In the August 2001 issue of AudioXpress, in a reply to a letter, Kenneth P. Miller referred to "optimum Class B" for a bipolar transistor output stage. It refers to a bias setting that produces the lowest distortion of anything other than a full class A bias.
He talks about added distortion as you increase bias beyond the optimum point (but still keep it less than full Class A), and "transconductance doubling" being the reason.
Has anyone else heard of this?
I know that Class A is the modus operandi for most of the amplifiers discussed in this forum, but for those of us building an AB amp it would be useful to have some verification and background on this issue. I can't see building my tri-amped system with 3 Class A amplifiers; natural gas is a much cheaper way to heat my house than electricity.
He talks about added distortion as you increase bias beyond the optimum point (but still keep it less than full Class A), and "transconductance doubling" being the reason.
Has anyone else heard of this?
I know that Class A is the modus operandi for most of the amplifiers discussed in this forum, but for those of us building an AB amp it would be useful to have some verification and background on this issue. I can't see building my tri-amped system with 3 Class A amplifiers; natural gas is a much cheaper way to heat my house than electricity.
Paul
Yes, optimum class-B biasing will give lower distortion than a class-AB design. If you want more details, borrow a copy of Doug Self's 'Audio Power Amplifier Design Handbook'. Chapter 5 explains the mechanism by which the distortion is increased in class-AB working.
Alternatively, have a look at the JLH Class-AB article on my website( http://www.gmweb.btinternet.co.uk/index.htm ), particularly the subsequent 'Letters to the Editor', which covers exactly the same points though with different terminology. This article was written over thirty years ago and shows that there is nothing new in this 'discovery' despite the impression given in various more recent publications. JLH makes a very valid arguement against the minimum distortion class-B arrangement, preferring the benefits of class-A working at low power levels with class-AB biased designs.
Geoff
[Edited by Geoff on 07-28-2001 at 04:14 PM]
Yes, optimum class-B biasing will give lower distortion than a class-AB design. If you want more details, borrow a copy of Doug Self's 'Audio Power Amplifier Design Handbook'. Chapter 5 explains the mechanism by which the distortion is increased in class-AB working.
Alternatively, have a look at the JLH Class-AB article on my website( http://www.gmweb.btinternet.co.uk/index.htm ), particularly the subsequent 'Letters to the Editor', which covers exactly the same points though with different terminology. This article was written over thirty years ago and shows that there is nothing new in this 'discovery' despite the impression given in various more recent publications. JLH makes a very valid arguement against the minimum distortion class-B arrangement, preferring the benefits of class-A working at low power levels with class-AB biased designs.
Geoff
[Edited by Geoff on 07-28-2001 at 04:14 PM]
Thanks, Geoff. It's interesting that Miller referred to "new research" for something that Hood knew about 30 years ago. I'll check out your site's articles.
I also e-mailed Rod Elliott (it's his 60W AB that I'm building) to see if he had any thoughts.
I've been hemming and hawing about buying the Self book. This is a good excuse to take the plunge.
I also e-mailed Rod Elliott (it's his 60W AB that I'm building) to see if he had any thoughts.
I've been hemming and hawing about buying the Self book. This is a good excuse to take the plunge.
pmkap said:
Actually, 'Self gave very good reasons for this distinction.....an unbiased SEPP stage infact operates in Class-C, as each of the complementary members are driven foward-active for slightly less than 180 degrees........and can only be, (if we follow this urgument through to its logical conclusion), considered to operate in class-B if each member is biased to conduct for as close to 180 degrees as practicable...
Class-AB bias can then be described with some precision, as that condition in which the SEPP is biased into class-A up to some arbitrary power output before one or other of the members in the SEPP arrangement cuts off...
It is well understood that a low bias AB usually gives better specs than a higher bias, but it is also seems that listeners generally prefer the sound of higher bias.
This conclusion is supported by several instances of single-blind comparisons of the same amplifier between low and higher bias.
This conclusion is supported by several instances of single-blind comparisons of the same amplifier between low and higher bias.
The reasoning is clear. In the A region of class AB, you have two active devices and hence a gm doubling, whereas the gain drops when class B mode is entered. Class AB with very high bias may still sound better because for normal listening levels, it is all A, and at higher levels the effect is so small it gets swamped.
What I haven't seen yet is a formula how optimum AB is really defined as a function of emitter resistance and load impedance.
What I haven't seen yet is a formula how optimum AB is really defined as a function of emitter resistance and load impedance.
I've built several class B amps over the past couple of years. Depending upon the components/topology used, I've noticed that some of them seem to literally have a bias null point. That is, when adjusting bias using a distortion analyzer, there appears to be a THD minimum (greater THD at both lower bias and greater bias). Other amps appeared not to have this minimum. Instead, they show greater THD at low bias, which decreases at some value, and then only further decreases at greater frequencies and greater powers, with a bias approaching a class AB designation. I therefore think each amp design needs to be examined for its own unique distortion characteristics with both measurements and listening tests.
I too, as NP explained, tend to go slightly above the just acceptable bias point if the heatsink-design goal will tolerate the greater setting.
As an example, one recent amp I made has an acceptable THD with a bias of about 50 mA. Going from 50 mA to 90 mA lowers THD from 0.0075 to 0.006%. Further increases, say up to 200 mA, slightly lowers the THD to about 0.0045% (all values for 1W into 8ohms). From this figure, THD doesn't seem to change even when increasing the bias to the +1A range. I cannot hear any difference when going above the 80 mA bias point. Since I wish to run the amp as efficiently as possible, I've settled on the 100 mA range for this particular design.
Regards, Robert
I too, as NP explained, tend to go slightly above the just acceptable bias point if the heatsink-design goal will tolerate the greater setting.
As an example, one recent amp I made has an acceptable THD with a bias of about 50 mA. Going from 50 mA to 90 mA lowers THD from 0.0075 to 0.006%. Further increases, say up to 200 mA, slightly lowers the THD to about 0.0045% (all values for 1W into 8ohms). From this figure, THD doesn't seem to change even when increasing the bias to the +1A range. I cannot hear any difference when going above the 80 mA bias point. Since I wish to run the amp as efficiently as possible, I've settled on the 100 mA range for this particular design.
Regards, Robert
paulb,
I have a brother who lives in Calgary!
Whether you get a distortion minimum at a particular bias current will depend upon the amp design and the devices used. It is a complicated system. It is quite possible that a low bias will measure better than a higher bias.
You need to listen to the design.
There are many distortion mechanisms at work, simultaneously, in a PP output stage. Some will improve with bias and some will get worse. It depends which you are measuring. You ears give you the net effect of these mechanisms and so the bias point that sounds best may well be different from the bias point that a particular measure predicts as best. Hence Nelson's observations.
I have a brother who lives in Calgary!
Whether you get a distortion minimum at a particular bias current will depend upon the amp design and the devices used. It is a complicated system. It is quite possible that a low bias will measure better than a higher bias.
You need to listen to the design.
There are many distortion mechanisms at work, simultaneously, in a PP output stage. Some will improve with bias and some will get worse. It depends which you are measuring. You ears give you the net effect of these mechanisms and so the bias point that sounds best may well be different from the bias point that a particular measure predicts as best. Hence Nelson's observations.
Re: The sweet spot!!!
I have seen no evidence that the 'sweet THD spot' is load dependent.....may be the case......however, it is clear that overbias is better than underbias, which is where your amp. is likely to drift, if you bias for the said 'sweet spot'.....better to go for class AB.....
Self's schemes for maintaining thermal stability over the long term are not very convincing......eventually bias will drift.....i would rather drift into Class AB than class C....
Tube_Dude said:
I have seen no evidence that the 'sweet THD spot' is load dependent.....may be the case......however, it is clear that overbias is better than underbias, which is where your amp. is likely to drift, if you bias for the said 'sweet spot'.....better to go for class AB.....
Self's schemes for maintaining thermal stability over the long term are not very convincing......eventually bias will drift.....i would rather drift into Class AB than class C....
I've never designed conventional audio power amplifiers (just one very unconventional one, see Electronics World February 1996), but my guess is that for a complementary emitter follower configuration, the optimum bias is the one which causes a voltage drop of a specific number of kT/q's across the emitter resistors, probably something close to ln(2)*kT/q across each emitter resistor.
In this equation, k is Boltzmann's constant (1.38065E-23 J/K), T is the absolute temperature and q is the electron charge (1.6022E-19 C). All of this assuming that the transistors are not in high injection in the quiescent point and that the influence of the emitter and base bulk resistances is negligible.
For MOSFET output stages and output stages with local feedback around two transistors in each half (CFP), I'd guess the optimum bias is not very reproducible, being dependent on the precise transistor parameters.
Marcel van de Gevel
In this equation, k is Boltzmann's constant (1.38065E-23 J/K), T is the absolute temperature and q is the electron charge (1.6022E-19 C). All of this assuming that the transistors are not in high injection in the quiescent point and that the influence of the emitter and base bulk resistances is negligible.
For MOSFET output stages and output stages with local feedback around two transistors in each half (CFP), I'd guess the optimum bias is not very reproducible, being dependent on the precise transistor parameters.
Marcel van de Gevel
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.