[Foxhood]

Took a detour from my TCB iterating efforts to experiment with sound involving one of the smaller AVR-Dx (AVR64DD28) and its built-in 10-bit DAC.

My reasoning is that if the original OPSC was a 72Mhz ARM processor juggling "6" channels of stereo, 16-bit, 44Khz PCM data from a uSD card via a SPI bus. Surely a 24Mhz AVR processor could juggle a few 8-bit mono channels. Benedini showed it was possible for AVR to do sound and that was with a older memory constrained ATmega8 type connected to slow Flash memory, no DAC and long before libraries like sdFat became the optimized beasts we know today. Only thing the AVR lacks is a DSP for floating point math so one has to cheat a little to keep mixing data within the ballpark of the ALU by using bit-shifts and multiplications for manipulating volume.

Approach is fairly simple. Two output buffers with one marked as active. A timer routine slowly goes through the active buffer pushing the values to the DAC output while the processor is prepping the other buffer. Grabbing pages of 8-bit signed 22Khz PCM via sdFat from files, manipulating volume and adding them to the buffer being worked on. Been running through benchmarks with pages of 1024 bytes and i've managed to get it to handle 9 files before it started to fail at preparing before the DAC Timer caught up to it. Which is very promising considering the old benedini only managed 2 channels and the OPSC could do 6. Honestly would have been happy with just 4 really. The main reason I'm getting this far is because this little AVR has 8KB of RAM which lets the sdFat library spend far less time seeking as it would with older ATMega.

Next is to get Logic running and test with actual soundsets rather than little test-waveform files, oh and see if i can optimize the SPI driver as currently it is using the default on within the core. Since I'm working with raw-files without a sound library i got a lot of freedom in how i deal with data. Can even do stuff like pre-processing files to make playback easy (e.g. behead them so you get RAW) and simply check the available variable to know when the last few pages are up for fading. Gonna take a lot of coding. But the progress thus far is fairly exciting.

I've also drafted a small board for it with the minimum components to see how small it actually is. It is really just the AVR, a generic LDO, uSD Slot and the MAX9768 (same as regular OPSC). BOM puts it at like 10$ plus PCB which makes it the cheapest sound-card yet, may be good for existing boards, though i am definitely looking at fitting the circuit unto my TCB Re-design as a co-processor. Perhaps even see if i can put it in charge of some 32A drivers...

[LukeZ]

Hi Fox, it's fascinating to see you working, and what is possible when a real engineer tackles a challenge. An inexpensive but capable sound card is definitely one of those things people are frequently asking for, and doesn't really exist presently. I will be following your progress here with great interest !

[IIHadesII]

Hello, I’m not a developer and I don’t really understand what you’re doing there, but it sounds very exciting. I think many people would love to have a kind of TCB sound board again (similar to the Teensy 3.2). It’s especially cool if you can define all the sounds yourself using a micro SD card.
and if its going to be cheap ( becouse i think thats what the most people need, good and cheap) than thats gonna be the perfect solution for everybody.

best regards Ilias

[Snipah]

Agreed whole heartedly, I need a few sound cards not just one so something compact, powerful, and flexible is long over due..

[jhamm]

and if its going to be cheap ( becouse i think thats what the most people need, good and cheap) than thats gonna be the perfect solution for everybody.

That's wishful thinking.
Look what happened to the Benedini sound modules.
Chinese copies caused Benedini to give up, the software is no longer being developed.
You can still get these copies for very little money...

[Foxhood]

Hello. Was a bit absent during december. Know how it goes. Lots of things to do, and good luck getting any packages around the holidays on time...
Progress was slowed down a bit because of it. But i'm back at it.

Hello, I’m not a developer and I don’t really understand what you’re doing there, but it sounds very exciting. I think many people would love to have a kind of TCB sound board again (similar to the Teensy 3.2). It’s especially cool if you can define all the sounds yourself using a micro SD card.
and if its going to be cheap ( becouse i think thats what the most people need, good and cheap) than thats gonna be the perfect solution for everybody.

best regards Ilias

The gist is that i'm trying to create something akin to the Benedini TBS in size and components, but with the same kind of functionality as the Teensy 3.2 OPSC and cheap enough in material that you could like get a bunch made at PCBway or something through the inclusion of compliant BOM and Centroid files.
Or hell. Maybe some chinese copycat starts manufacturing for even cheaper! For the sake of the Open Panzer project as a whole I'd consider that a good thing. Wouldn't mind recognition and a tip though if for some reason people do start building my offshoots ;P.

Agreed whole heartedly, I need a few sound cards not just one so something compact, powerful, and flexible is long over due..

I'll do my best. Are you looking for sound from RC inputs or OpenPanzer specifically? Cause latter is the priority at first. Former i don't have much information on what people are looking for, like channel count.
I'm trying to keep things in the realm of being possible to build oneself using a miniature reflow plate (e.g. Sequre T55 or Miniware MHP50) and compliant with assembly services.

and if its going to be cheap ( becouse i think thats what the most people need, good and cheap) than thats gonna be the perfect solution for everybody.

That's wishful thinking.
Look what happened to the Benedini sound modules.
Chinese copies caused Benedini to give up, the software is no longer being developed.
You can still get these copies for very little money...

That is kind of why I'm taking a stab at this. The only option atm for programmable sound are copycats making clone boards that are quite frank: horribly outdated and reliant on old pirated copies of what is now essentially Abandonware. An Open-source alternative is much needed. On costs: I'm avoiding stuff like double-sided components and overall complexity to push assembly costs down as much as possible. As stated above: Intent is for it to be feasible to get it however one wishes.

[Tom]

I love the original idea of the TCB sound card or something developed as it's equivalent as it has some great features that aren't available. Like track clank sounds that could scale in speed with movement etc. I'm not sure price is a huge an issue.. as to get the best quality people did buy the original Benedini and the TCB is now going for $ 175 but a great product? And of course if it is cloned and the price dropped..yes more people would buy

[Foxhood]

Milestone reached!

After hours of non-stop cacophony and trying out every single Filesystem combination possible in order to solve a specific annoying stutter (SD Cards, SdFat and filesystems be fickle).
I have successfully gotten 6 sound channels operational, mixing and getting output on a dinky speaker in good quality!! Along with repetition when needed and quick Fade-in/out. All with just the SD Card, an AVR64DD32 Microcontroller IC and a random Output Amplifier i had lying around.

This means that It looks like this crazy idea of mine is actually going to work! And possibly just as well as the original OPSC.
Now that i know that. I can start turning this from experimental to a full solution in both software (all that logic) and hardware. This is probably going to take a while though so don't hold your breath just yet.


On the Audio file specs. These are the parameters I'm currently working with and almost certainly will stick to:
=Sample-rate: 22Khz
=Bit-rate: 8-bit Signed PCM
=Channels: MONO
=Format: RAW (Header-Less)
This makes it so the processor can just grab data and immediately start multiplying based on volume and adding it in. Not as high in sample and bit rates as the original OPSC, but for mechanical sounds there is no real difference in quality honestly.

[LukeZ]

Congrats Fox on your milestone! Although I know the difference between a working experiment and a full solution is probably still 90 percent of the total effort, yet getting the experiment first to work is where the talent and genius come in, and where success is most gratifying to the engineer.

Your comment about "hours of non-stop cacophony" brought back some good (or maybe traumatic) memories.

If I recall correctly I think 8-bit sound at 22Khz is what Benedini used, and I agree that is more than adequate. We are far more constrained in terms of quality to the condition of the original recordings themselves, the relatively tiny size of the speakers that can be used in a model, poor accoustics, etc...

[Foxhood]

If I recall correctly I think 8-bit sound at 22Khz is what Benedini used, and I agree that is more than adequate. We are far more constrained in terms of quality to the condition of the original recordings themselves, the relatively tiny size of the speakers that can be used in a model, poor accoustics, etc...

Not just benedini. Pretty much everyone used that setup. Clark, Elmod, Taigen, Heng Long, games like War Thunder. 8-bit, 22Khz accross the board.
Only Open Panzer works with 16-bit. Which I can tell is mostly cause the Audio Library is rather particular about that. Like it only accepts 16-bit, 44khz for Wav files, though is a smidge more flexible with Raw and from Flash...

Ofcourse just copying what others do is bad engineering. So i also did some of my own digging into the sounds.
The Frequency Spectrum of RC sound-effects is dominated by low-frequency high-amplitude rumbling and booms, Mid-frequency mid-amplitude clangs and hisses and quite a bit of Hard-Clipping when it comes to explosions. The audible frequencies are well below the Nyquist Frequency if using 22Khz sampling (though 11Khz would be too low), while the lack of low-amplitude high-frequency audio renders no tangible benefit to using 16-bit, Not helping either is that unless you use a high quality Audio DAC you will only get an effective resolution of about 1 bit less than the DAC on chip (9-bit for AVR-Dx, 11 bit for teensy3).

From this the conclusion is that yeah: 8-bit, 22Khz is definitely the sweet-spot for this kind of stuff. Though the presence of Hard-clipping (which results in harsh-sound high-frequency odd harmonics) and the use of a on-chip DAC peripheral (processor noise) does render the use of a Low-Pass filter as probably being a good thing.


...Can you tell that I also dabble in Audio Electronics?

[LukeZ]

Yes... yes, I can. 

And that's just the person to be working on this project!