I haven't seen much discussion of NOS ADC's on diyAudio or really anywhere other than Altmann's site, so I'm curious if any forum members have built their own or have interest in doing this?
I understand the maximum practical NOS R-2R bit depth is around 18, although I've only been able to find 16 bits from Analog Devices (I believe they have/had an 18 bit r2r?).
Some chip offerings are capacitive ladders while others are resistive, should I care? Does one settle faster than the other?
Any recommendations for the simplest part count to get S/PDIF from the A/D and into toslink optical?
My ultimate goal being to build a very low part count NOS A/D-D/A with toslink for use with a MacBook... At this point I'm thinking outboard PSU or batteries makes the most sense as it is a portable.
Music studio's don't have access to this type of technology as there is no commercial NOS converter offering and while DSD sounds good, nearly all post production utilizes 16-24bit PCM (EQ/Filter/Dynamics/Etc). It's very likely most SACD's have been through a second DSD conversion after PCM edits occurred. Why does this matter? Well, you can hear it.
My conclusion is that sampling 192 Khz at true 16-19bits will give the best digital capture, post processing won't need decimation and finally something like Weiss Saracon could be used to convert to DSD.
Any help and discussion relating to simple NOS ADC and DAC designs will be both enjoyed and appreciated.
Btw, this inquiry is purely for building my own recording system and there are no commercial motives (outside of making good recordings).
Thanks,
Anthony Bisset
http://www.dspaudio.com
I understand the maximum practical NOS R-2R bit depth is around 18, although I've only been able to find 16 bits from Analog Devices (I believe they have/had an 18 bit r2r?).
Some chip offerings are capacitive ladders while others are resistive, should I care? Does one settle faster than the other?
Any recommendations for the simplest part count to get S/PDIF from the A/D and into toslink optical?
My ultimate goal being to build a very low part count NOS A/D-D/A with toslink for use with a MacBook... At this point I'm thinking outboard PSU or batteries makes the most sense as it is a portable.
Music studio's don't have access to this type of technology as there is no commercial NOS converter offering and while DSD sounds good, nearly all post production utilizes 16-24bit PCM (EQ/Filter/Dynamics/Etc). It's very likely most SACD's have been through a second DSD conversion after PCM edits occurred. Why does this matter? Well, you can hear it.
My conclusion is that sampling 192 Khz at true 16-19bits will give the best digital capture, post processing won't need decimation and finally something like Weiss Saracon could be used to convert to DSD.
Any help and discussion relating to simple NOS ADC and DAC designs will be both enjoyed and appreciated.
Btw, this inquiry is purely for building my own recording system and there are no commercial motives (outside of making good recordings).
Thanks,
Anthony Bisset
http://www.dspaudio.com
Unless you know of a non sigma delta audio dac, then any device you are likely to encounter may well need some form of format conversion.
anthonybisset said:
I think your two goals are mutually exclusive. For the lowest parts count, get a CODEC that has ADC, DAC, S/PDIF, and USB all in one chip. There are many to choose from; but that gives you delta-sigma coming and going. For best sound with 18-bits and at least 98 KHz you should make your own R2R ADC & DAC with discrete components.
Post #2 -
There are many NOS DAC's designs around and some have very low part counts like the DDDAC and similar TDA1543 layouts. For the first try (and many will cringe, I'm gonna clock both the DAC and ADC controllers from the incoming SPDIF signal. If the MacBook's SPDIF is too jittery then I'll go with an inline reclocker of some sort until I've developed brain circuits which can understand the nuances of PLL reclocking.
Regarding Post #3 -
Low part count, not lowest 🙂
With Napolean's hat tipped to the right (http://www.dhm.de/gifs/sammlungen/militaria/eh181.jpg)
I continue:
So far I'm thinking version 1 NOS_ADDA:
D/A:
- CS8412
- TORX79
- TDA1543 tower (no gain stage necessary, I like this simplicity for V1 anyway)
A/D
- CS8404a-CS
- TOTX79
- AD7690 (x2)
Regarding the AD7690 -
(http://www.analog.com/en/prod/0,,AD7690,00.html)
(http://www.analog.com/en/prodDesc/0,2895,AD7690_0,00.html)
Pinout shows: SDI, SCK, SDO, CNV || will these talk to the CS8404a directly?
And more importantly, how can we use two A/D's for stereo.. Do we need to MUX the two chips data lines. I'm lost on this angle. Maybe I'll build the first version as mono.. Most of my recordings focus on frequency and dynamics rather than stereo separation anyway.
References:
http://en.wikipedia.org/wiki/Successive_Approximation_ADC
Further questions:
If we feed an 18 bit signal to a CS84xx series chip, does it pad the last 6 bits with 0's, or must we again do some formatting to make sure the computer sees a 24bit word?
Anyone with an Altmann creation ADC know what chips are being used? I'd just buy one outright but don't have the money and was bothered it's only 16 bit's.
-anthony
There are many NOS DAC's designs around and some have very low part counts like the DDDAC and similar TDA1543 layouts. For the first try (and many will cringe, I'm gonna clock both the DAC and ADC controllers from the incoming SPDIF signal. If the MacBook's SPDIF is too jittery then I'll go with an inline reclocker of some sort until I've developed brain circuits which can understand the nuances of PLL reclocking.
Regarding Post #3 -
Low part count, not lowest 🙂
With Napolean's hat tipped to the right (http://www.dhm.de/gifs/sammlungen/militaria/eh181.jpg)
I continue:
So far I'm thinking version 1 NOS_ADDA:
D/A:
- CS8412
- TORX79
- TDA1543 tower (no gain stage necessary, I like this simplicity for V1 anyway)
A/D
- CS8404a-CS
- TOTX79
- AD7690 (x2)
Regarding the AD7690 -
(http://www.analog.com/en/prod/0,,AD7690,00.html)
(http://www.analog.com/en/prodDesc/0,2895,AD7690_0,00.html)
Pinout shows: SDI, SCK, SDO, CNV || will these talk to the CS8404a directly?
And more importantly, how can we use two A/D's for stereo.. Do we need to MUX the two chips data lines. I'm lost on this angle. Maybe I'll build the first version as mono.. Most of my recordings focus on frequency and dynamics rather than stereo separation anyway.
References:
http://en.wikipedia.org/wiki/Successive_Approximation_ADC
Further questions:
If we feed an 18 bit signal to a CS84xx series chip, does it pad the last 6 bits with 0's, or must we again do some formatting to make sure the computer sees a 24bit word?
Anyone with an Altmann creation ADC know what chips are being used? I'd just buy one outright but don't have the money and was bothered it's only 16 bit's.
-anthony
anthonybisset said:
Didn’t you say you wanted 18-bit R2R performance? The TDA1543 performs like a 14-bit DAC: Read the datasheet. There are plenty of 18-, 20-, & 24-bit R2R DACs that will outperform a tower of 1543s with a lower total parts count and much less power consumption.
Hello
I plan to built a non-os dac this year.
I did suggest to a friend to listen some non-os dacs, he just send me an email and say that;
''I have heard few non-oversampling DACs, yes. Most of them lose quality and resolution when the music gets really busy.''
Anyone know about any resolutions loss of that kind with non-os dacs ?
Thank
Gaetan
I plan to built a non-os dac this year.
I did suggest to a friend to listen some non-os dacs, he just send me an email and say that;
''I have heard few non-oversampling DACs, yes. Most of them lose quality and resolution when the music gets really busy.''
Anyone know about any resolutions loss of that kind with non-os dacs ?
Thank
Gaetan
Whatsup
Could you point out some non-sigma delta DACs that feature 20+ bits?
I've looked through all the Analog Devices DAC datasheets and the closest I can find is the
dual channel AD1865 with 18 bits precision and decent specs and it's being phased out, the recommended replacement is a sigma-delta design.
anthony
Could you point out some non-sigma delta DACs that feature 20+ bits?
I've looked through all the Analog Devices DAC datasheets and the closest I can find is the
dual channel AD1865 with 18 bits precision and decent specs and it's being phased out, the recommended replacement is a sigma-delta design.
anthony
anthonybisset said:
When did you start worrying about using discontinued parts? Didn't you know the CS8412 and TDA1543 were phased out years ago?
rfbrw,
Thanks for the list.. 1704 looks the best of the lot.
Now, if only there was a true 24bit ADC chip. I've been reading up on sign-magnitude conversion techniques but can't find a A/D employing it. Although really slow ones are probably inside digital voltmeters.
-anthony
Thanks for the list.. 1704 looks the best of the lot.
Now, if only there was a true 24bit ADC chip. I've been reading up on sign-magnitude conversion techniques but can't find a A/D employing it. Although really slow ones are probably inside digital voltmeters.
-anthony
anthonybisset said:
First, have you even attempted to read and understand the datasheets? Have you attempted to locate a source for the parts you are specifying in your design?
Second, the AD7690 you want to use is a rather new part. The datasheet is dated 2006 and audio is not one of the suggested applications for the part. Why then, would you assume it is appropriate for your audio application and why do you think it would talk directly to an obsolete audio DIT, such as the CS8404?
As I suggested, if you want an 18-bit R/2R ADC you are going to have to roll your own or use obsolete parts, if you can find them. The universe of high quality, audio ADCs is very small and there are few parts floating around. Maybe you could salvage something from a used DAT recorder.
What's up,
Thanks for your feedback.
I've been reading datasheets, schematics and forums maybe 3-6 hours each day for the past 2 weeks. I have very little previous experience with digital circuitry and so I'm doing my due diligence by asking questions here and elsewhere along with significant amounts of reading. I write audio software and modify/build circuits for analog synthesizers so I'm not a fish out of water, merely a tropical fish in the artic ocean (someday this may be the norm).
I need to reframe the discussion a bit so there's no confusion. I want to build a simple combined ADC/DAC with the focus being on the ADC's accuracy. The DAC's bit depth is not as important as the ADC although matching them makes for an easier circuit. Secondly which chips I use are not of concern as long as they are not sigma-delta and are along the lines of Sign-Magnitude, SAR or R2R. However, if you told me I should use Sigma Delta, I would be open to hearing why you think so and I wouldn't resort to quoting some math I don't understand. I have a hunch there is something to NOS recording and playback at greater than 44.1 khz.
It seems to me that NOS designs at 44.1 khz heavily involve the speaker and amplifier in approximating/filtering the digital steps. If we double or quadruple the sampling frequency then the speaker and amp will have less time to approximate information and can be more accurate to the waveform we are reproducing without wasting as much energy, although sometimes wasting energy sounds good.
If you have advice on which chips I should consider I'm listening closely.
cheers,
-anthony
Thanks for your feedback.
I've been reading datasheets, schematics and forums maybe 3-6 hours each day for the past 2 weeks. I have very little previous experience with digital circuitry and so I'm doing my due diligence by asking questions here and elsewhere along with significant amounts of reading. I write audio software and modify/build circuits for analog synthesizers so I'm not a fish out of water, merely a tropical fish in the artic ocean (someday this may be the norm).
I need to reframe the discussion a bit so there's no confusion. I want to build a simple combined ADC/DAC with the focus being on the ADC's accuracy. The DAC's bit depth is not as important as the ADC although matching them makes for an easier circuit. Secondly which chips I use are not of concern as long as they are not sigma-delta and are along the lines of Sign-Magnitude, SAR or R2R. However, if you told me I should use Sigma Delta, I would be open to hearing why you think so and I wouldn't resort to quoting some math I don't understand. I have a hunch there is something to NOS recording and playback at greater than 44.1 khz.
It seems to me that NOS designs at 44.1 khz heavily involve the speaker and amplifier in approximating/filtering the digital steps. If we double or quadruple the sampling frequency then the speaker and amp will have less time to approximate information and can be more accurate to the waveform we are reproducing without wasting as much energy, although sometimes wasting energy sounds good.
If you have advice on which chips I should consider I'm listening closely.
cheers,
-anthony
Cirrus has two chips that are SPDIF receivers and transmitters: CS8420 and CS8427. CS8420 is also an ASRC, so you may not want that. However, CS8427 is limited to 96 kHz, which may or may not meet your needs.
- Status
- Not open for further replies.