Obviously Petr has no idea what I did.
I took the derivative of the slope d(Vout)/dt, which is the most accurate way to derermine the slew rate.
Up to +/- 6Volt the slew rate was symmetrical using a square wave with a very fast slope.
Beyond this level, which will never happen in real life, the amplifier became non linear so that’s where I stopped.
So SSassen 90Volt/usec in both directions is a good achievement for your amp, way beyond what will ever be needed and especially because the square wave output signal is nicely critically damped with a very fast settling time.
Hans
I took the derivative of the slope d(Vout)/dt, which is the most accurate way to derermine the slew rate.
Up to +/- 6Volt the slew rate was symmetrical using a square wave with a very fast slope.
Beyond this level, which will never happen in real life, the amplifier became non linear so that’s where I stopped.
So SSassen 90Volt/usec in both directions is a good achievement for your amp, way beyond what will ever be needed and especially because the square wave output signal is nicely critically damped with a very fast settling time.
Hans
Last edited:
Quick comment: there is an extra required step to identify slewing issue: in simulation or measurements, determine the 10%-90% time for both the leasing and trailing edges of the amplifier step response, for various amplitudes of the input signal (in simulation you could go to the amplifier clipping, in practice don't try it, the output RC cell (and potential other components, may blow). If:
- 10%-90% time does not depend on the output amplitude, then what you are measuring is the amplifier rise time, directly related to the closed loop bandwidth.
- 10%-90% time does depend on the output amplitude, then you reached the large signal slewing limit, and you can calculate the slew rate.
Note1: the slewing waveform is usually NOT nicely damped, but the transition from the slope to the plateaus is abrupt.
Note2: if you cannot identify a slewing limitation as above, this means you are safe; the rise/fall time (that is, the amplifier bandwidth) are high (low) enough to mask any slewing effects, so slewing distortions will never occur. So limiting the amplifier bandwidth to a reasonable value (say, 100KHz) by (for example) using an input low pass filter) is a good protection against slewing distortions (assuming they may ever occur, which is always questionable for the low slewing audio signals).
Last edited:
Extreme signals like playbacked recordings of real instruments expects an upper flat bandwidth up to 100kHz at least due to phase related issues (count one decade) - the ear is susceptible to this very much in that range! I don't want a violin creeping in from behind my left or right ear. (Notice my attachment in #20.)
I'm bitter as this is the actual reality of recording habits. Crap are spikes, gold plated wall plugs and the lot. Like digging corpses being resurrecting digits. What we're actually talking about? Mooly: next split? Granted! But let's keep on topic.
The pic is an actual rig from a theatre, the audience swamped into another imaginary reality - it works, but please don't touch the wires or their temporarily reality evaporates before the curtain (aka very life) drops.
"No Mooly, no more splits I beg thou!" {Shakespeare: The Damned}
The pic is an actual rig from a theatre, the audience swamped into another imaginary reality - it works, but please don't touch the wires or their temporarily reality evaporates before the curtain (aka very life) drops.
"No Mooly, no more splits I beg thou!" {Shakespeare: The Damned}
He talks about very simple thing.
Last VAS stage of the original Sander's amp can be seen as an common emitter stage from lower rail loaded on the current source from the upper rail. The output point of this stage are loaded on some capacitance. Because of this the charge of capacitance for one slope (falling) are provided by the higher current of the lower transistor and thus faster while charge for another slope (rising) are provided by limited current from the current source and thus slower.
The common talk about nothing.
Yes, symmetrically driven VAS could be faster and symmetrically, but are we need such a speed and what's the price for symmetry.
Slew rate and rise time are similar but not the same IIRC.
Slew rate = max rate of change in output with all the LTP current provided into the second stage.
Rise time= is the 10-90% transition time with the LTP still operating in its linear region ie both LTP transistors conducting some current.
All audio signals are BW limited (in nature as well). A good rule of thumb spec is 1V/us per peak output volt, so (from memory) for a 100 W amp 80V/us
Slew rate = max rate of change in output with all the LTP current provided into the second stage.
Rise time= is the 10-90% transition time with the LTP still operating in its linear region ie both LTP transistors conducting some current.
All audio signals are BW limited (in nature as well). A good rule of thumb spec is 1V/us per peak output volt, so (from memory) for a 100 W amp 80V/us
Yes, they even have different dimensions of quantity...
While slew rate are measured in "V/s" then rise time is only "s".
I fully agree with Syn08's definitions but simplified them a bit.
As long as rise time does not change when increasing amplitude, slew rate increases with the same rate as the change in amplitude.
When doubling the amplitude no longer shows a doubling in slew rate, rise time has increased and slew rate starts to interfere.
With SSassen's amp this was the case at +/- 90V/usec.
That the output signal was still critically damped as shown was a bonus.
Musings on amp design... a thread split.
When considering 300Khz an ambitious BW limit for an 100Watt@8R amp, a slew rate of almost 80V/usec would be needed to produce 300Khz at full power, which as Syn08 remarked, will blow your Zobel filter to pieces.
In practice, less will be sufficient and not many commercial amps will even come close this 80V/usec.
On top of that, keep in mind that all this is the outcome of a simulation, and real life results may and probably will be lesser.
Hans
It is simple to correct for that, I'm just wondering how best to simulate slewrate limitations so small differences are quickly found and can be corrected.
Thanks petr_2009, can you point me towards the schematic of the Graham amplifier you refer to? I'd like to simulate it and compare it to my own design, as I'm always open to new ideas and there's a good chance I'll learn something in the process.
Going by your criterion, I think you meant 40V/uS and not 80V/uS for a 100W (40V Peak) into 8Ohms amplifier.
Simply use the method in LTSpice as was used here
Musings on amp design... a thread split.
Offer a square wave with 0.1usec rise/fall time, display the output signal V(out) and its derivative d(V(out).
By increasing the input voltage you will easily find the point where the slew rate stops.
And as a matter of fact, I don’t see a reason why slew rate should be symmetrical as long as the slowest of the two is fast enough like your 90V/usec.
Hans
Sander,
I hope this is your name, is it?
My feeling is that you still might have not enough confidence in finding the slew-rate, the way I mentioned.
That's why I have taken a smaller time range with a higher square wave frequency, enabling to show you in more detail the slew rate.
The image at the left is made with a +/-0.3V input signal, showing a symmetrical +/-90V/usec slew-rate.
When increasing the input signal to +/-0.5V, you see that the upgoing ramp still has still ca 90V/usec while going down is with a slew rate of ca 120V/usec.
Very simple to perform, isn't it.
However what is quite visible is the asymmetric behaviour that's encircled in green.
I would worry more about that part than the asymmetric slew-rate
Hans
I hope this is your name, is it?
My feeling is that you still might have not enough confidence in finding the slew-rate, the way I mentioned.
That's why I have taken a smaller time range with a higher square wave frequency, enabling to show you in more detail the slew rate.
The image at the left is made with a +/-0.3V input signal, showing a symmetrical +/-90V/usec slew-rate.
When increasing the input signal to +/-0.5V, you see that the upgoing ramp still has still ca 90V/usec while going down is with a slew rate of ca 120V/usec.
Very simple to perform, isn't it.
However what is quite visible is the asymmetric behaviour that's encircled in green.
I would worry more about that part than the asymmetric slew-rate
Hans
Attachments
Not critical. 40 or 80 us/V will be fine
Wide BW amps don’t blow Zobels because the source material is BW limited. When I test near full power 100 kHz square waves I disable the Zobel.
The question may well then be why do you need a wide BW amp?
No logical answer. My new Citroen C4 Spacetourer has a 3 cylinder 1500 cc engine and develops c 150 BHP. Some folks sneer at that. They need 5 liters (there’s an orange 5l Mustang just up the road from me).
If you want 20-20 kHz at +0 dB -0.1dB you will require a BW of at least 10x that.
Each to his own.
For Sanders amp, you could feed a sine wave in so the OP was close to the max pk-pk and increase the frequency until the sine wave triangulated. Then measure the rate of change at the zero crossing. You can also calculate it from it=CV where i is the LTP current, C the total Miller capacitance and V the peak swing in 1 us