That one looks curious, for a purely analogue circuit.
The leading corner, before the edge is more rounded than the trailing corner, as if (imbalanced) FIIR filtering had been applied to the signal.
With analogue circuits, this kind of behavior can be seen when all-pass filtering has been deliberately applied, but I have never seen normal, simple circuits anticipating the signal in such a way.
My experience is that rise time is generally taken to be 10-90% initial to final values.
Slew rate limit being a rate should be independent of initial and final values so it should not include the bandwidth effects, 2/3 generally works well for this as long as the signal remains in slew limit.
Elvee
Really neat observation, interested in your follow up (could having the outputs close to the rails contribute to the asymmetry?)
Thanks
-Antonio
Slew rate limit being a rate should be independent of initial and final values so it should not include the bandwidth effects, 2/3 generally works well for this as long as the signal remains in slew limit.
Elvee
Really neat observation, interested in your follow up (could having the outputs close to the rails contribute to the asymmetry?)
Thanks
-Antonio
I apologize for nit picking, but I dislike the the interchanging of rise time and slew rate. Rise time is a function of frequency response. Slew rate is due to a limitation in the maximum dv/dt, is usually a function of current charging a cap, not time constants. Again, sorry for the nit picking.
I have and do.
The sig generator may not be symmetrical
The amplifier may not be symmetrical between + & - halfs, this includes capacitance's. Feedback does not always clean this up...
_-_-bear
That is not really the problem here.
If you examine the waveform closely, you'll see that the level begins to change ~1µs before the 50% transition.
This means there has to be a delay of 1µs somewhere. 1µs is huge: it is the transit time through 200m coaxial cable for example.
Ooo boys, you have time to think and discuss about my waveform, nice. OK, what you see it is very fast amp response in extreme conditions when plot was taken. In normal signal conditions amp is actually even faster.
So the extreme conditions were hard clipping of the 1kHz signal, that is why the delay (latch) at begining of the ramp but the middle 80% of the ramp tells you enough about the speed (slew-rate). Propagation delay measured from input to output was only 33 ns.
So the extreme conditions were hard clipping of the 1kHz signal, that is why the delay (latch) at begining of the ramp but the middle 80% of the ramp tells you enough about the speed (slew-rate). Propagation delay measured from input to output was only 33 ns.
that can be another way of "cheating"
its possible for there to be "windup" on clipping recovery - the internal nodes overshoot coming out of clipping and can provide drive levels that couldn't be reached in pure linear operation
the most realistic measure of slew limiting audio is power bandwidth - the frequency where a near full scale sine, the slew limit causes some specified level of distortion - typ 1%
with margins of 10-100x over the dV/dt of realistic audio signals the measurement of these high slew rate limits is just playing with numbers - hence lots of games are played with the measurements
its possible for there to be "windup" on clipping recovery - the internal nodes overshoot coming out of clipping and can provide drive levels that couldn't be reached in pure linear operation
the most realistic measure of slew limiting audio is power bandwidth - the frequency where a near full scale sine, the slew limit causes some specified level of distortion - typ 1%
with margins of 10-100x over the dV/dt of realistic audio signals the measurement of these high slew rate limits is just playing with numbers - hence lots of games are played with the measurements
Power bandwidth and slew rate are two distinct, commonly used parameters for characterizing amplifiers.
If measurement conditions are constructed in such a way as to equate slew-rate to power bandwidth, we only need one parameter.
But then, I think something would be lost: the spirit of slew rate is to examine how an amplifier behaves under sort of "free fall" conditions, i.e. how an amplifier behaves when it is subjected to a grossly excessive step in speed, but still within its amplitude capabilities.
It normally isn't a realistic test, because any modern amplifier has sufficient sr and power bandwidth to reproduce any real musical signal, but it gives some insight into the capabilities and failure modes of the amplifier.
A test like the one presented by Lazycat is probably worthless, except for showing off, and even for that purpose, it is counter-productive: it shows the presence of storage mechanisms in case of saturation.
In summary, I think there is still room for power bandwidth specs (x% THD), as well as for slew rate spec (pure recovery speed, at 2/3rd of the waveform)
If I had to drop one of them, it would be the slew rate, obviously, but the ecosystem is large enough to accomodate both.
If measurement conditions are constructed in such a way as to equate slew-rate to power bandwidth, we only need one parameter.
But then, I think something would be lost: the spirit of slew rate is to examine how an amplifier behaves under sort of "free fall" conditions, i.e. how an amplifier behaves when it is subjected to a grossly excessive step in speed, but still within its amplitude capabilities.
It normally isn't a realistic test, because any modern amplifier has sufficient sr and power bandwidth to reproduce any real musical signal, but it gives some insight into the capabilities and failure modes of the amplifier.
A test like the one presented by Lazycat is probably worthless, except for showing off, and even for that purpose, it is counter-productive: it shows the presence of storage mechanisms in case of saturation.
In summary, I think there is still room for power bandwidth specs (x% THD), as well as for slew rate spec (pure recovery speed, at 2/3rd of the waveform)
If I had to drop one of them, it would be the slew rate, obviously, but the ecosystem is large enough to accomodate both.
I honestly don't know, but I suspect it has little to do with the actual, measured speed of the amplifier in question (I mean, in completely blind tests: if you know you're listening to a "fast" amplifier, that might influence your judgement).
A lot of "colour" originates from the phase shift in the mid to high frequencies. While sound in itself won't sound distorted, the resulting wave form is vastly different from the original.
An amp that does up to 20KHz (and a little beyond) without any phase shift either due to the input filter or due to Cdom limiting should sound "fast" and "transparent".
Just the fact that a simple percursion instrument like a hi-hat is comprized of many many MANY individual sinewaves ranging from sub 1KHz to over 50KHz (yes, metal produces frequencies well above what we can hear) it's easy to conclude that phase shift will colour the sound. It's the higher frequencies that our hearing is most sensitive to phase shift.
So unless you know how the original track sounds, you won't be able to distiguish.
The faster an amp, the later phase shift sets in.
I didn't read the whole thread, but there is an amplifier in a linear.com appnote that does 3000 V/us, measured in the real world (see AN-21 at linear.com, "Son of Godzilla" in Figure 12). So 500 V/us isn't all that fast, in comparison. Linear.com even has opamps that can do 2500 V/us (LT1818 maybe?).
What kind of evidence do you have to support that assertion?
Psycho-acoustic data tend to show that the ear is completely insensitive to phase shift at higher frequencies; in fact, phase shifts needs to reach multiples of 360° to begin to be audible.
In addition, amplifiers utilizing GNFB have very small phase shifts at audio frequencies.
As an example, here is the latest incarnation of Kenpeter's JLH:
At 20KHz, the phase shift is 40 millidegree.
For an airborne sound, this equates to 1.87µm. Could a tweeter be positionned with that accuracy?
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.