Updated 8x8 and 16x16 audio

Does the xlr board have enough gain for a dynamic mic(SM58)? I’d like to make a mixer for my mom. And use another board for possibly a powered speaker or monitor. So that would only need a balanced and unbalanced input.
 
The XLR board should have *close to* enough gain. The PGAs have 60dB of gain, so a vocalist with a strong voice will be fine, a whisperer not so much!

I haven't yet tested this in practice - the XLR connectors haven't arrived yet!

Most of my mics are condenser, but I have some Behringer XM1800 dynamics so they should do OK for testing even if they're not as sweet as SM58's. I can work out absolute sensitivity using my ECM8000 measuring mic, if I dummy up a phantom powering arrangement.

I'll post the results here and on the repo when they're available.
 
All good news … let’s hope the new boards work as intended!

Took a look at your linked repo, nice job on the documentation. Obviously it’s all a work in progress, but should there be a control driver one too? I was looking at your figures for CPU use, which seem high; this may be because you’re multiplying the AudioProcessorUsageMax result by 100, which is not applicable for the Teensy 4.x library (despite Paul banging on about maintaining backwards compatibility for any changes, he didn’t in this case, and also hasn’t documented it).
 
Jonathan,

Yes, there will be a separate driver repo linked to the hardware one (control_TLV320AIC3104). I'll move the driver doco across to it when it's there.

The mux code is incorporated, as Rob Tillaart's excellent library for the PCA9548 series was far more extensive than was required to control the mux in this instance.

I'm loath to release code until its properly tested on the final hardware.
 
Thank you! Looks very comprehensive and well-commented and documented, so it should be easy to use and figure out if (when) I do something wrong!

I assume the TDM objects are only needed pending Teensyduino 1.60, which should have all the (known) TDM bugs fixed? There’s a few items in there I don’t recognise and may be unused, but that’s to be expected in development code. If changes are needed then I’m happy to roll them into my multi-TDM, if you like, then the eventual TD1.61 can just have that plus your control suite for full support of your hardware.
 
I've uploaded the example files.

The 16-bit DMA-audio buffer data copying may not be required. I wrote it before I found the bug in the 32-bit code. The example files have no argument in the constructor, so I guess the code is unused .. it's been 6 months since I finished work on the driver.

The critical fix is line 105 in the input and the same code in the output. As you say, these should be fixed in 1.60.

I did have some code in there for inverting BCLK, but I intended to remove it if I haven't. I went for hardware inversion, so that the standard (fixed) TDM driver would work OOB.
 
That’s great, I’ll have a nose.

Sounds like early adopters will have plenty of choice for TDM, then: either your fixed one, the TD1.60 beta one, or my multi-TDM one. I also implemented the BCLK inversion in software, I’ll probably leave it in there “just in case” - it’s only a few lines of code.
 
Yes, leave in the BCLK inversion - it will make things easier for other hardware creators. I only removed that feature from my solution as I didn't know if that it would be included in 1.60.

BTW, thanks for the pull request with the fixes. Those typos do love to get through.

I left in the change to AudioProcessorUsage( ), however I seem to get a fractional rather than percentage return value (see the tests in the hardware repo's readme.md) . . . or . . . processor usage is so low that it's barely registering (which would be magnificent)! I'll write something that is heavy on CPU and see. In the meantime, I'm happy to accept that your version is correct.

Interesting that Paul's moved to using Sparkfun as an exclusive hardware wholesaler. I hope that means he'll actually have more time for new products and software. His announcement suggests that's his intention.
 
I just ran Paul's memory and CPU usage example and (unsurprisingly) you are correct. I will amend the CPU load figures.
 
The CPU load is astonishingly low for the number of channels, isn’t it! I suppose it’s just a bit of memory shuffling then letting DMA do its thing … the shuffling for multiple outputs got a bit mind-bending, though o_O

Yes, the Sparkfun announcement is good news, should free Paul’s creativity :D Though it would also be nice if he could devote a little time to website and repo maintenance, too … it’s all very overgrown and stale at the moment. It’s just about OK for the experienced user, but n00bs are really struggling as far as I can tell.
 
The connection shuffling issue is why I write the dynamic patching example. Looping through arrays is a wonderful thing!

The PCBA is complete and waiting for pickup. Monday next week looks like the likely arrival date. Testing should take almost no time - they'll simply work, or they won't as this version is little changed from the last - just smaller and with the ground planes intact!
 
New PCBA boards have just arrived. A quick test shows CODECs are being programmed and audio inputs / outputs are working.

More to follow after more thorough testing.

Jonathan, I'll post the PCBA files on the repo as soon as the boards are fully validated.

Tindie listings will follow for manufactured boards.
 
The board files are now uploaded to the hardware repository, with full JLCPCB PCBA files for the main PCB and gerbers for all the Wing boards. https://github.com/palmerr23/Teensy8x8AudioBoard

Apart from the main PCB, BOM files are indicative only. Parts should be verified before ordering. The XLR Wing board BOM file has both male and female XLR connectors listed. Only one set of four is required.

There have been some minor updates to the driver files - mainly to default to differential inputs as single ended configurations misbehave when differential signals are applied. https://github.com/palmerr23/control_TLV320AIC3104

I intend to have the Tindie listings done by the end of the week. I have 8 spare main boards in hand and several of each of the Wing boards. Sadly, the cheapest international postage I could get from Australia was around $AUD20. However, it doesn't increase much for additional products.

Have fun!
 
Last edited:
Awesome! How do you like it? Were you able to test it with a dynamic mic by chance? How's the noise? Great job on this, incredibly impressive!
 
Here's a test with an ancient Electrovoice RE80 that hasn't been dropped too often (despite mainly being used as a bass drum mic).

I've left the audio as recorded - Teensy peak indicators were at about 0.8 at the loudest part.

The red line in the spectrogram is the peak when I counted to 10. The black is just holding the mic. The mic has little treble response, so the drop off is not surprising. It looks like around 55dB S/N at low frequencies, and 40-45 above 3kHz.

So, not as good as a professional preamp, probably good enough for live sound.

Line level inputs are a quite different story. Noise is around -90dB. Distortion is 0.015% (weighted).
 

Attachments

  • test.zip
    314.7 KB · Views: 102
  • test.png
    test.png
    54 KB · Views: 94
Last edited:
I cant seem to download your file. Im also having some issues with the back components on your BOM at the jlcpcb website. Sorry, Ive never ordered a pcb before! Are you only getting the top side assembled?
 
There are no back components in the BOM or placement files. Everything is on the front for the standard set up.

You will need to purchase and mount all the connectors, as these are not cost-effective for PCBA.

Also, the cost to do one board could be prohibitive as there are set-up costs.

Or you can buy one or more from me on Tindie! I'll put the basics up in a day or so. I haven't done the final figures yet, but a fully-assembled PCB (all connectors mounted) will be < $US30 and the postage for 1 - 3 units is around $20USD.

That should help comparing the cost with JLCPCB's PCBA + sourcing the connectors.

Happy to help further if you want to go down the PCBA path.

I changed the test files, so refresh the screen for the new ones.
 
Ordered! I also did some PCBs intended for clock buffering, so we can see if a ridiculous I/O count is possible. I (probably) won't be doing rigorous noise / crosstalk etc. measurements, though. 15 PCBAs plus 50 buffer PCBs were $163 or £132, including shipping and VAT. That includes a $9 "coupon" so YMMV. Oh, and I've had JLCPCB come back in the past with surcharges; hoping that won't be the case this time as the VAT looks about right...
 
You must log in or register to reply here.
Back
Top