Let's talk of possible new accessory boards

I know it's not a board, but TBH, selling Teensies with PSRAM already attached, simple as it is, would be a huge win for those of us with shaky hands
As Paul has mentioned, Proto Supplies sells Teensy 4.1 with one or both memory slots soldered in. I've used them for the last couple of Teensy 4.1's that I've bought. While I can still solder through hole stuff, SMT soldering has become problematical (I've been having some muscle failure that I'm undergoing tests with the doctor that makes it harder to hold things with fine motor control).

You can choose:
  • 1 PSram chip soldered on
  • 2 PSram chips soldered on
  • 1 PSram and 1 flash chip soldered on (flash can be 16MB, 128MB, or 256MB)
 
While I can still solder through hole stuff, SMT soldering has become problematical (I've been having some muscle failure that I'm undergoing tests with the doctor that makes it harder to hold things with fine motor control).

Hi Mike

Sorry to hear that - but I know the feeling about not wanting to solder those small parts anymore. The eyes aren't the same, not as steady and not enough patience anymore.

Hope everything turns out ok.
 
For anyone who missed it, there's been a thread on larger PSRAM parts. These are 16Mbytes each, and seem likely to be the biggest that will ever be manufactured. The code to support these is in cores; associated changes are required in LittleFS, but are believed to be OK and just need Paul's approval and pulling in (possibly with minor tweaks).

All the signs are that Proto Supplies will be supporting these as soon as stocks become available.
 
Delta-sigma. 860KSPS, 16 bit, SNR about 90dB. In that range (speed and precision), why not use a SAR?
ADS8885 perhaps, 18bit, differential input, 400kSPS, settle to 18 bit in 1.2us (basically no latency), SNR 100dB, THD -115dB, 3.3V powered but admits upto 5V reference...

BTW ADS1115 is only 860SPS, not 860kSPS !
 
Last edited:
What about something for industrial applications:
- 24V power input, (12-36 volts ideally) with lots of protection and filtering.
- RS485, with selectable 120ohms termination, for modbus
- CAN, with selectable 120ohms termination, for CANOPEN
- 24V I/O using ST chips
- connectors for stepper motor driver.
- connector for DC motor driver.
- connectors for encoders.
- connector for TFT display (ILI9341, ST7789,...)
- connector for touch
- connectors for remaining serial, SPI, I2C interfaces
- Ethernet

But mainly, a board where connector placement is not random, in order to fit a enclosure with rear connectors and front TFT/buttons/led/encoder
Perhaps add a couple of SSR's or physical relays, LEDs on every I/O pin, replace the plethora of connectors for a universal expansion connector bus? Opto-coupled digital I/O for voltage-agnosticism? A lot of industrial sensors are 10V or 20mA current loop too - so many possibilities alas.
 
Perhaps add a couple of SSR's or physical relays, LEDs on every I/O pin, replace the plethora of connectors for a universal expansion connector bus? Opto-coupled digital I/O for voltage-agnosticism? A lot of industrial sensors are 10V or 20mA current loop too - so many possibilities alas.

Seems like a lot of the suggestions are application-specific "motherboards" that one might design for use with Teensy 4.x or Micromod. As a software person, I don't understand why there is not a variety of open source hardware in the same way that there is a vast array of freely-available software libraries.

Can someone say briefly why there aren't a zillion application-specific boards, with all of the files necessary to have them built by a PCB vendor, from which one could choose to use as-is, or make small changes for one's own requirements?
 
There probably are quite a few, but there are perhaps slightly more barriers to open hardware than there are to open software.
  • you can't just download a bare PCB, or an assembled one. Someone has to make them, and hold stock
  • even if the (Gerber) files are available, you have to have them made and possibly populated. The cost for this soon mounts up: it's either stupidly expensive for a few examples, or you end up paying a sensible unit cost, but for way more units than you actually wanted
  • you can't just change a few lines on hardware the way you can with software
  • often they're so application-specific that they're no use to you - even trival things like being the wrong shape, or more important ones like using pins you reserved for some other use
I imagine that's the reason there are so many single-function breakout boards on the market. Makes it easier to prototype, then you can make your own open-source application-specific hardware ... and put a few up on Tindie ... and sell hardly any ... and it goes out of production, or near enough.

It would be great if the usual suspects had an "open hardware corner", where they can stock and sell open-source hardware on consignment, and without support. It probably doesn't make business sense, though.
 
While I can still solder through hole stuff, SMT soldering has become problematical
Like others have mentioned, sorry to hear that. Hope all works out.

I know the feeling! I have never been overly coordinated and eyes, hands, patience ... Are not what they used to be.
Not that I was ever overly coordinated.

As my brother used to kid me, by saying that I was ambidextrous, equally uncoordinated with both hands.
 
Isn’t that “ambisinistrous”?!

I’m lucky enough to have been blessed with a steady hand and short sight (about -5.5 on average), makes small assembly jobs much easier. I’m not coordinated at higher speeds, though - rubbish at sports and forbidden to dance by my better half :giggle:
 
As I have almost no soldering skills :

A small "low height" board with at least 1 or 2 qwiic connectors and quite a few pins on different jst connectors. Adding a circuit so that we can program a keep powerbank alive cycle would be nice (mosfet + a few resistors). Also adding a "cheap" flash memory chip + spi display connector could be a bonus.

This can be a small micromod carrier board if that's lower height.
 
I have almost no soldering skills, but I married well, so he can solder - it was a condition of our marriage (kidding). I also have a Hakko.

That said, even with my resources, I've found myself gravitating towards kits with integrated screens and peripherals. That leads me to a lot of ESP32s and is the primary reason I like them. The freenove ESP32S3 Development Kit is a great example of one of these kits, and having a small kit like this would be pretty cool for the teensy. I saw a larger one on this site some time ago, but this is more compact, and kind of purpose built for multimedia.
FreenoveEsp32S3Devkit.jpg
 
Paul,

If you do go with a higher quality audio IC, I know that there are many choices.

For software support, my work with Teensy-based derivatives like Tympan mean that I have working, highly exercised, Teensy-Audio-style control classes for the TI AIC3206 and AIC3212.

I'm sure that others here also have AICs that they've developed control classes for.

Either way, you gave many good choices.

Hitting the rice point that you need to hit, that's the harder part. The AIC3206 and 3212 are probably too expensive.

(I'd be happy with a MIDI + Qwiic + LiPo shield!)

Chip
 
My main thoughts on a higher quality audio board revolve around ground isolation. I really do believe even with SGTL5000, ground loops are our main enemy of quality sound, not the codec's specs. My gut feeling is a codec with better specs would be pointless for most use cases if we keep the connection to Teensy's GND.

Optical or other isolation for the I2S signals will probably add more cost than the codec chip. How the isolated audio circuitry will get power is a good question. Adding a small DC-DC switcher, and plenty of filtering for its ripple+noise, will add even more expense. Maybe instead of all that cost the product would have jumpers to connect Teensy's GND and Vin/Vusb, and a barrel jack or other connector for external "clean" power and a diode to keep it from flowing back into Teensy if plugged in while the jumpers are still in place. I guess the big question is what can we do to keep cost "reasonable" and still deliver excellent quality audio for people who just plug it in and use it?

Choice of codec versus cost is also tricky. I'm seeing about $2 in volume for AIC3206, which doesn't seem too bad. But it's specs aren't going to impress audio specs enthusiasts, not even 100dB SNR on the ADC, and THD+N at only -87dB (or -70dB worst case). Then again, maybe nothing will impress some folks? (but my guess is some people who obsess much about specs probably worry little about practical problems like ground loops) I really haven't spent much time shopping for codec chips lately. Perhaps by the time (if ever) this board were to move forward, newer chips will become available? Anyway, my general though it I'd really like to see THD+N below -96dB and SNR well above 100dB so we can at least credibly say "better than 16 bits", even if that costs more.

Which specific codec chip, I really don't have strong feelings. My main concern is the practical usage stuff like ground loops and clean power so most people have a good chance of actually getting the rated specs when they actually put it to use in their system which we can't know in advance.
 
My main thoughts on a higher quality audio board revolve around ground isolation. I really do believe even with SGTL5000, ground loops are our main enemy of quality sound, not the codec's specs. My gut feeling is a codec with better specs would be pointless for most use cases if we keep the connection to Teensy's GND.
In my opinion that needs an urgent solution because the SGTL5000 of course producing data noise during signal processing and SPI use. What's about some kind of HF-Shield sheet around the focused circuit on the Audio Shield - Is it to realize? Has anyone mentioned this yet?

And @PaulStoffregen
Please forgive me this step to ask right here, but I need to talk to you urgent, it's about the GamesCom Cologne. Can we pitch it in a private Conversation?
 
Paul,

I agree that the current board design is likely underperforming, given the specs of the SGTL5000. I don't know why.

For our initial 3206 design, we did not have to take any of the special isolation steps that you mentioned, yet we still got decent performance.

Here's the version compatible with Teensy 3.6: https://github.com/Tympan/Tympan_Rev_C_Hardware

And here's a version compatible with Teensy 4.1: https://github.com/Tympan/Tympan_Rev_F_Hardware

As you can see, I don't believe that we did anything special for isolation. But, unlike the Teensy Audio Board, we can easily connect mics or line-level sources via our pink "mic" jack and the audio sounds fine. We don't get the ground-related buzz that novice users of the Teensy Audio board often encounter.

Perhaps the difference is simply that we use standard 3.5mm jacks for audio connections (for both input and output) and not the 0.1" headers used on the Teensy Audio board. By using standard audio jacks for our analog connections, the user always gets the correct signal lines and the correct grounds. Perhaps this leads to more reliable, cleaner connections?

Or, maybe the difference is that our device always has a battery attached. Perhaps the battery has the side effects that, even when attached to USB, the system still has a nice reservoir for smoothing power, thereby making for cleaner audio? (That seems more far-fetched)

Thank you so much considering these issues. Such consideration and such open discussion is what separated Teensy from so many others.

Chip
 
Last edited:
Paul, your comment about the AIC3206 not being "exciting enough" is probably true. It is a step up from the STGL5000 in that the AIC3206 can run at 96 kHz. It can also be an interface for PDM mics, which can radically improve one's noise floor. I use both features regularly and they're important.

(Remember, if you run at 96kHz, you get access to ultrasound up to the 40 kHz range. This isn't about audiophile cork-sniffing...there's so much fun that can be had with ultrasound: listen to animals! Listen to corduroy pants! Do weird ultrasound-based communication! Implement your own fancy distance sensing! It opens up so much nerdy fun)

But, other than its support for 96kHz and for PDM mics, yeah, the 3206 isn't especially impressive on paper. It'd be better if its ADC noise floor was another 5 dB lower. Or, if it was a 4-cannel chip.

Whatever IC you choose, please do allow for 96 kHz and for connecting PDM mics. For your board to support PDM mics, I'm *not* asking for you to include connectors ($$); these are digital mics, so just having through-holes for regular 0.1" headers is fine (power, clock, data, ground).

Chip
 
Last edited:
My assessment is this:

The audio adaptor doesn't have a separate clean (linear regulator) on its digital supply and a properly separate analog ground, so it picks up digital hash from the T4 processor driectly from the 3V3 and GND pins (the T4 is a _very_ noisy chip running at 600MHz and 100mA current levels, its all over the shop if you probe it with an SA).

This is made worse by not having a ground plane on the audio adapter so digital return currents may couple to the analog lines directly - the audio analog lines run through the I2S part of the headers.

Note that a lot of the high frequency digital hash will just be an RF annoyance, its the lower frequency components that fall in the audio band (and probably up in the over-sampling bandwidth of the ADC) that can impose on the SGTL5000 audio, especially if they get on the analog ground (pin 7 of the SGTL5000, which curiously is not shown in the audio adapter schematic - you have to look at the board to figure out how its connected).

The lower frequency components, especially currents in the ground, will directly add to analog paths.

There is no bulk decoupling of the 3.3V supply suitable for audio frequencies. In fact that would be prohibitively bulky I suspect, but its worth trying 100 to 1000uF as mitigation if you have audio adapter noise issues.

The trace carrying ground to the microSD card socket has significant high-impedance length shared with the AGND (pin 7 on SGTL5000), so the very heavy switching currents from an SDcard are directly imposed on the SGTL analog ground pin (SDcards can pull over 100mA in bursty patterns). A ground plane would greatly help with this to reduce the high inductance and relatively high resistance of this ground trace. The current to the SD card should not be routed anywhere near the SGTL chip!

I think this is the weakest point of the design - a microSD card in use will be pretty audible as far as I can tell looking at the layout. Certainly if you use the mic input...

The classic approach to mixed signal design would be full EMI filtering of the __5V__ and GND coming in, then two LDO regulators for SGTL (1.8 and 3.3) digital and analog supplies, and all completely separate from the SDcard and memory chip (which would be on a separate piece of groundplane). Crucial is avoiding sharing power and especially ground traces or planes for the SGTL5000 with anything else. Then IR voltages on the lines from the SDcard won't appear in series with the analog signals...
 
The lower frequency components, especially currents in the ground, will directly add to analog paths.

When/if a higher quality audio board is in the design phase, you'll be on my short list of people to evaluate how well prototypes actually perform.

And generally speaking, yes, I agree with your analysis. For a future board designed with these issues in mind, my hope is we'll have plenty of time to test several prototypes and really refine it so the end result is as good and free of interference as possible.
 
A different sort of module but I think this new Raspberry Pi Radio module is interesting from a couple of angles:
- could an adapter board be developed and integrated for teensy 4
- the module approach to ease certification is interesting for those considering making and selling in lower quantities
- the castellated pcb connections also interesting for low volume manufacture and diy
 
I favour the small ESP32 varieties, they offer the very useful ESP_NOW capabilities. Send data from ESP to ESP and it just works with built in handshaking. If data get's to target ESP32 then ok returned else if it doesn't false returned.
See here. They are the same width as a Teensy.
z.png
 
new Raspberry Pi Radio module is interesting from a couple of angles
This module already in use as:

SparkFun Thing Plus - RP2350 : https://www.sparkfun.com/sparkfun-thing-plus-rp2350.html

... this board includes their radio module for single-band 2.4 GHz WiFi 4 (802.11n) and Bluetooth® 5.2 ... The Raspberry Pi Radio Module (RM2) operates over an SPI interface,
---
https://datasheets.raspberrypi.com/rm2/rm2-product-brief.pdf : uses Infineon’s CYW43439 [ 1YN ]
... Supports 20 MHz channels with data rates up to 96 Mbps (PHY rate)
... Raspberry Pi Radio Module 2 will remain in production until at least January 2036

1DX uses the CYW4343W :: Same as in Arduino GIGA
while 1YN uses CYW43439 > This is the other chip on hand for the post GIGA recommended chip
 
It seems that there are two competing scenarios when it comes to WiFi and Bluetooth.
1) The fastest possible speed used with web pages, etc. Can be complicated and expensive for best performance.​
2) Simple non complicated safe communications for control purposes. For safe control needs to be simple and preferably low cost.​
I fall into the latter category.

I have a system turning on a Remote Radiator Valve which is controlled by ONE pin on the Teensy using an ESP32C3 for communication.
The process is:
1) ESP32C3 is normally in a sleep (dormant) state.​
2) Teensy asserts a pin to wake ESP32C3.​
3) After 20mS Teensy maintains the pin asserted to tell ESP32C3 to turn remote valve ON, or resets the pin to tell ESP32C3 to turn remote valve OFF.​
4) The ESP32C3 goes back to sleep upon completing it's task.​
5) Should an error occur (Unable to communicate with remote Sonoff ESP, for example, the ESP32C3 sets an error flag and waits Uart communication from Teensy.​
A Sonoff Basic WiFi relay can be had for under $5 and the ESP32C3 to interface the Teensy to WiFi and controll the Sonoff is about $2,50 (I say about because the Dollar is drifting DOWN).​
So for $7.50 the Teensy is able to control (On/Off) a Mains electricity device using WiFi.​
If only we had a Teensy - ESP32 library. I know @defragster has written some code to program an ESP32 using a Teensy as a write through device.​
It would be nice if we could re-program the ESP32C3 on the fly. In initial mode it would have automatic pairing capabilities, when it is paired ok, that part of the program could be "thrown away" and the main running code programmed into it.​
 
Back
Top