2
\$\begingroup\$

I was watching a video where Dave Jones of EEVBlog interviews Rod Elliot of Elliot Sound Products. Rod was discussing that he tends not to use digital chips in his designs because the chip will become obsolete 6 months after the design is published. He goes on to say that most analog components don't have this problem and goes on to give this quote:

Although transistor types and everything do change, ON Semi for example a lot of their new transistors, they're brand new. They're labelled as being older devices but they're not; they're modern technology. But I try to make sure that my circuits will work. If they update the transistors and they get faster, more linear, that never hurts anybody. If they get faster, usually that makes the amplifier more stable, not less stable.

I don't understand the last part of his statement. Given a standard audio voltage/power feedback amplifier, how does a faster transistor tend to improve stability? If anything, I would have thought that a slower device would improve overall stability. Doesn't more bandwidth offer greater opportunity for higher-frequency poles (caused by reactive passives and parasitics) to interact with the active device's open-loop gain and cause phase shifts which may lead to oscillation?

\$\endgroup\$
1
\$\begingroup\$

You need to think of an amplifier like a control system. Imagine that you are trying to rotate a piece of machinery to a certain angle and using feedback from an angular position sensor to drive the error out of the system.

If the machinery is lightweight, it requires less energy to get it rotating and conversely, it requires less braking energy to stop it once it has reached it's "destination". So there may be a little bit of overshoot; the machine slightly overshoots the target position but the control loop brings it back quickly and the error is removed.

Now if that machine were much heavier, it would take more energy to move it and more energy to stop it and that likely means a greater overshoot past the target position and it could lead to sustained oscillation at the desired position. That oscillation frequency would be lower than the oscillation frequency of a lighter machine should it have similar problems.

So, faster transistors are like lighter machinery and slower transistors are like heavier machinery.

You can't rule out that a faster transistor might produce nuances not seen in the slower transistor but the chances are they are at a much higher operating frequency and easier to get rid of (or even ignore).

\$\endgroup\$
  • \$\begingroup\$ This is the case for a well designed circuit...i.e., designed to specification rather than performance. Early in my career, I had to diagnose an oscillation in the 10MHz range that suddenly began to plague a previously stable product. It turns out the supplier updated the die with another with the same specs "or better." The circuit was inherently unstable at this frequency, but the lack of response in the transistor in this range kept it in control. Moral: design to spec, not performance. \$\endgroup\$ – Cristobol Polychronopolis Nov 30 '18 at 14:57
  • \$\begingroup\$ Thanks for the analogy, it helps. But I'm still having a hard time imaging how more gain at higher frequencies leads to more stability rather than less stability. Take an op-amp for example. As you know, op-amps are purposefully internally compensated to reduce their high-frequency open-loop response in order to promote stability. But your explanation suggests that the faster the op-amp, the more stable it will be in a given circuit relative to a slower op-amp. What am I missing here? \$\endgroup\$ – pr871 Nov 30 '18 at 15:18
  • \$\begingroup\$ (1) An op-amp's compensation is done before the output stage. (2) the output stage of an op-amp is always much faster than the signals it receives from the compensation stage. (3) the compensation stage is the slowest stage in an op-amp - everything else is designed to be faster to avoid introducing oscillations when feedback is applied. This bit is underpinned by the parameter "phase margin". \$\endgroup\$ – Andy aka Nov 30 '18 at 15:23
  • \$\begingroup\$ I think I might be getting it. So are you saying that substituting faster active elements for slower active elements tends to improve stability IF the magnitude of the overall loop gain is already limited by some other factor so that the high-frequency poles introduced by the fast active elements occur well beyond the frequency where the loop gain magnitude has dropped below 0 dB? \$\endgroup\$ – pr871 Nov 30 '18 at 16:20
  • \$\begingroup\$ Correct. Everything but the main single pole limiting circuit should be in the realm of at least ten times faster and, for op-amps, it's 100's to 1000's times faster. This prevent two poles occuring at a frequency where open-loop gain is greater than unity and turning neg f/b into pos f/b and making a circuit oscillate. \$\endgroup\$ – Andy aka Nov 30 '18 at 16:57
-1
\$\begingroup\$

Faster devices, in a power amplifier with 3 or 4 stages: differential inputs, cascodes to avoid Miller Effect, conversion from differential to single-ended, output stages with locally active feedback to minimize zero-crossing distortion, and the huge output devices themselves ---- reduce the phase shifts, and allow the BODE plots (gain and phase) to be simplified, perhaps needing less complicated compensation networks.

Or maybe not. With reactive loads ---- speakers ---- design is already exciting.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.