Power amp transistor speeds?

[AndrewT]

Hi,
all power amp output transistors are limited in maximum speed approx 2MHz to 45MHz.
but the Input, Vas, Predriver and Driver stages usually incorporate transistors that are much faster.

My thoughts are no better than guestimates but possibly:-
output - 6MHZ, Driver - 30Mz, prediver - 100MHz, VAS - 150MHz, LTP - 200MHz.
for faster outputs say 30MHz, then at least the drivers and maybe the predrivers need to be faster but how much faster.

Q1. Is there a rule of thumb to determine the minimum speed for each of the stages?

Q2. Can a speed ratio from outputs back through the stages to inputs be used as a guide for selecting suitable transistors?.

Q3. Is an amp that uses few stages AND fast transistors faster than one with slow transistors? i.e. less of a phase shift input to output.

Q4. Where should the current source/sinks fit in the speed stakes?

[lt cdr data]

interesting post, any thoughts out there to enlighten us?

[anli]

I think such consideration is not productive without involving a concrete schematics topology.

[ilimzn]

Originally posted by AndrewT
Q1. Is there a rule of thumb to determine the minimum speed for each of the stages?

Not really because it depends on the topology used as well as DC operating conditions for some transistor types. In fact, many fast output transistors do not have a constant gain-bandwidth product (which is what you refer to as speed) but it depends HEAVILY on the current through the transistor. There is similar dependence for smaller parts, just not this pronounced.

Q2. Can a speed ratio from outputs back through the stages to inputs be used as a guide for selecting suitable transistors?.

Yes, though depending on stage, you can chose faster or slower. There is no easy answer, but when finding replacements, you tend to look for similar to what was there. That being said, if you know wat the circuit is doing and how, yopu can improve it (sometimes quite substantially) by chosing different parts. There has been, after all, some progress in semiconductor technology

Q3. Is an amp that uses few stages AND fast transistors faster than one with slow transistors? i.e. less of a phase shift input to output.

Broadly speaking, unless you did something really wrong, yes - but then, when you start considering the number of stages, things get quite complicated. Topology is a heavy influence here, as well as the fact that you can usually trade speed for gain and vice versa. So, there is definitely a sweet spot, given parts to design an amp, and the required gain, between speed and number of stages.

Q4. Where should the current source/sinks fit in the speed stakes?

[/quote]

Just like any other active component, they have their own bandwidth. But they can be used in so many different ways in a circuit, it is impossible to give you a blanket statement about t hem. just like anything else, properly used, they can be a great benefit, improperly used, they are a problem.

[jcx]

The questions seem to be slightly off target, in feedback theory the point of preferring faster/lower phase shift devices isn’t for the sake of speed as an absolute good in itself but to allow “room” for more feedback gain at lower frequencies over the working band of the amplifier

Bode’s phase integral relation is fundamental to understanding allowable gain/phase vs frequency tradeoffs that are necessary for stability while maximizing feedback gain over the working frequency band – faster devices permit stability to be obtained at higher frequency gain intercept, increasing the available negative feedback gain*frequency area under Bode’s phase integral curve

Bode’s integral relations and related useful tools for reasoning about feedback amplifiers are only just introduced in senior level undergraduate control theory courses, as we all know you can’t be expected to really understand and apply material until the 2nd appearance in your courses – often requiring grad level courses or seriously motivated self study

adding to the difficulty academic control theory has adopted a mathematical proof style of presentation based in the state space matrix algebra representation of the system so getting a practical engineering perspective from modern textbooks is tough


I see we already have some mixed advice here, you can but should not want to “choose faster or slower” devices and higher speed allows higher gain - not a tradeoff

Generally the output devices will be the slowest, limiting potential performance from negative feedback, if technology permitted everything else in the amplifier should be much faster than the output, as fast as possible with diminishing returns for additional accumulated phase shifts of <~10 degrees at the gain intercept frequency from the additional circuitry

[ilimzn]

Originally posted by jcx
I see we already have some mixed advice here, you can but should not want to “choose faster or slower” devices and higher speed allows higher gain - not a tradeoff

I think I may have worded this strangely so there is some misunderstanding.

What i meant with chosing slower or faster, was intended for the purpose of choosing replacement devices for an existing design. At least i believe that's what the original poster was asking about.

Transistors are devices which come with discrete sets of charactristics, and we neither can get ideal transistors, nor can we always get exact replacements.

In simple terms, you can have a transistor with a very high ft and moderate hfe, or one with lower ft but high hfe - which one would you chose? If you require a current gain of 100, no matter how high ft is, if your part has a current gain of 10, you cannot make it higher. In this context, higher speed definitely does not alow higher gain.

However, in a given design, a part has an operating point. In amplifiers with feedback, replacing devices with similar ones, shifts operating points regarding gain vs bandwidth. In this context, there are many parts that could fit, because the operating points are still within tolerances and the complete amp will perform properly - even if some devices are faster or slower and/or have higher or lower current gains (and a ton of other different characteristics) than the originals. Designs need to be tolerant of this since semiconductors indeed have very large parameter tolerances. In this context, you can, within the confines of a single stage (and usually more than the single stage) 'trade' one parameter for another, but as I said, there is a 'sweet spot' - the example I have given above, a very high ft transistor with low current gain, put where high current gain is required, is definitely not within that sweet spot, in other words, it\s outside of the alowable tolerances.

On the other hand, for any given design the minimum gan and ft are 'given' for it to function properly. If one understands how the design works, it is possible to select parts such that the performance is improved - subjective to the law of diminishing returns. What devices are replaced and what the differences need to be with respect to the originals depends on their application - you cannot really give a blanket statement here without knowing the design in question.

Clearer now?

[jcx]

I’m interpreting the question to be more generally about amplifier design tradeoffs so we may be talking past each other to some degree

Originally posted by ilimzn

…
In simple terms, you can have a transistor with a very high ft and moderate hfe, or one with lower ft but high hfe - which one would you chose? If you require a current gain of 100, no matter how high ft is, if your part has a current gain of 10, you cannot make it higher. In this context, higher speed definitely does not alow higher gain.
…

I can see your point where limited to 1:1 replacement, but suppose (very loosely speaking here) you can replace the hfe 100 transistor with 3 very much faster transistors each having gain 10 – for a composite hfe ~1000 in a triple darlington or cfp for instance

Subject to many limitations it can be possible to utilize the extra gain to reduce distortions at low frequencies, while if fast enough there could be little effect on stability with appropriate compensation

[AndrewT]

Hi all,
good reading in the discussion so far. I wish someone could talk me through that Bode stuff (yet another area I'm deficient in).
I am not thinking about altering an existing amplifier. More about choosing components for the next series of builds.
If I were to suggest Krell Clone (KSA50) as the overall topology, would that allow some more specific suggestions?
To date various driver & output transistors are being put forward, all of varying speeds. Some members are suggesting that the distortion will be lower for the faster devices.
Another old amp design (Leach Clone) is being updated and faster components are again being suggested.
Nail your colours to the pole.