PMA said:
Hi PMA,
No offense meant, but this is 1khz, that's easy, I don't need Hawksford for that. Show us the curves with 20kHz, or even better 50kHz, and include the output signal from the correction circuit, that is the real indicator that it does anything. What output devices are you using in the sim?
Jan Didden
janneman said:
Hi Jan,
very pleased by your interest 😉 . Just wait no more than 2 weeks for the results of the real amplifier, the simulation is a closed chapter for now. /I prefer listenning to the amplifiers as the target of my effort/.
I was simulating with the same devices that are used for the real amp.
You are not quite correct that I do not need error correction for 1 kHz, even for this the improvement in THD is more than of one order (more than 20 dB).
I certainly do have simulated output of the correction circuit, but for the reason mentioned I do not feel like displaying it.
Last but not least - this is neither Hawksford, nor anything else. This is inspired by Nelson Pass, but using quite different signal input port, as you have already noticed before.
All these circuits are related, and they are all diddling the bias
voltage for one reason or another. In the 1975 Threshold amps,
it was being used to lower distortion and keep the output stage
transistors in forward bias?
Did it sound better? In retrospect I would say it sounded a lot
like an ordinary BJT follower biased so as to idle at about it's
rated power (pure Class A being biased at twice) which is pretty
much what it was. Attempts to improve the sound at lower bias
currents were not satisfying.
In Hawksford's and Cordell's case, the bias is adjusted strictly
for lower distortion figures, and it is very successful at that,
but my experience is that it still tends to sound like the same
amp operated at the same (constant) bias levels.
What's so magic about the bias and not the distortion numbers?
I cannae say. 😎
voltage for one reason or another. In the 1975 Threshold amps,
it was being used to lower distortion and keep the output stage
transistors in forward bias?
Did it sound better? In retrospect I would say it sounded a lot
like an ordinary BJT follower biased so as to idle at about it's
rated power (pure Class A being biased at twice) which is pretty
much what it was. Attempts to improve the sound at lower bias
currents were not satisfying.
In Hawksford's and Cordell's case, the bias is adjusted strictly
for lower distortion figures, and it is very successful at that,
but my experience is that it still tends to sound like the same
amp operated at the same (constant) bias levels.
What's so magic about the bias and not the distortion numbers?
I cannae say. 😎
fab said:
Of course. V3 is a Source of the M1 NMOS (upper half). V7 is a Source of the M2 PMOS (lower half). V21 is an output voltage. As source resistors R12, R13 are of 0.1 Ohm, 10*(v(3)-v(21)) is a current through M1 and 10*(v(7)-v(21)) is a current through M2. An addition of both currents is shown as well. The output is loaded by 8 Ohm resistor.
Low bias current
IMHO it is impossible to maintain something what I would call "a sound consistency" at low bias currents.
IMHO it is impossible to maintain something what I would call "a sound consistency" at low bias currents.
What is magic, Nelson ? It is that nature havn't distortion, 'cos in case if have it, should be own caricature 😎 and I can't do caricatures of sound 
.
.
Maybe a better question would be: what is the sound (reproduction)? Maybe some good answers would be helpful in creating caricatures.
PMA said:
OK, I understand. I had interpreted your curves as showing that the Hawksford error correction works. Now I understand you have the curves, but don't want to show it, and in any case it's not Hawksford, so I wonder what those curves DO show, except the currents in a class AB stage.
I would think that if you were going to implement an error correction, you would want to know for sure that it works as you expect, I mean if you are not interested to know if it works, what's the point? If you are convinced by some means that it works, why not show it?
Listening is OK, but of course you cannot determine if the error correction works by listening. After all, who knows how correct error correction sounds, or almost correct error corection, or completely wrong error correction. You may like the sound, but maybe you would like it even better with all those error correction things ripped out.
Jan Didden
Jan,
unfortunately I do not uderstand you. I have simulations, the circuit was shown here, and I want to compare the simulations with measurements on real circuit, this will happen very soon. Where is the problem?
But there is one thing I do not want to do. To take part in never ending nothing solving time wasting discussions about existence/non-existence of current feedback etc. This is from my point of view only a storm in a cup of water.
BTW - try to plot the same current curves for the standard AB output stages and you'll probably be very surprised what you will see - abrupt switching, non-symmetry, sudden slope changes.
unfortunately I do not uderstand you. I have simulations, the circuit was shown here, and I want to compare the simulations with measurements on real circuit, this will happen very soon. Where is the problem?
But there is one thing I do not want to do. To take part in never ending nothing solving time wasting discussions about existence/non-existence of current feedback etc. This is from my point of view only a storm in a cup of water.
BTW - try to plot the same current curves for the standard AB output stages and you'll probably be very surprised what you will see - abrupt switching, non-symmetry, sudden slope changes.
Hi Everyone.
I opened the Hawksford string, and saw the 'error cancellation' circuit.
I thought ' yeah - sure ! ' and did not give it a second thought.
__________________________________________________
On accidentally rechecking (because the Solid State Index page does not give any idea how long this string is) I find over 30 pages of postings airing some interesting commentry.
It is my impression that error cancellation will arise ONLY when the amplifier is resistor loaded, because it generates leading capacitor current generated 'correction' wrt output voltage slew.
When a real world dynamic loudspeaker is being driven, with momentary and dynamically generated back EMFs rendering resultant current flow considerably out of phase with amplifier output voltage around loudspeaker driver section crossover frequencies, the error correction capacitor will generate high frequency current compensation when the loudspeaker does not require the amplifier drive be corrected - i.e. the 'error compensation' will increase distortion in a way that is unique to the loudspeaker being driven.
I have not seen anyone mention this.
Cheers ............ Graham.
I opened the Hawksford string, and saw the 'error cancellation' circuit.
I thought ' yeah - sure ! ' and did not give it a second thought.
__________________________________________________
On accidentally rechecking (because the Solid State Index page does not give any idea how long this string is) I find over 30 pages of postings airing some interesting commentry.
It is my impression that error cancellation will arise ONLY when the amplifier is resistor loaded, because it generates leading capacitor current generated 'correction' wrt output voltage slew.
When a real world dynamic loudspeaker is being driven, with momentary and dynamically generated back EMFs rendering resultant current flow considerably out of phase with amplifier output voltage around loudspeaker driver section crossover frequencies, the error correction capacitor will generate high frequency current compensation when the loudspeaker does not require the amplifier drive be corrected - i.e. the 'error compensation' will increase distortion in a way that is unique to the loudspeaker being driven.
I have not seen anyone mention this.
Cheers ............ Graham.
Graham, could you please specify what you mean by words "the error correction capacitor" /I would like to see it in a real schematic/.
BTW - speaker and its model, Back EMF and resonant circuit model /acoustic-mechanic-electric transformation and 1 model including all of those/ doing the same - I guess we were already solving this. And it can be proven both in simulation and real world measurements.
BTW - speaker and its model, Back EMF and resonant circuit model /acoustic-mechanic-electric transformation and 1 model including all of those/ doing the same - I guess we were already solving this. And it can be proven both in simulation and real world measurements.
Hi PMA
There is only one capacitor on the referenced .gif I downloaded.
What point are you making in your second comment, is something wrong ??
Cheers ......... Graham.
There is only one capacitor on the referenced .gif I downloaded.
What point are you making in your second comment, is something wrong ??
Cheers ......... Graham.
Hello Graham,
the reason why I asked is that I also displayed an error correction circuit in this thread:
http://www.diyaudio.com/forums/showthread.php?postid=425708#post425708
There is no error correction capacitor and I assume that Hawksford circuit also does not use "an error correction capacitor".
Regards, Pavel
the reason why I asked is that I also displayed an error correction circuit in this thread:
http://www.diyaudio.com/forums/showthread.php?postid=425708#post425708
There is no error correction capacitor and I assume that Hawksford circuit also does not use "an error correction capacitor".
Regards, Pavel
Hi Pavel,
Saw your circuit.
Is it a follower with smoothed transconductance between two devices operating in class-A and only one in class-B. This would be automatic bias correction rather than error correction.
Cheers ....... Graham.
Saw your circuit.
Is it a follower with smoothed transconductance between two devices operating in class-A and only one in class-B. This would be automatic bias correction rather than error correction.
Cheers ....... Graham.
Hi Graham,
good point. It really drives FET bias voltage according to signal magnitude and polarity. And the change for the N-MOS half is different from that for the P-MOS half.
Regards, Pavel
good point. It really drives FET bias voltage according to signal magnitude and polarity. And the change for the N-MOS half is different from that for the P-MOS half.
Regards, Pavel
PMA said:
Hi Pavel,
Well, I was a bit disappointed, you know, I thought hah, here we are onto someting. I know from experience sims often can be deceiving. I was hoping that if you show the output currents at higher freq they would more clearly show the abrubt changes without error correction you mention, and then looking at the output of the error correction when put in, we would see those abrubt changes being filled in.
Depending what devices/models you use in the output stage sim, the curves you showed at 1kHz often look too good to be true - and that is what they are.
Jan Didden
PS Indeed, save us from the CFB thread - although I must confess some guilt for that. But wasn'ty it a great learning experience also??
Graham Maynard said:
Hi Graham,
Conceptually, the Hawksford error correction s described by Malcolm in his original paper is specifically meant for output stages that already have a gain close to one.
The correction circuit basically is a voltage controlled current source. The voltage input is the difference between the input and output of the output stage, i.e. the deviation of the gain of one. The output current adds to the driving current to to increase the input drive so that the output level is what it should have been with a gain of one. Often you see a couple of bipolar transistors, with their emitters to the output stage output and their bases to the output stage input (or vice versa, I'm too lazy to check, but you get the point), (these is the voltage input), while the resulting the collector current pulls down (or releases) the input current coming from the Vas. Note that for this to work the collector current must be exactly correct to develop the correction voltage over the Vas output impedance. If for instance the Vas output resistance is 1k ohms, the collector current variation must be exactly 1mA per volts deviation of the output stage input/output voltage.
There are some more details but this is the concept. You will see that if the thing works as advertised, the output impedance of the amp is exactly zero and the load has no influence on the output signal.
The nice thing is that often the two bipolars double as bias generator so they are there anyway.
Jan Didden
janneman said:
Jan,
you are quite right - the circuit without error correction suffers from the illness mentioned. The error correction cures it quite well even for the higher frequencies. I have also simulated some of the well known MOSFET amps without error correction and the non-symmetry of the output device currents and abrupt changes were the obvious problem.
I am just testing the circuit on a test bench. Having the results collected I will let know.
Regards, Pavel
- Status
- Not open for further replies.