"Speed distortion" seems to be just another name and spin on Graham Maynard's (RIP) "first cycle distortion", or whatever he called it, but such sudden discontinuous change doesn't exist in the real world of audio, for example a CD audio is heavily LP filtered above 20 kHz so sudden discontinuities would be smoothed out such there wouldn't exist any components above 20 kHz anyway, at least as far as I can see.
CD audio is mostly filtered, some of which is done digitally in modern Sigma Delta DACs. In addition to low level high speed glitches there is also some RF mixed in with dac analog outputs. Some of that RF is looks like a clock signal above 1 MHz and a spectrum of RF noise above 20kHz. The RF and any glitches are attenuated by analog active and passive filters in the DAC. Using an AK4499 evaluation board as the source, an oscilloscope measurement was taken at the of the output of a SOA low distortion Neurochrome headphone amp turned up to full gain. I was able to see an approximately 2.5MHz square wave at the amplifier output connector.
Are you saying that phase response to audible frequencies is amplitude dependent?
Is this another slew rate argument?
Digital audio is filtered using over-sampling to avoid the phase distortion that linear filters can not avoid.
The success of audio compression algorithms implies that precise waveform details are not important as long as the audio spectrum frequencies remain.
Is this another slew rate argument?
Digital audio is filtered using over-sampling to avoid the phase distortion that linear filters can not avoid.
The success of audio compression algorithms implies that precise waveform details are not important as long as the audio spectrum frequencies remain.
Last edited:
I'm not convinced that a (well designed) amplifier's time delay is dependent on the amplitude. What may be happening here is crossover distortion. If present, it exhibits as a lower gain at low amplitudes. That is well known to increase at low volumes. But assuming an amplifier has minimised or eliminated this effect, the response should be proportional.
It may also just be that dV/dt is causing some internal oscillations which damp out quickly, and don't appear as an output, but may be triggered by crossover effects if that is not properly adjusted.
It may also just be that dV/dt is causing some internal oscillations which damp out quickly, and don't appear as an output, but may be triggered by crossover effects if that is not properly adjusted.
The success of audio compression algorithms implies that precise waveform details are not important as long as the audio spectrum frequencies remain.
Maybe not exactly the right explanation. Compression works to make a song stand out on the radio in relative terms as compared the song before and the song after. It does that by raising the average volume level. Louder always sounds better in relative terms, at least it does up to the point of discomfort.
Given a choice to forgo the loudness wars and just turn up or down the volume a little as needed for different songs, it may be that more people would favor less compressed sound. Also now that we now have pretty good digital perceptual VU meters, hopefully there will be a trend in mastering and perhaps other processing of digital audio that doesn't require compression in order that average volume level can be perceived as competitive.
Last edited:
I think he is referring to lossy compression like MP3 and AAC. Actually we could all learn a lot from those efforts as to what is most important and what is least important in the audio stream.
An amplifier with significant phase modulation with level should show distortion artifacts (sidebands etc.) from the modulation.
An amplifier with significant phase modulation with level should show distortion artifacts (sidebands etc.) from the modulation.
Jan, what about Bob? He is healthy?
Do you agree with the statement of Kyrill Hammer?
Can you distinguish the difference between distortion in steady state operation and when the signal frequency or amplitude changes?
Filter processing softens the effect of speed distortion, but does not eliminate it. If the amplifier had enough bandwidth of 30 kHz with an output voltage slew rate of 5 V / μs, then today amplifiers with a bandwidth of 5 MHz and higher would not be designed.
Personally, I agree with Kyrill Hammer and with my tests I try to show that he is right, that's all.
Well I started to drag myself past the marketing stuff in that Hammer piece until I hit his basic misunderstanding of feedback and time delay, then I gave up.
But let's stick to your graphs and what they mean. Interested in your reply to my questions.
Jan
"Speed distortion" seems to be just another name and spin on Graham Maynard's (RIP) "first cycle distortion", or whatever he called it, but such sudden discontinuous change doesn't exist in the real world of audio, for example a CD audio is heavily LP filtered above 20 kHz so sudden discontinuities would be smoothed out such there wouldn't exist any components above 20 kHz anyway, at least as far as I can see.
Fully agree, it also reminded me of Grahams' stuff. I had several discussions with him, he didn't seem to understand that a signal that has a sinewave start with no bandwidth limiting is a signal with infinite bandwidth. Amplifiers do funny things with infinite bandwidth signals, but does that have anything to do with how they reproduce audio?
Jan
I looked at the article posted. The ‘speed’ distortion the loop making a rapid adjustment at either frequency or amplitude or a combo of both occurs.
You can see this type behavior by comparing a fast rise time square wave signal with a slower one and looking at the diff amp output.
To minimize this, limit the BW and a obviously the compensation design has to be spot on.
I’m not a mathematical sort of guy, but if you take a 20 kHz signal and change its amplitude instantaneously as the OP has done in the test, you are going to get signal artifacts >> than the stimulus signals themselves.
The amplifier is behaving normally.
You can see this type behavior by comparing a fast rise time square wave signal with a slower one and looking at the diff amp output.
To minimize this, limit the BW and a obviously the compensation design has to be spot on.
I’m not a mathematical sort of guy, but if you take a 20 kHz signal and change its amplitude instantaneously as the OP has done in the test, you are going to get signal artifacts >> than the stimulus signals themselves.
The amplifier is behaving normally.
Last edited:
In my post above I meant to say artefacts >> higher in frequency than the stimulus signal (assumed to be a 20 kHz sine wherein the amplitude is changed instantaneously)
Jan: "Also, to judge the performance for audio, the input signal should first be run through an input low pass filter at, say, 30kHz or so."
Jan, I first of all evaluate the parameters of amplifiers, in particular such a speed parameter as speed distortion, and I evaluate the sound quality with my ears. As for the application of a filter for signal processing, I think it would be foolish to measure the slew rate of the output voltage using a meander of a processed low-pass filter with a cutoff frequency of 20 kHz, like here.
I think there is no need to explain which amplifier has more high-speed distortion
Best regards
Petr
Jan, I first of all evaluate the parameters of amplifiers, in particular such a speed parameter as speed distortion, and I evaluate the sound quality with my ears. As for the application of a filter for signal processing, I think it would be foolish to measure the slew rate of the output voltage using a meander of a processed low-pass filter with a cutoff frequency of 20 kHz, like here.
I think there is no need to explain which amplifier has more high-speed distortion
Best regards
Petr
Attachments
The issue is that, as pointed out above by others, if you have an instantaneously changing signal you introduce very high frequencies. In the limit, if your signal changes infinitely last, that signal will have an infinite bandwidth.
It is nonsense to test an audio amplifier with an infinite bandwidth signal and then think you can give a meaningful comment on how they will perform as an audio amplifier.
The reason to send the input signal through an input low pass is exactly to let you do meaningful comments on how the amplifier works for audio. Is that not what you want?
Jan
It is nonsense to test an audio amplifier with an infinite bandwidth signal and then think you can give a meaningful comment on how they will perform as an audio amplifier.
The reason to send the input signal through an input low pass is exactly to let you do meaningful comments on how the amplifier works for audio. Is that not what you want?
Jan
Why do we talk as though amplifiers have ideal brick wall 20kHz filters at the input? Do they really?
Indeed, why should an audio amplifier "be capable" of handling crap noise spitting DACs and the likes up in the MHz region without ahem.. producing "speed distortion" to qualify as "HiFi" like the concerned stuff mentioned in your posts #10240 and #10242 which is not part of the audio information, or is it rather your concealed DAC FUD marketing peddling?
Put a strong radio TX next to an audio gear and you can make almost any of them behave oddly.
Put a strong radio TX next to an audio gear and you can make almost any of them behave oddly.
Last edited:
Markw4: Why do we talk as though amplifiers have ideal brick wall 20kHz filters at the input? Do they really?
No they don't. And I was not aware that we assumed that; I certainly do not.
I want to also draw attention, for an example of a signal that exercises the amp at high frequencies and high slewrates, without making it so unrealistic that the results are useless: the DIM 30 test signal:
DIM stands for Dynamic InterModulation distortion. It is a technique used to measure the non-linearity of a device, and it’s designed to be particularly sensitive to distortions produced during transient conditions typical of audio program material. In DIM measurements, a square wave at a frequency of 3.15 kHz is low-pass filtered and then linearly combined with a sine wave at a frequency of 15 kHz. DIM 30 and DIM 100 use single-pole low-pass filters with cutoff frequencies of 30 kHz and 100 kHz, respectively.
Jan
No they don't. And I was not aware that we assumed that; I certainly do not.
I want to also draw attention, for an example of a signal that exercises the amp at high frequencies and high slewrates, without making it so unrealistic that the results are useless: the DIM 30 test signal:
DIM stands for Dynamic InterModulation distortion. It is a technique used to measure the non-linearity of a device, and it’s designed to be particularly sensitive to distortions produced during transient conditions typical of audio program material. In DIM measurements, a square wave at a frequency of 3.15 kHz is low-pass filtered and then linearly combined with a sine wave at a frequency of 15 kHz. DIM 30 and DIM 100 use single-pole low-pass filters with cutoff frequencies of 30 kHz and 100 kHz, respectively.
Jan
All music signals are bandwidth limited. Microphones, preamplifiers and recording equipment all have finite bandwidth.
If you have a 20 kHz sine wave at say and you instantaneously increase/decrease its output (anywhere on the waveform) you will introduce HF components well above the stimulus frequency.
So, indeed this type of waveform does test for amplifier behavior in the presence of HF content, and it should behave under these conditions - easy to accomplish through good compensation design and with band limiting. However, you don’t get instantaneous changes in music signals precisely because that are BW limited.
I don’t think this has anything to do with DAC/ADC either because you’re supposed to BW limit the input signal to ensure good behavior.
If you have a 20 kHz sine wave at say and you instantaneously increase/decrease its output (anywhere on the waveform) you will introduce HF components well above the stimulus frequency.
So, indeed this type of waveform does test for amplifier behavior in the presence of HF content, and it should behave under these conditions - easy to accomplish through good compensation design and with band limiting. However, you don’t get instantaneous changes in music signals precisely because that are BW limited.
I don’t think this has anything to do with DAC/ADC either because you’re supposed to BW limit the input signal to ensure good behavior.
Put a strong radio TX next to an audio gear and you can make almost any of them behave oddly.
I would be more concerned if a close by radio receiver or a dac caused problems. The real world is not ideal, that's all, nothing more to it.
Ditto regarding your excessively magnified DAC FUD problems which only seems to exist in your imaginary world.
I don’t think this has anything to do with DAC/ADC either because you’re supposed to BW limit the input signal to ensure good behavior.
You anyway need to band-limit the ADC input to avoid aliasing.
Jan
- Home
- Amplifiers
- Solid State
- Bob Cordell's Power amplifier book