SoundWing Audio I/O Board

The replacement project for this is under way here.

After much struggling and researching, I've been unable to figure out how to use I2S as a bidirectional multi-slave bus for audio. Maybe it can be done, and maybe it can't; I'm just not able to figure it out.

I've thus decided to set this project aside and turn my attention to an Arduino Shield that combines this with the EchoWing. That will bypass the need for I2S, taking the processor out of the audio flow altogether and leaving it to control the effects and other functions.

I'll supply a link in a separate log here once I've started that project here on Hackaday.

I've become aware of two major problems with my current design, one fundamental and one simple hardware.

The fundamental one first: as you've probably gathered from the title, I'll be switching the I/O protocol from I2C to I2S. I'd been under the impression that the I2S-related pins on Feather boards weren't broken out, and only learned in yesterday's Hackchat that it is possible to attach I2S to a Feather (as some people clearly have done). I'm still trying to figure out which pins to use, as help in that regard is fairly sparse, but once I figure it out I'll be switching not only this board but also the two other audio-related Feathers (the TunerWing and the EchoWing).

The other problem is with the connector blocks for IN and OUT. It turns out that the footprints for the TE Connectivity ones I'd chosen are, relative to the board itself, huge -- and of course they don't fit the spots that I'd set aside for them. I'm going to switch to something else -- probably Molex, though I'm still wishing to put something there that can accommodate bare wires.

On a related note, I'm switching to Kicad. Once I worked out how to install a Template for FeatherWing, I found it fairly easy to use -- not much harder than Fritzing -- and quite flexible. The files I'll be uploading in the future will be from Kicad, replacing the Fritzing ones I've been using.

I just got word that my PCB layout is all wrong: the MAX4468 and PCF8591T are the wrong size. So, I'm going to have to call the PCBs already printed up a loss, and get new ones done.

Thankfully, Riley at PCBWay is very helpful and will walk me through the process.

I'll change the files here (the JPG and the FZZ) once I've finalized what I'm sending (probably later today).

I just put in the order for the prototype for this board. With all costs, coupons, etc., the total price is $60; for a single, fully-assembled prototype (with four board-only siblings) coming from China, I don't think that price is too bad.

The bad news is that the coronavirus outbreak has caused some delays, as the production facility had to be shut down for a couple of weeks. The lead time was already kind of long (about a month, if memory serves), but I won't be expecting this until around mid-May.

I'm not too worried about that, though, since I still have to get other parts and stuff in order to test this board, and that'll take about that long. So, it all works out!

After working with Riley, my assigned customer service representative at PCBWay, I've reduced the number of assembled boards in my order from 5 to 1. This saves quite a bit off the bill; the tag is now $62. That seems like a steep price for a Feather-sized PCB, but given that it's a single prototype (and I still get 4 boards without the parts), that's not too bad. I can let out a more formal run, with a lower cost per assembled board, once I'm sure it works correctly.

Between my own finances and delays at PCBWay because of the coronavirus outbreak, I'm waiting until at least 5 March to actually pay for and order the board.

(As a side note, while I'm not clear on whether Riley is male or female, this individual is really incredibly helpful and dedicated to doing a good job. That's something I've noted with Chinese companies in general, and I've been told is simply a part of Chinese culture. Many of my fellow Americans could learn a thing or three from them in this regard. I'm not always happy with the product itself, but customer service is top-notch.)

I just got the estimate from PCBway for the new configuration: $95 for five boards. That's $19 each, which is not a bad price at all.

I don't think I'll be needing any more revisions; I just need to wait to find out what happens with the contest, and for my bank account to have that much space in it. If I end up buying them myself, it'll probably be at least February 5, and possibly as late as March or April.

(I may be wasting money by buying assembly for five boards. PCBway only makes boards in lots of 5, or I'd get just one or two. I may revise my order and get only one assembled, especially if I have to revise the board design.)

Note to self: Always double-check the pull-up resistors for the I2C bus.

(I inadvertently deleted them when I cleared out the two separate I2C chips to replace them with just one. It's fixed now.)

Working out some details, I noticed that the PCF8591T only works at the 100kHz speed on the I2C bus, and I'm not sure how well audio will transmit both ways well through the bus at that speed (especially since the device I'll mainly be using this for also uses an EchoWing, which also drives audio along the I2C bus both directions, making four signals along that bus). I took a look around for something that could go on the 400kHz speed, and found the Analog Devices AD5593R will do that.

The bad news is that the AD5593R has many bells and whistles (it's an 8-channel configurable chip, with each channel configurable to ADC, DAC, or GPIO) that I won't use, but that drive the price up another $2-3. Given that I'm trying to keep the price down, I think I'll hold off on that configuration, unless someone advises otherwise.

(I did put together the layout, though, so it'll be ready just in case.)

After the price shock I got this morning, I went on a more dedicated search for a less expensive, single-chip option for an ADC/DAC. I think I found one, with the NXP Semiconductor PCF8591T. It has four analog inputs, but it's not a problem to just tie the extra three to Ground; the $2-3 price tag also makes the board much, much less expensive, as well as much more open in terms of space.

I'm going to just hold on to this for a while before I send another quote request to PCBWay. Just from that one change, I do think I'll end up saving about $25 from that $138. I probably can also eliminate a few other components; I'll have to study the data sheets a bit more than I have. With any luck, I'll be able to bring the total with them to under $100 (meaning less than $20 per board).

Barring some calamity, I hopefully will have something truly presentable by the time a decision is made on the Take Flight With Feather contest (this is the version I'm now promoting for it -- unless, of course, it's too late to do something like that).