R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 [2]  All   Go Down

Author Topic: AD/DA converters question  (Read 14484 times)

Jim Williams

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1105
Re: AD/DA converters question
« Reply #15 on: December 13, 2006, 10:52:51 AM »

I suppose you all could line up 32 Vishay medical grade bulk foil resistors at .005% tolerance and very tight temp grades. Then you could realize they cost over $10 each.

Or, you can get clever and smart. Most of these chip makers offer demo pcb's to show off the product. Get a Crystal CDB5381 from digi key for around $400. Add a 5 and bipolar 12 volt supply and go. For the tweakers, pop off the 5532 opamps for something better.

For DAC's, they offer the CDB4398 demo pcb with the same opamps and powering requirements around the same price range. This is the best 1 bit chip out at the moment.

For the anal retentive like myself, only the Burr Brown PCM/DSD1792 will do for playback. This chip set offers -129 db THD+noise and with the co-sine operation it reaches -132db, within 8 db of the theoretical, magical 24  bit dynamic range.
BB offers this demo pcb for around $900, I got mine from the engineering dept. as a comp. Add some stellar opamps and a great psu and I know of no DAC with this dynamic range and very low THD. Due to parasitics, I don't know if a discrete design could achieve these results. Certainly not for this price range.
Logged
Jim Williams
Audio Upgrades

zmix

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2828
Re: AD/DA converters question
« Reply #16 on: December 13, 2006, 03:52:28 PM »

I had an interesting conversation with the designer of a piece of equipment that was 're-issued' 20+ years after the original. The device was a sort of delay based effect and in the re-issue the feedback path was done internally in the DSP, but in the original the feedback path was all analog. I asked the designer about the change, and he stated that the only converters available today were sigma delta and that the original used a successive approximation converter. He said that the newer converters didn't maintain nearly the fidelity of the successive approximation converter, so they decided to go DSP.

It's interesting to think about how audible the anomalies of a converter become after so many regeneration cycles.

Back in the 1980s there was an Opamp listening test which specified 10 devices in series.

Anyone ever listen to 10 converters in series?

Consul

  • Full Member
  • ***
  • Offline Offline
  • Posts: 226
Re: AD/DA converters question
« Reply #17 on: December 13, 2006, 06:06:04 PM »

Bruno Putzeys wrote on Wed, 13 December 2006 04:51


Measure them. Open them up and look at the circuit boards. Check the schematics.


I honestly don't think I'll be able to get schematics for any of the devices in question. That means reverse-engineering. I don't think I have the skill for that. In fact, I know I don't.

Quote:

At any rate, you'll need to do quite a bit of analysis and tests. Do not trust a manufacturer's verbal claim unless they back it up with hard data. Claiming to pay attention to "low jitter" and "circuit board layout" is de rigueur these days. Nobody will market a product without saying these two things. But do they walk the walk? Only the actual hardware can tell you that.


Well, I do thank you for your reply. It's confirmed what I've suspected the whole time: unless someone steps up to the plate, spends the money, does the reversing, and reports the findings on a variety of budget devices, selecting a set of AD/DA converters will remain a shot in the dark unless one can spend at least $500 per in and out. I would really love to be the one to gain all the fame and adoration by stepping up to said plate myself, but finances simply will not allow that, not to mention my own skill with electronics, which is existent but not quite expert.

I guess I should point out that I'm not interested in DSD right now. Regular old PCM is where I'm stuck at the moment.

I really hope I don't sound like I'm whining, even though, really, I am. Wink

A friend of mine suggested making a discrete D/A by trimpotting the output resistors, then wrapping the whole thing in an oven. This would add a lot of current noise, though, right? What would happen if one ran the trimpots in parallel?
Logged
Darren Landrum

"Never be afraid to try something new. Remember that a lone amateur built the Ark. A large group of professionals built the Titanic." - Dave Barry

bruno putzeys

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1078
Re: AD/DA converters question
« Reply #18 on: December 14, 2006, 04:47:39 AM »

Consul wrote on Thu, 14 December 2006 00:06

I guess I should point out that I'm not interested in DSD right now. Regular old PCM is where I'm stuck at the moment.

I'm not a proponent of using 1-bit signals in recording and processing (which is what the trade name DSD really means). I am, however, very pleased with the results I get using 1-bit conversion in discrete AD/DA converters. After the AD you'd strap a decimation chain to get PCM and before the DA an upsampler and a digital sigmadelta modulator produce a 1-bit signal. This is how most AD/DA chips work these days, except that they'll usually use two to five bits worth of conversion.
So PCM work does not condemn you to using a ladder type converter. The data format and conversion method are two separate issues. For either type you need digital filtering (unless you want to build analogue brick-wall filters) so then the number of bits on the converter end becomes a separate decision.
Consul wrote on Thu, 14 December 2006 00:06

A friend of mine suggested making a discrete D/A by trimpotting the output resistors, then wrapping the whole thing in an oven. This would add a lot of current noise, though, right? What would happen if one ran the trimpots in parallel?

One would certainly place them in parallel. If the resistors are 0.1% devices and the trimpots are, say 1000 times the value of the resistor, the pots will not introduce noise.

Again I'd stress that one should choose the conversion topology and data format on their own strengths. Which means low-bit on the converter side and PCM on the processing side.
Logged
Warp Drive. Tractor Beam. Room Correction. Whatever.

Affiliations: Hypex, Grimm Audio.

Consul

  • Full Member
  • ***
  • Offline Offline
  • Posts: 226
Re: AD/DA converters question
« Reply #19 on: December 14, 2006, 05:07:41 PM »

Thank you for the info, Bruno! I'm definitely directing my friend to this thread now. Very Happy Maybe we'll even come up with a design...
Logged
Darren Landrum

"Never be afraid to try something new. Remember that a lone amateur built the Ark. A large group of professionals built the Titanic." - Dave Barry

Consul

  • Full Member
  • ***
  • Offline Offline
  • Posts: 226
Re: AD/DA converters question
« Reply #20 on: December 16, 2006, 05:34:36 PM »

Would something like this work for the digital filtering?

http://www.wavefrontsemi.com/index.php?id=11,13,0,0,1,0

It's an inexpensive DSP chip good up to 50khz clock speeds, 50 MIPs. I just wish they made a DIP version.

EDIT: You don't really have to answer if you don't want to. I'm not out to be a competitor or anything. Wink I just love to learn about as much as I possibly can. Trying to make my own D/A would be a genuine challenge, though I don't know if it's one I want to take on just yet.
Logged
Darren Landrum

"Never be afraid to try something new. Remember that a lone amateur built the Ark. A large group of professionals built the Titanic." - Dave Barry

bruno putzeys

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1078
Re: AD/DA converters question
« Reply #21 on: December 17, 2006, 09:42:42 AM »

The processing capacity allows for some fairly decent filtering but the I/O only runs at the word clock rate which in turn is limited to 50kHz. Quite annoying if one wants to make an upsampling filter.

SMD soldering is easier than you think. Get a decently fine soldering tip and more importantly, get fine solder. Once you're used to it you'll never want to go back to thru-hole.
Logged
Warp Drive. Tractor Beam. Room Correction. Whatever.

Affiliations: Hypex, Grimm Audio.

Consul

  • Full Member
  • ***
  • Offline Offline
  • Posts: 226
Re: AD/DA converters question
« Reply #22 on: December 17, 2006, 07:51:01 PM »

Thank you for the reply. I guess my question now is, why would one want to build an up-sampling filter?

I must sound like a real newbie, huh? Unfortunately, this kind of information seems incredibly difficult to search the Internet for. I'm a mechanical engineering student just starting out, so I have no professors I can bother with my silly questions, either.
Logged
Darren Landrum

"Never be afraid to try something new. Remember that a lone amateur built the Ark. A large group of professionals built the Titanic." - Dave Barry

bruno putzeys

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1078
Re: AD/DA converters question
« Reply #23 on: December 18, 2006, 03:12:49 AM »

Newbiness is permitted Smile

You'd build your own upsampling filter if you're unhappy with the performance of existing ones. There aren't any commercially available upsamplers using windowed filters, all are halfband based and none have strong bandpass dither for linearising multibit DACs.
Logged
Warp Drive. Tractor Beam. Room Correction. Whatever.

Affiliations: Hypex, Grimm Audio.

Consul

  • Full Member
  • ***
  • Offline Offline
  • Posts: 226
Re: AD/DA converters question
« Reply #24 on: December 18, 2006, 10:49:03 PM »

Okay, now you're just showing off.  Razz Wink

Time to hit Google again...
Logged
Darren Landrum

"Never be afraid to try something new. Remember that a lone amateur built the Ark. A large group of professionals built the Titanic." - Dave Barry

bruno putzeys

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1078
Re: AD/DA converters question
« Reply #25 on: December 19, 2006, 03:23:55 AM »

I'm referring to the fact that for economical reasons upsampling and decimation filters usually don't fulfil the Nyquist criterium, but are -6dB at fs/2 in order to make all even coefficients trivial (halfband).
They also don't ring out smoothly at the ends but stop in a small "spike" that's the result of numerically optimising the filter to fit a certain spec with the least number of coefficients (equiripple). I wouldn't consider this a problem in  filters with very low in-band ripple (in halfband filters this corresponds to a large stop-band rejection), because then the tails are so deep down that any irregularity there can't possibly be audible.
Further, ladder DACs benefit from quite heavy dither (at -20dB or so) to smooth out the low level nonlinearities. Of course you'd put this dither at the end of the spectrum after upsampling, ie in the 100kHz region.
Logged
Warp Drive. Tractor Beam. Room Correction. Whatever.

Affiliations: Hypex, Grimm Audio.

Consul

  • Full Member
  • ***
  • Offline Offline
  • Posts: 226
Re: AD/DA converters question
« Reply #26 on: December 19, 2006, 09:51:47 AM »

Sorry, I was only making a joke about my own inadequate knowledge. I hope I didn't come across harshly.

I have more reading to do, I think.
Logged
Darren Landrum

"Never be afraid to try something new. Remember that a lone amateur built the Ark. A large group of professionals built the Titanic." - Dave Barry

gentlevoice

  • Newbie
  • *
  • Offline Offline
  • Posts: 7
Re: AD/DA converters question
« Reply #27 on: April 11, 2008, 09:26:00 AM »

Hi PSW forum,

I hope to connect with Bruno Putzeys through this forum - or somebody else who knows about DSD A/D conversion technology.

I'm considering building (as part of a DIY - collaboration) a DSD A/D converter operating at 256*fs. As I believe sample rates will go up in the future, this is to make it reasonably future proof.

I am aiming for a design with a very high sound quality.

Briefly investigating the effects of noise-shaping and higher order modulators, I find a second order modulator interesting. The reason for this is to obtain "benign" noise-shaping effects. I realize there are some trade-offs to this - like e.g. a lower theoretical SNR.

Before venturing into this project I am interested in feedback from people with experience in this field: Could there be problems with this approach?

Should you reply to this post, please consider me a layperson, although one with a very good generalist technical understanding.

Thank you  Smile

Jesper (Denmark)






Logged

bruno putzeys

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1078
Re: AD/DA converters question
« Reply #28 on: April 14, 2008, 03:51:07 AM »

gentlevoice wrote on Fri, 11 April 2008 15:26

Could there be problems with this approach?

Yes. The second is choosing your sample rates higher than necessary. The greatest enemy of sigmadelta converters is intersymbol intermodulation. In a discrete implementation that means you should make temporally separated pulses (return to zero) with enough time for DAC to settle fully back to zero. Settling time issues translate as "intersymbol intermodulation" (i.e. where the area under two subsequent ones is not exactly equal to the area under two ones with a zero in-between) and this in turn causes the dominant tone to fold back into the audio band. What you hear is a frequency-modulated whistle that becomes very audible at low signal amplitudes. If you have good experience with RF layouts, you can get to 128fs (approx 6MHz) with good results. 256fs is highly doubtful. A discrete 256fs converter exists but it has not exactly met with universal acclaim.

The first is making design choices before you've gotten your target specs lined up. Before you make any design choices, scrutinize your wish list from a "black box" perspective. Suppose you have a pair of converters connected back-to-back and you're listening through them. There's no way you will say "oh this is a 1-bit at 128fs with a 4th order loop" or "this is a 24-bit ladder" or "this is a 5-bit at 32fs with a 6th order loop" etc. So these items should be no part of a requirement spec and can't possibly be a choice to be made up-front.

So, try to separate performance requirement specs from design choices. The performance spec comes first and is based on the user experience (sound, operatability, anything you can determine without taking the lid off). Then don your engineering hat and see how you will achieve, technically those requirement specs. The design choices you make are then made strictly on the specs. If the user intervenes and tells you he won't buy the box if it doesn't use his favourite opamp, circuit topology or numeric representation, shut the user up.

As soon as a project has more input variables than it has outputs, specs may directly conflict. In your case, the extra input variable is fun factor. It's a DIY project. Perhaps all your technical (and hence sonic) needs are fulfilled straight away by buying an off-the-shelf chip (or even a complete box), but it's not nearly as much fun as knocking something together that's only half as good. In this case choosing your priorities well will save you a lot of disappointment. Nevertheless, at least heed the second item Smile
Logged
Warp Drive. Tractor Beam. Room Correction. Whatever.

Affiliations: Hypex, Grimm Audio.

gentlevoice

  • Newbie
  • *
  • Offline Offline
  • Posts: 7
Re: AD/DA converters question
« Reply #29 on: April 14, 2008, 02:00:48 PM »

 Laughing  Laughing   Thanks Bruno for the explanation and design "input" - you made me laugh with the link to your 2 "design paths". I would say that my approach really is both - because my design goals are often aimed at perfection (No. 2), when I realize it can't and/or is difficult to do, I revert to the first one and then juggle between them ...  Confused  

I have also found that my ways of designing often transgress "normal" engineering approaches which makes specifying goals in words or specs somewhat elusive ... E.g. I have mounted some of my amplifier designs on balsawood, thereby eliminating the PCB, something that, as I perceive it, caused the design to sound more open, liquid, and less "noisy" (smile).

However, my reason for researching this as a collaborative DIY- project is that I have not found a commercially available solution that satisfies my wishes.

I know that grimmaudio has the AD1 available which appears interesting and seems to be a very special design (although too expensive for my budget and not entirely convenient for me - I'm looking for something portable). Also, now the KORG MR-1000 with a 5,6 MHz DSD sampling rate is in the market - although maybe in its base design it is not in the same league as the AD1.

There are other designs as well but I've never heard of a 256*fs design. Can you share who/which company makes this?

BUT - even more important - my intuition keeps telling me that the high-frequency output power of delta-sigma modulator designs (mostly at 64*fs) will have an (negative) effect on me, personally (BTW there is a study indicating that although a person may not actually hear a sound/tone e.g. above 20 kHz, brain wave measurements indicate that the brain waves change when the tone is present). I reckon this noise will also influence electronics and speakers.

I consider my intuition to be correct in this respect which is why I'm looking for ways of overcoming this - e.g. by designing a delta-sigma at 256*fs, or, if possible, at an even higher fs, and as a low-order design. I know that very high frequency delta-sigma modulators have been made, yet my technical know-how does not allow me to discern if they are applicable to high-end audio.

An alternative is a very high sample frequency 20-24 bit PCM converter, e.g. ~ 400 kHz fs (I have my reasons for this, should be possible to disable any digital filters), but I haven't yet found such a chip/converter suitable for audio. Maybe you know of such a converter/design?

I have had contact with a Danish audio manufacturer (considered high-end) and he told me that for classical music the SACD format is positively preferred. Others say something different, though - my personal experience is that PCM and DSD each seem to have sonic imprints that may be somewhat unique to them (I realize implementation has a lot to say - and not being humble, no, I may not be able to discern close designs from eachother ... ).

In the marketplace, what I notice is that SACD title issues now have reached >5000 and DVD-A is declining - as far as I know. So maybe SACD is what will happen in the future. Also with high capacity discs available, such as blu-ray, I consider it likely that recording rates will go up in the near future.

Well, yes, that's where I am with this. I'd appreciate if you/others have comments ...

All the best to your evening, Bruno! Here in Denmark, right now the sun is setting - slowly painting the sky orange/reddish - and the birds are singing. It's a good experience   Rolling Eyes

Regards,

Jesper
Logged
Pages: 1 [2]  All   Go Up
 

Site Hosted By Ashdown Technologies, Inc.

Page created in 0.079 seconds with 18 queries.