Chris
They were doing ‘modern engineering analysis’ buck then in the hard way.
They were aware of the problems and limitations even before WW2.
This is evident from reading the old patents.
And they kept bothering during the stereo era because this was what they had to work with.
When the time came and technology allowed economically feasible mass production of new formats, they had no reason to stick on developing better band aids for vinyl.
I enjoy the photo analogy.
I tap my foot.
Mary had a little lamb
aboriginal George
They were doing ‘modern engineering analysis’ buck then in the hard way.
They were aware of the problems and limitations even before WW2.
This is evident from reading the old patents.
And they kept bothering during the stereo era because this was what they had to work with.
When the time came and technology allowed economically feasible mass production of new formats, they had no reason to stick on developing better band aids for vinyl.
I enjoy the photo analogy.
I tap my foot.
Mary had a little lamb
aboriginal George
It might be worth spending some time working through the Shannon-Nyquist theorem.
The most optimistic data on timing acuity is ~5us under very specific conditions. jneutron can give you the reference for that.
Why do people keep asking if digital can resolve details which, by design, do not make it through the anti-aliasing filter (except in highly-smoothed form)? It is the same as those who insist that their CD player/DAC can reproduce square waves, when the maths/science requires that they cannot!
Hi, and a turntable cartdrige can reproduce a square wave recorded on a LP ?
just asking ... i do not know but still curious.
Thanks and regards,
gino
And what use would that be!
Reproducing 'square waves' is like a chocolate fire guard IMO.
Are the analogue electronics able to handle the high frequency content, is the layout, are the cables, is there a ground plane under all signal traces etc etc.
What rise time of square wave would be acceptable, 1ns-20ps?
Reproducing 'square waves' is like a chocolate fire guard IMO.
Are the analogue electronics able to handle the high frequency content, is the layout, are the cables, is there a ground plane under all signal traces etc etc.
What rise time of square wave would be acceptable, 1ns-20ps?
Hi, i asked because someone mentioned the poor performance of digital circuits with square waves. Let's say the digital and analogue circuit are equally poor ?
Thanks and regards,
gino
Again, it depends on the frequency and what's cut. Phono cartridges ring (so do cutter heads). Digital systems are bandwidth limited.
If you look at a 1kHz square wave from a phono cartridge (that's all my test records have or I've seen published), there's overshoot at leading and trailing edges, and ringing on the top and bottom comprising cartridge resonances and cutter head resonances. A 1kHz square wave from a 16/44 system will only show Gibbs phenomenon. Since a digital system can easily incorporate prospective filters, the Gibbs waveshape can be altered quite a bit and result in a pretty clean looking waveform. A 15 kHz square wave will show up as a sine wave because of bandwidth limiting. I can only imaging what a phono system would look like since this is not a waveform cut into any test records of which I'm aware.
All irrelevant to actual audio signals, of course.
If you look at a 1kHz square wave from a phono cartridge (that's all my test records have or I've seen published), there's overshoot at leading and trailing edges, and ringing on the top and bottom comprising cartridge resonances and cutter head resonances. A 1kHz square wave from a 16/44 system will only show Gibbs phenomenon. Since a digital system can easily incorporate prospective filters, the Gibbs waveshape can be altered quite a bit and result in a pretty clean looking waveform. A 15 kHz square wave will show up as a sine wave because of bandwidth limiting. I can only imaging what a phono system would look like since this is not a waveform cut into any test records of which I'm aware.
All irrelevant to actual audio signals, of course.
Hmm.. an argument taken out of context.
Consider a digital stream converted back to analog using a simple sample and hold. And then, consider the same conversion using an IIR output filter with 8, 16, 64, whatever, numbers of samples (Z). It is trivial to reconstruct to picosecond levels as long as Z is sufficiently large, and impossible if Z=1.
Why is it assumed that this high level of time accuracy can be achieved with an input s/h with single bit sampling on the front end?
Presentation of the math is not consistent with actual circuits in that argument, as the math does not consider the truncation up front.
I wouldn't worry about absolute delays as long as all content gets the same delay.
Below a microsecond, I wouldn't even consider it regardless of what it alters.
Jan Nordmark demonstrated 1.2 uSec dithered capability back in 72. Also, undithered limited bandwidth 5 uSec undithered from 500 hz out to maybe 2Khz.
Others have since confirmed, I just do not have names in my head. edit (other than JJ of course, as speedskater had pointed out earlier, and Greisinger)
jn
Last edited:
I'm not following you here. Possibly your use of Z is confusing, do you mean Z domain? I know that, for example, a simple two pole two zero IIR filter can do RIAA with extremely low phase error and a small (sub 1 sample) fixed latency. This operates on only three samples at a time.
I could dig out the exact numbers.
Sorry.
Z is the delay element in digital filters.
edit: by low phase error, do you mean below 5 uSec in the 500 to 5Khz band where humans are most capable of discernment interaural?
jn
Attachments
Last edited:
Yes, the IIR filter is derived using the Z transform to solve for the coefficients in a recursive finite difference equation similar to what you show.
The answer below uses no more than two previous samples and has a smooth error curve that is only 333nsec at 20k. The error being from the exact minimum phase answer. There is an exact mapping from the difference equation to a continuous time equation in f/fs. I would have to run some software to get a fit from 500-5k but consider for a moment component matching in an analog RIAA, the digital version would be perfect L to R.
EDIT- To get these numbers you need to use ~64bit math and using some least squares tweaking you could dial in the 500-5k phase to be your minimum error range.
Last edited:
Hey, I just opened papoulis to find a pretty picture..
We agree the digital is less susceptible to component value accuracies...
Well at least it's easy..
Do you know if all brickwalls uses 64 bit?
I guess we can assume that we agree that independent clocks are out?
jn
I think all modern boxes like mini-DSP uses a large number of bits to eliminate the overflow/shift/error propagation problems you see when using older DSP's. At least their Excel spread sheet gives many floating point digits of resolution for you to enter (I know nothing about mini-DSP).
I was thinking more of of modern CPU based solutions. Sox for one has no problem using double precision floats for several filters in series in real time. BruteFIR has a shocking throughput on a fast machine.
One clock for all!
I was concerned that the first try would not appear, since it took some time to do so, so I posted it here as well. I DO find the article both readable and fascinating in its implications. And 16 years has gone by since that article was published. Whole books are now written on the subject. The second 'page' was taken from one that I attempted to buy yesterday. Unfortunately, I accidentally bought an E-book and it was incompatible with the operating system on my older MAC, and I might have lost it in the process. You should look into it too. You might have your eyes opened to the 'new applied physics' nanotechnology.
What implications, John?
Why is it that whenever you post stuff like this, you NEVER follow through with anything? NEVER. It just tells me that you don't have the the faintest clue about any of it and you just throw it up as a smoke screen to make it appear that you do.
So again, what are the implications, John?
se
- Status
- Not open for further replies.