...present?
This question is obviously still not investigated. Also not in relation to different Hfe values.
Both at low and higher idle currents.
Both at single ended and push-pull buffer versions.
Who knows more about this?
What spread range would still be just acceptable?
Please note, the emitter resistors in most cases lower than 1 Ohm, sometimes even not exist.
Some guys tell me, the only way to avoid unwanted THD/IM effects through the spread of Ube voltage values and the AC current gain factor "hfe" never use paralleling output power devices (no matter whether BjT's or MOSFET's; only paralleling complete amplifiers include the front end is possible without unwanted selecting of transistors and unwanted THD enhancing at the same time. This is often to observe by IC amps like TDA7293 or LM3886/LM4780. But by discrete amps this I have never see.
An other way it could be to use a large amount of paralleling (small/medium signal) transistors with emitter resistor of 10 - 30 ohms. But I am not shure. By simulation unfortunately I am not able to make a variation of Ube voltage and hfe.
This question is obviously still not investigated. Also not in relation to different Hfe values.
Both at low and higher idle currents.
Both at single ended and push-pull buffer versions.
Who knows more about this?
What spread range would still be just acceptable?
Please note, the emitter resistors in most cases lower than 1 Ohm, sometimes even not exist.
Some guys tell me, the only way to avoid unwanted THD/IM effects through the spread of Ube voltage values and the AC current gain factor "hfe" never use paralleling output power devices (no matter whether BjT's or MOSFET's; only paralleling complete amplifiers include the front end is possible without unwanted selecting of transistors and unwanted THD enhancing at the same time. This is often to observe by IC amps like TDA7293 or LM3886/LM4780. But by discrete amps this I have never see.
An other way it could be to use a large amount of paralleling (small/medium signal) transistors with emitter resistor of 10 - 30 ohms. But I am not shure. By simulation unfortunately I am not able to make a variation of Ube voltage and hfe.
Last edited:
I doubt there is any significant THD-difference.
For most of the time VGS of mosfets is matched to within 100mV to guarantee that all mosfets turn on. Source degeneration resistors take care of current sharing. For bjts this is pretty pointless due to tiny variations in VBE.
Matching is not always done to decrease THD. It's mostly done for the peace of mind
- and in some distortion cancellation circuits (differentials...)
But that's only the technical side of things, so the stuff you can actually measure. Others may give complementary input.
Have fun, Hannes
EDIT:
I dare to disagree. Even if there was data posted here, most would plainly ignore it. Matching is poor man's circuit optimization and I don't have a problem with that.
For most of the time VGS of mosfets is matched to within 100mV to guarantee that all mosfets turn on. Source degeneration resistors take care of current sharing. For bjts this is pretty pointless due to tiny variations in VBE.
Matching is not always done to decrease THD. It's mostly done for the peace of mind
- and in some distortion cancellation circuits (differentials...)But that's only the technical side of things, so the stuff you can actually measure. Others may give complementary input.
Have fun, Hannes
EDIT:
I dare to disagree. Even if there was data posted here, most would plainly ignore it. Matching is poor man's circuit optimization and I don't have a problem with that.
Last edited:
If parallel BJTs have slightly different Vbe or current gain then the main effect is that one will do more work than others. Give the typical local feedback around the output stage there will be little effect on distortion.
If the opposite sides of a P-P output have different gains then the main effect will be a small rise in even-order distortion. Local feedback will deal with most of this in a well-designed amp, then global feedback will finish the job.
The reason nobody seems to have done any work on this is that it is really a non-problem.
If the opposite sides of a P-P output have different gains then the main effect will be a small rise in even-order distortion. Local feedback will deal with most of this in a well-designed amp, then global feedback will finish the job.
The reason nobody seems to have done any work on this is that it is really a non-problem.
Don't think Vbe matching is critical. Emitter resistors are needed to share the current and if the current gets to be high in one device the voltage drop increases to force the other(s) to take more current.
If it is a problem you probably have not got large enough emitter resistors.
Hfe matching is needed to reduce 2nd harmonic distortion. I recommend hfe matching as any amplifier usually benefits, but by how much depends on the design: simpler/older circuits benefit more if they have lower open loop gain than more recent designs. You can still make an improvement e.g. from .002% to 0.001% though...
John
If it is a problem you probably have not got large enough emitter resistors.
Hfe matching is needed to reduce 2nd harmonic distortion. I recommend hfe matching as any amplifier usually benefits, but by how much depends on the design: simpler/older circuits benefit more if they have lower open loop gain than more recent designs. You can still make an improvement e.g. from .002% to 0.001% though...
John
Member
Joined 2009
Paid Member
I think it depends on what kind of output stage you're talking about. A class AB output stage implies that the BJTs are optimally biassed so that there is roughly 26mV across each emitter resistor. I believe it's a widely held opinion that under-biassing in Class AB is bad news but although D. Self says over-biass is also bad, most don't find this an issue. So, if the Vbe of the devices is all over the map you have the danger of some of them producing bad cross-over distortion. If you bias the whole thing high enough to avoid that problem I would expect mostly a large benefit from reduced current swing in each device compared with a single device.
You may be able to simulate this by using different emitter resistor values in Spice even though each BJT is identical they will have different turn-on voltages and bias current requirements.
You may be able to simulate this by using different emitter resistor values in Spice even though each BJT is identical they will have different turn-on voltages and bias current requirements.
This fact I can confirm by various attempts. By enhancing the value of emitter resistors an additional advantage I note: The increasing of stability of idle current.
Are there any experiences, from which ohm value of emitter resistor thermal compensation for idle current is no longer necessary (perhaps between 3 and 10 ohms) ?
The aim of the question is follow: I want to build three output power stages (to investigate differences in various kind, among others the sound quality and reliability):
One with one NPN/PNP high power BjT pair,
one with three medium power NPN/PNP pairs and
one with a massive number of parallel pairs in the to126 outline.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.