I know there has been a lot of discussion about the optimum class AB bias current. According to the books of both Cordell and Self the optimum bias current for a class AB (BJT) push pull amp is with approx. 26mV over Re.
I have repaired several commecial amps and remembered the service manual stating much lower Re voltage for bias adjustment. I checked some service manuals:
Sony TA-N77ES 15mV // 0.47E
Sony STR-DA50ES 6mV // 0.22E (SAP15)
Denon AVR-1909 2mV // 0.47E/2
Rotel RB1090 5mV // 0.22E
Rotel RMB1095 5mV // 0.22E
Harman Kardon AVR366 26mV // both emitter resistors (0.27E * 2)
I assume thos manufacturers have access to decent measurement equipment, but they all seem very underbiased.
I can imagine they take some temperature rise into account, but f.i. the Harman Kardon uses the SAP15 which has internal diodes and is expected to be more stable over temperature.
Als checked some Marantz with SAP17 transistors, which show similar figures.
Any explanation for this... of course higher Iq is more expensive due to heatsink requirements, but f.i. a Sony TA-N77ES was their top model and still seems under biased.
Who knows why ?
Is the optimum AB bias in the majority of cases not the optimum ?
I have repaired several commecial amps and remembered the service manual stating much lower Re voltage for bias adjustment. I checked some service manuals:
Sony TA-N77ES 15mV // 0.47E
Sony STR-DA50ES 6mV // 0.22E (SAP15)
Denon AVR-1909 2mV // 0.47E/2
Rotel RB1090 5mV // 0.22E
Rotel RMB1095 5mV // 0.22E
Harman Kardon AVR366 26mV // both emitter resistors (0.27E * 2)
I assume thos manufacturers have access to decent measurement equipment, but they all seem very underbiased.
I can imagine they take some temperature rise into account, but f.i. the Harman Kardon uses the SAP15 which has internal diodes and is expected to be more stable over temperature.
Als checked some Marantz with SAP17 transistors, which show similar figures.
Any explanation for this... of course higher Iq is more expensive due to heatsink requirements, but f.i. a Sony TA-N77ES was their top model and still seems under biased.
Who knows why ?
Is the optimum AB bias in the majority of cases not the optimum ?
It would be an interesting exercise for those with the necessary distortion measuring equipment to actually see if the commercial amps really did perform better when the bias was at the theoretical optimum rather than the value stated in the service manual.
I suspect practical commercial constraints are at work in what you observe. Firstly I doubt whether any critical listener could identify a moderately under biased from a correctly biased stage.
Heat build up is a killer for electronics and manufacturers know from experience that customers will place magazines and so on over ventilation slots and enclose the amp in an unsuitable rack...... the list goes on. An amp idling at 120ma with 90v rails gets a whole lot hotter than the same at idling at 20ma.
So they play safe, it allows them to use smaller PSU and heat sinking and it also minimises warranty repairs.
I suspect practical commercial constraints are at work in what you observe. Firstly I doubt whether any critical listener could identify a moderately under biased from a correctly biased stage.
Heat build up is a killer for electronics and manufacturers know from experience that customers will place magazines and so on over ventilation slots and enclose the amp in an unsuitable rack...... the list goes on. An amp idling at 120ma with 90v rails gets a whole lot hotter than the same at idling at 20ma.
So they play safe, it allows them to use smaller PSU and heat sinking and it also minimises warranty repairs.
and those can be different topology.
An EF is quite different from a common emitter (CE). Go back and read D.Self.
...Not to mention that some types of output transistors have built-in emitter resistance so the calculation based on the drop over Re is off.
To add to the confusion, and very dependent on topology, the reactivity of the load also has it's contribution, and we know real speakers vary in this respect with almost any parameter you care to think of. I have found it is good practice to use a 'speaker simulator' in addition to a dummy load to adjust Iq using the distortion residual as a guide.
To add to the confusion, and very dependent on topology, the reactivity of the load also has it's contribution, and we know real speakers vary in this respect with almost any parameter you care to think of. I have found it is good practice to use a 'speaker simulator' in addition to a dummy load to adjust Iq using the distortion residual as a guide.
Depending on the circuit topology and parts ( it is not enough to state if its EFP or CFP or EF2 or EF3 ) there is quite a few other things that might be involved.
If you remove the thermals from the picture ( let us suppose that no amplifier has a small or marginal heat sink ) there is many things to observe to make a comparison between bias A next to to bias B except produced temp.
In reality from measuring aspects expect distortion ( many types of it ) to change may be slew rate and content of harmonics ....change doesn't always mean for the best For example you might get less distortion and more harmonics that you don't like.
Expect though that for a given amplifier these results could partially be simulated and for real results these should be measured on the given construction /wiring, arrangement and so on .... ( everything will eventually play some role )
Repair experience from lets say 600-700 amplifier per year will add 2 more cents ...
In words of saying often we used our clients as "lab rats" to gain non polarized feedback.
.... We intentionally gave replacement amplifiers to the clients while their amplifier was under repair from the "wrong" region ...
---We gave someone that was listening class B all his life a class A amplifier and the opposite...
---We gave intentionally an amplifier that was originally underbiased from factory, after the repair with the bias pushed as much is possible on the edge of thermal limit ...
---we overbiased or underbiased ( in reference with the original schematic )many amplifiers for clients and gave them away for evaluation .
---We gave to clients that had low bias mosfet amplifiers high bias BJT amplifiers ...as replacement amplifiers
Except intentional ideas made, we have suggested/asked clients please listen to this or that evaluate and give us your feedback That is also done many many times ....
All these are done to gain feedback and healthy non polarized samples from clients listeners
Results are like that :
--Listeners that like low instrument count music IE a piano and a voice a sax and a voice will be very happy with Class A and consider a slight boost of bias us an upgrade to their system
--Listeners that go for electronic music or high instrument count will be happy enough with amplifiers biased like the English man describes ( which actually a terrible way to bias amplifiers ) and in general with marginal low bias
---Higher bias tends to favor middle area, low instrument count ,and lower speed, while lower bias all the opposite...
kind regards
Sakis
If you remove the thermals from the picture ( let us suppose that no amplifier has a small or marginal heat sink ) there is many things to observe to make a comparison between bias A next to to bias B except produced temp.
In reality from measuring aspects expect distortion ( many types of it ) to change may be slew rate and content of harmonics ....change doesn't always mean for the best For example you might get less distortion and more harmonics that you don't like.
Expect though that for a given amplifier these results could partially be simulated and for real results these should be measured on the given construction /wiring, arrangement and so on .... ( everything will eventually play some role )
Repair experience from lets say 600-700 amplifier per year will add 2 more cents ...
In words of saying often we used our clients as "lab rats" to gain non polarized feedback.
.... We intentionally gave replacement amplifiers to the clients while their amplifier was under repair from the "wrong" region ...
---We gave someone that was listening class B all his life a class A amplifier and the opposite...
---We gave intentionally an amplifier that was originally underbiased from factory, after the repair with the bias pushed as much is possible on the edge of thermal limit ...
---we overbiased or underbiased ( in reference with the original schematic )many amplifiers for clients and gave them away for evaluation .
---We gave to clients that had low bias mosfet amplifiers high bias BJT amplifiers ...as replacement amplifiers
Except intentional ideas made, we have suggested/asked clients please listen to this or that evaluate and give us your feedback That is also done many many times ....
All these are done to gain feedback and healthy non polarized samples from clients listeners
Results are like that :
--Listeners that like low instrument count music IE a piano and a voice a sax and a voice will be very happy with Class A and consider a slight boost of bias us an upgrade to their system
--Listeners that go for electronic music or high instrument count will be happy enough with amplifiers biased like the English man describes ( which actually a terrible way to bias amplifiers ) and in general with marginal low bias
---Higher bias tends to favor middle area, low instrument count ,and lower speed, while lower bias all the opposite...
kind regards
Sakis
Last edited:
This subject seems immortal; but surely the aficionados of over-biasing - to or beyond the 'knee' of the drivers, must consider the source of the bias current. Taken as it is from the intermediate stage (VAS), where the current source is substantially limited, over-biasing will serve the opposite purpose intended, by introducing current limitations where it would best be used - in maintaining good VAS linearity. If even moderate over-biasing is desired, would it not make sense to source the required current separately ?
I looked closely at this optimal bias business in the case of an op amp based amplifier with a CFP output stage and massive feedback.
LTSpice simulations did not show me any optimum biais.
Changing biais changes the tdh20khz versus power curve, but there is no clear optimum, all can be seen is more a matter of taste.
Because thoses thds are in the 0.00x % i doubt there is any audible optimum.
So, I dug the original Oliver's paper to see what's all about.
The reasoning is flawned, the over tutted optimum is not defined, it is just handwaved and replaced by some equivallence.
Then comes equations and finally approximations to pop some results. This has no better value then LTSpice simulations which are using the same bjt equations with no approximation.
So I disregard the Oliver story and consider my simulations far more meaningfull. This in the case of the topology I have chosen.
Try to find an optimum biais about your favorite topolology, this should be better than faith in the"optimum Oliver biais".
LTSpice simulations did not show me any optimum biais.
Changing biais changes the tdh20khz versus power curve, but there is no clear optimum, all can be seen is more a matter of taste.
Because thoses thds are in the 0.00x % i doubt there is any audible optimum.
So, I dug the original Oliver's paper to see what's all about.
The reasoning is flawned, the over tutted optimum is not defined, it is just handwaved and replaced by some equivallence.
Then comes equations and finally approximations to pop some results. This has no better value then LTSpice simulations which are using the same bjt equations with no approximation.
So I disregard the Oliver story and consider my simulations far more meaningfull. This in the case of the topology I have chosen.
Try to find an optimum biais about your favorite topolology, this should be better than faith in the"optimum Oliver biais".
Last edited:
I agree.
I also agree ltspice is very useful, I am sure Oliver would have welcomed the benefits of spice
Look at the thd curves with different signal levels, there is some peaking going on usually. The place where the peaking occurs is dependent on bias, so that may lend towards an optimum point for your application. But to me the larger the class "A" region is usually the better, but at a cost of heat, that increase costs. Like designing with latfets.
I always understood the "Oliver Criteria" as a rule of thumb more than any sort of rigorous bias point. In other words, it's probably a bit high bias compared to whatever "optimum" would be, but being a bit high is far safer than accidentally ending up underbiased. Also, SPICE can get you close, but unless you have very well characterized transistors around the operating parameters, it just going to give you a ballpark value rather than any optimum. Gotta tweak from there (or leave it a bit hot and not sweat it)
Also CFP's tight feedback loop is going to mess with biasing in any way, shape, or form.
Also CFP's tight feedback loop is going to mess with biasing in any way, shape, or form.
It seems that most would agree that 'over-biasing' does sound better than optimal - and I must say, over biasing sounds better to me too.
So the question would be: to what extent should one over-bias ?
I've read several accounts of people setting the level to just beyond the 'knee point' so that the transfer of signal goes from linear region to linear region - this to ensure (hopefully) the best, (smoothest?) sound quality.
I am almost persuaded to fit a front panel control with which to adjust bias either infinitely or in discrete steps.
A question to DPH: Daniel, can you say why CFP output would mess with the biasing ?
BTW: Who is Oliver ?
Another point is, would it make more sense to source the over-bias currently separately, ie: NOT from the VAS CC supply ?
So the question would be: to what extent should one over-bias ?
I've read several accounts of people setting the level to just beyond the 'knee point' so that the transfer of signal goes from linear region to linear region - this to ensure (hopefully) the best, (smoothest?) sound quality.
I am almost persuaded to fit a front panel control with which to adjust bias either infinitely or in discrete steps.
A question to DPH: Daniel, can you say why CFP output would mess with the biasing ?
BTW: Who is Oliver ?
Another point is, would it make more sense to source the over-bias currently separately, ie: NOT from the VAS CC supply ?
Last edited:
It's impossible to separate the bias of the driver from the output in a CFP, plus the local feedback shrinks the overlap region substantially. In general, the overall distortion minima (or at least at max power) for a CPF output stage has notably less bias than an equivalent EF2. This has major implications at lower power outputs, though.
Oliver: Bernard M. Oliver - Wikipedia
I think the degree of overbias is up to the user's preference and/or the thermal/PSU limitations of one's setup. But I'd also not obsess with it to a level that drives you mad (aka bringing it out to the front panel).
Can't say one way or another to the latter question.
Oliver: Bernard M. Oliver - Wikipedia
I think the degree of overbias is up to the user's preference and/or the thermal/PSU limitations of one's setup.
But I'd also not obsess with it to a level that drives you mad (aka bringing it out to the front panel).Can't say one way or another to the latter question.
In fact I had seriously considered CFP output, exactly because it so reduces the X/O region, requires something like a tenth of the bias normally applied to EF stages, and produces less X/O distortion too. A class A amplifier I used to own some years ago had CFP output stages - and was probably the best sounding amp I've ever owned.
BTW, I was kidding about the front panel bias control .....
BTW, I was kidding about the front panel bias control .....
I'm certainly not saying that I'm anti-CFP, just that the bias-distortion curve looks different. And the VBE doubling region changes dramatically. Class A obviates that though.
If Geddes is to be believed (and I am willing to buy in), then it's really much more important to leave the bias fat regardless of whether we have ever-so-slightly higher peak power distortion. (Whoop de doo, IMO)
Here at DIYAudio, there's enough people that would have been dead serious about the bias tweaker that I didn't know.
If Geddes is to be believed (and I am willing to buy in), then it's really much more important to leave the bias fat regardless of whether we have ever-so-slightly higher peak power distortion. (Whoop de doo, IMO)
Here at DIYAudio, there's enough people that would have been dead serious about the bias tweaker that I didn't know.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.