Multichannel Audio DSP Field Mixer Recorder

4 preamp cards and 4 balanced line drivers using OPA1637.

The Vcom and higher +/-9v rails for OPA1637 buffer are supplied through the stackable headers on the preamp cards. They are 1mm to long so not sure I will be able to squeeze all 8 preamps in there using those. 

The 6mm brass standoffs worked fine for the line drivers and I send the VQ down one line of brass standoffs and it works. 

I've started work on the case again and will print another in CF and will glue on some thin aluminium panels in recesses in the CF.

I am making plates for connectors that bolt to the chassis. so far only 2 plates , 1 for the mini XLR inputs / outputs and 1 panel for the AES IO mini XLR and 1/4" headphone sockets.

I made some slots in the digital and analogue main board mounts so I can slide in some EMI shielding between the 2 boards.

this is a very rough 3d thing of the assembly. you can click to make a cutaway.

https://grabcad.com/library/bbbox-1

I've started work on the case again and will print another in CF and add some aluminium panels.

I am making plates for connectors that bolt to the chassis. so far only 2 plates , 1 for the mini XLR inputs / outputs and 1 panel for the AES IO mini XLR and 1/4" headphone sockets.

I have no idea if this 3d grab cad page will embed but its a rough idea!

https://grabcad.com/library/bbbox-1

hopefully final video update. The filter is for half mclk frequency of 12.288 ( 6.144mhz ) 

some interesting notes on buffer filtering for ADC here..

https://www.ti.com/lit/an/sboa114/sboa114.pdf?ts=1625582064839

I was looking at frequency ranges just above audio but the filters job is to cutoff anything from 6mhz and higher ! 

update simulation using opa1637 as the preamp buffer to ADC.

better results in simulation at least.

Why I can't use the OPA1632 to attenuate the PG2500 with +/-5v power rails

OPA1602 attenuation and low 6mhz low pass filter

The main case is just one part with top and bottom access panels.

I'm making connector panels that can be changed so if somebody wants for instance 4 x full XLR

and a DSUB for the rest of the channels its just 4 bolts and an FPC connector to change it! 

I will make a panel for the AES sockets next. 

There is a recess for the oled confidence screen and Time Code clock battery. 

Also a recess for BT module and programmer connector.

I'll need to learn how to make a nice sliding panel for the battery change as I hate swing out doors that just snap off like a lot of sony pro cameras etc.

Later down the line when I'm happy with its functionality and I work out better waterproofing I will work on beveling the edges and work on the horrible sharp corners! The bottom edges have a routed channel for a waterproof strip the will mate with the bottom panel to form a water seal.

The preamps are working great with the new buffered VQ ( vcom ) going down a bolt that holds them in place.

I can't say the same for the headphone preamp at all !! 

I had not really thought through the bias injection to the buffer properly ( tried to inject it into each input leg of the negative rail which has the negative feedback loop. the positive was going to GND  ) wrong. 

ignore the note about 100uF decoupling ! 10uF c1,c2,c3,c4 should be enough to remove the DC bias :)

Now I need to incorporate this 50khz corner frequency filter in the negative feedback loop of my 4 channel OP4A134 or OPA1604 if this basic version works.

from a cirrus logic app note : 

A two-pole Butterworth with a corner
frequency of 50 kHz attenuates the signal at
20 kHz by approximately 0.1 dB and has nearly
ideal phase linearity within the audio band

As I type this I realise it may have been more sensible to breadboard the buffer with the filter and test it 1st before sending off for the pcb.. 

I can test the filter with a sweep  generated by the DSP from say 15khz -> 100khz and make sure the filter is doing the job.

I assume these ultra high frequency's can effect the lower ones harmonically if they are not removed? anyone know why i am actually making this filter ? it's just the recommended one in the datasheet !

I connect the differential DAC audio signals to the non inverting inputs of the 4 channel op amp IC and bias it with the buffered VQ (VCOM) form the DAC. I think if I need some attenuation of the DACS outputs I'll need to use voltage dividers on the inputs as the non-inverting inputs since any non-inverting op amp input has a minimum gain of 1 ? in other words in this configuration I cannot use the buffer for attenuation.

Please drop a message in the comments section if I am wrong there? thanks in advance :)

EDIT:  I just realised I can set the output level of the DAC using I2C registers , so would never need to set attenuation anyway.

I beefed up the upper lower case mounts where there was available space.

Also added some pcb mounts that don't require spacers. 

Good bad and the ugly, 

This will have to do while I work on more important updates with analouge boards.

Will try carbon firbre nylon composite later. 

As you can see this is my 1st go with a 3d printer! It's actually my 2nd go at the lid. The 1st lid was completely wrong , always measure twice !

my specially ordered miniXLR sockets with the nuts on the outside of the case are stupid. I realised the pcb on the back is worth 1$ so it's far better to put the mini xlr sockets in through the holes ( should be D shape 11mm ) then tighten the nuts / washers from the inside the case before soldering the pcb onto all the pins. need to add thread lock glue to those nuts I think or you would end up with a bunch of washers and nuts flapping about in between the sock and pcb eventually.

cramped ugly nuts on the exterior of the case.. no ! got to go :)

top right corner it a whole for the programmer cable.

the uSD cards need more thumb room , it's a tweezer job to get them in and out !

The whole case needs beveled edges , way too sharp everywhere!

The bulge in the case is caused by the 40pin analogue board connector from ADC DAC card

that protrudes over the edge of the board edges. easy mistake to forget that those IDC connectors are much wider than the box around the footprint on the PCB !

for now I will just cut a whole in the case so it doesn't warp the box as I want to change the connector for light weight FPC ribbon cable , so its a problem that will eventually go away :)

Well I guess this is the great thing about having a 3d printer , you just keep refining the case until your happy then get it made professionally. I didn't bother with vase mode or whatever is the best way to print. I used sun PLA+ which is slightly stronger than the usual PLA. But could well have used anything. 

time to do some work on the preamps and line driver cards now I have learned more ! 

Originally I tried to avoid using buffers between the ADC and preamps and DAC and line drivers. 

not really my best idea ! I think I avoided them since I just didn't really know what they did :)

.. since then I have learned the PGA2500 vcom is in fact an input !!! pretty basic information but something I had completely overlooked.

There is an advantage of having the PGA's audio signals swinging around 0V so I had connected VCOM to ground.

BUT... the ADC has to swing around about 2.5v or near since it just used +5v. The exact bias voltage is available on VQ the pin.

you can't just hook up VQ to VCOM on the the pga or ADC buffer driver ( 10uA maximum current sink for VQ )

OK , I need a VQ buffer so I can send this bias reference across all 8 preamps.

why not just add a buffer / proper filter as recommended in the data sheet and keep the PGA2500 VCOM connected to 0V with decoupling capacitors in between the buffer biased at VQ ( buffered )

the buffered VQ bias ref voltage has to get around all the preamps !!

Staring at the cards and backplane it seemed like VQ was going to make a massive trip all over the design and pick up who knows what EMI. 

Ages ago I thought at some stage I will need to add a threaded bar right through all the preamps to stop them flapping about in the header sockets. then I though , just send the VQ voltage down that bar.

then I briefly thought I had a clever idea , followed by another moment when I realised they have very likely been doing tricks like this since the days of vacuum tubes.. followed by another thought of , maybe the massive 2m threaded bar will pick up even more noise than if had just supplied the buffer with 2.5v from voltage dividers of the the 5V rail. The same 5V the ADC uses anyway.

I don't know what's inside the ADC IC , how is VQ derived ? Its ridiculous to think it probably is just voltage dividers !   

finally pretty much back where i started before the stm32f7 to H7 upgrade.

I still have an annoying bug in the BT app -> Micro where I need to press a button 3 times before it makes it to the app.. the sliders are probably only sending every 3rd value to. Its obviously a buffer or just the annoying fact I have no EOL character being recognised in the UART between stm32 and HC06 module.

I'm sure there is an obvious fix. I'm using DMA and streaming out tons of meter value data on the same UART. 10 characters go one way and 3 come in for device control. I did try and align the buffer of TX and RX but it didn't help. It could also be the way the App Inventor code sends the characters!

[edit]  

Fixed ! as soon as i finished writing this log , I tried disabling the FIFO buffer on the UART and bingo.

one press on a button and instant reaction in the hardware. The sliders feel more accurate too since there not skipping values.

It's working with an interrupt that happens when 4 character arrive in DMA. my incoming commands are a fixed 4 bytes long. the 1st byte is just a marker to check I am receiving a real command and not garbage.

I don't send back any acknowledge byte yet to the app. 

[end edit]

The individual channel power switching is working. I will need to force a mute on the powered down preamp channels as there is some small noise there. I can do that in the ADC or the DSP , not sure which is better , possibly better to mute the relevant ADC channel so it's quiet when powered back on. Now there is a large thud ( boom ! )

I don't really like the 1.27mm headers , pins can easily snap off and the connections are not as solid as 2.54mm. looking forward to the day I can just solder those preamp cards onto the backplane !  There is no way I would have got 8 preamp channels in such a small box with the larger headers so no choice ! I do wonder if I would have been better off using flat flex ribbon cables like in a mobile phone.

I think the next job after fixing the BT uart will be ltc timecode IO which will come through the TI Codec on SAI3. I'm really looking forward to that from a C code perspective. Then right after that I have to wake up the RTC TXCO and figure out how to jam sync incoming timecode to it :)

Way off into the future but hopefully before next year I will get the Xmos chip awake and get AES3 IO in and out of the DSP ! Hopefully I can setup TDM8 on the XMOS to sync up the the DSP as a slave without to much trouble. 

AVB seams like pipe dreams now , but the gigabit ethernet PHY IC is there waiting and connected to XMOS.

I doubt USB C sound card mode will work at the same time as AES3 and AVB but who knows ! that will be decided by XMOS computational power and my coding ability in the XMOS IDE i think.