Sy,
First, what this broadband noise will be interesting to determine.
Second, I entered Dimitiri's shown circuits into my Simulation Software. This showed a very minimal difference in noise, much, much less that either Scott's Measurements or Dimitri's simulations suggest. Also, Dimiti's simulation is for 1uF/5GOhm, not 50pF/5GOhm. So I would careful using this Sim as evidence, it may better to run your own.
I have performed a simulation much closer based on Scott's circuit, with added RC Networks based on Bob Pease's Capacitor Soakage article (Mica - estimated) and based on the numbers in Dimitri's Sim for Polystyrene. The noise shows so little difference between ideal, RC ladder simulated Polystyrene and RC Ladder simulated Mica that I omitted labelling the traces...
Simulations are a notoriously dodgy area, however if take cognisance of the origins of Johnson/Nyquist noise, we find that if we parallel additional RC Circuits where the R is much greater than the main R in the circuit and C is much smaller, the noise is essentially unaffected, Thevenin, Kirchoff, Nyquist, Johnson and all, nothing more than EE201 should be needed.
Of course, if one knows not the underlying theory and instead yokes oneself to questionable believes, it is easy to jump on false evidence seeming supporting one's viewpoint, but that is hardly better than clutching at straws.
Past that, as I pointed out, that certain types of capacitors (or resistors for that) are noisier than others (though it only shows up in extremely low level and/or high impedance situations which are rare) is hardly news, nor is that certain capacitors are much more microphonic than others.
Ciao T
First, what this broadband noise will be interesting to determine.
Second, I entered Dimitiri's shown circuits into my Simulation Software. This showed a very minimal difference in noise, much, much less that either Scott's Measurements or Dimitri's simulations suggest. Also, Dimiti's simulation is for 1uF/5GOhm, not 50pF/5GOhm. So I would careful using this Sim as evidence, it may better to run your own.
I have performed a simulation much closer based on Scott's circuit, with added RC Networks based on Bob Pease's Capacitor Soakage article (Mica - estimated) and based on the numbers in Dimitri's Sim for Polystyrene. The noise shows so little difference between ideal, RC ladder simulated Polystyrene and RC Ladder simulated Mica that I omitted labelling the traces...
Simulations are a notoriously dodgy area, however if take cognisance of the origins of Johnson/Nyquist noise, we find that if we parallel additional RC Circuits where the R is much greater than the main R in the circuit and C is much smaller, the noise is essentially unaffected, Thevenin, Kirchoff, Nyquist, Johnson and all, nothing more than EE201 should be needed.
Of course, if one knows not the underlying theory and instead yokes oneself to questionable believes, it is easy to jump on false evidence seeming supporting one's viewpoint, but that is hardly better than clutching at straws.
Past that, as I pointed out, that certain types of capacitors (or resistors for that) are noisier than others (though it only shows up in extremely low level and/or high impedance situations which are rare) is hardly news, nor is that certain capacitors are much more microphonic than others.
Ciao T
Attachments
The CD isn't more acoustically transparent ... it has problems IMO.
No I have very good CD play back. My bluray player has the ESS Sabre32 Reference DAC.
Our CD playback can sound very good, and I really enjoy it .. but when comparing a very good sounding recording on CD like Natalie Merchant's Tiger Lily to the LP 45 rpm you hear how limited the CD is.
Dynamic Comparison of LPs vs CDs - Part 4 - page 2 — Reviews and News from Audioholics
That's his subjective opinion. I've heard others say the opposite. I mean if you can't hear the affect that A/D and D/A conversion has on the signal then something like a Blowtorch preamp would not interest you.
.
The equalization used on the LP reduces the surface noise preserving the dynamic range of the recording.
I really cannot see how CD and LP can have a broadly similar dynamic range. I think you have to include the playback performance in the assessment (everything), and if you do that LP's are usualy not better than about 65dB (a first class pressing). I believe CD is considerably better than that. Try Stravinsky's 'Firebird Suite' where the strings fade to zero. I think CD would win on that one.
Bonsai said:
The dynamic range of the LP improves as frequency increases.
The narrow grove of an LP is more accommodating to high frequencies since they require less excursion of the cutting lathe. Hence the need for attenuating the low frequencies to fit in the groove. At least, that's how I understand it ... not any kind of expert on cutting an LP.
Hi,
It is not a factor. Just take any decent multibit DAC and measure it...
The lowest signal level is one that is equivalent to a sinewave at -87dBfs, if measured using classic analogue methods..
These tones are only measurable with dither AND effectively, in terms of analogue measurements do not exist because they are below the noisefloor.
Yes, absolutely, IF you apply identical measurement techniques. Without dbx or dolby and really low noise tape a decent "halve track" mastering tape machine can approach -80dB for noise WITH over 14dB headroom remaining.
However, if we measure below the noisefloor things go very far down. There is eventually a granularity to the magnetic dipoles in the tape, but the levels are very low.
What about it?
I am not championing Analogue Recording Methods, I merely insisting on keeping things REAL, that is realistic, comparable and demonstrable, instead of using made up numbers and unfounded claims in any debate concerning the relative merits and to not "build strawmen". Basically the same as before I was sinbinned for "disruptive behaviour" (disruptive only to "Strawmen" and unfounded claims, I would add).
Okay, so CD Digital audio does not work without degrading it's dynamic range by at least 6dB. Is that what you claim? If so, I will tend to agree. So the real dynamic range for CD then becomes 84dB. Do you agree?
The 96dB Figure is not correct. It is an advertising "Milkmaid calculation" derived number. The real number is 90dB without correcting for the RMS/Peak error at low levels and with 1lsb Dither it becomes (around) 84dB.
I am very familiar with the paper. However, "dither" is just a fancy way for saying "adding noise" (or rather to avoid admitting that one adds noise) and "noiseshaping" is just a fancy way of saying "adding fuzzy distortion" (or rather to avoid admitting that one adds fuzzy distortion).
The AD was designed by Burr Brown on a Chip Level. The supporting circuitry is at least as good or better than BB recommends in their application notes and demo boards.
It does sound closer to the real recording than an EMU 1616M (which is a "textbook" design and has notionally superior specification).
So I am on the other hand am inclined to wonder if the "no difference between CDR and LP" was a combination of wishful thinking and the heavy weighting towards null results present in the small scale ABX test?
So "We wait and we wonder"....
Ciao T
It is not a factor. Just take any decent multibit DAC and measure it...
The lowest signal level is one that is equivalent to a sinewave at -87dBfs, if measured using classic analogue methods..
These tones are only measurable with dither AND effectively, in terms of analogue measurements do not exist because they are below the noisefloor.
Yes, absolutely, IF you apply identical measurement techniques. Without dbx or dolby and really low noise tape a decent "halve track" mastering tape machine can approach -80dB for noise WITH over 14dB headroom remaining.
However, if we measure below the noisefloor things go very far down. There is eventually a granularity to the magnetic dipoles in the tape, but the levels are very low.
What about it?
I am not championing Analogue Recording Methods, I merely insisting on keeping things REAL, that is realistic, comparable and demonstrable, instead of using made up numbers and unfounded claims in any debate concerning the relative merits and to not "build strawmen". Basically the same as before I was sinbinned for "disruptive behaviour" (disruptive only to "Strawmen" and unfounded claims, I would add).
Okay, so CD Digital audio does not work without degrading it's dynamic range by at least 6dB. Is that what you claim? If so, I will tend to agree. So the real dynamic range for CD then becomes 84dB. Do you agree?
The 96dB Figure is not correct. It is an advertising "Milkmaid calculation" derived number. The real number is 90dB without correcting for the RMS/Peak error at low levels and with 1lsb Dither it becomes (around) 84dB.
I am very familiar with the paper. However, "dither" is just a fancy way for saying "adding noise" (or rather to avoid admitting that one adds noise) and "noiseshaping" is just a fancy way of saying "adding fuzzy distortion" (or rather to avoid admitting that one adds fuzzy distortion).
No offense intended, but you're criticizing digital using your own digital designs as proof. I'm inclined to wonder whether there might have been a bit of self-fulfilling prophecy here. Or, are you saying that you designed the phono stage only, and not also the 24/96 A/D? Maybe you just need 192 kHz.
The AD was designed by Burr Brown on a Chip Level. The supporting circuitry is at least as good or better than BB recommends in their application notes and demo boards.
It does sound closer to the real recording than an EMU 1616M (which is a "textbook" design and has notionally superior specification).
So I am on the other hand am inclined to wonder if the "no difference between CDR and LP" was a combination of wishful thinking and the heavy weighting towards null results present in the small scale ABX test?
So "We wait and we wonder"....
Ciao T
Sy,
Or perhaps it is a simulation error, caused by running the simulator under conditions it is not designed to correctly handle? Simulatons are just that, math based on certain simplified models.
So perhaps is is a better model if you want to believe into your thesis, which, as it cannot be sufficiently reconciled with basic electronic theory, would seem extraordinary....
Well, why don't you tell me what the correct value should be, given my stipulations given in my original post, if you disagree.
Ciao T
Or perhaps it is a simulation error, caused by running the simulator under conditions it is not designed to correctly handle? Simulatons are just that, math based on certain simplified models.
So perhaps is is a better model if you want to believe into your thesis, which, as it cannot be sufficiently reconciled with basic electronic theory, would seem extraordinary....
Well, why don't you tell me what the correct value should be, given my stipulations given in my original post, if you disagree.
Ciao T
Agree that this may be the case on the actual LP - but I think you need to look at the system dynamix range here - so, the LP (with its surface noise etc), the RIAA equalizer, where the topology will have a significant impact on the dynamic range.
Anyway, I am still trying to understand how an LP can have a similar dynamic range to a CD. Interesting link to AA above BTW.
Hi,
I mean the actual modulator core in the DAC is capable of representing a limited number of actual states or levels it can represent by oversampling rate and inherent design.
The late 1980's TDA1541 can represent directly 2^16 levels for any given sample.
Modern "low-bit" DAC's can usually represent 6 - 64 levels * 128 - 256 OS (for 44.1KHz) or in other words 9.5 to 14 Bit. The only current production DAC's that manages or exceeds true 16 Bit equivalent core performance is the ESS9018 in 2-Channel mode and the BB PCM1704.
Ciao T
I mean the actual modulator core in the DAC is capable of representing a limited number of actual states or levels it can represent by oversampling rate and inherent design.
The late 1980's TDA1541 can represent directly 2^16 levels for any given sample.
Modern "low-bit" DAC's can usually represent 6 - 64 levels * 128 - 256 OS (for 44.1KHz) or in other words 9.5 to 14 Bit. The only current production DAC's that manages or exceeds true 16 Bit equivalent core performance is the ESS9018 in 2-Channel mode and the BB PCM1704.
Ciao T
Actually, it's Scott's thesis, it is plausible considering standard physics, and he's presented experimental evidence. You've offered no alternative, nor any analysis as to why his thesis is incorrect.
Hi Sy,
Ahh, you are right, it was originally 100m, I upped it to 100G so I could quickly run a set of AC transfer curves (to be sure which trace was which).
It was at 100mOhm for the traces shown in my post and indeed should be when running the traces with any other Sim...
Ciao T
Ahh, you are right, it was originally 100m, I upped it to 100G so I could quickly run a set of AC transfer curves (to be sure which trace was which).
It was at 100mOhm for the traces shown in my post and indeed should be when running the traces with any other Sim...
Ciao T
Thanks, Thorsten.
I'd appreciate two more bits of clarification:
First, I would expect three lines on the sim chart (I'd run it myself, but I have about zero ability)- the ideal, the polystyrene (which should run fairly close to the ideal), and the mica. I only saw two. Is it the resolution of my laptop screen? The scale?
Second, what equations did you use to derive the values in the model? Yes, you said, "Pease," I've read his stuff on soakage, but were your values derived by taking into account the experimentally observed DA behavior of actual mica caps? If so, could you explain your calculations? It would be educational for us, and I would be personally appreciative.
Given my limited experience with simulation, I am agnostic about the accuracy and applicability of that model, especially at high frequencies, but could certainly be convinced one way or the other.
I'd appreciate two more bits of clarification:
First, I would expect three lines on the sim chart (I'd run it myself, but I have about zero ability)- the ideal, the polystyrene (which should run fairly close to the ideal), and the mica. I only saw two. Is it the resolution of my laptop screen? The scale?
Second, what equations did you use to derive the values in the model? Yes, you said, "Pease," I've read his stuff on soakage, but were your values derived by taking into account the experimentally observed DA behavior of actual mica caps? If so, could you explain your calculations? It would be educational for us, and I would be personally appreciative.
Given my limited experience with simulation, I am agnostic about the accuracy and applicability of that model, especially at high frequencies, but could certainly be convinced one way or the other.
Hi,
There are three lines, two are so close they only visibly diverge at extreme zoom factors.
First, the general equations for scaling are on the screenscrape by Dimitri. I scaled Bob Pease's Mylar model, as he shows mylar as pretty much "as bad" as Mica.
I find that simulations generally work acceptably for "generic stuff" (AC Frequency Response, Gain, DC Operating points), but are at best "ballpark" for HD, Noise and other more complex stuff. In general this is due to the fact that neither models nor simulator implementation seem very "realistic".
I have been using PSpice since the mid 90's and Electronic Workbench prior to this (and have made many custom models for PSpice over the years based on curve tracing real devices) and have had plenty of occasion to compare simulation to actual circuit.
I would say that I find Sim's good for first cuts, to try out if something stands a chance to work at all. But I also find that it is constantly essential to perform reality checks on the results, often small mistakes can give drastically unrealistic results. It is one of the reasons why I rarely if ever publish results from Simulations, not because I cannot, but because I know they are unrealistic...
I usually find that real circuits need manual fine tuning from Sim's, often quite substantial, to perform as intended.
Ciao T
There are three lines, two are so close they only visibly diverge at extreme zoom factors.
First, the general equations for scaling are on the screenscrape by Dimitri. I scaled Bob Pease's Mylar model, as he shows mylar as pretty much "as bad" as Mica.
I find that simulations generally work acceptably for "generic stuff" (AC Frequency Response, Gain, DC Operating points), but are at best "ballpark" for HD, Noise and other more complex stuff. In general this is due to the fact that neither models nor simulator implementation seem very "realistic".
I have been using PSpice since the mid 90's and Electronic Workbench prior to this (and have made many custom models for PSpice over the years based on curve tracing real devices) and have had plenty of occasion to compare simulation to actual circuit.
I would say that I find Sim's good for first cuts, to try out if something stands a chance to work at all. But I also find that it is constantly essential to perform reality checks on the results, often small mistakes can give drastically unrealistic results. It is one of the reasons why I rarely if ever publish results from Simulations, not because I cannot, but because I know they are unrealistic...
I usually find that real circuits need manual fine tuning from Sim's, often quite substantial, to perform as intended.
Ciao T
"Modern "low-bit" DAC's can usually represent 6 - 64 levels * 128 - 256 OS (for 44.1KHz) or in other words 9.5 to 14 Bit. The only current production DAC's that manages or exceeds true 16 Bit equivalent core performance is the ESS9018 in 2-Channel mode and the BB PCM1704."
This does not sound right to me. Please take a look at some of the Stereophile reports.
This does not sound right to me. Please take a look at some of the Stereophile reports.
- Status
- Not open for further replies.