need drivers to be faster than outputs (all bjt)?

[h_a]

I got the hint from another member here that - to avoid oscillations - one needs to ensure that the drivers have a greater gain-bandwidth product than the outputs.

Well obviously the VAS needs then to be faster than the drivers too (all bipolar).

While it sounds reasonable I cannot understand that with Bode-analysis, so I would love to hear your experiences!

Have fun, Hannes

PS: by burning bandwidth with frequency compensation everything is possible - so lets say the amplifier is not (peculiarly) compensated.

 

[Andy L. Francis]

Need drivers to be faster than outputs?

Briefly: yes.

 

[h_a]

Thank you for your reply!

Is this your own experience with amps? As far as I know you're happily building circuits so I guess your experience is first hand?

Applies the same rule also to the VAS? In the sense VAS faster than drivers benefits stability?

Of course if there's an explanation I'm always keen to hear it - don't worry I don't feel bored hearing a bit theory.

Have fun, Hannes

EDIT: just for the sake of completeness I found this thread
http://www.diyaudio.com/forums/showthread.php?s=&threadid=57713&highlight=
It didn't came very far unfortunately.

 

[myhrrhleine]

As a 'rule-of-thumb' use the best part you can get.
Higher Ft if available.
More stable Hfe (stays the same for different currents)
Higher Hfe
etc.
This device performance is often limited in devices with higher voltage, current, or power, which is why we end up with drivers faster than outputs, Vas faster than drivers, LTP faster than Vas, etc.

 

[h_a]

Thanks for the reply Myhrrhleine, I'm perfectly aware of this!

I know that voltage+current requirements usually lead to this choice - however I'm more interested how that affects stability.

I do not know an amp at the moment which shows exactly this ft-stage decline, but I could imagine that such an amp - with sufficiently simple topology introducing only little phase shift - could do without any frequency compensation at all?*

Have fun, Hannes

*just as a couple of mosfet amps, where the device parasitics are sufficient to give stability.

 

[myhrrhleine]

The frequency compensation is just for convenience.
Each device adds to the overall phase error.
Using a compensation capacitor gives a simple control.
Whereas no compensation puts the dependancy on the transistors which have varying capacitance depending on many factors.
you could end up with an amp which is stable at only one frequency orpower level , or maybe not stable at all, ever.

 

[h_a]

People what are your experiences??

So many pro-amp guys here! C'mon please share your experiences!

Does this speed order help generally to get an amp stable?


@myhrrhleine

Each device adds to the overall phase error.

That's the reason I assumed a simple topology where the devices alone do not accumulate 180° phase shift. Per definition this is stable under all conditions. So I was interested if in this special case one could cause instabilities by choosing slower transistors for the VAS (slower than drivers, drivers slower than outputs).

As far as I know this case is not covered by Bode-analysis, so I would like to hear the experiences of the people here which are more lucky than me and have far more real building experience.

If I had the time and a oscilloscope at hand, I would do it on my own - but that's not possible at the moment

Have fun, Hannes

 

[wg_ski]

In many situations, it is necessary to have the drivers faster than the outputs. With ANY type of compound-feedback arrangement it is required that the first transistor in the pair (or triple) be fast with low output capacitance. It's that output capacitance and the feedback that causes trouble. With emitter followers, it's not uncommon to have drivers that are about as fast as the outputs, and not have any trouble. In an EF, each "stage" (even in a triple) has its own local feedback - it is not combined. You just need to keep an eye on the total phase shift. When I build amps myself, I take the lead capacitor from the VAS output and not the outputs. With triples, that buys you some phase margin as the HF feedback doesn't have that extra phase shift.

VAS stages are actually *slower* than outputs. They are intentionally slowed down usually with Miller capacitance and become the slew rate limiting factor. This is desirable, as if you try to switch the outputs too fast they go into cross conduction and get hot. So why do you want a transitor with low Cob here? Because the capacitior inside the transitor is nonlinear and causes HF distortion. The miller cap you add swamps it out so the total doesn't change much with applied voltage.

 

[myhrrhleine]

People what are your experiences??

So many pro-amp guys here! C'mon please share your experiences!

Does this speed order help generally to get an amp stable?


@myhrrhleine


That's the reason I assumed a simple topology where the devices alone do not accumulate 180° phase shift. Per definition this is stable under all conditions. So I was interested if in this special case one could cause instabilities by choosing slower transistors for the VAS (slower than drivers, drivers slower than outputs).

As far as I know this case is not covered by Bode-analysis, so I would like to hear the experiences of the people here which are more lucky than me and have far more real building experience.

If I had the time and a oscilloscope at hand, I would do it on my own - but that's not possible at the moment

Have fun, Hannes

They will accumulate the 180° of phase shift.
It's a matter of controlling when that happens and the overall gain when that happens.

 

[EWorkshop1708]

VAS stages are actually *slower* than outputs. They are intentionally slowed down usually with Miller capacitance and become the slew rate limiting factor. This is desirable, as if you try to switch the outputs too fast they go into cross conduction and get hot. So why do you want a transitor with low Cob here? Because the capacitior inside the transitor is nonlinear and causes HF distortion. The miller cap you add swamps it out so the total doesn't change much with applied voltage.

Since the transistor's internal capacitance varies, you say the miller cap makes it more stable. Besides VAS, I would imagine this applies to output transistors and drivers as well. On an amp I built some time ago, I put 1nf caps from B-E of drivers, and it really helped a lot. Although some people here disagree with that approach, for me, it worked, and the amp sounds good, that's all I care about.

BTW
I also wonder if the Zobel has a similar effect to a miller cap.

I have a theory...........Instead of the capacitance of the output stage and speaker cable varying with frequency, a zobel cap would average it out, making it more stable, and being it's in series with a resistor, it's not like the amp's output stage has to actually drive a capacitor directly, it's more like driving a small tweeter in comparison. I've used 0.22uf for zobel instead of the usual 0.1uf with much better results, although once again, people disagree, and think it hurts performance.

 

[h_a]

They will accumulate the 180° of phase shift.

Ok, I should have been more clear. I meant 180° phase shift with gain. There are a couple of mosfet amps out there that don't have any form of frequency compensation! I have not yet seen a bipolar amp without compensation.

I would be interested how fT affects stabilit. Overall I have the impression that compensation is largely an empirical thing. That's ok, it makes my life a bit more complicated however, since I try to sort of design stability by trying to choose the right parts - without actual building and measuring (I can't do that at the moment).

All the best, Hannes