Delta LSB
Ok is possible by bad mode to evaluate the delta of bit of audio word
if we think the 0 db at value 2^16=65536 , the value of an example audio track
start at -20 dB this means at 0.1 of maximum range about 6553.6.
The spectrum at 15khz is at -45dB this means 0.0056234 of maximum range this
means a delta 368.54 .
This analyis don't consider the envelope but only attempts the maximum possible delta
on bits
Ok is possible by bad mode to evaluate the delta of bit of audio word
if we think the 0 db at value 2^16=65536 , the value of an example audio track
start at -20 dB this means at 0.1 of maximum range about 6553.6.
The spectrum at 15khz is at -45dB this means 0.0056234 of maximum range this
means a delta 368.54 .
This analyis don't consider the envelope but only attempts the maximum possible delta
on bits
Attachments
Good question about the settling time definition. I believe this value gives the time needed to reach the analog output value that is within +/-1LSB of the ideal value. So if you mute the output for 0.2 us after the conversion time, you could eliminate this error.
I think muting of the unwanted part is better than sample-and-hold of the wanted part.
ok about S/H
But the setling time is a variable and not fixed it depends from delta lsb
But the setling time is a variable and not fixed it depends from delta lsb
ok about S/H
But the settling time is a variable and not fixed it depends from delta lsb
But the settling time is a variable and not fixed it depends from delta lsb
Yes, but it will have some maximum value, therefore, so long as the muting interval is => than the maximum settling time it is possible to mute all settling activity with a uniformly timed trigger pulse. However, without a S/H circuit, the result will be greater net glitch energy, but it would no longer be correlated with the signal. I believe that JVC had utilized JFET switches to mute DAC chip settling activity as part of their famous K2 technology.
Last edited:
Intersting on muting stage to have the jfet.
Is Clear that stage can to be very low noise ( on switching ). I think
Muting part of a sample-and-hold DAC output will modify the audio frequency response. By how much will depend on what fraction of the bit time is muted.lcsaszar said:
Do you have any mathematical equation for this? I understand that an infinitely short pulse will have no rolloff, and a zero-order S/H will have sin(x)/x frequency response.
I don't have the equations to hand, but any good book on sampling theory or digital audio ought to either include them or at least point you in the right direction. From memory you basically take the Fourier transform of the DAC output pulse shape and that gives you the frequency response. As you say, an impulse output gives flat response and the usual s/h gives a sinc response. This is fortunate for NOS DAC fans, as the sinc response reduces the nasty effects of ultrasonic images.
Other reconstruction methods may be better than zero order hold, at least on a subjective ground. I think John (ECDesigns) reported better sonic results with a circuit he calls the reactor. There is an oscillogram somewhere in that long thread (ultimate tda1541).
Other methods may look superficially better (e.g. linear interpolation between points) but actually cause an even worse deviation from flat frequency response, so more correction is needed - but this brings phase problems. The best solution is impulses into a perfect reconstruction filter. We can't do that, so next best is s/h into a good reconstruction filter. Curiously, that is what almost everyone does - apart from those who design with their ears instead of their brains.
Results of tests
I have tested Delta LSB on a some parts of music tracks .
Is Right 500LSB is a good point to start a dac project here my results by Gaussian Calculus
Medium value 204LSB
3 sigma value 457LSB
I have tested Delta LSB on a some parts of music tracks .
Is Right 500LSB is a good point to start a dac project here my results by Gaussian Calculus
Medium value 204LSB
3 sigma value 457LSB
Tests results
Test is on CD's tracks
I have calculate only by delta LSB from sample(k) and sample(k-1)
When you apply the oversampling you use a spline algorithm this increases the difficulty
in the calculation DELTA LSB .
There is many type of spline algorithm I don't know whether to appraise
Test is on CD's tracks
I have calculate only by delta LSB from sample(k) and sample(k-1)
When you apply the oversampling you use a spline algorithm this increases the difficulty
in the calculation DELTA LSB .
There is many type of spline algorithm I don't know whether to appraise
What on earth have splines to do with digital audio?
Splines are curves which join points so they look smooth to the eye.
Digital reconstruction filters recover the original audio which left the anti-aliasing filter by removing ultrasonic images. A reconstruction filter may employ a mix of analogue and digital filtering.
Splines are curves which join points so they look smooth to the eye.
Digital reconstruction filters recover the original audio which left the anti-aliasing filter by removing ultrasonic images. A reconstruction filter may employ a mix of analogue and digital filtering.
Oversampling
The modern dac using a digital interpolation read this
http://yehar.com/blog/wp-content/uploads/2009/08/deip.pdf
The modern dac using a digital interpolation read this
http://yehar.com/blog/wp-content/uploads/2009/08/deip.pdf
That appears to be the result of a student project. His second paragraph seems to show his lack of real understanding, when he immediately starts talking about interpolation when he should be talking about reconstruction. He admits that most of his (unreferenced) sources were on the internet, not peer reviewed journals or textbooks. I would not attach any signifcance at all to that paper. You will have to do better than that.
A bit more
This is a official example on video but for audio is the same
IEEE Xplore Abstract - Cubic splines for image interpolation and digital filtering
This is a official example on video but for audio is the same
IEEE Xplore Abstract - Cubic splines for image interpolation and digital filtering
No; audio is not the same. Please don't spread misinformation. As I said, splines are used to make things appear smooth to the eye. For the ear you have to think about Fourier components so you need a good reconstruction filter to remove all images.
You really must start listening to what we say, or refer to relevant sources. Why not try reading a good textbook on digital audio?
You really must start listening to what we say, or refer to relevant sources. Why not try reading a good textbook on digital audio?
- Status
- Not open for further replies.