Charles,
I don't think you have. Looking at your oscilloscope pictures (I really suggest using a spectrum analyzer), it's hard to see any difference between the two pictures. Yes, the third picture might have some extremely small HF ringing, but as it is above Nyquist for a 44.1 kHz signal, and above human hearing range, it won't be audible anyway. So so far you have shown that the output of 2 different ADCs might possibly *look* minimally different on an oscilloscope screen. What does that show about either the sound quality or the measurements of your ADC?
I don't think you have. Looking at your oscilloscope pictures (I really suggest using a spectrum analyzer), it's hard to see any difference between the two pictures. Yes, the third picture might have some extremely small HF ringing, but as it is above Nyquist for a 44.1 kHz signal, and above human hearing range, it won't be audible anyway. So so far you have shown that the output of 2 different ADCs might possibly *look* minimally different on an oscilloscope screen. What does that show about either the sound quality or the measurements of your ADC?
Well, if you say so... But you have definitely not shown to us that that is the case.
Original question (post 220):
Revised question (post 225):
I would advise you to stop digging, as the hole you are in is getting deeper. Someone who appears to be a stranger to Shannon and Fourier really should not be 'designing' commercial ADCs.
I answered the question originally posed.Charles said:
Revised question (post 225):
The answer to the revised question is any source with signal components above 96kHz.Charles said:
No. You have shown that an anti-aliasing filter causes Gibb's phenomena (pre and post ringing) on a sharp transition, whatever ADC technology is used. The lack of such a filter with your ADC means that the ringing is absent, but instead you will get aliasing. This might not be obvious in the time domain (oscilloscope) but a spectrum analyser will show it. If you don't like anti-aliasing filters then you don't have to use them; just don't do any sampling!Charles said:
I would advise you to stop digging, as the hole you are in is getting deeper. Someone who appears to be a stranger to Shannon and Fourier really should not be 'designing' commercial ADCs.
As stated earlier, I made this filterless SAR ADC, because I wanted to find out how alias distortion would sound, when it would appear.
After I had finished the ADC, I recorded vinyl, recorded with microphones, recorded my old Philips CD-player, recorded with my Attraction DAC as input, and I never got any alias distortion, all I got was great sound.
But one day, I connected my Squeezebox (this has a sigma-delta DAC inside) to my SAR ADC, and then I got a really noisy hiss on the recording, because all the high frequency garbage of the sigma delta DAC's output folds down right into the audible range if you do not filter it before it enters the ADC.
So the only instance when I was able to get alias distortion with my filterless ADC, is when a sigma-delta DAC pollutes the signal.
After I had finished the ADC, I recorded vinyl, recorded with microphones, recorded my old Philips CD-player, recorded with my Attraction DAC as input, and I never got any alias distortion, all I got was great sound.
But one day, I connected my Squeezebox (this has a sigma-delta DAC inside) to my SAR ADC, and then I got a really noisy hiss on the recording, because all the high frequency garbage of the sigma delta DAC's output folds down right into the audible range if you do not filter it before it enters the ADC.
So the only instance when I was able to get alias distortion with my filterless ADC, is when a sigma-delta DAC pollutes the signal.
No, you never noticed any alias distortion. That is the problem with experience; when unconstrained by hard theoretical facts it can mislead people.Charles said:
Try recording an FM tuner output at 44.1kHz sampling rate. You will probably be able to hear some mush around 6.1kHz - the alias of the stereo mux signal at 38kHz which has made it through the tuner's own filtering. With a good tuner this might be 30-50dB down; a bad tuner could be much greater. Shannon really was right!
Hi Trevor,
I have shown that a sigma-delta ADC causes distortion when fed with a HF-click and that a zero-filter SAR ADC (i.e. my Creation ADC) does not create distortion, btw. the test wav I used is available for download on my website. So this is something everybody can try out himself and I need not add anything to pure facts, as they stand for themselves
In the examples on my websites I have chosen this mixed signal which contains a square step added to a sine wave, so that you can clearly see, that the step as well as the sine contains no distortion when recorded with my Creation ADC, but does get corrupted when recorded with any sigma-delta ADC
But your idea of sampling a 20kHz square wave is good. I have the DAC that can play back a true square-wave and I have the ADC that can sample a true square wave, and the output will again be a pure square wave.
Try this with a sigma-delta ADC/DAC and you will get a huge amount of distortion. To say that this distortion can be explained by the theory of Fourier transformation will not remove the distortion.
The definition of distortion is that the output is different than the input.
Charles
Nonsense !!
With your ADC try and sample a 20KHz squarewave at 96KHz and I can assure you there will be alias components at 36KHz and 4KHz as a result of the 3rd and 5th harmonics of the square wave and you cannot get rid of these artefacts once they have been sampled that way.
And btw the sigma delta ADC's and DAC's are not adding any non linear distortion. The so called linear distortion you are seeing is coming from the anti aliasing filters used which is the correct way to do it. There is no free lunch when it comes to sampling theory. What you are trying to do is to rewrite the laws which are well established.
Unfortunately, in the audiophile world it seems to be all too easy to disregard them
but they forget who sets the objectives on any new audio standards and hardware. It is usually done by engineers and not by audiofools. They don't come up with anything new !!
Last edited:
Indeed. There is unfortunately nothing Special or Magic about a properly engineered product. Anyone can buy one, while it takes a Real Audiophile (who has read the right reviews in the Right Magazines) to recognize a product created by an Artist rather than an engineer. But even then you can always make it even more Magical and Special by adding the right cables and tweaks...
Actually it can be surprisingly easy to disregard maths and physics; you just need to be sufficiently ignorant or arrogant.TNT said:
The big snag is that maths and physics will never disregard you. The Nyquist sampling limit is no respecter of persons; it always operates perfectly and will generate aliases when given the opportunity. Fourier always applies to music waveforms, as they have a finite (zero) number of finite (zero) discontinuities. Causality requires a brick-wall filter to cause ringing, not by adding anything but by what is taken away - the 'ringing' was already there in the (artificial) waveform but was hidden by higher frequency components which the filter has removed.
All this unavoidable maths and physics keeps biting the unwary. The sad thing is that often they don't even notice!
hey what happened to charley ??
he demonstrates on his own website why you need proper antialiasing filters for adc's and reconstructions filters for dacs
check it out
Mother of Tone - The CD Format
is this dude just trying to take the **** out of everyone ??
he demonstrates on his own website why you need proper antialiasing filters for adc's and reconstructions filters for dacs
check it out
Mother of Tone - The CD Format
is this dude just trying to take the **** out of everyone ??
It might have been helpful if he had noted that analogue filters and IIR digital filters use an infinite number of input samples, and so may be good approximations to a reconstruction filter. It might also have been helpful to note that an inadequate reconstruction filter produces some combination of frequency response errors for the wanted signal (no non-linear distortion) and unwanted higher frequency images. The resultant signal waveform may look unnecessarily alarming.
For example, the envelope modulation seen on a high frequency tone tone is simply a result of the signal beating with the image.
However, that page is a reasonable attempt at trying to grasp, and present, some understanding of digital audio. It is a pity that some of his products do not take account of this.
For example, the envelope modulation seen on a high frequency tone tone is simply a result of the signal beating with the image.
However, that page is a reasonable attempt at trying to grasp, and present, some understanding of digital audio. It is a pity that some of his products do not take account of this.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.