an audio delay somewhere, we first wanted to see if it was really going to be a problem in the overlap areas, anyway.

It wasn't.

The only thing that we did do was observe the audio phase at and below 1 kHz and then, using the ability to select either a 0 or 180 degree audio source, set them as close as we could.

So, what does it sound like in the overlap areas?

First of all, the coverage of the sites and their geography meant that about the only significant overlap areas were in canyons to the east of the Salt Lake area where signals from either transmitter would already be subject to multipath, anyway.  As it turns out, traversing these area it's rather difficult to tell where the coverage of one transmitter begins and the other ends - and it often goes both ways.  In those area that do have severe overlap the contention between the two transmitters sounds little different than typical mobile flutter - perhaps slightly "faster" than typical 2-meter flutter but not as fast as what might be heard on a 70cm repeater in an area with severe multipath!

In the next part, a bit of "nerdy" technical information about how the various parts work...


This page stolen from