Hey guys! I'm Jos, an embedded hardware engineer from The Netherlands. I am into vintage audio gear by hobby, but also really like a modern smart home. Right now I can turn on my audio setup with a smart power outlet, but I don't have many means to change the volume or input. It just feels wrong to get a modern surround processor connected to classic systems.
Here is the solution I came up with:
In summary:
The idea is to use a TI TPS65130 symmetrical switching converter to change the 5V USB input to a +/-12V output which is cleaned up by a TPS7A49/TPS7A30 low noise LDOs for super clean +/-10V for the opamps. We can then steal the 12V for the trigger output. The remaining 3.3V we can take from the ESP32 module.
The end result is this circuit board:
A mockup for a 3D printable enclosure is also getting closer:
Finally, I have a lot of code to learn. ESP-IDF with the LVGL graphic library got me this for now:
I am gonna pull the trigger and order the PCB next weekend or so. I have attached the schematic. Does anyone have any feedback or ideas for now?
Here is the solution I came up with:
In summary:
- Full analog audio path with opamp amplification and a relay based attenuator.
- ESP32 DEVKITC module for relay control and wireless MQTT integration with Home Assistant server.
- Every channel wired up to ESP32 ADCs to switch output based on what input is playing.
- 12V Trigger output for semi modern power amps.
- Two rotary encoders and a round smart watch IPS screen as the user interface.
- Neopixel output for good measure.
- USB C power input with switching converter and linear regulation.
The idea is to use a TI TPS65130 symmetrical switching converter to change the 5V USB input to a +/-12V output which is cleaned up by a TPS7A49/TPS7A30 low noise LDOs for super clean +/-10V for the opamps. We can then steal the 12V for the trigger output. The remaining 3.3V we can take from the ESP32 module.
The end result is this circuit board:
A mockup for a 3D printable enclosure is also getting closer:
Finally, I have a lot of code to learn. ESP-IDF with the LVGL graphic library got me this for now:
I am gonna pull the trigger and order the PCB next weekend or so. I have attached the schematic. Does anyone have any feedback or ideas for now?
You are a talented guy but why use those dreadful (crackling, bad contacts. L/R crosstalk etc.) 3.5 mm input/output jacks and not the better and standard RCA connectors? RCA are on vintage gear and modern devices so they must do something good.
And a switcher feeding a switcher feeding 2 linear regulators. If you add pads for 4 diodes and 2 caps you can still move over to a small transformer. Recently I got into the same regulators and found out they do not output the exact same voltages when using identical valued setting resistors. You could add a pad for a parallel resistor to make them the same voltage. Or a potentiometer.
Outputs have a muting relay but the contacts are in the signal path. Changing it to "shorting outputs to GND" will make the relay contacts non intrusive to signal quality.
And a switcher feeding a switcher feeding 2 linear regulators. If you add pads for 4 diodes and 2 caps you can still move over to a small transformer. Recently I got into the same regulators and found out they do not output the exact same voltages when using identical valued setting resistors. You could add a pad for a parallel resistor to make them the same voltage. Or a potentiometer.
Outputs have a muting relay but the contacts are in the signal path. Changing it to "shorting outputs to GND" will make the relay contacts non intrusive to signal quality.
Last edited:
Thanks for the reply, Jean Paul.
The "shorting outputs to GND" idea is great. I will design that one in! The 3.5mm jacks and switcher are indeed a design of compromise...
I'm really looking for a single board solution, so a 4 layer board is necessary if I want any kind of proper grounding scheme. At JLCPCB the 4-layer PCB prices increase an insane amount if you go over 10cmx10cm. So limited by that size, I would have to give up one or two inputs in order to fit RCA connectors.
With the 3D printed case, a fully integrated linear power supply is just not really safe to do for thermal reasons. It would also get complication to get the seperate 12V@100ma for the trigger out.
I will definitely use an active probe with my oscillocope to measure the power supply and post back the noise results though.
The "shorting outputs to GND" idea is great. I will design that one in! The 3.5mm jacks and switcher are indeed a design of compromise...
I'm really looking for a single board solution, so a 4 layer board is necessary if I want any kind of proper grounding scheme. At JLCPCB the 4-layer PCB prices increase an insane amount if you go over 10cmx10cm. So limited by that size, I would have to give up one or two inputs in order to fit RCA connectors.
With the 3D printed case, a fully integrated linear power supply is just not really safe to do for thermal reasons. It would also get complication to get the seperate 12V@100ma for the trigger out.
I will definitely use an active probe with my oscillocope to measure the power supply and post back the noise results though.
Power consumption will be very low, likely below 3VA. If you add the pads you can use an off board 5…10VA transformer in case the switchers do what they nearly always do.
You’ll regret the 3.5 mm jacks for sure. It is the micro USB of analog audio.
You’ll regret the 3.5 mm jacks for sure. It is the micro USB of analog audio.
There are smd test pads on every power supply rail. If all fails, I can inject power from a linear supply there.
Even at 70mm, RCA won't fit. So that leaves the choice between giving up an input channel or leaving the jacks..
Even at 70mm, RCA won't fit. So that leaves the choice between giving up an input channel or leaving the jacks..
Everyone can get around with 3 inputs 🙂 Some get around with just 1 set of inputs.
Better make something really good and no regrets because of design choices. It will be expensive anyway.
BTW I think you overestimate the loss to heat. If that device needs so many cooling slots it identifies as power amplifier.
Better make something really good and no regrets because of design choices. It will be expensive anyway.
BTW I think you overestimate the loss to heat. If that device needs so many cooling slots it identifies as power amplifier.
Interesting design, I'll follow this thread 🙂
@jan.didden made one like this, you can see it here: https://diyaudiostore.com/products/linear-audio-silent-switcher?variant=48030808260
@jan.didden made one like this, you can see it here: https://diyaudiostore.com/products/linear-audio-silent-switcher?variant=48030808260
I finished the design and ordered the PCB from JLC. It came in way sooner than the 20 business days promised by the cheapest shipped method I selected.
11 euros shipped for 5pcs 4 layer 10x10cm PCB is crazy cheap. Just don't go over that size or the price explodes...
I will be slowly soldering and validating this board in my work breaks the coming work week or two. Also gotta prime the 'ol 3D printer for a case prototype. Software wise, the rotary encoders and graphics are working. Gotta spend a weekend designing a gui.
Watch this space😎
11 euros shipped for 5pcs 4 layer 10x10cm PCB is crazy cheap. Just don't go over that size or the price explodes...
I will be slowly soldering and validating this board in my work breaks the coming work week or two. Also gotta prime the 'ol 3D printer for a case prototype. Software wise, the rotary encoders and graphics are working. Gotta spend a weekend designing a gui.
Watch this space😎
As far as I can read the LM4562 datasheet, this SNR difference only comes into play at around the 5V RMS mark.
That 14V pk-pk. No power amp input is gonna come close to redlining this thing even at +-12V supply.
Interesting project. Watch out for the ESP32 which generate electronic noise when maintaining the WiFi connection. The ESP32 and other MCUs don't have dedicated analog grounds so the internal ADCs are noisy. External ADCs are needed if accuracy and reliability are important. A DS1804-50 would work for volume control and it can be set by simple code from the MCU.
Yes , ESP 32 generates lots of noise. best to have it on it's own board , separate from the analog. I (am !!) / would build everything linear and port
(an analog) 5V to a separate board for the ESP. A little 30-50VA toroid/caps don't take much more room than your typical 3A china smps.
PS- really noisy 0V (below).
(an analog) 5V to a separate board for the ESP. A little 30-50VA toroid/caps don't take much more room than your typical 3A china smps.
PS- really noisy 0V (below).
Attachments
It is likely only a 10VA that is needed (or even smaller). If that ESP32 feeds through its garbage on the 5V line then it makes even more sense to have a transformer based PSU with a dedicated 5V regulator and ferrite beads/CLC etc. for the ESP32.
Last edited:
It would be worth considering shielding the ESP in a metal case and making the antenna external. Maybe a PIC would be better, but then you don't have wifi.
You could even truly isolate the I2C (opto) from the analog. The 5V feeds an LT1117 (3.3v) on the devC board. Doubt the noise could "backtrack"
through 2 regulators. I'm using super Jung analogs for my +/- 12V. Nearly noiseless. Why would I have an ESP anywhere near those ??
I just use a 7805 hooked to the unregulated supply for a separate ESP circuit.
through 2 regulators. I'm using super Jung analogs for my +/- 12V. Nearly noiseless. Why would I have an ESP anywhere near those ??
I just use a 7805 hooked to the unregulated supply for a separate ESP circuit.