I think that your results were more-or-less as would be expected. That is why, I imagine, that when broadcasters used valve-era professional three-channel ISB receivers for reception of HF relays for rebroadcast, they mostly used one or other of the sidebands but as far as I know did not attempt to combine the two after demodulation. The results seemed to have been good enough given that the technique was apparently quite widely used. (See the ISB Receiver thread;
https://www.vintage-radio.net/forum/...d.php?p=660609.) When demodulating a single sideband, the reinserted carrier phase less important. But synchronous demodulation of a double sideband signal requires that the reinserted carrier phase be closely aligned with that of the original carrier. If the sidebands are demodulated separately with a non-phase aligned reinserted carrier and then recombined, then I suspect that the effects of the carrier phase error would reappear, compounded by whatever other errors were occasioned by the separate processing of the sidebands.
But the fully synchronous PLL approach, when done well, is definitely better for fading MF and HF signals. With full carrier AM, the carrier filter appears not to be needed, as the PLL itself acts as a very narrow band filter. And separation of the sidebands after demodulation could be done by the phasing method rather than using separate sideband filters. Thus switching between DSB, USB and LSB was done fully in the audio domain. With later solid state three-channel ISB receivers using PLL carrier recovery, the carrier filter was still used, I imagine for noise reasons given that the pilot carrier was 16 or 26 dB down.
Way back when I had the opportunity to discuss this topic with the designer of the Phase Track Liniplex HF receivers, which probably had one of the best implementations of PLL synchronous demodulation. At the time I did not have the knowledge to ask as many questions as I would today, and nor did I take any notes. The difficulties of attempting carrier and sideband separation using IF filters were mentioned though, as well as the need for a tracking PLL. Receiver synthesizer phase noise was an issue that showed up with USB and LSB operation, which is why the original Liniplex design used a crystal-controlled oscillator. An interesting comment was that with DSB AM, even better results could be obtained by using as Costas loop, because this regenerated the carrier based upon the sideband information, and so had it phase-aligned with the whole signal. The obverse of that, though was that sideband separation was reduced in the USB and LSB modes.
The group delay problems introduced by separate sideband filters could have been why Crosby did not attempt to separate the sidebands in his exalted carrier receiver proposal. I had written the following a couple of days back, intending to edit it and post it before long, but as it fits here I’ll include it as is, including any redundancies.
Related to this topic, I recently came across an article by M.G. Crosby about exalted carrier AM reception, in “Communications” magazine for 1945 February (available on the excellent AmericanRadioHistory.com site).
Here the primary objective was to minimize the effects of carrier fading distortion on MF and HF reception. In essence this was achieved by extracting the carrier (at 2nd IF in a double-conversion receiver) from an AM signal using a very narrow bandwidth crystal filter, amplifying and limiting it, and then after appropriate phase adjustment either recombining it (at exalted level) with the full bandwidth IF signal for conventional diode demodulation or using it as the reference input to a product demodulator whose other input was the full bandwidth IF signal. Also included was an AFC system driven from the narrow bandwidth carrier channel.
Insofar as reconditioned incoming carrier was used as the reference, I think that this would qualify as quasi-synchronous demodulation, at least when a product demodulator was used.
Actually though the Crosby receiver looks much like the point-to-point SSB/ISB receivers that were originally developed in the 1930s. These also separated and reconditioned the carrier for use as a reference after extraction with a very narrow band width filter. The sidebands were also filtered out separately and then individually demodulated in product demodulators that used the reconditioned carrier as reference. Usually provision was also made for using a locally generated carrier for use with those SSB transmissions that had fully suppressed carriers. Carrier-driven AFC was the norm for these receivers. They could be and were used for the reception of AM signals in order to minimize the effects of carrier fading distortion. In that case one or the other sideband was used according to which had the least interference.
So one may ask what did the Crosby circuit bring to the table beyond what existing SSB/ISB receivers could do. Use of both sidebands simultaneously – assuming neither was suffering from undue interference – would have reduced the amplitude variations due to selective fading in the recovered audio, as typically only one sideband at a time is affected as the fade combs through. My first-hand experience with the Liniplex F2 confirmed that this was a non-trivial gain, at last for SW programme listening (it showed up in my ad hoc BBC WS Play-of-the-Week test; DSB, where one could use it, was definitely better than LSB and USB). Also, there would have been no loss of lower audio frequencies. The sideband filters in SSB/ISB receivers cutoff points varied with type, say 300 Hz for voice filters (300 to 3400 Hz) and 100 Hz for broadcast relay filters (100 to 6000 Hz), so there was always some loss, and the 300 Hz number was very inappropriate for broadcast listening.
Crosby also proposed that his receiver could be used for reception of phase-modulated transmissions in fading conditions, for which purpose both sidebands would have been required. This would have required a π/2 phase difference between the reference and signal feeds to the product demodulator, rather than the zero difference required for AM demodulation.
Whether the Crosby receiver was ever realized in practice I do not know. One candidate for consideration was the Press Wireless exalted carrier diversity receiver, mentioned in the BBC description of its Tatsfield monitoring station, see:
http://www.bbceng.info/Operations/Re...rch%201961.pdf. Although I have not found any information specific to the Press Wireless receiver, the associated patent provides some clues. It appears that the Press Wireless approach was to use a locked oscillator to generate the local reference carrier, this normally being locked to the incoming IF. The oscillator output was mixed with the full bandwidth IF before demodulation. When the incoming carrier level was too low to allow locking of the oscillator, the latter was suppressed, and normal AM demodulation took place. That suggests that the demodulator was of the rectifying type, not the product type. Be that as it may, it looks as if the Press Wireless receiver had fully synchronous rather than quasi-synchronous demodulation, and it did not have a narrow bandwidth carrier filter.
Cheers,