I respectfully disagree with the main point of your premise, to paraphrase: "The uBitx is ok because it has less spurious emissions than are allowed for a commercial amateur transceiver operating at 100 watts or a transceiver/amplifier operating at 1500 watts". There are two things wrong with this premise, 1) Modern commercial transceivers are specified to have emissions at least 50 dB below carrier level; many are specified at 60 dB or more. Such transceivers have spurious emissions far below those of the uBitx even when operated at 100 or 200 watts. 2) Unfortunately that is not how the rules are written or applied by the respective governing bodies. The allowed level of emissions must be 43 dB below carrier level, 50 dB in most countries other than the USA, regardless what that carrier level may be. That may not seem logical but that is the rule that is a condition of our license and that we committed to uphold when we received it.
The compliance of my uBitx with these rules is summarized here:
<Screen Shot 2018_08_10 at 2.51.25 AM.jpg>
You have addressed only the recently discovered harmonics issues. And my unit, as you have noted, is not disqualified by excessive harmonics on bands above 17 meters. Unfortunately, on those bands, it is disqualified by the previously discovered "45 MHz minus carrier" spurs.
You say "First: the trouble seems to be more with CW than SSB harmonics. We can reduce the CW level by decreasing the CW drive level."
Perhaps, but I think you will find that the harmonics on CW are far more a function of putting the harmonic rich square wave output of the Si5351 directly into the transmitter signal chain without filtering. The reason the CW versus SSB issue disappears in the high bands is because the circuitry itself acts as a filter "rounding" out the square waves at these higher frequencies. I don't think your proposed drive level change will affect CW harmonics but it is worth a try.
You have written "Second: the alignment as it comes out from HF signals will show far better harmonic suppression than being reported here. For instance, the -38 dbc on 20 meters and -30 db on 40 meters will almost disappear if you balanace it out with the bias on the IRF510s. The factory alignment works like this : First crank up both the IRF510s for 100 ma standing current on each, then tweak one of them to null out the harmonics. It is like balancing out the carrier on diode modulator."
The harmonics causing the problem in my screen shots are odd harmonics; 3rd, 5th, 7th, 9th etc. The alignment procedure you recommend only affects even harmonics due to the cancellation effect of the push-pull PA. It has no effect whatsoever on odd harmonics and, in fact, had been optimized just as you describe before my testing.
You say: "Third: below 14 MHz, the IRF510s are distorting with too much drive. If you back off the drive to adjust to a leve of 7 watts, the harmonics will climb down to be within the spec."
Sorry, but not so for my tests. As noted in my text, all harmonic screenshots were taken with drive reduced to result in a consistent power output for each band of +33 dBm, 2 watts. This was done precisely to avoid overdriving any stage of the transmission chain. At maximum drive, particularly on 80 and 40 meters the harmonics are considerably worse.
Your filter designs and components as supplied are adequate for the PA filtering application. Removed from the board and laid out in a linear fashion with no relays, the filters exhibit well over 55 dB of attenuation at the 3rd harmonic and higher for each band. The primary limiting factor is the strategy of running the input and output of each filter through at least one common relay; lower bands do this with two or all three of the relays. Input to output crosstalk within the relay becomes the first and most substantial contributor to the "blow-by" of harmonics bypassing the filters. The second factor is the "daisy chaining" of the low frequency signals through multiple relays. Finally, a high power PA harmonic filter must have extremely high overall isolation between the input and output of the assembly. The long traces and circuitous routing on the board to connect all of these relays contributes to lack of input/output isolation. When laid out on the board and routed through the relays the filters exhibit on the order of 25 dB attenuation at harmonic frequencies; far too little to be compliant.
The relays per se are not the problem.
It may be possible to make the uBitx compliant with only better PA filter layout: No sharing of relays between filter inputs and outputs, no daisy-chaining of relays, more straightforward PCB layout. However, due to the "45 MHz - carrier" spur issue such filters would need to be BPF rather than LPT. Alternatively, a 4-6 filter assembly, switched in conjunction with the PA filter and placed between the first mixer and the drive chain would likely clean up the signals to the point that the existing filter arrangement may work.
Based on the discussion here and a parallel one on the Facebook uBitx page, I do not believe my results are "sample of one" or the result of a defective unit. That, however, remains a possibility until someone publishes data assembled with similar care to refute them.
I love the concept of uBitx. I want it to work and work well. I sympathize with the pragmatic concept that "it makes less spur power than an high power commercial rig" but that is not the rule. And you can already see the number of people who connect the uBitx to linear amplifiers on the assumption they are starting out with a compliant radio but they are not.
I cannot, in good conscience, put my unit on the air as it stands. Each amateur radio operator needs to be guided by his or her conscience in this matter. But if the decision is made to go ahead it should be made with the full knowledge that it is a violation to do so.
Thanks for soliciting responses and for listening.