I'm quite sure they were not added to make the QUAD 33 sound more musical and that Peter Walker would turn in his grave if he heard anyone suggest otherwise, but still, if low-pass filtering makes the amplifier sound more musical for whatever reason, you could use them for that - if a low-order filter at 10 kHz also does the trick.
I love your efforts to dispel slew rate BS, but I'm not sure I understand this statement. Square waves are easy to use because you can use any frequency and easily find the edge and measure the time derivative. You can do that with sines that degrade into triangle waves too, but I don't see why that would be superior. As you can see below, both methods yield the same result.
That said, the idea that slew rate DEPENDS ON waveform is, of course, completely wrong.
High power load with squares, Zobel resistor likely to be burnt, small signal x large signal nonlinearity. Jan definitely has a point.
And your triangle example is not any good. Maximum dV/dt of the sine wave is at zero crossing. The slew induced distortion starts much earlier than the sine looks like a triangle. Make a derivative of the output signal and you will see. If it is a pure sine, derivative would be a cosine, so no change in shape. SR distortion would lead to abrupt shape change.
And your triangle example is not any good. Maximum dV/dt of the sine wave is at zero crossing. The slew induced distortion starts much earlier than the sine looks like a triangle. Make a derivative of the output signal and you will see. If it is a pure sine, derivative would be a cosine, so no change in shape. SR distortion would lead to abrupt shape change.
Last edited:
Yes, in principle you can use any signal, but to eyeball it you would want a signal that shows a clear change of shape at the onset of SR limiting. I believe that is better detected with sinewave signals than with square waves.
At any rate, eyeballing it is very inaccurate and badly repeatable.
You wouldn't 'measure' THD by just eyeballing the signal either.
One of the takeaways from Walt Jung's work that I linked to before is that SR limiting can be reliably detected by a sudden increase in 3rd harmonic distortion.
Jan
I am not, speaking about my designs. But when testing commercial, unknown amplifiers, I am afraid. I have almost destroyed Cello Encore power amp when I was asked to test it with a square wave. It went very fast, you know (crossconduction??). And I do know that lower repetition frequency means less accumulated energy in the Zobel resistor. It is a task for a primary school student.
Oh, I know. He's got about 1000 times the knowledge I have.
Not sure I get that. My triangle is at a frequency so high that dV/dt is maxed out all along the waveform, except the very peak. This means that the slew rate is approximately the same across the entire slope. The difference between my previous measurement and a narrow band around the is negligible.
If I lower the frequency to just the point where it starts distorting and measure around the zero crossing, I get the same value. But maybe I missed your point.
I have yet to burn any Zobel resistors, but I do see why what could happen.
But at the same time you must use a rise time that is fast enough to cause SR limiting.
There are smarter ways to measure SR limiting reliably.
But I know you know that 😎
Jan
Hence the "low risetime". The lower the risetime, the faster the rising edge.
I'm not sure I'm overly worried that my Zobel resistor would burn out. The current spikes are about 5us in duration and SPICE tells me that the RMS power is only a bit over a watt across my 5W resistor. Given the thermal mass of such a large resistor, it shouldn't be an issue. I'll try it on a physical amplifier when I get a chance. It'd be interesting to know how hot the resistor gets! 🙂
Real world, large signal, 10kHz square response, 60Vp-p, 3.9 ohm load. Well, I am not afraid to measure my designs with a square wave. Please note that small ringing is due to inductance of wires from amp output to the load.
An easy way to address slew rate limiting is to view certain driving voltages within the GNFB loop with a scope. When the loop can't respond fast enough to input changes, such a driving voltage will show overshoot (not visible at the output). So one only has to determine the driving voltage just before or at clipping, notice the value of that and compare it to the value of the overshoot pulse when applying a small square wave input signal. That overshoot pulse can't get higher than the noticed voltage just before or at clipping.
The best place to do that is at the front end LTP collector currents. That will tell you everything you need to know about the GNFB loop - slewing, clipping etc.