You have to skip the first 1/10 of time scale as it contains no valid signal if you use the "FOURIER ..."
See attached picture of a 20kHz fft of v(x)
Attachments
Astx, your simulation still has too much DC drift, hence the FFT noise floor obscuring the harmonics. You can use the Hann FFT window to fix this, but you will need to make all the DC blocking caps 1F or so to reduce the drift so that the .four command reports accurately. The CCS capacitors may matter too.
Do not give the sine source a delay in the plotted data! It should start at the same time as simulation or before. The abrupt start creates a large impulse which obfuscates the FFT.
Do not give the sine source a delay in the plotted data! It should start at the same time as simulation or before. The abrupt start creates a large impulse which obfuscates the FFT.
Clean your optics! Check your statements!
< 500µV DC offset and < -90 db harmonics @20kHz . Real life FFT with hp3585A shows nearly the same smooth frequency response.
Sim starts with 1 cycle DC and 9 cycle sinus so it is necessary to skip 1/10 of sim otherwise you get garbage
BR, Toni
astx, the noise floor in simulation is meaningless. All the obscuration in post 583 is entirely unique to simulation, regardless of the degree it may appear to mimic real life. Lower the computational noise floor so that it doesn't obscure the harmonics.
Delaying the sine wave for one cycle will do no good as long as the silence is included in the FFT; the abrupt start will create an obscuring impulse with close to the same period as the silence, which will be much too close to the period of your fundamental.
My lenses appear to be clean.
Delaying the sine wave for one cycle will do no good as long as the silence is included in the FFT; the abrupt start will create an obscuring impulse with close to the same period as the silence, which will be much too close to the period of your fundamental.
My lenses appear to be clean.
Not any more unrealistic than a fake noise floor caused by a sine starting impulse.
Actually your new FFT AGREES with the .four analysis. I see about -106db distortion, which is about .0005%.
What mystifies me is that the .four command seems to randomly choose whether or not it will ignore circuit drift.
Last edited:
Sorry but my sim reflect the real life measurements nearly exactly. As you have the original asc files please tell us how to correctly sim fft.
If that is the case, then you have unwittingly sabotaged your simulation in order to match it to real life.
You don't need to use the Hann window to see this. Set the sine start delay to 0 and make all DC blocking or bootstrapping caps 1F. The DC drift will no longer obscure the harmonics.
EDIT: Now I see you exclude the sine start in your FFT. I found this was extra effort when I could do basically the same thing in the .tran statement. Still, the noise floor is caused by drift.
You don't need to use the Hann window to see this. Set the sine start delay to 0 and make all DC blocking or bootstrapping caps 1F. The DC drift will no longer obscure the harmonics.
EDIT: Now I see you exclude the sine start in your FFT. I found this was extra effort when I could do basically the same thing in the .tran statement. Still, the noise floor is caused by drift.
Last edited:
As I do not want to add 1F caps and other magics to fake the measurements I think it is better to skip one cycle ...
"Hann" was your Idea ...
Believe me - there is no energy from crossover distortion but about 10 to 15 ppm driving THD21k@200W@8R in real life ...
Michael's criteria for goodness?
If you would enlighten us unwashed masses ... I'm at a loss to understand your point of view. I note
Surely you are not just posting to point out that everyone else is cerebrally challenged?
Michael, thank you for your erudite opinion.Your naiveté and ignorance of elementary analogue electronics is simply shocking.
If you would enlighten us unwashed masses ... I'm at a loss to understand your point of view. I note
- you spurn all 'real life' evidence
- your SPICE examples all show poor stability & terrible THD. If I've missed one that doesn't, please point it out.
- every other person's circuit you 'modify' immediately takes on these characteristics that most of us try to avoid
Surely you are not just posting to point out that everyone else is cerebrally challenged?
Yes, Hann was my idea. I never said it was the problem.
The 1F caps don't add any magic. They won't change the harmonics. They will do absolutely nothing unless your DC operating point was solved very badly.
Thanks for this Toni.
I can't go any further until I can sim the level of oscillation you show. I only get some of this behaviour WITH the inductance of the ribbon cable and Clean earth link to Star earth. The small measures you've tested for me show I'm on the right track but not nearly there.
My sims have both ribbon inductance AND capacitance. They show MORE oscillation and no THD change with longer ribbon so they are missing some major factor(s).
Toni, are these are sim results?astx said:
Regarding the Cherry version .. as my sim results show only 2dB improvement over TMC and the measured THD for all versions is 10dB worse, there is really no point pursuing Cherry until the sim shows the wriggles on slew limiting. In theory, the main advantage of Cherry over TMC is less xover and your measurement techniques don't show any. Something else is the dominant THD mechanism.
I'm certain the issues are related and have a feeling I'm on the verge of a breakthrough in understanding big Power Amps with IPS & VAS on a separate PCB.
Again thanks for all your hard 'real life' work Toni.
PS I've made a bid for that broken PA amp so if successful, this beach bum may take you up on your spare PCBs.
_____________
For those simulating THD, I'll repeat my recommendation for Eugene Dvoskin’s Total Harmonic Analyzer from the LTspice Yahoo Group. Bear in mind the various caveats I've mentioned in previous posts. It takes out all the guesswork and fudging to get a good 'noise' floor. Close study of the code will reveal what is necessary to get sensible results eg the correct delay etc. The amp must have little LF roll-off but this is easy to arrange without mucking up the sim.
You can extend it from the default 10 harmonics to as many as you want. The harmonic plots I've shown are very useful to pinpoint the dominant THD mechanisms. And so is the residual.
I didn't say 'unconditional' is the same as 'conditional'.
You'll have to read my 200 pg explanation
I did have an Idiot's Guide[1] on this website but can't seem to find it.Are you referring to his continuous 9dB/8ve 45 degree Loop Gain function? You gonna do this on your amp?
Impossible to get even close to mythical Bode with 'pure Cherry' or plain Miller.
However, TPC & TMC have enough flexibility for a good approximation. I've done this with TPC in da old days[2].
Never tried evil TMC for real.
__________________
[1] Of course, 'Idiot' refers to the author and NOT to Guru Zan
[2]Also a version which replaced the Cherry cap with a 2-pole network. The TPC version worked well compared to plain Miller. But the 'pure Cherry' amps already reached the limits of my THD gear at that time so couldn't see if 'TP Cherry' was an improvement.
Might do some more work if I can find a collaborator for the 'real life' stuff.
'TP Cherry' is double differentiating (2nd order) making Enhanced VAS + EF2 (4th order) into a 2nd order system which re-enters the realms analysed by Prof Cherry. Should be more stable if done right and also help IPS THD
Got it. Attached a fft picture of the second half of a 1ms sample to get DC stabilized. Now we can see a more detailed fft.
Thanks again - there is a lot to learn!
Attachments
Yes. My discrete darlington version hasn't seen the real life - maybe this summer there will be a breadboard version for testing ...
you are welcome!
Simply send me your address by PM and you will get some spare PCB's for testing.PS I've made a bid for that broken PA amp so if successful, this beach bum may take you up on your spare PCBs.
Bode's actual solution is a bit more complex and lot more subtle than that.
9dB/8ve is just an over-simplification that has been repeated until the misconception has become accepted.
I do plan to do as close as I can to the real deal in my amp.
Exactly my point
Best wishes
David
Last edited: