Hi,
It is. TDA1541 with DEM Reclocking.
Actually the "Delta Sigma" part (or more precisely the equivalent PWM "bits") is included in my 13 bits.
The other techniques are noise shaping, or if we want to express it negatively noise modulation. While noise shaping allows nice measured sinewaves when using large numbers of cycles averages (the essentially random noise is filtered out), there is no "average of 4,192 transients" in music. 😀
No, I am saying that these "other means" do not in any real and relevant way extend the core resolution of the device.
Given how many people listen to Non-Oversampling TDA1543 DAC's and are happy these 13 Bit are probably good enough most of the time, but it's still a downgrade from a real 16-Bit DAC. 😀
Ciao T
It is. TDA1541 with DEM Reclocking.
Actually the "Delta Sigma" part (or more precisely the equivalent PWM "bits") is included in my 13 bits.
The other techniques are noise shaping, or if we want to express it negatively noise modulation. While noise shaping allows nice measured sinewaves when using large numbers of cycles averages (the essentially random noise is filtered out), there is no "average of 4,192 transients" in music. 😀
No, I am saying that these "other means" do not in any real and relevant way extend the core resolution of the device.
Given how many people listen to Non-Oversampling TDA1543 DAC's and are happy these 13 Bit are probably good enough most of the time, but it's still a downgrade from a real 16-Bit DAC. 😀
Ciao T
Last edited:
Hi,
I am sorry, but NOTHING can make a 120dB Dynamic Range (not even A-Weighted) 16Bit mixdown. It's like a Mozzy sucking on a mummy. Forget it. Not happening.
I am familiar with the process.
BTW, did you ever try recording/mastering in 88.2 or 176.4 24Bit, then taking the master, scaling the signal with the cleanest algorithm so the higest peak equals 0dbfs and then "resample" to 16/44 by discarding the extra samples and wordlength and THEN play back with a filterless DAC (getting back close to what Peter Baxandall heard all these years ago)? It is quite an interesting experiment.
Okay, even the company I work for uses the impulse response for marketing Non-OS.
In my view this is NOT what makes the "sound" of Non-OS, but the stuff that really happens is by far to esoteric for most degreed engineers (I have alluded to it in my own writings on prior occasions, in conversations I get so many blank stares, I gave up ages ago), never mind consumers.
And if you have a product to sell you need nice pat slides that sales people can say: Look at this - this is ours, this the others" so that the potential buyers can appreciate what makes "your" product better. Everyone does it, from Coca Cola to Lockheed and everyone in-between.
Ciao T
I am sorry, but NOTHING can make a 120dB Dynamic Range (not even A-Weighted) 16Bit mixdown. It's like a Mozzy sucking on a mummy. Forget it. Not happening.
I am familiar with the process.
BTW, did you ever try recording/mastering in 88.2 or 176.4 24Bit, then taking the master, scaling the signal with the cleanest algorithm so the higest peak equals 0dbfs and then "resample" to 16/44 by discarding the extra samples and wordlength and THEN play back with a filterless DAC (getting back close to what Peter Baxandall heard all these years ago)? It is quite an interesting experiment.
Okay, even the company I work for uses the impulse response for marketing Non-OS.
In my view this is NOT what makes the "sound" of Non-OS, but the stuff that really happens is by far to esoteric for most degreed engineers (I have alluded to it in my own writings on prior occasions, in conversations I get so many blank stares, I gave up ages ago), never mind consumers.
And if you have a product to sell you need nice pat slides that sales people can say: Look at this - this is ours, this the others" so that the potential buyers can appreciate what makes "your" product better. Everyone does it, from Coca Cola to Lockheed and everyone in-between.
Ciao T
Simple answer; I guess I feel like that limits me too much. Can't I do better than the DAC chip used in the SA963? Also, I would like to build one of these for my brother too. He does not have the 963. He has the SD9200. I also have some hope that I can make this somewhat modular, allowing later upgrades and/or changes. Example, the preamp output stage could be a Alpeh P clone to begin with. Later, I could replace that PCB with a tube stage.
Right now, I am thinking it will be composed of 3 main PCB's.
1) preamp output and volume control
2) DAC and I/V stage
3) uP control board (remote control for volume and mute)
Ok, so the TDA1541 is great for NOS applications. What about the PCM1704?
isn't this a true 20+ bit converter? without noise modulation, etc. ?
isn't this a true 20+ bit converter? without noise modulation, etc. ?
Do you have any "good" discrete design you can share with me?
Also, what is the "Right" kind of IC? can you give me an example? do you mean something like the diamond transistor or just choosing the right op-amp?
Hi,
Well, one could make the PCB pretty universal. The ASRC's can convert from one format into another and usually take any input you like. By getting rid of SPDIF as interface you remove so many headaches and problems, it really tells.
Untill some recent extremely serious work on cracking and eliminating the whole SPDIF jitter thing WITHOUT ASRC I have never been able to make a DAC that rivalled my modded or groundup designed one box players.
With a One-Box player you simply stick a set of 74AC74 or similar in front of your DAC's and clock them directly from the clock that you use to drive the rest of the Player, then send this clock, if needed suitably devided to the transport and you eliminate any jitter issues.
Of course, this can also be done externally, but then you still need to modify the transport to be able to send a clock back.
My latest "not quite commercial" DAC I am playing with uses asynchronous USB (so same situation - clock is on the DAC side, not on the Computer side - so it is like clock-linked Transport & DAC).
And even with a mediocre Hybrid DAC (one from Burr Brown with 13 Bit real resolution) it is about as good as something much more excotic that uses SPDIF (and not the WM Receiver Chip). It will be fun to see what happens when I pull my PCM1704 samples off the shelf and interface them to the USB core and start playing 176.4/24 Recordings by Prof Johnson instead of my HDCD Rips.
Ciao T
PS, my transport is a custom build fanless silent touchscreen PC, nowadays I would just use one of them 700 USD touchscreen nettops instead. Plays up to 24/192 and more software comes on line every day.
Well, one could make the PCB pretty universal. The ASRC's can convert from one format into another and usually take any input you like. By getting rid of SPDIF as interface you remove so many headaches and problems, it really tells.
Untill some recent extremely serious work on cracking and eliminating the whole SPDIF jitter thing WITHOUT ASRC I have never been able to make a DAC that rivalled my modded or groundup designed one box players.
With a One-Box player you simply stick a set of 74AC74 or similar in front of your DAC's and clock them directly from the clock that you use to drive the rest of the Player, then send this clock, if needed suitably devided to the transport and you eliminate any jitter issues.
Of course, this can also be done externally, but then you still need to modify the transport to be able to send a clock back.
My latest "not quite commercial" DAC I am playing with uses asynchronous USB (so same situation - clock is on the DAC side, not on the Computer side - so it is like clock-linked Transport & DAC).
And even with a mediocre Hybrid DAC (one from Burr Brown with 13 Bit real resolution) it is about as good as something much more excotic that uses SPDIF (and not the WM Receiver Chip). It will be fun to see what happens when I pull my PCM1704 samples off the shelf and interface them to the USB core and start playing 176.4/24 Recordings by Prof Johnson instead of my HDCD Rips.
Ciao T
PS, my transport is a custom build fanless silent touchscreen PC, nowadays I would just use one of them 700 USD touchscreen nettops instead. Plays up to 24/192 and more software comes on line every day.
Hi,
Depends heavily on the chosen DAC, there is nothing that comes even close to Universal for solid state, it is possible to make something that has very little compromise and is pretty universal using Tubes though.
Again, it depends heavily on the specific DAC.
Take as an Example a differential Voltage Output DAC like the WM874X series from Wolfson (or similar DAC's from AKM, Burr Brown and Cirrus logic). In this case what you need most is speed, so you need some of Video Op-Amp that can handle al that very high supersonic noise all these chips put out lineary, while also having low noise with sensible impedance and so on.
For example I find the LM6171/72 to be a good choice with Delta Sigma DAC's, if the circuit impedances can be kept low enough to not cause too much noise from the noise currents. Few if any people use them though.
The Diamond transistors are another good choice, but not if you need a filter/balanced-2-se converter and so on.
Ciao T
Depends heavily on the chosen DAC, there is nothing that comes even close to Universal for solid state, it is possible to make something that has very little compromise and is pretty universal using Tubes though.
Again, it depends heavily on the specific DAC.
Take as an Example a differential Voltage Output DAC like the WM874X series from Wolfson (or similar DAC's from AKM, Burr Brown and Cirrus logic). In this case what you need most is speed, so you need some of Video Op-Amp that can handle al that very high supersonic noise all these chips put out lineary, while also having low noise with sensible impedance and so on.
For example I find the LM6171/72 to be a good choice with Delta Sigma DAC's, if the circuit impedances can be kept low enough to not cause too much noise from the noise currents. Few if any people use them though.
The Diamond transistors are another good choice, but not if you need a filter/balanced-2-se converter and so on.
Ciao T
Optional 2nd PLL
The dynamic performance of the D/A conversion system is
influenced by clock jitter and the sensitivity of this jitter is
mainly given by the architecture of the D/A converter. The
EVM-1702 uses a low jitter digital audio interface receiver
(CS8412), in conjunction with a multi-bit DAC (PCM1702),
to achieve superior performance without the use of a re
clocking circuit.
The EVM-1702 has an optional 2nd PLL circuit (not in-
stalled) which can be populated if the user so chooses. Refer
to Figure 1. The clock selection can be controlled by JP1
through JP3, normally set to side 1.
http://www.datasheetarchive.com/pdf/Datasheet-02/DSA0023347.pdf
maybe there were no codec chips and mainboard wonders back then, though.
The dynamic performance of the D/A conversion system is
influenced by clock jitter and the sensitivity of this jitter is
mainly given by the architecture of the D/A converter. The
EVM-1702 uses a low jitter digital audio interface receiver
(CS8412), in conjunction with a multi-bit DAC (PCM1702),
to achieve superior performance without the use of a re
clocking circuit.
The EVM-1702 has an optional 2nd PLL circuit (not in-
stalled) which can be populated if the user so chooses. Refer
to Figure 1. The clock selection can be controlled by JP1
through JP3, normally set to side 1.
http://www.datasheetarchive.com/pdf/Datasheet-02/DSA0023347.pdf
maybe there were no codec chips and mainboard wonders back then, though.
Last edited:
Hi,
Yup. It needs great care in implementation though, it is very sensitivie to power supply noise, needs the DEM system clock linked to the audio clocks (with low jitter) and special care for the inputs and so on.
The PCM1704 has the ability to take in 24 Bit words and it actually does something with all bits.
But the actual analogue performance in terms of Signal/Noise ratio and/or dynamic range is around 112dB (it does do 120dBA "typhical" A-Weighted SNR/DNR for the K-Grade chips).
So realistically, if we take the old chestnut of 96dB = 16 Bit, then we have around 18.5 Bit or so equivalent dynamic range. If we use the PCM1704 in parallel/differential configurations we can one more bit every time we quadruple the Chips.
So, for arguments sake, using two groups of 8pcs parallel PCM1704 in a differential arrangement (the issues of driving so many IC inputs becomes non-trivial BTW, as does re-clocking) we get another 12dB and thus another 2 Bit, so we would end up with > 20 Bit "real" resolution.
All in all the current situation with only the PCM1704 remaining as "direct conversion" DAC is subideal. No-one seems to work on anything to challenge it.
BTW, by direct conversion I mean that a given numerical input leads to a given analogue output, not to a bunch of noise the average of which is something not entierly unlike what the correct analogue representation of the numerical input may be, if the noiseshaping algorythms and digital filters actually work anything like they should but rarely do.
BTW, more recent FPGA's can run easily at speeds of 100MHz+, even for 192KHz one could get 512 times Fs as basis for PWM use, this would give a real 9 bit equaivalent resolution in PWM.
Using fairly simple to trim R2R DAC's hung directly of the FPGA outputs with good power supplies for each bank used as "DAC" can produce 14 Bit precision easily, with some effort 16 Bit can be approached. If one uses several switched elements for the MSB(s) one can forgo trimming and get a similar level of precision as the TDA1541 attains (16 Bit without trimming).
So a DAC capable of accepting and rendering 24 Bit without noiseshaping, filtering etc. could be made in tru DIY fashion. Modern FPGA's also have enough resources to allow memory buffers and SPDIF receivers to be integrated.
So if we have a FPGA wizzard (Deep Black Art level needed) who would like to the needed work for free (open source) it may be possible to break the current deadlock (e.g. all DACs designed include noiseshaping and low bit PCM cores as this is the only way to get the measured performance to compete with all the other chip makers with the silicon area budget accountants and shareholders allow).
Ciao T
Yup. It needs great care in implementation though, it is very sensitivie to power supply noise, needs the DEM system clock linked to the audio clocks (with low jitter) and special care for the inputs and so on.
The PCM1704 has the ability to take in 24 Bit words and it actually does something with all bits.
But the actual analogue performance in terms of Signal/Noise ratio and/or dynamic range is around 112dB (it does do 120dBA "typhical" A-Weighted SNR/DNR for the K-Grade chips).
So realistically, if we take the old chestnut of 96dB = 16 Bit, then we have around 18.5 Bit or so equivalent dynamic range. If we use the PCM1704 in parallel/differential configurations we can one more bit every time we quadruple the Chips.
So, for arguments sake, using two groups of 8pcs parallel PCM1704 in a differential arrangement (the issues of driving so many IC inputs becomes non-trivial BTW, as does re-clocking) we get another 12dB and thus another 2 Bit, so we would end up with > 20 Bit "real" resolution.
All in all the current situation with only the PCM1704 remaining as "direct conversion" DAC is subideal. No-one seems to work on anything to challenge it.
BTW, by direct conversion I mean that a given numerical input leads to a given analogue output, not to a bunch of noise the average of which is something not entierly unlike what the correct analogue representation of the numerical input may be, if the noiseshaping algorythms and digital filters actually work anything like they should but rarely do.
BTW, more recent FPGA's can run easily at speeds of 100MHz+, even for 192KHz one could get 512 times Fs as basis for PWM use, this would give a real 9 bit equaivalent resolution in PWM.
Using fairly simple to trim R2R DAC's hung directly of the FPGA outputs with good power supplies for each bank used as "DAC" can produce 14 Bit precision easily, with some effort 16 Bit can be approached. If one uses several switched elements for the MSB(s) one can forgo trimming and get a similar level of precision as the TDA1541 attains (16 Bit without trimming).
So a DAC capable of accepting and rendering 24 Bit without noiseshaping, filtering etc. could be made in tru DIY fashion. Modern FPGA's also have enough resources to allow memory buffers and SPDIF receivers to be integrated.
So if we have a FPGA wizzard (Deep Black Art level needed) who would like to the needed work for free (open source) it may be possible to break the current deadlock (e.g. all DACs designed include noiseshaping and low bit PCM cores as this is the only way to get the measured performance to compete with all the other chip makers with the silicon area budget accountants and shareholders allow).
Ciao T
Hi,
I think the oscillator and much of the circuit matches what Nelson used in D1.
This was 1997 BB pre the TI aquisition, when they actually still cared enough to make good products, oh those old golden days.
It is a bit prehistoric though (but few DAC's anyone currently can buy would outperform this 13 Year old concept if the 2nd PLL is used).
Jos van Eijndhoven - DAC2
This shows something more modern. There are further ways to improve the concept which I cannot discuss.
Ciao T
I think the oscillator and much of the circuit matches what Nelson used in D1.
This was 1997 BB pre the TI aquisition, when they actually still cared enough to make good products, oh those old golden days.
It is a bit prehistoric though (but few DAC's anyone currently can buy would outperform this 13 Year old concept if the 2nd PLL is used).
Jos van Eijndhoven - DAC2
This shows something more modern. There are further ways to improve the concept which I cannot discuss.
Ciao T
Ok, but based on your earlier arguments for NOS, the dynamic range of 16/44 recordings does not require even 18 bits. So isn't the PCM1704 pretty darn good?
Sorry, I don't know an FPGA designer with that much free time or skill. Sounds pretty cool though. You would think there would be at least one audiophile out there with the necessary skill set.
Last edited:
ok, i've got a lot of good information. I am gonna need to go off and think about all of this some as I am starting to get confused.
I think I am clear on the analog stages and power supplies, but DAC chip selection is still giving me trouble.
If I do not use the PCM1704 or the TDA part. What is left? PCM1794 or similar should all be about the same?
Perhaps I should just use PCM1794 with ASRC (allowing it to be bypassed)
and take great care of the analog stages and power supplies. Then this will be good enough to get a good result.
Anyway, I almost have a prototype of this already. maybe I will finish it up and see what I think of it.
I think I am clear on the analog stages and power supplies, but DAC chip selection is still giving me trouble.
If I do not use the PCM1704 or the TDA part. What is left? PCM1794 or similar should all be about the same?
Perhaps I should just use PCM1794 with ASRC (allowing it to be bypassed)
and take great care of the analog stages and power supplies. Then this will be good enough to get a good result.
Anyway, I almost have a prototype of this already. maybe I will finish it up and see what I think of it.
The pcm1704 (K!!!) with 112dnr is pretty darn good for CDQ playback already, if you parallel 2 or 4 you shouldnt do it because of better numbers (spec game) , rather the +-1mApp provided by 1 chip is quite low, you see the pcm1702 evkit has 4 in parallel ... The 120dB I mentioned is not a requirment , it just wouldnt hurt. But mastering engineers would be insane to exploit this, even if they could, I dont think they would . You can be sure at least 19.5 bits can be obtained by careful mastering ( 19.5 bits, where it counts) .
Last edited:
OK, so from what I've heard so far it seems like the PCM1704 would be a very respectable chip to use. I will look into this a bit more now.
If I use the PCM1704. Is the DF1704 the only DF that can be used to get left and right data out? Would SM5842 also work? I haven't had a chance to look at the datasheet yet.
Hi,
For CD-Replay I would consider PDM100 or PDM200, from Pacific Microsonic, long discontinued, hard to get but can be occasionally found on e-bay. Apart from that, subjectively the best filters you can get for CD if you do not want to use Non-OS.
NPC also makes okay sounding filters, I'd avoid the BB Filters. Even BB's own engineer avoided them in their eval boards, take a hint... 🙂
Using a ASRC as "digital filter" as suggested by others may also be an option, it can also deal with format conversion and all.
Ciao T
For CD-Replay I would consider PDM100 or PDM200, from Pacific Microsonic, long discontinued, hard to get but can be occasionally found on e-bay. Apart from that, subjectively the best filters you can get for CD if you do not want to use Non-OS.
NPC also makes okay sounding filters, I'd avoid the BB Filters. Even BB's own engineer avoided them in their eval boards, take a hint... 🙂
Using a ASRC as "digital filter" as suggested by others may also be an option, it can also deal with format conversion and all.
Ciao T
Well,
there is a rumour from some PhilipZ engineers (not Eindhoven-headquarter): They want to publish the first CD-Player with 14-bit DAC TDA1540 without OS. Same with TDA 1541. The marketing-crew insist of applying OS...go figure.
Rumour no. 2: the current production smd 1704 isn´t that good.
Try FET Audio Store "PMD100 NOS: $35 each – ask to check stock"
Yamaha YSF210B is another sleeper....
A TDA1541 based player with broken laser cost less than a sixpack.....
LG Carsten
PS: I have a Mission PCM II here. More or less a FilipZ CD650 with some aditional boards between Saa7220 and DAC containing logic-chips. Without drawing a schematic ... does this player include the DEM-thingie? If someone has the schematic, please tell me about it.
Ok, so I looked into the secondary PLL technique a bit more and I am confused.
If you look at the phase noise performance of a typical VCO like the VC7225ALZ from RALTRON:
-70@10Hz
-95@100Hz
-120@1kHz
-140@10kHz
-150@100kHz
Integrating the SSB phase noise density over the audio bandwidth seems to yield jitter in the 100's of pico-seconds RMS. So even if the PLL bandwidth is kept really low (a few Hz) the jitter of a VCO is much worse than a quality crystal.
So it seems to me the only way to get really low (<50psRMS) DAC sampling clock jitter is through the ASRC technique.
What am I missing here? I have read that the close in phase noise is less audible, is this why the high phase noise at low frequency offsets is not a problem?
So I guess it is still a question of the lesser evil? You can get lower jitter with the ASRC, but it has other problems.
Does anyone know of a higher performance VCO for use in a secondary PLL?
If you look at the phase noise performance of a typical VCO like the VC7225ALZ from RALTRON:
-70@10Hz
-95@100Hz
-120@1kHz
-140@10kHz
-150@100kHz
Integrating the SSB phase noise density over the audio bandwidth seems to yield jitter in the 100's of pico-seconds RMS. So even if the PLL bandwidth is kept really low (a few Hz) the jitter of a VCO is much worse than a quality crystal.
So it seems to me the only way to get really low (<50psRMS) DAC sampling clock jitter is through the ASRC technique.
What am I missing here? I have read that the close in phase noise is less audible, is this why the high phase noise at low frequency offsets is not a problem?
So I guess it is still a question of the lesser evil? You can get lower jitter with the ASRC, but it has other problems.
Does anyone know of a higher performance VCO for use in a secondary PLL?
I take that back. I just re-calculated and it seems I was wrong. 🙁
If I am doing it right this time, it looks like the total integrated jitter for the above SSB spec. ends up around 8ps RMS. I would say this is very good.
Anyone measure a secondary PLL circuit? Do you get jitter this low in practice?
If I am doing it right this time, it looks like the total integrated jitter for the above SSB spec. ends up around 8ps RMS. I would say this is very good.
Anyone measure a secondary PLL circuit? Do you get jitter this low in practice?
BTW, why is the jitter across the SPDIF link so bad to begin with? Seems like reclocking the data stream in the transport right before sending out the cable would clean it up nicely and you wouldn't need to worry about it much in the receiver.
Most DIY set-ups would use a high quality short cable so that jitter degradation through the cable would be low.
Most DIY set-ups would use a high quality short cable so that jitter degradation through the cable would be low.
- Status
- Not open for further replies.