low-biasing or high biasing in Class-AB amps

kASD

Member
2004-12-25 12:24 pm
N/A
Hi guys....,

every time you build amps especially Class-AB , biasing plays an important role in the sound quality of amplifiers in terms crossover distortion elimination... many times we came accross good designs which are both low-biased ones and some are high-biased ones..what i refer to low Biasing is the idle current of output devices between 10mA to 65mA and above 65mA upto 200mA as high -biasing....
Biasing helps in elimination of cross-over distortion of output devices.....whether its a BJT , Mosfet or an IGBT.... it certainly plays an important aspect of sonic perspectiveness of a Class-AB amp.....

So guys what are your comments regarding the biasing of an amplifier , does it should be high or low....

regards,
Kanwar
 
Generally the higher the better! In a class A power amp the amount of biasing is still critical, not so much as to eliminate x-over distortion since in class a there is really none but the amount of biasing still affects the overall sound of the amp. Wade through the KSA-50 thread for more info about biasing...... its at 240+ pages right now......

Mark
 
I prefer low bias. The lower I can get away with the better. Increased bias has diminishing returns for sound quality (and in fact distortion increases above the optimum value for BJTs). Lower bias increases efficiency, and I hate wasting power.

It all depends on the situation and the amp itself. Some amps don't allow any extra biasing because opf limitations discussed below in the next paragraph. So yes, in these amps the lower the better, but many high end amps actually perform better and measure better at higher bias settings. Some factorys tend to be conservative on this and set them lower for greater reliability. They have a warranty to provide ya know.



If you go with Graham's definition, you can bias up any Class AB amp to function as a Class A amp.

Not totally true unless you biased a class AB amp and kept it at a very low power level. Keep in mind that most class AB amps would not stand being biased at class A. The power supply would be too small, ditto for the heat sinks and output stage. All of these are generally alot smaller for class AB than for class A. Threshold used to make two versions of the S-200. One was 50 watts RMS class a and the other was 100 watts RMS class AB high bias.

When I refer to the higher the better it is usually determined by looking at harmonic content on a spectrum analyzer while adjusting the bias level itself. In most cases it takes more bias to achieve low even order harmonics if that can be achieved at all. Some commercial amps are not really clean at any bias setting so standard manufactuirers settings are then the best to follow.

Mark
 
Nice thread Kanwar!

"I prefer low bias. The lower I can get away with the better. Increased bias has diminishing returns for sound quality (and in fact distortion increases above the optimum value for BJTs). Lower bias increases efficiency, and I hate wasting power."

- Well said and commendable , Mr Evil.


A good design eliminates the need for wasteful high bias, and I typically aim for 25-50 mA while meeting other design requirements. IMHO it is an intrinsically poor design that needs to waste considerable power to achieve listenable quality.

Consider a 50V rail amplifier (maybe 100W/8ohms) that has poor crossover characteristics and/or poor PSRR so output stage/ power supply commutation products appear in the output ( usually a spray of even harmonics, i.e. 2nd, 4th,6th,8th,10th...),
the product doesn't sound real good so they decide to turn up the bias so it runs Class A to 4W (about the crest to mean for programme just about to clip a 100W amp). So they need to run the output stage at 1amp continuously! The output stage retaliates by dumping 100W of heat into the environment via the heatsink.

Does it sound better? probably. But it hasn't eliminated the problem if it's commutation because it will simply occur at a higher level. So keep it turned down.

An intrinsically better design wouldn't need such a prop.

Greg
 

kASD

Member
2004-12-25 12:24 pm
N/A
Thanks to all of you for the wonderful comments and opinions!


Hi Greg,

I completely agree with you about the designing of amplifiers plays a vital role in the way in which it is designed. A good design in terms of PSRR, Cross-over characteristics would be perfectly alright with the low-biasing indeed.

In normal pushpull Class-AB the cross-over characteristics very much define the nature of the sound quality and low- biasing just acts as a catalyst to get the right thing at the output......

What is your opinion.....

regards,
Kanwar
 
I'm not sure high vs. low is meaningful in all cases.

In L-MOSFET higher is better until it all melts down in a pool of molten slag.

In a CFB configured BJT appearently the there is a very definite optimal bias point that can't be deviated from too far with out increasing distortion. Hopwever, thermal tracking is acomplished fairly easily. It is probabaly best to neither under or over bias.

In a EF configured BJT the first concern is to prevent thermal runaway since survival is a precondition to hearing anything.:D . Accurate thermal tracking is a non-trivial exercise (and pain in the neck!) and the usual practice is to under bias with the idea that is the safest course. As thermal conditions change the lag in thermal tracking will actually result in over bias some of the time. EF biasing seems like it should be a real mess, but there is a considerable tolereance in both directions from the optimal bias before the effects become appearent. Under real conditions EF biasing works better than one thinks it should.
 
Sam9 "I'm not sure high vs. low is meaningful in all cases.

In L-MOSFET higher is better until it all melts down in a pool of molten slag. "

A good design tailored around such devices with a consumnate understanding of their attributes and limitations need not be compromised. I can use IRFP240/9240 at 25mA and achieve < 0.005% - as a starting point! By design, then finesse it.

It might use one or two more 10c transistors than the simplest designs but compared with some of the more obscene offerings out there .....

Run, don't walk away from any suggestion you need to consume 200W-900W of power to produce 100W of 'nirvana'.
 
Hi all,

Douglas Self has studied this question. His conclusion was, that for the most linear take-over of the NPN to the PNP part and vice versa, you need a bias that depends on the emitter resistors of the output stage. In other words, the important issue is not the bias current, but the bias voltage across the Re's.

If you think about it, that makes sense, because the voltage drop across the Re's determines the cutt-off characteristics.
I don't remembert exact figures (I am at work), but I think for a CC output stage the figure was 20-30 mV across each Re, and for a CE output stage about 60 mV across the Re's.

Jan Didden
 

djk

R.I.P
2001-02-04 4:23 am
USA
I agree with janneman and Self.

John Curl has also made the same point, and had a reference from HP to back him up.


I did my own tests and came up with the same results. I had to use an IM analyzer as the THD was too low to measure. The distortion decreased as bias increased up 'till about 10~20mV across the emitter resistors, then started going up as the bias was increased further.
 
amplifierguru said:
Well he would know?

What if I proliferated endless graphs and came to some lame conclusion? Pays well a series in EW, yes?


Well, I don't know if I understand your point. Isn't the object to get the most linear transition from B to A? I mean, if you have no clear objective, anything goes. BTW, 20mV across 0.22 ohms is about 100mA bias, so I think my memory is off, because I'm sure it was less.but I will check tonight.

The arguments in the article were very clear and convincing to me. If you dont agree with Self, which I'm sure many people don't, please tell us why so we can also learn.

Jan Didden
 
amplifierguru said:
pls disregard #12,, too explicit.


100mA is not High Bias by any stretch, but good to see the detractors come out of the woodwork. What is this nit-pickers corner.

I must remember the more published guff the more cred...

Ho Hum.

???? Publishing your stuff for all to see and pick on puts you under scrutiny of those that are your peers, that really know what they are talking about. You really have to have a strong case otherwise you are shot down in flames. So I can't see why that is bad.

Anyway, I didn't want to imply that 100mA is strong bias, just that I think it is a bit high for the customary class AB we mostly see, and I am sure it is higher than Self recommends. It all depends on your definition of course, one man's ceiling is another man's floor as Paul Simon used to sing. But let me check first this afternoon.


Jan Didden
 

Rudy

Member
2003-02-27 10:16 pm
Belgium
An advantage in high bias is temp stability off the devices/heatsink, the difference in power dissipation between idle and playing music is none or very little, so the devices or not/less thermical stressed. And this is a big advantage in term off livespam for the output stage.

Rudy
 
Rudy said:
An advantage in high bias is temp stability off the devices/heatsink, the difference in power dissipation between idle and playing music is none or very little, so the devices or not/less thermical stressed. And this is a big advantage in term off livespam for the output stage.

Rudy



Rudy,

I think with competent design nowadays thermal stability is no longer an issue. As to thermal stressing, I would love to have some info on that. I have never seen an amp failing because of thermal stressing/cycling, so I wonder if that would be a reason to select a certain bias.

Jan Didden