ADC and DAC Noob Questions

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

buildafriend

Well-known member
Joined
Jun 30, 2009
Messages
1,392
Location
Omnipresent NYC
Hello,

I'll start by saying my understanding of A/D and D/A is just that of the operator. My technical understanding is more like that of a tech than an engineer. I can read schematics and I can interpret a decent amount of datasheet info but I understand much more about the analog realm of things than I do about the digital realm. That being said, I would like to start experimenting a bit with ADC and DAC chips. There are so many! There are entire threads about which ones certain companies use.. it's a lot to dig through. I could just look through Mouser for good specs but I would feel more confident buying some chips ( that I will probably end up breaking.. lol ) that are recommended by a fellow tinkerer or maybe by someone who has a lot of experience with them.

Do you guys have any advice for getting started?

Respectfully,
-JP
 
AD and DA performance for audio is determined as much if not more by clock integrity, power supply, PCB layout and the analog circuitry than the actual converters themselves.

I've been slowly schooling myself on the topic and I find Eval boards and their documentation very helpful.  I have boards from TI, Arda and soon ESS.  The board layouts ware well done, simple and easily understood.  Cheap and fun to play with.

It's interesting how many commercial converters don't realize the performance of the chips used.

 
ruairioflaherty said:
AD and DA performance for audio is determined as much if not more by clock integrity, power supply, PCB layout and the analog circuitry than the actual converters themselves.

I've been slowly schooling myself on the topic and I find Eval boards and their documentation very helpful.  I have boards from TI, Arda and soon ESS.  The board layouts ware well done, simple and easily understood.  Cheap and fun to play with.

It's interesting how many commercial converters don't realize the performance of the chips used.

The chip selections do seem to vary greatly..

Maybe some of you are willing to share some literature worth reading that relates to the subject?

Which Eval board/s do you find useful?
 
buildafriend said:
ruairioflaherty said:
AD and DA performance for audio is determined as much if not more by clock integrity, power supply, PCB layout and the analog circuitry than the actual converters themselves.

I've been slowly schooling myself on the topic and I find Eval boards and their documentation very helpful.  I have boards from TI, Arda and soon ESS.  The board layouts ware well done, simple and easily understood.  Cheap and fun to play with.

It's interesting how many commercial converters don't realize the performance of the chips used.

The chip selections do seem to vary greatly..

Maybe some of you are willing to share some literature worth reading that relates to the subject?
+1 the PCB layout can make a signifiant difference. Study the manufacturers application notes and preferred layouts.

Some trial and error may be involved that can be difficult with multi-layer PCBs.

JR
 
+1 the PCB layout can make a signifiant difference. Study the manufacturers application notes and preferred layouts.

Some trial and error may be involved that can be difficult with multi-layer PCBs.


I remember you guys mentioning that the layout style becomes more of a consideration at higher frequencies when I was working on some regulator and tube PCB's.

I started studying the CS5381 and it seems like a pretty high quality chip.

_____________________________________________________________________________________

Maybe it would make more sense if I built 1 or 2 ADC projects first? I found 2 that seem interesting. This way I can start to get an idea of what the layouts look like and I can gain a general understanding of the parts and connectivity.

http://www.beis.de/Elektronik/ADDA24QS/AD24QS.html

http://www.diyaudio.com/forums/group-buys/176416-diy-cs5381-analog-digital-converter-pcb-group-buy.html

I have to learn a lot of new definitions. I don't understand the need for so many clock connections. I also don't understand what's going on in the power down and mode settings. This is gonna be a mission!  :)

__________________________________________________________________________________________________________________________

Maybe this thing would be useful?
http://www.ti.com/lit/ug/sbau128c/sbau128c.pdf

Maybe something like this is what was being referred to by ruairioflaherty?
https://www.intersil.com/en/tools/reference-designs/hs-adc-eval-kit.html

So much info!  :eek:
 

Attachments

  • chip overview.JPG
    chip overview.JPG
    77.1 KB · Views: 33
I've been following this thread and am appreciative of your questions as I have wondered the same. I have been very interested in building an ADC and DAC for awhile but was just overwhelmed by the amount of info.  I'm also interested in that beis ADC kit and would like to see more reviews/comparisons. Best of luck.
 
I've recently started back into this and am building Raphaels (smd) AD/AES boards...but I also pulled out my old TI PCM 4222EV board and began to play with it...a lot of the data makes no sense until you actually try to plug it in and make it work...I have the older AKM stuff but found the user interface via serial port a bit cumbersome...with my TI eval board I can dump into my computer via spdif or even AES to see what is happening as well as take a signal into a separate DAC board I picked up (a really cheap DAC headphone board from china that actually has burrbrown stuff in it, sounds amazing)...

Anyway the TI stuff is really well documented but like I said it helps to be able to tweak around with it...
 
buildafriend said:
I have to learn a lot of new definitions. I don't understand the need for so many clock connections.

I suppose that there is a nomenclature issue surrounding the many clock connections. Let's parse it.

All of the converters have a data interface. For audio, one standard is the Philips "Inter-Inter-Sound" interface, or I2S. It consists of three or four signals.  (There are others, like TDM, which is often used for chips with more than two converters. There are older parallel interfaces, too.) Pretty much all converter data sheets will have a timing diagram detailing the I2S port operation.

I2S is a synchronous serial interface, similar to SPI. There are either one or two data lines, depending on whether you are talking to only an ADC, only a DAC, or both. The data lines are commonly called SDO and SDI, for Serial Data Output and Serial Data Input, respectively.

NB: the names are usually, but not always, with respect to the converter chip. So SDO means "serial data out from this chip," which means the chip is an ADC. But always RTFDS to clarify. I2S supports data words up to 32 bits long, and it is a two-channel protocol, so for each sample time 64 bits are shifted.

As noted, it's a synchronous port, meaning a clock is provided by some master. That clock is often called BCLK, for bit clock (it might also be called SCLK, for shift clock).  Bits will shift out on SDO on the falling edge of BCLK, and will be shifted in via SDI on the rising edge of the clock. This clock runs at 64 times the sample frequency, because it has to transfer one pair of samples per sample period. So for a 48 kHz sampling frequency BCLK will run at 3.072 MHz.

Now "something" has to generate BCLK. Often it is something like a DSP or other processor which acts as a master. In some cases, the converter can generate it, too. Again, RFTDS, and what you choose depends on what else is in your design.

There is a second "clock" for I2S, called LRCLK. It runs at the sampling frequency, and a lot of people confuse it with word clock. It really isn't a clock at all. It is simply a frame indicator (and sometimes it's called a frame clock or frame sync). It is low for 32 BCLKs, which indicates that the data bits are for the left channel, and it is high for 32 BCLKs, indicating that the data bits are for the right channel.

Jitter on LRCLK is not really interesting, because as noted it is a frame indicator.  Jitter on BCLK is not really interesting either, except to the extent that it's so bad that it clocks in the wrong bits (which is quite difficult with a period of ~32 us).

Finally, there's the modulator clock, or MCLK, which is what actually runs the conversion. This clock runs at (usually) 256 or 512 times the sample frequency, and it is very sensitive to jitter, because it is the clock on which the audio is sampled.

Hope this helps (a bit).

-a
 
Hi ...

Reading the thread I'm thinking "where do you want to go with it"? Do you want to build something you can use for e.g listening purposes or do you want to delve deeper into the realm of knowing how to design a DAC or ADC? And also: Which level of quality do you want to reach (technically and/or audibly) and what is your economic scope?

You link to a couple of ADC projects that have already been made (beis; diyaudio) - others could be Acko's ES9102 project (diyaudio) or an eval board based on AKM's AK5394 ADC. This board is the board that 1audio (Demian Martin; diyaudio) has found to measure best of the boards he has tried (to my knowledge; about - 129 dB 2H).

DAC's could be TI's PCM17xx series which seems to be quite well regarded (diyaudio, hifiduino, others) not least since the I/V conversion can be very simple (a resistor). Something like the DDDAC (diyaudio) might be a suggestion here (I would suggest using lower resistor values though for much lower distortion. Output levels - DC or AC - below ~ 0.7 volts). Another suggestion based on an R2R DAC could be soekris engineering's DAM1021 which seems to be attracting quite some interest (diyaudio, hifiduino, others):

http://www.diyaudio.com/forums/vendors-bazaar/259488-reference-dac-module-discrete-r-2r-sign-magnitude-24-bit-384-khz.html

On the other hand if you are interested in delving deeper into the "intricacies" of AD/DA converter design I could suggest that you look into what "marce" on diyaudio has posted of links to literature on the subject. In his daily life he's a very-high-speed PCB designer and posts regularly on literature related to high(er) speed PCB design. IMHO the PCB design is indeed important/key to a fine result in designing an ADC or DAC - spec-wise or sound-wise.

Personally I've found these suggestions to be very informative, helpful - and also surprising as there were phenomena I didn't really know about.

Some of what he has suggested has been:

High Speed Digital Design: A Handbook of Black Magic (this is a book)
(Howard Johnson also have a website where he has answered various questions related to digital design:
https://www.sigcon.com/

Henry Ott: Electromagnetic Compatibility Engineering (also a book - seems to be a reference on the subject)
IMO a very useful book which I would assume is also accessible to people without a highly specialized technical background. Price-wise not cheap though (i.e. to my economy ) but I have found it in the library.

Other texts on the subject could be:

http://www.ewh.ieee.org/soc/emcs/acstrial/newsletters/spring09/designtips.pdf

http://www.mouser.com/pdfDocs/Tektronix_Fundamentals_of_Signal_Integrity.pdf

https://www.google.dk/webhp?sourceid=chrome-instant&rlz=1C1VFKB_enDK631DK631&ion=1&espv=2&ie=UTF-8#q=Precision+Analog+Designs+Demand+Good+PCB+Layouts  (powerpoint - quite interesting I think)

http://www.analog.com/library/analogDialogue/archives/39-09/layout.pdf
A practical guide to high-speed printed circuit board layout (could be a starting point).

And if you are interested in trying to simulate some of the solutions you could be considering then marce suggests this (free) software:

http://www.saturnpcb.com/pcb_toolkit.htm

It can tell about inter-track inductive & capacitive coupling; track-to-ground impedances and much more.

Well, hope this may be useful ...

Good luck with your endeavours  ;)

Jesper


 
Thank you Andy. There is a plethora of information there, and I highly value your input.

I'm still moving along here guys, currently my curiosity is moving towards what makes the high end clocks so high end other than the fact that they make your converters sound better and that they are more accurate. How does one "supe up" a clock and I'm curious what the most desired chips are. What is 2N3904 of clock chips and what is "high end?"

Thanks regardless,
-JP
 
It is not so much about the chip as the passive stuff around it.

For things required to lock to external references, VCXOs usually beat VCOs (but have very narrow lock ranges requiring separate units for 44.1 multiples and 48 multiples), keeping the noise out of the power and control voltage nodes is critical (Charles Wenzel has some good notes on this) and in the better stuff you see extensive use of ferrite and really LOW noise regulators dedicated to the oscillator and PLL components.

Search on 'Low phase noise' crystal oscillators for extensive commentary. 

When designing, start with a 4 layer board and move up from there, getting good performance in two layers is pretty much impossible.

Regards, Dan.
 
buildafriend said:
I'm still moving along here guys, currently my curiosity is moving towards what makes the high end clocks so high end other than the fact that they make your converters sound better and that they are more accurate. How does one "supe up" a clock and I'm curious what the most desired chips are. What is 2N3904 of clock chips and what is "high end?"

When you start talking converter clocks, it's the oscillator for the modulator clock that matters. And the offerings from the usual sources, such as Epson and Fox, all have very low phase noise for not a whole lot of money. What makes life complicated is that you have to support two separate "families" of clock frequencies, mainly the 44.1 kHz and its multiples family, and the 48 kHz and its multiples. This means you need a 22.5792 MHz oscillator and a 24.576 MHz oscillator, and a mux or some other way to choose between them. Plus you have to manage your power supply so that the oscillator outputs are as clean as possible.

Another option is the Silicon Labs devices. They have oscillators which can switch between two pre-programmed frequencies by pulling a select pin high or low. And they have others that will let you select the output frequency over a very wide range via I2C port.

Then there is the word clock problem, which requires synthesizing the modulator clock to the low-frequency word clock input. This generally uses a PLL and a VCO, a VCXO, or even direct digital synthesis. For purposes of a standalone DAC or ADC, you can save yourself a lot of aggravation by foregoing word-clock sync and just using a standard oscillator or one of the SiLabs parts.

-a
 
I always wondered why people didn't simply use S/PDIF to lock. Coming at a faster rate, it's so much easier to PLL from a 3MHZ+ source rather than a 44.1kHz!

How on earth we're expected to lock on solidly to a 44.1kHz source and create 24MHz+ clocks is beyond me.
 
Rochey said:
I always wondered why people didn't simply use S/PDIF to lock. Coming at a faster rate, it's so much easier to PLL from a 3MHZ+ source rather than a 44.1kHz!
You could, but that would have to be an unmodulated S/PDIF signal (akin to AES Black). Modulated data adds jitter, which is not what you'd want here. (See also this post by Dan Lavry).

Rochey said:
How on earth we're expected to lock on solidly to a 44.1kHz source and create 24MHz+ clocks is beyond me.
Why would that be a problem? At work I routinely lock a 10MHz OCXO to the 1Hz PPS output of our timing GPS receiver. The main issue is characterizing phase noise (or, longer term, drift) of the reference and target sufficiently well to design a good loop filter.

JDB.
 
If the bandwidth of the PLL is sufficiently low (sub-Hz) you can easily lock directly to the digital input. To avoid large-signal issues (dropouts upon large low-frequency jitter) a data buffer is however needed/advisable.

Samuel
 
buildafriend said:
Maybe it would make more sense if I built 1 or 2 ADC projects first? I found 2 that seem interesting. This way I can start to get an idea of what the layouts look like and I can gain a general understanding of the parts and connectivity.

http://www.beis.de/Elektronik/ADDA24QS/AD24QS.html

http://www.diyaudio.com/forums/group-buys/176416-diy-cs5381-analog-digital-converter-pcb-group-buy.html

I realise this is an old-ish thread, but I was wondering if you went ahead and tried one of the kits? I'd also like to learn more about how DACs work and have always found "doing, then reading" a much easier way to get my head around things than the other way around.  I was looking at the Beis kit... :)
 
My life has been a little hectic over the last 6 months but it's all falling back together now. I haven't had much time between live sound, media wrangling, and chasing a computer sci degree. I have a quad tube pre to finish and then I'll get back on the horse with this. It may be some time. PM me if you want to tackle this learning curve together. Maybe we can help push each other a bit.

All the best,
-JP
 
Back
Top