• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Audio analyzer project

syn08

Senior Member
Joined
Aug 9, 2019
Messages
376
Likes
445
Location
Toronto, Canada
I started this project over one year ago; went through several false starts (AK4499 MIA due to the AKM fire, relied on the TI ADS127L01 ADC (which proved to be improper to audio due to the high noise), tried the AKM AK5572 ADC, distortions were too high, plus the ever annoying (for measurements) noise shaping, the ES9822 datasheet hell, the ES9311Q chip hell, etc...).

Now that I retired I obviously have more time to assign to this project - and here's the first prototype, ES9038PRO DAC, ES9822PRO ADC, XMOS USB 384k interface. DAC/ADC are both dual mono in this prototype. So far, this is nothing but at at best a glorious USB sound card on three boards, there's no analog I/O yet, no autoranging, etc..., DAC and ADC are in an analog loop. A single board (4.4" x 3.9") version is in the factory, due next week, (with both DAC and ADC in stereo mode), the analog I/O design is on the whiteboard... it will require some magic to match the digital side performance :).

See the attached first measurements using REW (Vrms was not calibrated). No gimmicks, no filtering, no distortion trimming (I concluded it is a useless exercise, being so dependent on level and frequency; I may eventually consider lookup tables in the firmware, though), no notch filters, just the raw data, therefore this is pretty much the worse case performance (up to the analog I/O).

Even on this first prototype the design target for the analog loop was met (THD<-120dB, THD+N<-105dB), for both metrics the limitation being mostly the ES9822 ADC. I am expecting the single board version to get slightly better.

IMG_2320.JPG


meas.jpg


The Arta measured values are in pretty good agreement with REW although, for whatever reason, the spectral components are visually slightly lower. But I concluded some time ago that anything around -120dB is essentially very hard to reproduce exactly. Of course, some golden ears claim they can hear such differences :rofl:.

meas_arta.jpg


Some schematics were already published on the other side, where I was banned from. As soon as the project is completed, it will be uploaded as Open Source, both hardware (schematic, Gerbers) and software. There's no IP to protect here, other than the good layout practices and the effort to navigate through the ESS documentation nightmare.
 
Last edited:

Stinius

Member
Joined
Jan 14, 2022
Messages
25
Likes
23
Very nice project Ovidiu.
I'm looking forward to see what's coming.
Stein
 
OP
syn08

syn08

Senior Member
Joined
Aug 9, 2019
Messages
376
Likes
445
Location
Toronto, Canada
Since I know some are interested in the 20KHz performance, here's the raw result @192KHz sample rate

1659909974327.png


As a reference, here's the Focusrite Scarlett 2i2 3rd Edition (a low cost decent USB audio sound device) under exactly the same conditions.

1KHz, @48k sampling
1659911119877.png


20KHz, 192k sampling
1659910828067.png


Note the noise shaping...
 

Attachments

  • 1659909069617.png
    1659909069617.png
    88.8 KB · Views: 31
  • 1659910798084.png
    1659910798084.png
    113.3 KB · Views: 31
Last edited:
OP
syn08

syn08

Senior Member
Joined
Aug 9, 2019
Messages
376
Likes
445
Location
Toronto, Canada
Single board digital side completed... As expected, performance went a little better (actually it remains about the same as above, but this version uses only one DAC/ADC chips, so noise should be theoretically 3dB worse). This is the most complex board that I ever designed and soldered myself, manually (I was never an EE professional).

Now, full ahead for the analog side, that will be fun, with lots of prototyping...

1660670414511.png
 

al2002

Active Member
Forum Donor
Joined
May 18, 2016
Messages
172
Likes
110
Good to see your work here syn08. Subscribed, looking forward to future posts.
 

Dave Zan

Active Member
Joined
Nov 19, 2019
Messages
163
Likes
458
Location
Canberra, Australia
This is the most complex board that I ever...soldered myself, manually....
I am beyond impressed, more like in awe.
So, I have waited but now you have this much done, I have to ask, any idea of costs?
I understand that you don't plan to sell this yourself but BOM cost?
Or, even better, approximate cost from a PCB factory that does assembly?
Obviously it would be just a ball park estimate until you have the other board finished.
But any spread estimate would be really appreciated.

Best wishes
David
 

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,603
Likes
7,687
Location
Prague
Amazing job, @syn08 . We can get similar and probably even better results these days with available components like Topping DACs + E1DA Cosmos ADC, at very reasonable price.

D10s_1kHz.png


D10s_E1DA_0dBFS_noise_and1k_262k.png


D10s-OPA1622-E1DACosmos_19+20k.png


BTW would you mind showing the spectrum below 100 Hz down to 20Hz as well, with the X-axis not being cut at 100Hz?
Have you found SMA connectors necessary?
 
Last edited:

Zaireeka

Active Member
Joined
Nov 23, 2020
Messages
264
Likes
271
Location
fRAMCE
Impressive work, soldering all that stuff must have been quite tricky! I really love those coax connectors, so sexy...
 

restorer-john

Grand Contributor
Forum Donor
Joined
Mar 1, 2018
Messages
10,374
Likes
29,950
Location
Gold Coast, Queensland, Australia
OP
syn08

syn08

Senior Member
Joined
Aug 9, 2019
Messages
376
Likes
445
Location
Toronto, Canada
So, I have waited but now you have this much done, I have to ask, any idea of costs?
I understand that you don't plan to sell this yourself but BOM cost?
Or, even better, approximate cost from a PCB factory that does assembly?
Obviously it would be just a ball park estimate until you have the other board finished.
But any spread estimate would be really appreciated.

At the current (insane) semi prices, I estimate the BOM at about $400. Half of that is in the ESS, XMOS chips and the Crystek oscillators. But I am ready to admit that the whole implementation is over engineered - it could probably successfully take a few simplifications, in particular in the power supplies. But then this is the way I like to do things for my own experience, a no compromise approach. Because I can :).

About assembly costs, I have no idea, but from what I am reading it doesn't seem too bad. However, my estimate is that the logistics around such an approach is much more time consuming than manually soldering a couple of boards (which is what I am usually doing). To add insult to injury, last time I've checked board assembly companies don't stock specialty parts (like the ESS chips). Shipping two chips to China by courier doesn't make any sense, while in house hot air soldering a DFN 0.4mm pitch chip on an otherwise fully populated board is not fun, unless you enjoy watching 0603 parts flying around, and the smell of melted plastic.

So I believe assembly houses are a good way for a medium to large series of a commercial product, which is not what I am doing for my own fun and learning experience.
 
Last edited:
OP
syn08

syn08

Senior Member
Joined
Aug 9, 2019
Messages
376
Likes
445
Location
Toronto, Canada
Amazing job, @syn08 . We can get similar and probably even better results these days with available components like Topping DACs + E1DA Cosmos ADC, at very reasonable price.

BTW would you mind showing the spectrum below 100 Hz down to 20Hz as well, with the X-axis not being cut at 100Hz?
Have you found SMA connectors necessary?

I would not expect my implementation to be much different from any competent implementation using the same chipsets. At the end of the day, there is no secret sauce, it is the silicon dictating the performance. I would expect mine to be 6db better noise wise (which is already obvious, from what you posted, in the noise floor), due to the ES9038PRO (4 DACs in parallel for each channel). Since the SINAD performance is here mostly defined by the noise, I would expect the THD+N to also go down 4-5dB once that I adjust the gain chain to use the FS on both the DAC and the ADC. Do I need to mention the ES9822 ADC so-called data sheet doesn't say a pip about the chip FS level or at least the internal reference voltage value? Also, don't forget I am not using any 1KHz FS distortion compensation (for the reasons already mentioned) - this is something the E1DA ADC heavily relies on. Not much help outside 1KHz and FS...

Excuse my ignorance, but is it possible to control both the Topping DAC and the E1DA ADC from a single REW software instance? I thought no, therefore such a setup is good for point measurements, but e.g. the sweep features won't be available?

I'll post the spectra down to a few Hz (this is the current LF bW) as soon as I'm back to the setup, but I can tell, there's nothing outstanding under 100Hz. No mains frequency artifacts, everything is balanced, after all.

You may also want to estimate the 19+20KHz IMD at 192KHz sampling frequency, you may be surprised by the crud you will find up to 96KHz. Not that we can hear it, but it is definitely relevant for the instrument capabilities. If we should be concerned only about what's audible, then the FocusRite interface above is what we would ever need for audio measurements.
 
Last edited:
OP
syn08

syn08

Senior Member
Joined
Aug 9, 2019
Messages
376
Likes
445
Location
Toronto, Canada
One thing that I was unable to make work is the ES9822 ADC @384KHz, with 49.152MHz MCLK. Everything is by the book (read: ESS datasheet), the ADC should work at 24.576MHz bit clock (and the MCLK divided by 2), 384KHz settings are by no means different than the settings for 96KHz and 192KHz (only different values), there's even a table in the data sheet with the registers settings for each mode (no need to put the ADC in 2x mode for 384KHz).

Doesn't work, the noise floor raises to absurd values (some -70dB). The DAC works beautiful at 384KHz... I gave up until a better (or luckier) detective than myself could identify a solution. This kind of detectivistic work is the only "secret sauce" of any ESS based implementation...

And a side note, as much as I hate the XMOS controllers (from almost every perspective, starting with the weird programming model, documentation that is everywhere and nowhere, licensing model, prices, and ending with the power intake) that's the only option for us mortals. Any other UAC2 solution is a best kept secret, I could rather get some information from the Pentagon than from those manufactures in Asia (CMedia, ComTrue, etc...) while a reliable and complete open source solution on a generic controller like the STMs doesn't really exist, and won't, since UAC3 is around the corner, eliminating most of the UAC2 mess (that is, the interface polling mechanism).
 
Last edited:
OP
syn08

syn08

Senior Member
Joined
Aug 9, 2019
Messages
376
Likes
445
Location
Toronto, Canada
The J-test at log and lin scale. No USB isolators, just good quality CCHD-957 Crystek clocks. Dedicated to the Golden Ears in the Jiterrati (©Chris719) team. Comment, the overall bandwidth (DAC output stage + ADC input stage) is about 150KHz.

1660937291533.png


1660937358341.png


And the 32 tones test (by REW)

1660943298099.png
 
Last edited:

julbo

Member
Joined
Jan 10, 2022
Messages
32
Likes
33
Location
Mediterranean
Single board digital side completed... As expected, performance went a little better (actually it remains about the same as above, but this version uses only one DAC/ADC chips, so noise should be theoretically 3dB worse). This is the most complex board that I ever designed and soldered myself, manually (I was never an EE professional).

Now, full ahead for the analog side, that will be fun, with lots of prototyping...

View attachment 224599
I was always an EE professional, neither me nor my brothers in (HPC) farms have ever soldered such an intricate SMD board by ourselves w/o the need of our trusting technicians to do the rework. Chapeau to you dear sir :D
 
Top Bottom