Budget 192khz A/D

Help Support GroupDIY:

Gold

Well-known member
Joined
Jun 23, 2004
Messages
2,747
Location
Brooklyn
The Q402 looks like it’s headed in the right direction. XLR’s on the interface is a good start. More measurement ranges is also a good addition.
 

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
3,124

Thats brand new out from TI , comes with software called the pure path console , inputs are fully configuarble either for analog or digital mic or bal/unbal line , even input impedence can be set through software 2.5 ,5k ,10k,20k , 50 db of programable gain plus calibration ,eq, compression and DRE, mic beamformer processing , the board apppears as an audio source in USB mode and it uses our friend the Xmos chipset . There are other 4 and 6 channel eval boards running on the ACMB motherboard , same price but it seems more limited in terms of sample frequency , might still make a great multichannel A/D for a small mixer with the programable inputs . The board will also communicate via I2s which means a broad range of concetivity with other audio hardware . The pure path consoles software although I havent tried it , looks very much audio orientated and should be a breeze for anyone familiar with audio terminology to set up .

$199 dollars is great value for what you get here , seems like a perfect front end for audio testing with DSP functionality which can be customised to our needs for measurement or for that matter to match a particular audio source.

https://www.ti.com/lit/ug/sbau359a/sbau359a.pdf?ts=1625346477852 data sheet

Specs of frequency responce at higher sample rates shows a progressive degradation in out of band performance relative to the lower rates , factors of around .45(48khz) vs .1(768khz) Fs for +/-0.05db . Should be easy to program an eq button to counteract the fall off upto several hundred khz for frequency response measurements.
 
Last edited:

moamps

Well-known member
Joined
Sep 24, 2005
Messages
1,835
Location
Croatia, HR
It may be a bit of a topic but I wonder if anyone has considered using an oscilloscope with a waveform generator (or has any experience already) to measure the frequency range up to 100kHz or more? 70 or 80dB of dynamic range seems quite acceptable to me for this area. Keysight EDUX1052G looks very good for the price below 800 € (VAT included), R&S is too expensive.
This oscilloscope looks to me like a very good and universal solution for a small DIY audio lab.
 

Attachments

  • keysight.png
    keysight.png
    82.2 KB · Views: 9

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
3,124
The point is that with REW and other similar programs we have the chance of creating an automated sequence of tones to test many aspects of signal integrity all at the one time and capture and display the results , much more than you get with a scope and a sig gen in isolation .

Having a usb audio interface that doubles up for measurement/cal purposes ticks a lot of boxes for me , you can easily 'ring out' your signal path including cable with tones and compensate in software for the extra capacitance of a long cable run on a mic or change matching impedence or add extra gain etc . Because theres no pots like on a regular interface and everything is set digitally you should have 100% repeatabillity , exactly what you want for measurement.

looks fairly idiot proof .
 
Last edited:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
It may be a bit of a topic but I wonder if anyone has considered using an oscilloscope with a waveform generator (or has any experience already) to measure the frequency range up to 100kHz or more?
Actually I have a Hantek DSO4102C, that I bought specifically for that, particularly because it advertised a "Bode plot" functionality which turned out to be inexistent! No sweeps, no Bode display. 8-bit resolution makes the curves quite hairy, to the point I deemed it unusable and put it to retirement.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
inputs are fully configuarble either for analog or digital mic or bal/unbal line , even input impedence can be set through software 2.5 ,5k ,10k,20k , 50 db of programable gain
Actually, it seems very much oriented to MEMS/ECM mics. Line input looks like a collateral. :)
$199 dollars is great value for what you get here
Actually, the EU ex-VAT price is 224€ ($266).
Specs of frequency responce at higher sample rates shows a progressive degradation in out of band performance relative to the lower rates , factors of around .45(48khz) vs .1(768khz) Fs for +/-0.05db . Should be easy to program an eq button to counteract the fall off upto several hundred khz for frequency response measurements.
I couldn't find any datasheet for the chip. All links point to the EVM.
 
Last edited:

moamps

Well-known member
Joined
Sep 24, 2005
Messages
1,835
Location
Croatia, HR
Actually I have a Hantek DSO4102C, that I bought specifically for that, particularly because it advertised a "Bode plot" functionality which turned out to be inexistent! No sweeps, no Bode display. 8-bit resolution makes the curves quite hairy, to the point I deemed it unusable and put it to retirement.
Thanks. The worst thing is always that you can't usually try the real thing before you buy it. This Keysight looks very good to me, it’s not too expensive, but I haven’t tried it so I’m not 100% sure. My "problem" is that I don't know at the moment any other software/hardware that is affordable for me that could measure the frequency range up to somewhere around 200kHz and give some *.csv or such file as a result.
 

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
3,124
The four channel TI ADC6140EVM board has the abillity to combine inputs for a total SNR of 119db without DRE and also an extra 3 db reduction in THD compared to the 6120 . The 6 ch board PCM6260Q1EVM-PDK does 110 db snr in single channel mode with thd at -95db , The two and four channel cards are 2 and 1 volt respectively for balanced and unbalanced input where the 6 channel is 10 and 5 volts at the input , they all do upto 768khz .


heres the spec sheet on the ADC chip itself


I found pure path console software here but sign up and hardware are required to go any further ,

 
Last edited:

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
3,124
I was wondering if the EQ section on board the card could be used to taylor response at high frequencies. Ok it probably will bring up extra noise and distortion but were not so worried about that when doing bandwidth measurements .
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
I was wondering if the EQ section on board the card could be used to taylor response at high frequencies. Ok it probably will bring up extra noise and distortion but were not so worried about that when doing bandwidth measurements .
The filters are so steep there would probably be enormous noise build-up to achieve a minor improvement (less than 1/3 octave).
 
Last edited:

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
3,124
The filters on the GUI are destined for use within the audio band but there seems to be a way to program custom filters through code on the pure path console then upload it to the EVM board . The blocks of code are already pre written we'd simply need to change some parameters for the right frequency ,gain and bandwidth .
As far as I can tell in terms of programing the DSP it doesnt care if you tell it boost at 20khz or 200khz, its just a number you fill in in a line of prewritten code .

The pure path app automatically installs thesycon Xmos usb driver on the host system , so no problem using asio in any of the usual measurement or recording software environments.

Even as an audio interface its good value for what you get, 4 fully configurable/'cal'able analog front ends ,onboard DSP, optical I/O and 2 ch output over I2S, 4 ch or more with TDM , in terms of input SNR it comes out at 119 in four channel mono mode with the DRE adding an extra 10db noise reduction.

In reference to Abbeys point , the response does nose dive very sharply but could we sacrifice maybe 10 or 20 db thd and noise margin (by raising the gain at high frequencies) purely for bandwidth investigation ?
There is mention in the chipset data sheet about the bi-quad filters being able to be set the any desired LP characteristic.


Looks like a very comprehensive set of tools . Setp ups (including analog input configuration ,filters and dynamics)are uploaded to the EVM before the ADC is powered up for recording/testing . You might have one set up for bandwidth investigation ,then another for audio recording or even tune for a particular audio source . It looks like a very powerfull piece of audio gear in itself with all the processing included , my guess is what ever processing you do between the analog input and the output of the device happens with very low latencies ,not only because of the higher sample rates but also because your signal doesnt need processing through the OS/DAW in the conventional way , of course the delay in the convertor itself has to be factored in also .
As well as the EVM passing the signal onto the DAW/software a DAC per channel could bring the processed signal back into the analog world for low latency input monitoring post DSP or off tape signals . Adding the DIYINHK IO board which appears to support TDM would offload any output processing from the TI .



I was starting to visualise a small mixer/ interface based on the TI and HK board , it cuts down on so many knobs and switches , you can do all your front end gain,EQ,DYN and create a digital monitor mix if you want. Suppose you want to add in a tube mic amp with +12dbu or more output you apply passive attenuation to get you down to the 2 volts sensitivity of the card input and your slugging the A/D right in the swingers .
For that matter why not match a step up transformer directly to the ADC for mic inputs, the CMRR afforded by the transformers(in the mic and the preamp) as well as the filtering effect at high frequencies should give very good noise performance , even if you have to add a few db of digital gain and maybe HPF to get the levels just where you want its a super clean signal path .

Reference page 14 (how to program biquad filters) and page 6 (How to generate N,D..... coeffecients with a filter design package) of the document posted above

Couple of other interesting bits here ,

https://www.ti.com/lit/an/sbaa494/sbaa494.pdf?ts=1625613509283

about sample rates and their supported features

and here about connecting other devices in parralel


Turns out none of the DRE,biquads, the summing or mixing capabillity works in 768khz mode its all reduced to a single channel with a decimation filter
 
Last edited:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
I would take the litt with a dose of defiance.
"First order high pass filter with specified gain, specified cutoff frequency,maximally flat passband and stopbandr esponse. Stopband frequency responsehas a -10 dB /decade slope."
Same error with 2nd-order filter.

"Gain; All pass filter at the specified gain"
??????????????


Low Pass Variable Q 2: The quality factor is the center frequency divided by the passband width.
How do they define the passband width of a HPF? By definition, it's infinite, even in digital where infinity is the Nyquist frequency.
 

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
3,124
Copied from TI forum as link didnt work.

On the decimation filter side, this really boils down to the design tradeoffs in making a more complex filter vs. things like package size, cost, and power consumption. Generally for audio applications we have found that the 70-80dB of stopband rejection is sufficient for a few reasons:


1) These systems are generally bandwidth limited in multiple other ways. These devices are only intended to be paired with microphones that might extend up into the ultrasonic range and the out of band noise that these microphones might pick up/generate would be very low level such that 70-80dB is enough to bring this noise well below the measurable level. Any playback path is also going to have similar levels of stopband rejection.


2) The front end PGA of our device has a bandwidth of ~80kHZ, so any frequencies higher than this are going to be inherently filtered by the PGA prior to reaching the decimation filters.


3) These are highly integrated devices so there generally aren't a lot of ways that this higher frequency noise would be able to couple into the signal path. It essentially boils down to noise on the supply (for which these devices have a typical PSRR of ~100dB) and noise on the input. If there is the potential for large amounts of high frequency noise to couple in to the input then an anti-aliasing filter can always be used.


4) You could potentially have some ground interference issues, but generally this can be avoided with proper ground plane management and by following the layout guidelines.


All of this is really to say that we have found that for most consumer audio applications these devices were designed for, 70-80dB stopband rejection is sufficient to not pass significant amounts of higher frequency noise that might degrade performance, and as long as this noise isn't mixing with something else it would be well outside of the audible bandwidth anyways. We do recognize that some people like to use these devices outside of traditional audio applications and we try to make them as flexible as possible to support that, but they are still designed with the audio market in mind first.

I put in a request for the paper relating to CMRR , maybe Bill might like to take a look .
 

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
3,124
I got an email back about the missing CMRR report from TI , said should be there early next week , anyway I gave a check on the site just now and the document has re-appeared ,

Application Report Input Common-Mode Tolerance and High CMRR Modes for TLV320ADCx120 Devices

 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
In the hope of using 768kHz SR, I ordered said EVM and a not too expensive DAC, that I intend to use with the FlexASIO wrapper.
Will report when (if?) there is actual progress.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
13,030
Location
Marcelland
I received the EVM this morning.
In order to simply make it work, one needs to download their Pure Path Console 3 app , which is subject to approval from a supreme entity.
They really want us to deserve them... At my age, I really hate hoops and rings.
 

Tubetec

Well-known member
Joined
Nov 18, 2015
Messages
3,124
Ahahaha , a supreme entity
Ive always hated it , corporate entities nowadays have more hooks and loops than a strip of velcro , I had my browser virtualised for ages , worked a treat , thwarts the entities leaving a little dump in your system files ,
sandboxie was my go to , then it stopped working , a percieved reduction in threat online and I stopped using it , online threats are only getting worse ,
 
Last edited:
Top