After removing the 470uf the results were an improvement in THD (as expected)..
1kHz, 3.15w with 470uf = 3.14%, without 470uf = 1.6%
1kHz 5w with 470uf = 4.8%, without 470u = 2.6%
10kHz, 3.15w with 470uf = 3%, without 470uf = 1.98%
10kHz 5w with 470uf = 4.5%, without 470u = 3.3%
1kHz, 3.15w with 470uf = 3.14%, without 470uf = 1.6%
1kHz 5w with 470uf = 4.8%, without 470u = 2.6%
10kHz, 3.15w with 470uf = 3%, without 470uf = 1.98%
10kHz 5w with 470uf = 4.5%, without 470u = 3.3%
Seems fine but my hearing stops at 10-12k, but just checked at 16 and 20kHz:
3.15w, 16k = 2.2%, 20k = 2%
5w 16k & 20k = 3.5%
3.15w, 16k = 2.2%, 20k = 2%
5w 16k & 20k = 3.5%
I am somewhat amazed at how these two changes have benefitted the sound quality.
There is greater clarity from all instruments but with no listening fatigue - overall a very pleasant sound. (mid/woofer driven by my Krell Class A clone)
LR8 CCS next.
There is greater clarity from all instruments but with no listening fatigue - overall a very pleasant sound. (mid/woofer driven by my Krell Class A clone)
LR8 CCS next.
Now fitted a 10m45 to one channel and ran the tests again. The spot frequency measurements below were done with Picoscope (REW's generator)
The fft was done with REW which shows a loss of about 1db at 20Khz which is not relevant.
1kHz, 3.15w with 470uf = 3.14%, without 470uf = 1.6% with 10m45 = 1%
1kHz 5w with 470uf = 4.8%, without 470u = 2.6% with 10m45 = 2.1%
10kHz, 3.15w with 470uf = 3%, without 470uf = 1.98% with 10m45 = 1.9%
10kHz 5w with 470uf = 4.5%, without 470u = 3.3% with 10m45 = 4.12%
See schematic for voltages - blue with the 10m45 fitted. C1 and R3 were removed.
I think the results are pretty good but no listening tests until the other channel has been converted.
Also, I can to increase the feedback to lower the gain by -6db.
The fft was done with REW which shows a loss of about 1db at 20Khz which is not relevant.
1kHz, 3.15w with 470uf = 3.14%, without 470uf = 1.6% with 10m45 = 1%
1kHz 5w with 470uf = 4.8%, without 470u = 2.6% with 10m45 = 2.1%
10kHz, 3.15w with 470uf = 3%, without 470uf = 1.98% with 10m45 = 1.9%
10kHz 5w with 470uf = 4.5%, without 470u = 3.3% with 10m45 = 4.12%
See schematic for voltages - blue with the 10m45 fitted. C1 and R3 were removed.
I think the results are pretty good but no listening tests until the other channel has been converted.
Also, I can to increase the feedback to lower the gain by -6db.
Changed Rfb (R12) from 2k to1k and now recorded 0.73% at 1KHz and 3.1watts, 1.65% at 5watts. Gain is now the same as the unmodded chnnel - 0.7v for 3.1w and ~ .9v for 5w so call it 1v RMS for max output with < 5% THD (5.5w)
At 10K, 1.8% at 3.15w and 4.7% at 5w.
So a further improvement except for 10kHz 5watts.
At 10K, 1.8% at 3.15w and 4.7% at 5w.
So a further improvement except for 10kHz 5watts.
Just check square wave for stability if you increase the FB. You may find CCS of 2.5ma is better, U2 plate voltage of 227v maybe on the high side for clipping. Maybe R7 820k.
I'll check square waves tomorrow.
I also noticed that the 3rd harmonic distortion is worse - worse than 2nd now. What might have caused this?
I also noticed that the 3rd harmonic distortion is worse - worse than 2nd now. What might have caused this?
The simulation does show the third harmonic overtaking the second but only just before clipping.
The cascode stage for sure can be tweaked ... voltages across tubes , current , amount of negative feedback , so on . This takes time . The pleasure of designing something pretty new 🙂
There is some ringing on a 1kHz square wave but I forgot to do loopback test to see what Picoscope might be adding.
Both channels have now been modded and listening tests commenced.
Both channels have now been modded and listening tests commenced.
The Picoscope should add that much. ;-) That's going to translate into a lot of ringing at 10kHz and above, maybe even some instability. Looks like your OPT isn't taking kindly to the feedback. Try a phase lead cap across the feedback resistor if you haven't already.
Any amp needs a small compensation cap across feedback resistor R12 for perfect square wave at 1KHz . Not surprising at all even if the scope is not perfect
Yes, I plan to add a cap across the fb resistor - thinking of 1000pf across 1k?
Is a start , usually the feedback resistor is much higher ... high resistor = lower value cap and the other way around
The simulation needs 1nF across the 1K FB resistor (lead compensation) and 470p in series with 4k7 across the current source (dominant pole - lag) but its only a guess on the transformer.