• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DIY WiSA Home Theater system

yes. ffmpeg send data to cavernpipe, can see processing, then how to get pcm?
 
yes. ffmpeg send data to cavernpipe, can see processing, then how to get pcm?
FFmpeg can't send data to CavernPipe, as it's a custom protocol. You will get an error if you try that. It needs implementation in players.
 
FFmpeg can't send data to CavernPipe, as it's a custom protocol. You will get an error if you try that. It needs implementation in players.
Hello @VoidX ,

Could you clarify this part ?
As part of my pipeline setup, I have been trying to forward RAW AC3 /E-AC3 signal to CavernPipe using FFmpeg. Am I missing something?
For example AC3:
[CavernPipeClient] Connecting to CavernPipe...
[CavernPipeClient] Output: 2ch, 48000Hz, 16-bit
[CavernPipeClient] Found pipe at: /var/folders/0l/j482c15x3wscjgfssb87dyt00000gn/T/CoreFxPipe_CavernPipe
[CavernPipeClient] Connected to CavernPipe.
[CavernPipeClient] Connected to CavernPipe.
[CavernPipeClient] Handshake sent.
[CavernPipeClient] First input chunk: 4096 bytes
[CavernPipeClient] First 16 bytes: 0B-77-E5-35-64-40-2F-84-2B-E2-1B-E2-B9-0D-F5-77
 
Some updates for the thread:
- Thanks to Cavern, I have implemented a Codec for MPV which relies on CavernPipe and now render Dolby AC3, EAC3 and EAC3-JOC for Atmos but unfortunately not TrueHD. This works for both file based and streaming files from online !
MPV with Cavern codec

1775372887805.png


On the other side, I have also received my Class D Amp based on TAS5828M (see discussion here) which is not my own design shared previously but pretty close !
I am quite impressed by the outpout reaching up to 90dB in stereo with only 12V input (optimal input is 12-24V, max. is 26.5V)

The next step is the wireless part Snapserver to Snapclient ESP32-C5 and I2S to the Class D amp which I already have some draft implementation !
 

Attachments

  • 1775373379540.png
    1775373379540.png
    465.4 KB · Views: 51
Last edited:
Thanks, I have PuraPath as well. I would not have managed to create tas5805m driver without it. I will create an equivalent driver for the tas5825/28 as well, hopefully somewhere before the end of this year. I will publish a new board based on the TAS5825M. Reach out when you're ready with the software part, I'll do my best to help with hardware
View attachment 474007
@sonocotta and @Glider95 - if y'all are using the ESP32, be advised that Espressif strongly urges "keep outs" around the module's PCB antenna.

1781044972427.png
 
Thank you @Pete Lee.
Pitty to see that the technology will be open only to manufacturers :/

With the increasing DIY Speaker projects and the new design appearing every weeks (partially thanks to 3D Printing ! ), an open to public wireless high quality audio transmission would have been great ;D.

On my side, I will try to continue to explore the use of the older WiSA HT (but same issue in term of board procurement, only available from scrapped products) or look into the new ESP32-C5 which onboards 5GhZ WiFI to create an equivalent of what WiSA E is achieving... (but that's huge amount of work).
A combination of these technologies + Cavern + Class D amps would be amazing !

@Glider95 - Okay, access to WiSA E Tx / Rx might become a thing for the DIY community - I'll keep you and others at ASR informed if we come up with an alternative plan that I think people are going to like. I will be the first to say that this is not the easiest thing we've tried to figure out...

A bit more on the WiSA E module lineup here: https://www.wisatechnologies.com/wisa-at-ces-2026

As someone who has had a long-running interest in DIY, amateur radio, and open source, I very much want to support experimenters of all stripes!
 
I regret not doing more thorough research before starting the WiSA project, I believe almost similar flow could be achieved with a combination of all these projects. Only one points will never beat WiSA is the latency: a combination of the all above will never reach <50ms (more like 200ms) but is it actually important today where streaming services or Media center always include a video delay feature ?

All the same, I applaud your efforts on this: we have patents and trade secrets around how we accomplish this - and let me tell ya, it is NOT easy at all.

BTW, sorry for the long absence: I needed to take a pause from the company to take care of my dad's health matters: I returned a few months ago. It really is a joy to be back with my engineering team!
 
I almost did not see your messages @Pete Lee as I did not receive any emails.

I am thrilled to read all the news you brought from WiSA / Data Vault - both the possible opening to the DIY community but as well the line-up of new WiSA products, it looks very exciting !
I still have WiSA SoundSend so if the new line-up is compatible with it, I am almost ready to adopt ahah
I am quite "proud/happy" to see that we went down the same route with receivers based on ESP32-C5 !
1781532377276.png

Note: Is it a typo or there is WiSA E Falcon products with different specs?

I have also slow down due to work obligations, and just working on this project here and there.... But with the latest AI model for codings, it really helps to do that on the side by small chunks :)
 
@Glider95 Unfortunately, the new lineup won't be compatible with the old SoundSend: that's HT only. There's a new SoundSend E.

In any event, we are noodling on selling some surplus WiSA HT transmitters (USB dongle) and how we can provide some amount of support for those with substantial engineering experience.

One of our principal engineers on HT noted that two USB transmitters can be combined to do a 9.1.6 system (which is interesting, see pics below)...

1781544214290.png
1781544186883.png
 
Back
Top Bottom