Thursday, February 19, 2009

IMD??? What IMD

The theme song from the Soprano's goes, "woke up this morning got a blue moon, got a blue moon in your eyes." When the lead singer in A3 sings it however it sounds like "woke up this morning got a boom boom, got a boom boom in your eye", and that kind of fits with the Soprano's theme. It also fits with this theme, because the data I'm about to present, though not definitive should expose precisely how the ARRL testing methodology has held back Ham radio. Hopefully it will put a boom boom in that eye.

The ARRL test originally put 2 strong signals in the passband of a receiver. The idea was to increase the amplitued of the signals until distortion occurred. From that you could graphically determine the imaginary IP3 of the radio. The test was jiggered form the start. They measured from S-5 to distortion, which tended to over state the IP3. Most radios of that era were up conversion radios, that is the 1st mixer ran with synthesizer way above 30 mhz, and that was mixed with the HF spectrum and gave a continuous representation of the1.8-30mhz spectrum that was presented to the IF filters. The 1st IF was then down converted again to around 455khz and then demodulated to audio. This gave a nice way to shift the 2 filters against each other giving a kind of continuous band pass tuning. The first mixer was in the 45 mhz range.

Ten Tec however designed the Omni-6 with the first IF in the 9mhz region. The radio was known for its strong signal handling and its clean VFO because it did not have to contend with phase noise from the synthesizer, crudding up the pass band. The engineers at TT got a wild idea. Instead of just one wide filter in the 9mhz region they expanded that to multiple filters. This was the birth of the "roofing filter" It was basically impossible to design a filter at 45 mhz that could do the job of 8 or 10 poles at 9 mhz. This meant the TT radio became even stronger. It basically was a radio that was custom made to pass the ARRL test. If the test spacing was 20khz you would have an incredibly high IP3 compared to radios that overloaded in the upper IF region, because they did not have the advantage of the "roofing filter". The name of that radio was the Orion. TT banged the heck out of that in their advertising and I can't say as I blame them. The Orion is basically the Omni-6 taken to its end. It is part Omni-6 and part digital radio. I won't go as far as to call it an SDR, at least in the Flex radio sense, but I will call it a FDR. Firmware defined radio. It had some flexibility in its design but to get new functionality you had to download proprietary firmware and flash the radio. This worked OK until the processor inside the radio ran out of memory. Basically at that point the radio was fixed in stone.

It still was and is a very strong radio however. What the Omni-6 lacked in features the Orion delivered, AND it still had that roofing filter that gave it super strong numbers on the ARRL test. Best in the business at the time. Next came the Orion-II which was the Orion with little more memory in the processor and narrower roofing filters. Then came the K-3. The Orion suffered from some quirkiness, and Elecraft set out to do the Orion right. They have the basic roofing filter design of the Omni-6 plus a much better derivation of the FDR, and the radio is modular so you can make it as basic or deluxe as the feature set allows, keeping with Elecraft's peddling of the old Heathkit model of selling kit radios that gave the "builder" the experience of being some kind of wild and woolly home brewer. What a wonderful marketing ploy. I used to read the Elecraft reflector and it was amazing how many posts I saw "Its Alive" as some builder completed his K-2 following the directions and lit it up and it worked. That is not to take away from their fun, its just to note if you build a kit according to the instruction it will probably work. The new radio is more complex and Elecraft made this radio modular so there is still that kind of "feel" of being wild and woolly. The K-3 has some really good numbers on the ARRL testing methodology. It has good numbers because it has roofing filters down to 200hz, and once again if you put one signal in the pass band and one signal out of the pass band, that knocks the mixing products way down not much mixing will occur. So this radio was not only made for the test, but the test was made for it. A marriage made in ham radio heaven.

So the question, what happens when you put 2 signals INSIDE the pass band? If you read my previous posting regarding the 160 contest you will recall I was working signals 40hz apart with multiple other signals of S9+ all within maybe 150hz of those 2 signals. In other words all the signals were INSIDE the pass band where the "roofing filter" couldn't jigger the results.

Here are some picture of some testing that was done and published on the CQHAM forum site. CQHAM is a Russian ham radio magazine or web-zine or what ever. I can't vouch for exactly how these tests were run, but from what I can tell looking at the pictures, 4 modern high buck radios were tested. What was looked at was the linearity of the in-band filter response with no signal, and how the radios respond to 2 strong signals INSIDE the pass band, not protected by the "roofing filter" The 2 pics with the signals appear to be one with the preamp on and one with the preamp off.

Here is the pic of the passband:

IMD-passband no signal

Note the brick wall filters exhibited by the F5K, and the lack of mixing products. Note the mixing products in the K-3 pass band even without signals present. My guess is these signals were moved up and down on the screen of the analyzer so the differences in the pass bands and stop bands could be easily discerned.

IMD-2 close spaced strong signals no preamp

One pic is worth a thousand words. The F5K has virtually no mixing even with S9+50 signals only 300 hz apart, and the pass band remains flat. There is no added energy or non linearity that gives the pass band a "bowed" appearance. This is a good example of why the F5K is so quiet on RX.

IMD-preamp on

OMG!!!! the bottom has dropped out of the K-3, often hawked as MBR (man's best radio!!!!) and now you see the fallacy of the ARRL testing methodology. If you stuck that s9+50 signal OUTSIDE the roofing filter you would probably have something closer to the first picture, but when 2 signals are competing INSIDE the pass band, the RX folds up. Look also at the F5K. You can work stations all the way down to MDS (minimum discernible signal) which is on the order of -140 dBm. This is my experience of the SDR-1000 vs. the Orion, and why the Orion went in the ditch when it was subjected to multiple S9++ signals competing with puny weak noise hugging DX signals in my contest experience. Note also how flat the pass band is in the F5K. I believe this is largely due to crud in the radio from the non linearity of the various multiple hardware stages in the legacy radios. The F5K on the other had is direct conversion and does all its magic in software via equations and not components, and as such the F5K represents a crud free zone.

My experience of the F5K is like this as well. I can get right next to any CW station, and as long as that station does not violate my pass band with clix or some other garbage, I have no trouble copying signals at the noise. 20 or 30 hz separation is more than enough. I have never experienced any blocking on the radio. It does not mean it can't be done but at least I haven't had the experience. I can often filter out stations that call on the transmit freq of the DX in a pileup, and eliminate them completely with the filtering in the F5K. I can't get rid of them if they are zero beat, but if they are 30 or 40hz down the road they are toast. I can see the offending signal popping up strong as heck on the panadapter but they are not readable in the audio pass band.

So there you go. I don't have the exact testing conditions etc etc (I don't read Russian) but you can figure pretty much what is going on just by looking at the pictures. One thing for sure it does put the boom boom in the eye of the ARRL testing methodology, and this game played with roofing filters. If you can get F5K kind of in band performance in this simple kind of test, why does the ARRL not test like this? To me the answer is pretty obvious, but I'll leave you to make your own conclusions.

One thing of note however. If you "can" do F5K kind of performance is it this "test" that is holding back advancement of the state of the art? Is what is going on simply ham radio manufacturer after manufacturer making yet another iteration of a radio that is not designed for best performance but designed so it can claim bragging rights with high test results?

I used to teach Physics at the college level. Many of my students were the "pre-med" crowd. It was my distinct impression they were not interested a whit in physics, but they just wanted me to teach them to the test, because "they were going to med school". The reason it was my distinct impression is because I was told that more than once. Teaching to the test has left a bad taste in my mouth ever since. Teaching to the test NEVER will advance the state of knowledge, the same as jiggering the Ham radio tests will never advance the state of ham radio.