Previous post was what I was going to say except to add that you might want to check in to the question of Clas AB vs. pure Class B for the output stage. AB sounds like a good idea, but there is body of opinion that has an essential flaw that offsets the benefits with the result that pure B output stages workout a bit better.
Hi sam9,
Notice that I used a small a for class aB.
In my book :
class B has no bias and exhibits bad crossover distortion.
class aB has optimum bias for minimum distortion.
class AB means running part of the output deliberately in class A,
above the requirements for aB - I don't really want to go into
the distortion consequences here as it wasn't part of the question,
and also depends on how much is A and how much is B.
We are really talking about what is pure class B and how
class B differs from class C.
I tend to go with historical acceptance, I believe its Doug Self
that claimed unbiased class B was really Class C, but this ignores
the fact that classical class C in radio circuits only conducts for
half ! the cycle, so I only agree with him in the pedantic sense.
In the historical sense class B usually refers to bipolar motor
controllers with no bias, i.e. very low static dissipation.
Whatever, I say class aB, you say pure class B, same thing.
class A enhanced class B wasn't part of the question (yet!),
so let's leave it there ?
/sreten.
Notice that I used a small a for class aB.
In my book :
class B has no bias and exhibits bad crossover distortion.
class aB has optimum bias for minimum distortion.
class AB means running part of the output deliberately in class A,
above the requirements for aB - I don't really want to go into
the distortion consequences here as it wasn't part of the question,
and also depends on how much is A and how much is B.
We are really talking about what is pure class B and how
class B differs from class C.
I tend to go with historical acceptance, I believe its Doug Self
that claimed unbiased class B was really Class C, but this ignores
the fact that classical class C in radio circuits only conducts for
half ! the cycle, so I only agree with him in the pedantic sense.
In the historical sense class B usually refers to bipolar motor
controllers with no bias, i.e. very low static dissipation.
Whatever, I say class aB, you say pure class B, same thing.
class A enhanced class B wasn't part of the question (yet!),
so let's leave it there ?
/sreten.sam9 said:
Class-b can produce lower total distortion than a/b but since a/b pushes the crossover region higher up the curve it produces less distortion at those critical low levels than class-b. Also for very low distortion in class-b the bias setting is fairly critical requiring tight control of Iq, over-biased a/b is more forgiving in this respect
Heavy class AB is best in my opinion, and I design the majority of my power amps that way. There 'can' be a problem with the crossover from class A to class B. I first found this out 35 years ago, when I used too large of value emitter resistors in my first personal amp design. Heck, I thought that 1 ohm with a .5A standing current was OK, but NOOOOO! Actually I should have used a .05 ohm resistor or so, for best transition. This is because an emitter follower, when voltage driven, prefers a small emitter resistor that equates to about .015-.025V when the amp is idling. The problem with this solution, is that thermal runaway of the output stage is possible, UNLESS you do your bias thermal compensation right. Still, it is possible to get an almost perfect transition from class A to class AB if you do it right.
Anyway getting back to the original question, here are two kits described as class AB
http://info.hobbyengineering.com/specs/DIY-k48.pdf
or
http://www.aksaonline.com
Actually, I'm not sure there are all that many expressly Class-AB diy projects published although there are exables of Class A or Class B amps biased into AB (deliberately or otherwise). AB seems a lot more common anong tube amp builders.
http://info.hobbyengineering.com/specs/DIY-k48.pdf
or
http://www.aksaonline.com
Actually, I'm not sure there are all that many expressly Class-AB diy projects published although there are exables of Class A or Class B amps biased into AB (deliberately or otherwise). AB seems a lot more common anong tube amp builders.
millwood said:
Have a look at http://www.diyaudio.com/forums/showthread.php?postid=247900#post247900
Steven
Steven said:
thanks, Steven for the reference.
I do have a quick question. The article seems to be concerned about open-loop operation of a class b amp. Will the same hold true when feedback is present?
I did a quick simulation on a modified citation 12 (with a t driver section that works in class a). lowest thd is obtained at about 29mv (130ma iq over 0.22 ohm emitter resistor). so I suppose that it holds true in closed loop as well.
Hi millwood,
Yes, the same holds true for amps with feedback. Maybe even more. Discontinuities in the crossover region of pushpull output stages manifest themselves in higher order harmonic distortion. The sharper the discontinuity, the higher the order of the harmonics in the distortion spectrum. Feedback is a circular mechanism, the output signal is fed back to the input and has to travel along the forward path of the amplifier again, and again, and again... The amplitude of the lower order harmonics will decrease by feedback, but the amplitude of the higher harmonics may even increase, because harmonics of the harmonics are created every time the signal is subjected again to the forward path non-linearity. In general this is the big problem of using feedback in audio amplifiers. While low order harmonic distortions are very well acceptable, since our own ears suffer from low order harmonic distortion too, higher order harmonic distrortions tend to be very annoying, especially the odd harmonics from the fifth upwards.
So, also in amplifiers that use feedback, it is very desirable to keep the open loop distortion as low as possible to avoid the creation of many high order distortion products by the feedback mechanism itself.
Steven
Yes, the same holds true for amps with feedback. Maybe even more. Discontinuities in the crossover region of pushpull output stages manifest themselves in higher order harmonic distortion. The sharper the discontinuity, the higher the order of the harmonics in the distortion spectrum. Feedback is a circular mechanism, the output signal is fed back to the input and has to travel along the forward path of the amplifier again, and again, and again... The amplitude of the lower order harmonics will decrease by feedback, but the amplitude of the higher harmonics may even increase, because harmonics of the harmonics are created every time the signal is subjected again to the forward path non-linearity. In general this is the big problem of using feedback in audio amplifiers. While low order harmonic distortions are very well acceptable, since our own ears suffer from low order harmonic distortion too, higher order harmonic distrortions tend to be very annoying, especially the odd harmonics from the fifth upwards.
So, also in amplifiers that use feedback, it is very desirable to keep the open loop distortion as low as possible to avoid the creation of many high order distortion products by the feedback mechanism itself.
Steven
If your going to tweak then the bias voltage should remain
fairly constant. Optimium Vbias changes slightly with different
Re's and consequently Iq's, but not by much.
So varying the emitter resistors will also vary Iq.
If you keep Iq constant and vary the resistors you will be
varying Vbias.
Also true for keeping the resistor the same and varying Iq.
/sreten.
fairly constant. Optimium Vbias changes slightly with different
Re's and consequently Iq's, but not by much.
So varying the emitter resistors will also vary Iq.
If you keep Iq constant and vary the resistors you will be
varying Vbias.
Also true for keeping the resistor the same and varying Iq.
/sreten.
Curl is right. There is a theoretical optimum. And there is an "artful optimum" that is higher.
For voltage-drive, in silicon, theory says around 25mV-30mV, depending on the exact transistor.
Lower leads to distortion and, in practice, thermal instability.
Higher sounds better to the ear.
1V is way too high. (John's not the only one to go too far.) 1V can make a sub-optimum Class A stage, but sucks for any Class "B" operation.
50mV is a nice semi-safe trial value for experimenters. If the bias junction(s) is well coupled to the output devices, then bias stability is fine, idle current is on the safe/better side of the theoretical optimum, and the ear likes this side of the "optimum" better than the theoretical optimum.
> should I adjust the bias-current, or... change resistor size?
Figure the idle current you want. Higher is almost always better. Limit is how hot you want the beast to idle. You usually do not want it either hot nor cold, just warm. (You certainly do not want to run the devices down where Beta or Ft is dropping rapidly.) Calculate the idle current from the idle-power. The emitter resistors are then 0.05V/Ibias. In smoke-test, start at much lower bias and let it cook. You probably have humm-busting and other chores to do first. When it basically passes OK audio, sneak up to half the calculated current, 0.025V per resistor. Try that a while, then try 0.05V per resistor. The difference should be subtle.
Changing resistor sizes is really a re-design, and it won't be easy to make meaningful comparisons with such major change.
For voltage-drive, in silicon, theory says around 25mV-30mV, depending on the exact transistor.
Lower leads to distortion and, in practice, thermal instability.
Higher sounds better to the ear.
1V is way too high. (John's not the only one to go too far.) 1V can make a sub-optimum Class A stage, but sucks for any Class "B" operation.
50mV is a nice semi-safe trial value for experimenters. If the bias junction(s) is well coupled to the output devices, then bias stability is fine, idle current is on the safe/better side of the theoretical optimum, and the ear likes this side of the "optimum" better than the theoretical optimum.
> should I adjust the bias-current, or... change resistor size?
Figure the idle current you want. Higher is almost always better. Limit is how hot you want the beast to idle. You usually do not want it either hot nor cold, just warm. (You certainly do not want to run the devices down where Beta or Ft is dropping rapidly.) Calculate the idle current from the idle-power. The emitter resistors are then 0.05V/Ibias. In smoke-test, start at much lower bias and let it cook. You probably have humm-busting and other chores to do first. When it basically passes OK audio, sneak up to half the calculated current, 0.025V per resistor. Try that a while, then try 0.05V per resistor. The difference should be subtle.
Changing resistor sizes is really a re-design, and it won't be easy to make meaningful comparisons with such major change.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.