PMA said:
There are no different kind of overshoots and NFB has no relevance here. Overshoots are a characteristic of a second order system, when a pole and a zero are in close proximity. The fact that NFB moves the poles in the complex plane (and therefore it can bring a pole close enough to a pre-existent zero) doesn't change the nature of the overshoot. Butterworth filter have poles and zeroes in proximity without any NFB. With or without NFB, a similar distribution of poles and zeroes and therefore an identical pulse response, doesn't have any impact on stability. You may think of stability in terms of sensitivity, but this is again a different story.
AndrewT said:
Here's a very good discussion about the pole-zero distribution effect on the step response:
http://me.queensu.ca/courses/MECH350/documents/MECH350_lecture13.pdf
Discard everything after page 262, it's mostly irrelevant for audio.
Edit: There is a famous 1974 paper from R. Meyer and P. Gray about this issue: "Relationship between the frequency response and settiling time of operational amplifiers". Can't post it here, but I can email it to whom is interested.
I have always found settling time to be a good indicator of audio quality. It shows both low level and high level behavior at the same time, and can be very hard to improve. It can show even small thermal anomalies. However good settling time is not enough for good audio, any more than low distortion or flat response.
TPC? Better safe then sorry
Hi Dimitri,
I really don't know whether this kind overshoot has a sonic impact. But from a technical point of view I simply don't like it. Better safe then sorry. But that's not all, we don't even need TPC it at all, as there are better ways to reduce the distortion.
Audible or not, overshoot is also a kind of distortion. Why not keeping the transmission as clean as possible?
Opposed to the typical overshoot from TPC (caused by the phase dip at around 10kHz), I know of one case that overshoot was audible, though this was caused by lack of sufficient phase margin. IOW, that was a slightly different story (a serious case of instability).
Cheers,
Edmond.
dimitri said:
Hi Dimitri,
I really don't know whether this kind overshoot has a sonic impact. But from a technical point of view I simply don't like it. Better safe then sorry. But that's not all, we don't even need TPC it at all, as there are better ways to reduce the distortion.
Audible or not, overshoot is also a kind of distortion. Why not keeping the transmission as clean as possible?
Opposed to the typical overshoot from TPC (caused by the phase dip at around 10kHz), I know of one case that overshoot was audible, though this was caused by lack of sufficient phase margin. IOW, that was a slightly different story (a serious case of instability).
Cheers,
Edmond.
Demian,
1% settling band is 0.09 dB, 0.1% is 0.009 dB. Does is above the relosution of human ear?
Edmond,
Yes, overshot is linear distortion, causing nonuniform group delay. The transition frequency is of order of closed loop, which is 10-100 times larger than audio band. Bob showed that there is neither ultra sound content nor fast slewing signal in audio. Why should I be aware of the overshot, if there is no way to excite it?
1% settling band is 0.09 dB, 0.1% is 0.009 dB. Does is above the relosution of human ear?
Edmond,
Yes, overshot is linear distortion, causing nonuniform group delay. The transition frequency is of order of closed loop, which is 10-100 times larger than audio band. Bob showed that there is neither ultra sound content nor fast slewing signal in audio. Why should I be aware of the overshot, if there is no way to excite it?
dimitri said:
Two observations- your right, on the face of it the error is small. But only if the errors are directly harmonically related. The internal phenomena that causes the error may have a completely independent time constant to the signal and show up in some other place where it is audible. However, more important, is that settling time is an indicator of the care and detail focus of the the design. Everything from thermal tails to cap nonlinearities will contribute to the settling time.
dimitri said:
Dimitri, I can't refute your reasoning. However, that other example I was referring to (low phase margin), was even ringing in the MHz range. Agreed, most unlikely that an external source excited it, yet some people claimed (not me) that they could hear it. After the ringing was fixed, it sounded better.
>Why should I be aware of the overshot, if there is no way to excite it?
Perhaps an internal source might excite it, for example crossover spikes with high HF content.
Cheers,
Edmond.
A CB (remember those?) or a "family radio" is a good way to find those internal instabilities. There is a valid question about the audibility of a small (1 dB?) peaking in the response above audibility. if the peak is at 250 KHz or 500 KHz is it a problem? Is it audible? The feedback I have had is that it tends to be preferred to a perfectly damped or overdamped response. If the response in the audio band (up to even 100 KHz) is within .1 dB how could it be audible? With the exception of mistracking on a phono cartridge what content is in an audio chain above even 15 KHz (look at the response curves for the microphones used for recordings, not for measurements)? But users have clear preferences and no knowledge of what the actual source of the differences are.
Most microphones have at least a 12dB/octave rolloff that starts pretty early, sometimes in the 10KHz region, that evolves to 24dB/oct in the ultrasonic region. Exceptions are seemingly only on measurement 1/2 and 1/4 in microphones that are especially constructed to keep going to 4OKHz or more. I think that this is what Demian is talking about.
I'm referring to the acoustic rolloff of a 1" or 1.5" diaphragm microphone for example, or the low pass effects of the coupling transformer in it or the subsequent audio processing devices before you get to the digital audio link that is so common today. No brick wall filters but pretty fast rolloffs just from the acoustics. if you look at the off axis response curves of a 1" B&K they are far from flat or even benign. B&K 4145
Why don't they use 1/2" or 1/4" mikes for recording? Look at the noise floor and the dynamic range of them. And they don't do the nice things that recording engineers like to hear from the mikes. U47 U87 are two of the most respected condenser mikes in use. And the Coles 4038 is one of the best ribbon mikes in use.
The point being that flat response to above audibility is not the primary point of these mikes.
With direct connection from instruments you will get more extension but its arguable whether there is a "correct" response for a direct connection to a source. I would argue that the reproducing system becomes part of the artistic expression of the musician with a direct synthesizer source for example.
Why don't they use 1/2" or 1/4" mikes for recording? Look at the noise floor and the dynamic range of them. And they don't do the nice things that recording engineers like to hear from the mikes. U47 U87 are two of the most respected condenser mikes in use. And the Coles 4038 is one of the best ribbon mikes in use.
The point being that flat response to above audibility is not the primary point of these mikes.
With direct connection from instruments you will get more extension but its arguable whether there is a "correct" response for a direct connection to a source. I would argue that the reproducing system becomes part of the artistic expression of the musician with a direct synthesizer source for example.
- Status
- Not open for further replies.