Warning
The schematic in post #118 contains a dangerous flaw. During startup, the drivers as well as Q9 and Q12 are inactive. As a result, the gates of the O/P devices are exposed to way too high voltages and subsequent destructive drain currents. Adding R15 (see below) simply cured the startup issue.
The second picture shows the drain currents, the sum of gate voltages and output voltage during startup.
Cheers,
E.
The schematic in post #118 contains a dangerous flaw. During startup, the drivers as well as Q9 and Q12 are inactive. As a result, the gates of the O/P devices are exposed to way too high voltages and subsequent destructive drain currents. Adding R15 (see below) simply cured the startup issue.
The second picture shows the drain currents, the sum of gate voltages and output voltage during startup.
Cheers,
E.
Attachments
Member
Joined 2009
Paid Member
In the past (simulations) I've had to add a capacitor to the current source to slow down how quickly it turns on and allow the output to settle. It's unsettling to find a circuit that looks brilliant suffer from turn-on issues that require all sorts of messing around. Nice to see you have a simple and elegant solution.
just out of curiosity
On request, here is a schematic of an amp without global NFB. I've simmed this thing only out of curiosity. So, please don't draw the conclusion that I belong to the camp of global NFB haters. Instead, quite the opposite.
The front-end is fully symmetrical and the VAS current is stabilized by means of one of the the most simple common mode control loops (Q7, Q8, Q13, Q14). VAS-Iq can be adjusted by means of trimmer X3.
The first FB loop encloses the input stage and VAS. The feedback signal is taken from the VAS emitter followers via R29 and R30.
The second FB loop enclose the OPS and half of the bias circuit (Q23...Q26).
To my own surprise, this amp clips gracefully (no ringing or sticking), even without Baker clamps etc.
Here some figures:
THD20Hz: 2.7ppm
THD1kHz: 3.5ppm
THD20kHz: 47ppm
Slew rate: 100V/us
PSRR: 80...100dB
Output impedance within the audio band: <= 1 mOhm
Output offset voltage depends on how well the input trannies are matched.
Second picture: harmonic components at 20kHz and 200W into 4 Ohms.
Third picture: Output voltage, drain currents and VAS current during startup.
Fourth picture: Clipping at 20kHz.
Cheers,
E.
On request, here is a schematic of an amp without global NFB. I've simmed this thing only out of curiosity. So, please don't draw the conclusion that I belong to the camp of global NFB haters. Instead, quite the opposite.
The front-end is fully symmetrical and the VAS current is stabilized by means of one of the the most simple common mode control loops (Q7, Q8, Q13, Q14). VAS-Iq can be adjusted by means of trimmer X3.
The first FB loop encloses the input stage and VAS. The feedback signal is taken from the VAS emitter followers via R29 and R30.
The second FB loop enclose the OPS and half of the bias circuit (Q23...Q26).
To my own surprise, this amp clips gracefully (no ringing or sticking), even without Baker clamps etc.
Here some figures:
THD20Hz: 2.7ppm
THD1kHz: 3.5ppm
THD20kHz: 47ppm
Slew rate: 100V/us
PSRR: 80...100dB
Output impedance within the audio band: <= 1 mOhm
Output offset voltage depends on how well the input trannies are matched.
Second picture: harmonic components at 20kHz and 200W into 4 Ohms.
Third picture: Output voltage, drain currents and VAS current during startup.
Fourth picture: Clipping at 20kHz.
Cheers,
E.
Attachments
Member
Joined 2009
Paid Member
Nice ! - I like to see exploring of no gnfb. Having heard my first no feedback amp I very much like what I hear. And as we know, a little feedback is the worse of all worlds - you use it properly or don't use it.
It's an impressive result indeed if a real amp could deliver this kind of low distortion at 200W into 4Ohms without feedback. But if the distortion gets large enough to be audible it looks to be dominated by the 5th harmonic - which would be unfortunate.
It's an impressive result indeed if a real amp could deliver this kind of low distortion at 200W into 4Ohms without feedback. But if the distortion gets large enough to be audible it looks to be dominated by the 5th harmonic - which would be unfortunate.
Hello Bigun
One thing I have found is that in practice you will have different dominant harmonic distortion than the simulations show because I suspect of the transistor models used in the simulation.
Arthur
I don't like the 5th harmonic either. The big question is can you hear it?
In the meantime I managed to bring the THD20k down by half:
THD20Hz: 2.7ppm
THD1kHz: 2.8ppm
THD20kHz = 23ppm (in stead of 47ppm)
This was possible by using a kind of two pole compensation (TPC). BTW, I'm not sure whether we should call it TPC or TMC. The point is that we are dealing here with a feed forward compensation that bypasses the error FB stuff, rather than a traditional Miller compensation or its derivatives.
If you want even lower distortion, then revert to global NFB or class heat, though the latter defeats the whole objective of the auto bias circuit.
Below the original compensation, modified version (added: C16, C17, R33 & R34), and the harmonics of 20kHz.
Cheers,
E.
PS: I just simmed a 'blameless' amp (with GNFB) and guess what? The 5th harmonic was just as large (or small).
Attachments
Member
Joined 2009
Paid Member
I've never actually done any listening experiments for particular harmonics, I guess it would be possible with some kind of digital synth. My concern of the 5th is based on popular opinion.
It's quite a nice idea to have a benchmark to compare against (i.e. the blameless) since this might possible allow the results to be crudely 'calibrated' against something built.
If I understand, your results with no feedback are as good as a blameless with feedback. Does this imply that the gnf approach and your approach are roughly equivalent wrt distortion ?
complexity
BTW, the LT1166 uses 43 trannies (instead of 10)!
I suppose it is meant as a complement.Wow, Pentium I internal schematic
BTW, the LT1166 uses 43 trannies (instead of 10)!
Hello Edmond
The clipping waveform you posted can you tell me what ohm load it is into.
I think that clipping waveforms should be into worst case load 2ohms.
Regards
Arthur
clipping
Hi Arthur,
That was with a load of 4 Ohms. At 2 Ohms same picture, though the output clips at an 1V lower voltage.
BTW, I wonder whether in real life the clipping behavior will be just as neat.
See the first subject in this comments: http://www.diyaudio.com/forums/soli...ogy-construction-troubles-17.html#post2268194
Cheers,
E
Hi Arthur,
That was with a load of 4 Ohms. At 2 Ohms same picture, though the output clips at an 1V lower voltage.
BTW, I wonder whether in real life the clipping behavior will be just as neat.
See the first subject in this comments: http://www.diyaudio.com/forums/soli...ogy-construction-troubles-17.html#post2268194
Cheers,
E
Hello Edmond
In real life the clipping on two ohms is not so neat , the amp I have (which you are familiar with) as the load goes from 8, 4 , 2 ohms the clipping behavior worsens on a single pair of BJT outputs.
If you have any ideas on how to improve this it would be very helpful.
By the way when you simulate phase margin do you do this into a worst case load like 2ohms for example.
Can you also explain to me why you prefer the complimentary circuits in your designs, I am very interested in you opinion, I very interested in trying a new front end.
Regards
Arthur
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.