I have been trying to understand for more than a year now why is it that amplifers that post lowers THD are not class A ?
Even AB amplifiers like Bryston and even Class H amplifiers like QSC Audio tend to have lower THD numbers than many high end class A amps.
the first answer i got to this question was that Class A amps don't use as much NFB and therefore have higher THD.
i then asked why not make class A amps with a lot of NFB and got no good answer but in a paper by Pass read that NFB does create certain distortions while reducing others so there is SOME logic to using less NFB in class A amps.
and yet NFB overall still reduces distortion, even the high order distortions which supposedly are what make class A superior in the first place ( having less of them that is ).
so my question remained unanswered then Montekay over at AVS pointed out that class A draws a lot of current at idle which makes his Usher R 1.5 amplifier hum. then in some document from Bryston i read that class A amplifiers have current related distortions or some such.
so i got bits and pieces of answers but no real answer.
and the question is of course - why can't i find a class A amplifier with distortion specs equal to or better than Bryston ?
Even AB amplifiers like Bryston and even Class H amplifiers like QSC Audio tend to have lower THD numbers than many high end class A amps.
the first answer i got to this question was that Class A amps don't use as much NFB and therefore have higher THD.
i then asked why not make class A amps with a lot of NFB and got no good answer but in a paper by Pass read that NFB does create certain distortions while reducing others so there is SOME logic to using less NFB in class A amps.
and yet NFB overall still reduces distortion, even the high order distortions which supposedly are what make class A superior in the first place ( having less of them that is ).
so my question remained unanswered then Montekay over at AVS pointed out that class A draws a lot of current at idle which makes his Usher R 1.5 amplifier hum. then in some document from Bryston i read that class A amplifiers have current related distortions or some such.
so i got bits and pieces of answers but no real answer.
and the question is of course - why can't i find a class A amplifier with distortion specs equal to or better than Bryston ?
There are like e.g.
http://www.diyaudio.com/forums/showthread.php?t=96853
You can!
The reason why there are not many high NFB class A amps is simply that low THD-numbers can also be achieved without ClassA output stages. Why running the output gleamingly hot if it does not bring any benefit?
At the low THD-levels of today's high NFB amps (non ClassA) other problems limit even lower distortion, like wiring, pcb-layout, NFB-takeoff point and whatnot.
0.0001% at full power can be achieved this way. An ClassA output stage maybe could lower that further, but that is becoming hard to measure then ;-)*
Have fun, Hannes
*that amp will also need a lot of power to actually show the distortion, otherwise it is drowned in the noise. Not practically doable in ClassA.
The THD specification (please note: total harmonic distortion) and the individual sound impression does not correlate in most cases
In my experience, there are only three kinds of distortions, which are impressed as bad sonic quality especially in the upper frequency range arround 5-20KHz
1) crossover distortion (not present by pure PP class A and Single Ended output power)
2) TIM distortion (not present by use of NFB together with only one voltage gain stage) there are mostly two gain stages and NFB for extemly low THD at lower frequencies.
3) Distortion through compression effects (to small power supply and / or not perfect layout and earth wire management
If I choise by simulation for the THD-Check the frequencies 1KHz and 10 KHz, I can't say in advance about the resulting sound charakter, because the comlex mixed music signal represents an entirely different condition.
Therefore, I investigated the possibilities to eliminate this lack
I came to the solution, to select a fundamental frequency of 100 KHz or 1 MHz (for the THD measurement), and 199 KHz + 200 KHz (for the measurement of IM)
If I get here still yet usable values for THD and IM (arround 1% to 2%), I am shure that I also get the best results in sonic quality.
Don't I get good "real live" results nevertheless, there are only mistakes present by periphery (eg layout, wiring management or to small power supply), but certainly not at the circuit topology.
With 98% of commercial power amplifier circuits, that I prepare for simulation, I get only error messages by such high frequencies (message: access violation) - for me the confirmation of not optimal engineering
If I scale down the frequencies to the usually 1KHz for THD and 19/20KHz for IM, I get approximately the same results as by the diagrams in the audio press
Most developers say to me that it is total nonsense for audio amplifier with such high frequencies to work for THD/IM measurement. But I claim only at such high fundamental frequencies I observe a match between measurement THD/IM results and sonic impression.
In my experience, there are only three kinds of distortions, which are impressed as bad sonic quality especially in the upper frequency range arround 5-20KHz
1) crossover distortion (not present by pure PP class A and Single Ended output power)
2) TIM distortion (not present by use of NFB together with only one voltage gain stage) there are mostly two gain stages and NFB for extemly low THD at lower frequencies.
3) Distortion through compression effects (to small power supply and / or not perfect layout and earth wire management
If I choise by simulation for the THD-Check the frequencies 1KHz and 10 KHz, I can't say in advance about the resulting sound charakter, because the comlex mixed music signal represents an entirely different condition.
Therefore, I investigated the possibilities to eliminate this lack
I came to the solution, to select a fundamental frequency of 100 KHz or 1 MHz (for the THD measurement), and 199 KHz + 200 KHz (for the measurement of IM)
If I get here still yet usable values for THD and IM (arround 1% to 2%), I am shure that I also get the best results in sonic quality.
Don't I get good "real live" results nevertheless, there are only mistakes present by periphery (eg layout, wiring management or to small power supply), but certainly not at the circuit topology.
With 98% of commercial power amplifier circuits, that I prepare for simulation, I get only error messages by such high frequencies (message: access violation) - for me the confirmation of not optimal engineering
If I scale down the frequencies to the usually 1KHz for THD and 19/20KHz for IM, I get approximately the same results as by the diagrams in the audio press
Most developers say to me that it is total nonsense for audio amplifier with such high frequencies to work for THD/IM measurement. But I claim only at such high fundamental frequencies I observe a match between measurement THD/IM results and sonic impression.
Last edited:
I wanted to reply to that several times and never took the opportunity. A memory access violation is an error in memory managment and has absolutely zero to do with the validity of your simulation. It just shows that your simulation program is buggy.
True, but only for your simulation tool, not for the simulation itself.
I am a bit surprised that you take your simulation results so seriously, the simulations I did in the past all deviated significantly in some aspects from reality (even if a few like THD and spectra were similar).
Of course one can get real insight from simulation, but in my experience one has to know very well how the circuit works and the limitations of the models in use.
Have fun, Hannes
Member
Joined 2009
Paid Member
This is a very interesting statement. There is a lot of skepticism regarding the use of simulations although most people do recognize the value of them. I have heard people correlate simulations with measurement (e.g. D. Self for one) but I don't remember anyone attempting to correlate simulations with listening impressions. Can you elaborate on your observations - what kind of sounds correlate with what kinds of simulation results did you observe ?
Member
Joined 2009
Paid Member
I agree with the general comment that THD is a poor metric to use when choosing an amp. It's a subject that has been discussed in this forum with some passion in the past - consensus is strongly in favour of not relying on this performance measure.
I wouldn't spend big bucks without listening to the amps myself, but generally high end amps sound very good with the differences often being subtle and down to personal preferences (your ears, the music you like, the kind of sounds you have been exposed to for the past decade etc.).
If you search the internet for some comparisons between Pass Labs and and Bryston you get some interesting comments:
"Since they are basically the same price, could I choose one amp over the other? Nope. " [http://www.proaudioreview.com/article/408]
"The exercise convinced me that both preamps are much more alike than different."
[http://www.audiocircle.com/index.php?topic=30854.0]
Both brands are very highly regarded.
I certainly don't claim to be a Nelson Pass but I think the following can safely be said (or has been said).
* THD is a nearly useless metric. It is still used for historical reasons and the fact that it was a standard, not because it is particularly useful.
* Crossover distortion CAN be highly audible, but Class A/B amplifiers can be designed such that it is inaudible.
*The differences between modern well-designed amplifiers is one whereby subtle differences exist. Assuming controlled conditions, they are difficult to pick in blind testing (assuming driving the correct load with the proper amount of power).
* Enthusiast will continue to debate the merits of various amplifier designs well past the point where it actually makes a difference.
* The only disadvantage that I can think of for Class A are the obvious ones. It is expensive, consumes lots of energy and it is difficult to design high-power solutions. Other than that, it removes any chance of having crossover distortion so it lets people sleep better at night knowing they have slayed at least one dragon.
* THD is a nearly useless metric. It is still used for historical reasons and the fact that it was a standard, not because it is particularly useful.
* Crossover distortion CAN be highly audible, but Class A/B amplifiers can be designed such that it is inaudible.
*The differences between modern well-designed amplifiers is one whereby subtle differences exist. Assuming controlled conditions, they are difficult to pick in blind testing (assuming driving the correct load with the proper amount of power).
* Enthusiast will continue to debate the merits of various amplifier designs well past the point where it actually makes a difference.
* The only disadvantage that I can think of for Class A are the obvious ones. It is expensive, consumes lots of energy and it is difficult to design high-power solutions. Other than that, it removes any chance of having crossover distortion so it lets people sleep better at night knowing they have slayed at least one dragon.
The easist way to do this is the compasion between pure Class A and normal AB (i. e. 30 -50 mA) The difference is clearly to see both by simulation and in real life measurement in nearly the same kind. High order distortion goes down significant by Class A. Sonic impression is more smooth and harshless goes down. For heavy metal fans it must not be necessarily an improvement.
More difficult is the quantitative detection of TIM in the simulation. But the basic evidence of TIM is easy to investigate. If I check the range between 100 kHz and 1 MHz, I have either low-pass character (wanted), or the conversion of a sine wave signal into a sawtooth signal (unwanted).
But I must clearly say, that it is not possible, to expect from the simulation all design errors by an amplifier probant. Please note, by simulation all resistors, capacitors, voltage sources (power supply) and input generator are idealized. Furthermore, uncertainty remains about the accuracy of the P Spice modeling (I cannot verify because I haven't the posibility of P-Spice analysis).
Nevertheless, I have good evidence about the circuit behavior in practice regarding the tendency in sound quality.
In general I can say, that the visual conversion of a sine wave signal into a sawtooth signal at arround 500 KHz by simulation clearly results in a sharp sonic impression, especially at voice playback. This happens mostly by two voltage gain stage and NFB. When I reduce the open loop gain of first stage (mostly differential amp) to unity gain (through emitter resistors) this effect disappears complete (I get still sine wafe, even by 10 MHz, but heavy reduced levels). Bass transmission now not so clean and tight, but by the upper range perfect sonic transmission).
Through simulation I have learned, that the royal way for best sonic results is the use of bi- or multi- amping systems (for single amping modes a certain threshold in the level of sonic quality can not be exceeded).
Class AB amps with high open loop gain for the lower frequencies below approximately 300 Hz (TIM and high order THD not notable audible here) and "Single ended" pure Class A (like Aleph, "First Watt" or ZEN - low open loop gain) for the higher frequencies above 2000 Hz (high values of damping factor not nessecary). For the midrange (3-way) I still searching the ideal circuit, but I think, it could be the "X"-topology from Mr. Nelson Pass
I have released some simulation results regarding good known amplifiers - have a look to follow weblinks:
http://www.diyaudio.com/forums/showthread.php?threadid=147745
http://www.diyaudio.com/forums/showthread.php?threadid=147811
http://www.diyaudio.com/forums/showthread.php?postid=1899831
Last edited:
Member
Joined 2009
Paid Member
I have never heard a bi-amped speaker (I guess I could try it but I'm loathe to open my speakers and disconnect the XO network) but I am willing to accept that this is an excellent approach. It gets my vote.
But I love the sound of a single full-range driver. I've been scratching my heard as to how to achieve bi-amping with a single driver, a combination of amps. It seems that it requires filters after the amplifiers to recombine the signals and this seems self-defeating.
I have just started to get into Nelson's work (the past 3 days) so I will look at your Aleph sims with interest.
I am using bi amp in my home....there are two Precision 1 amplifiers
having input filters..their output goes to a speaker that has two cables, one to the woofer and the other cable goes to midrange driver (squawker) and tweeter.
It is the fastest way to adjust to your room...if your room reflects sounds too much, in special treble range, then you reduce treble...if absorsive then you increase treble.
Also bass level can be adjusted too, you can produce a non perfect loudness contour increasing bass and treble.... it is "naaaissss" (Borat nice).
Each amplifier has it's own volume control.
A perfect way to listen music in US, while doing a cultural learnings of America to make benefit to the glorious nation of Casakistan...ahahahahaha!...bora is nice!
regards,
http://www.youtube.com/watch?v=jENuS9nc9p4&feature=fvst
having input filters..their output goes to a speaker that has two cables, one to the woofer and the other cable goes to midrange driver (squawker) and tweeter.
It is the fastest way to adjust to your room...if your room reflects sounds too much, in special treble range, then you reduce treble...if absorsive then you increase treble.
Also bass level can be adjusted too, you can produce a non perfect loudness contour increasing bass and treble.... it is "naaaissss" (Borat nice).
Each amplifier has it's own volume control.
A perfect way to listen music in US, while doing a cultural learnings of America to make benefit to the glorious nation of Casakistan...ahahahahaha!...bora is nice!
regards,
http://www.youtube.com/watch?v=jENuS9nc9p4&feature=fvst
Last edited:
i would not rule out the possibility that what you're doing might make some sort of sense.
why not use a more reasonable frequency like 50 khz though ?
thats great. but i still want my amp with 0.0001% THD. its what i need to be happy.
I certainly don't claim to be a Nelson Pass but I think the following can safely be said (or has been said).
* THD is a nearly useless metric. It is still used for historical reasons and the fact that it was a standard, not because it is particularly useful.
* Crossover distortion CAN be highly audible, but Class A/B amplifiers can be designed such that it is inaudible.
*The differences between modern well-designed amplifiers is one whereby subtle differences exist. Assuming controlled conditions, they are difficult to pick in blind testing (assuming driving the correct load with the proper amount of power).
* Enthusiast will continue to debate the merits of various amplifier designs well past the point where it actually makes a difference.
* The only disadvantage that I can think of for Class A are the obvious ones. It is expensive, consumes lots of energy and it is difficult to design high-power solutions. Other than that, it removes any chance of having crossover distortion so it lets people sleep better at night knowing they have slayed at least one dragon.
once again - no argument. i know IM and other kinky distortions are much more important than THD. but any truly accurate amp will also have low THD.
also i don't care about what is good enough. my QSC PLX is very nearly good enough for me, but i want something 100 times more pure just because i believe it should be possible to achieve.
What is usually largely ignored in THD-and distortion spectra discussions is the fact that speakers generally have a THD inbetween 1-10% at elevated power levels. Some produce also tons of higher order components, which are by many considered to be the worst and yet nobody comlains about their sound
For a 50W amp I can imagine more than 100dB SNR (to random noise floor) if done properly.
Have fun, Hannes
For a 50W amp I can imagine more than 100dB SNR (to random noise floor) if done properly.
Have fun, Hannes
Member
Joined 2009
Paid Member
Because class A amps don't need so deep NFB to sound clean. That means, they don't need so low THD to sound clean. That means, THD does not correlate with audible errors.
Let me correct that: for a 50W amp you can achieve (with the needed dedication) true 100dB SNR - not to random noise floor (the usual marketing gimmick*), but in comparison to ripple and whatever else humm components there may be.
Have fun, Hannes
*in commercial amps the 50/100 Hz components are sometimes 20dB and more above noise floor
That's the reason if an on-paper perfectly quiet amp humms happily...- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.