Budget 192khz A/D

[Tubetec]

I was looking around for a small A/D board for Rew measurements and other general audio purposes .

https://www.audiophonics.fr/en/devices-hifi-audio-adc/stereo-adc-board-wm8782-i2s-24bit-192khz-p-14897.html

i2s interface , set up with jumpers and switches ,two clock sources ,  filter components on the inputs which could be re jigged for better HF extension compared to  usual comercial interfaces  . Specs dont look outstanding only manages 100db s/n and -93 THD, still might be a cheap way of making a wideband input to REW .


https://www.ebay.ie/itm/XMOS-USB-Digit-Interface-I2S-IIS-to-Fiber-Coaxial-HDMI-SPDIF-AES-Output-Board/312787430205

that board takes i2s input and outputs it in several digital audio formats .
All comes in at less than 40 euros not including psu.

Most of the usb/i2s interface boards are designed as outputs , minidsp's streamer does the trick but its a tad pricey , suggestions welcome .

WM8782 has a pair of opamp inputs , the spec sheet says
Passband  +/-0.01db  .4535fs  -6db .4892fs @96khz , so flat to 43.5khz , no measurements for 192khz given.

 

[Tubetec]

Ive a couple of questions about this little ADC board/chip  ,
Would it be it possible to use an unbalanced stereo ADC as a balanced input mono source ?

In other words can I feed both channels with the same signal , then in software combine Left and Right channels ,flipping the phase on one channel so noise cancels and wanted signal adds , could I gain S/N ratio in this case or maybe help reduce inherent distortion ?

https://www.cirrus.com/pubs/proDatasheet/WM8782_v4.8.pdf

 

[Rocinante]

Ive a couple of questions about this little ADC board/chip  ,
Would it be it possible to use an unbalanced stereo ADC as a balanced input mono source ?

In other words can I feed both channels with the same signal , then in software combine Left and Right channels ,flipping the phase on one channel so noise cancels and wanted signal adds , could I gain S/N ratio in this case or maybe help reduce inherent distortion ?

https://www.cirrus.com/pubs/proDatasheet/WM8782_v4.8.pdf

One would think adding a THAT op amp (whatever your flavor) would deliver. Maybe use the Wayne Kirkwood design. At any rate following with great interest.

 

[Tubetec]

http://nihtila.com/2020/02/02/adc-ak5572-very-high-performance-audio-ad-converter/

that looks interesting ,
needs +/-15v ,5v supplies and i2s to usb interface ,
despite the nicely detailed spec and youtube circuit description from the designer no mention of what happens above 20khz , although he does mention the board was originally concieved for measurement purposes.

 

[Bo Deadly]

Page 67 of the AK5572 datasheet shows a circuit for running a mono signal into both ADCs.

  https://www.akm.com/content/dam/documents/products/audio/audio-adc/ak5572en/ak5572en-en-datasheet.pdf

Looks like it improves SNR by 3dB. I would think it would do the same for the WM8782. The AK5572 board in Tubetec's link uses OPA1632 as a fancy differential out op amp for ADCs. It has a mono mode as well but it doesn't bother to invert one of the inputs since OPA1632 is fully diff. It's just running parallel for 3dB noise improvement.

These are interesting for custom applications but for a test rig it's kinda hard to justify not buying a conventional USB audio interface since even an inexpensive one will outperform just about any circuit you might want to test.

 

[ruffrecords]

http://nihtila.com/2020/02/02/adc-ak5572-very-high-performance-audio-ad-converter/

that looks interesting ,
needs +/-15v ,5v supplies and i2s to usb interface ,
despite the nicely detailed spec and youtube circuit description from the designer no mention of what happens above 20khz , although he does mention the board was originally conceived for measurement purposes.

Blurb says simple high pass and low pass at the input but no mention of turn over frequencies.  Later FFT plots of 10KHz sine wave show harmonics up to 50KHz so it looks like it goes a fair way up.

Cheers

Ian

 

[ruffrecords]

Also just noticed this uses the LM4562 in the signal path. Not come across this one before but looks like a souped up NE5532.

Cheers

Ian

 

[Tubetec]

Audio interfaces vary but still showed a marked roll off on higher sample rates , for audio it makes little odds or even unwanted but for circuit testing in REW we might want to look at 100khz or more bandwidth , I think a dedicated front end for test purposes is justified . I might just get one of the 15 euro 192 boards as a proof of principle and see if we can achieve the 87.5khz bandwidth the maths suggest. With the 40 euro Xmos i/o and the i2s to Digital audio board interface , still only comes to 75 euros , and you have the option of adding a dac later if you want too .

All the other complications with the AK5572 board makes it a bigger job to get going , I might try contact the designer and see if he has any other measurements beyond 20khz .

 

[Rocinante]

Also just noticed this uses the LM4562 in the signal path. Not come across this one before but looks like a souped up NE5532.

Cheers

Ian

Yep.
I've just recently seen one in a DAC. I believe they are interchangeable with NE5532.

 

[rock soderstrom]

Also just noticed this uses the LM4562 in the signal path. Not come across this one before but looks like a souped up NE5532.

Cheers

Ian

They say the LM4562 and LME49720 are identical.

 

[mhelin]

I think AKM converters are now unavailable because of the AKM factory fire.

Another one to consider would be the PCM4222EVM.

 

[Tubetec]

I dug a little deeper into the AK5572 spec sheet and they give -0.1db ,-1db ,-3db and -6 db HF points for each value of sampling frequency . The jumper set digital LPF seems to be unused in 384 and 768khz mode .192khz gives around 90khz -3db , 384khz mode gives a -3db at 137.63khz and 768khz only manages 144.5khz.

The new version of this board also has an option for trimmer pots to calibrate the inputs .

 

[Whoops]

Thanks for bringing this up Tubetec.

I'm new to all this so could I please ask you what is the i2s interface?

To make this an input interface you would need that ADC board that has an i2s out and then you need another board that is i2s to USB? is this correct?

Thanks

 

[Whoops]

I dug a little deeper into the AK5572 spec sheet and they give -0.1db ,-1db ,-3db and -6 db HF points for each value of sampling frequency . The jumper set digital LPF seems to be unused in 384 and 768khz mode .192khz gives around 90khz -3db , 384khz mode gives a -3db at 137.63khz and 768khz only manages 144.5khz.

The new version of this board also has an option for trimmer pots to calibrate the inputs .

Seem like really good figures for a testing jig.

this are the prices from the website:

ADC AK5572 (v1.3)

Very high-performance audio AD-converter.

At the moment boards come with screw terminal and 2×4 I2S header soldered but XLRs (which are fully black unlike in the photo below) and 2×8 header unsoldered.

    Stereo board £87 – 11 in stock
    Mono board £76 – 2 in stock

 

[Tubetec]

https://www.diyinhk.com/shop/audio-kits/170-xmos-32bit-384khz-dxd-dsd256-usb-tofrom-i2sdsdreclock-spdif-pcb.html

That board takes an i2s signal from the AK5572 convertor and outputs it over usb.
Most of the xmos boards are destined for use as dacs but it seems by changing the firmware they can add inputs as well . The basic 39 euro i/o board only goes upto 384khz . It looks like the 89.99 multichannel board could be reprogramed to accept a 768khz i2s input. I need to enquire about it .

Its all down to an SDK you can get from thesyscon if you pay for the respective liscence , it allows customisation of the parameters of the Xmos chip and the drivers , you can also program the chip to do processing tasks ,

https://www.thesycon.de/eng/tproc_usb_audiodriver.shtml

The frequency responce figures I quoted were the manufacturers spec for the AK5572 alone ,
Nihtilas circuit may have its own limitations outside the audio band due to filtering , although his circuits dont add any signifigant THD or noise to the base level specs of the AD chip itself and the op amps have bandwidth measured  in Mhz.
I emailed him to find out more , 

 

[Tubetec]

https://www.youtube.com/watch?v=4FxB8RBPtKU

Xmos provide a programming suite for their chips- Xtimecomposer , the  code is beyond me but its useage in the context of measurement are obvious . It seems thesyscon have a liscenced software package that takes all the coding out of it , in other words a person familiar with audio can set it up the routing /ins/outs processing etc .
DIYINHK do an xtag programmer for 18.99 so it might be worth adding to the order in anycase

theres a lot going on xcore.com relating to audio processing , beam forming arrays and especially smart speaker or alexa type (A.I.sh!t) boxes  ;D

What would be nice is if the processing power of the xmos chip could be used from within REW to for instance implement a filter on the incomming signal .
Ive gotten one reply already from the DIYHK crowd , I'm not sure their translate is very effective  though.
Ive now asked about the possibility of them making a custom firmware without any output processing or mixing facillities as no Dac is needed in this case  ,  just an i2s input over usb at 768khz with the possibility of processing on the input signal  .
On the website the say they are working on the worlds first 768khz standalone recorder (w/ AK5572), so even though none of the current xmos boards on offer have 768khz i2s input they appear to have it in the pipeline along with its acompanying analog input board . 
more to follow ,

 

[abbey road d enfer]

Very interesting! It would be a much cheaper alternative than the Spectraplus souncard.
The big question is when will there be a complete usable system.

 

[Tubetec]

The DIYINHK guy got back to me , he would only say that very soon the I2s 768khz input board becomes available from the store , by the looks its going to be a special firmware for their DXIO 32 boards .

I was also looking at some of the newer ESS AD chips , they say a low noise floor bandwidth of 200khz , what ever they mean by that . They also have programable filtering with 8 modes ,so possibly a mode with wider bandwidth. I couldnt find a full spec sheet .

http://www.esstech.com/files/1816/0767/3501/ES9822PRO_DS_brief_v0.2.6.pdf

For those on tighter budgets who would like to investigate i2s A/D boards , theres two on ebay one based on the PCM1802 another on the PCM1808 ,only cost a few euros each and do upto 96khz.

 

[ruffrecords]

What about noise shaping? This is a 32 bit AtoD so presumably at some point this will be dithered down to 24 bits at which point there is the opportunity to include some noise shaping. Usually this is arranged to improve the S/N in the audio band at the expense of higher noise out of band. Fine if you want to make your specs look good but bad news if you are looking to see significantly beyond the audio band.

Cheers

Ian

 

[Tubetec]

Just trying to get more details on the ESS range of ADC's , turns out I need to sign a non disclosure agreement just to get the specs on the product , even if I did I might not even be allowed to share the finer details here .

Filters on the AK5572 are set by jumper and as far as I can see around 150khz -3db is about the best we can expect from 768khz . ESS ES9822 pro or ES8920 have a programable filter with 8 modes as well as on board processing , these chips are probably a lot more costly than the AK ones .

Even the ADC's with 32 bit are switchable to 24 bit , not really sure about dither/noise shaping its not mentioned in the specs.
But yeah I get what your saying about the filter playing a part in measured THD/and signal to noise , the higher sample rates on the AK5572 appear to perform worse than 44.1 or 48 khz mode. ESS ES98xx series are more or less brand new(Nov 2020) , cant even get a price one.