I will post the availability and price of the 3 Way DSP Amp boards here.
Read more about the boards here.
https://www.diyaudio.com/community/threads/3-way-dsp-amp.415065/#post-7734899
The price of one 3 Way DSP Amp board is $95.
The price of one I/O board is $15.
Delivery to Europe and America is at my expense. Delivery to islands or other distant places is negotiated separately.
Order more than one 3 Way DSP Amp board and get $5 off.
Payment via Payoneer.
We currently have four 3 Way DSP Amp boards and four I/O boards with ESP32 available.
Read more about the boards here.
https://www.diyaudio.com/community/threads/3-way-dsp-amp.415065/#post-7734899
The price of one 3 Way DSP Amp board is $95.
The price of one I/O board is $15.
Delivery to Europe and America is at my expense. Delivery to islands or other distant places is negotiated separately.
Order more than one 3 Way DSP Amp board and get $5 off.
Payment via Payoneer.
We currently have four 3 Way DSP Amp boards and four I/O boards with ESP32 available.
Last edited:
Hi Uriy,
How is your experience with sqeezelite-esp32?
I've used it in similar setup, however stability of the firmware was not great. It crashed on airplay/spotify streaming after pause/play/pause, crashes if you skip tracks in lms etc. I checked a year ago, maybe it has got some improvements...
How is your experience with sqeezelite-esp32?
I've used it in similar setup, however stability of the firmware was not great. It crashed on airplay/spotify streaming after pause/play/pause, crashes if you skip tracks in lms etc. I checked a year ago, maybe it has got some improvements...
The first time it was a little unusual. But when I figured out the structure of the organization of maintenance and management, it turned out to be quite a convenient thing for streaming.How is your experience with sqeezelite-esp32?
I tested the operation of sqeezelite-esp32 on the WROVER V4 board with an external antenna connected and under such conditions I did not have a single interruption in audio transmission. A couple of times the LMS server did not broadcast sound, but it was not the fault of sqeezelite-esp32, but my crooked hands when setting up the LMS. As I understand it, for stable operation of sqeezelite-esp32 it is very important to have a stable Wi-Fi connection, otherwise there will be problems.I've used it in similar setup, however stability of the firmware was not great. It crashed on airplay/spotify streaming after pause/play/pause, crashes if you skip tracks in lms etc. I checked a year ago, maybe it has got some improvements...
Very impressive! Guess I have some reading to do on sqeezelite-esp32, LMS and the phone apps. I need to verify that you can broadcast to multiple clients (i.e. powered Left/Right speakers) and they are in-synch.
LMS supports transmitting signals to several devices with sqeezelite-esp32, but LMS can only adjust the volume in the device that is selected as the main one. It is not comfortable. Also, the synchronization there is very conditional, sometimes the difference in the channels will be 0.5 ms, and sometimes 0.2 ms. I would not recommend using an LMS for stereo setup based on inconvenient volume control. The DSP itself has an SPDIF output, which allows you to organize stereo using one sqeezelite-esp32 receiver.Very impressive! Guess I have some reading to do on sqeezelite-esp32, LMS and the phone apps. I need to verify that you can broadcast to multiple clients (i.e. powered Left/Right speakers) and they are in-synch.
I would like to draw your attention to the fact that in this configuration, which I showed above in the figure, there will be an automatic switching between WiFi and SPDIF on the left speaker, priority is on WiFi, when there is audio transmission via WiFi, the board itself will select the I2S input when there is no stream data via WiFi, the board will automatically switch to SPDIF input.
Theoretically, there is no problem to use an external ADC that has an I2S or spdif output. At the same time, the adau1452 has a very good ASRC that eliminates the need to synchronize the frequency of the external ADC and the board's DAC.Theoretically it would not be a big deal to Use a ADC (inversion of DAC ) to gain analog inputs, would it?
Yes, the board itself monitors the signal at each of its outputs, and if no signal is received from the adau1452 at any of the board outputs, the board automatically switches off this channel, thereby reducing energy consumption and radiator heating. With all this, you will have a third channel in reserve for experiments, you can connect another speaker to it and transfer additional functions to it at your discretion. I know a project in which two channels were sent to a low-frequency speaker, one low-frequency speaker served the entire low-frequency band, and the second low-frequency speaker closed a narrow part of the low-frequency band in which the room suppressed the signal. That is, they removed the hole in the frequency response in the low-frequency range, and all the power of one speaker was given to this hole. I do not recommend doing this, I am simply describing a situation when the "extra" channel was given a job if desired.(somenone mentioned, that running it only with 2 channels should be no problem
What Powersupply do you recommend for the amplifier board?
This board is not particularly demanding to the power supply, the main thing is that the power supply at its output has a more or less stable voltage and can provide the peak power that you need. I tested the board on different topologies of stabilized switching power supplies, and on all power supplies the board showed itself the same. So you can use any stabilized switching power supply that you like best, the main thing is that this power supply can give the peak power that you need from the amplifier.
Thank you for the answer. I am a total layman, so please allow two follow-up questions:
- It says DSP in the title - does it have a DSP like the Minidsp integrated? If so: What software do I need for programming?
- If I need ~100W total outputpower (I recon 80W for my woofer and 20W for my tweeter): Can you recommend a specific power supply? Is this sufficient: https://meine-leds.com/Netzteil-Meanwell-NDR-120-24-DIN-Rail-Hutschiene-24V-120W ?
Yes, the board has a DSP of the ADAU1452 type. To program it, you will need a programmer called USBi and free software sigma studio.If so: What software do I need for programming?
You can take a ready-made programmer. USBi Aliexpres or USBi
Or make it based on CY7C68013A board like here. https://www.diyaudio.com/community/...rammer-using-cypress-cy7c68013a-board.269111/
Sigma studios can be taken here https://www.analog.com/en/resources...e/software/ss_sigst_02.html#software-overview.
Ok. If you need the full power of the channel, then 32V 5A for each board will be enough for you. - AIMA 32V 5ACan you recommend a specific power supply?
You can take one for 36V, it has output voltage regulation of +/- 10%, i.e. 36V can be reduced to 32.4V, which is quite acceptable for the board. - 36V 5A
Of course, you can use a 24V 5A power supply, but the maximum power you can get on a 4 Ohm load will be 50W per channel. Which is quite sufficient for home use of the speaker system in the near field, and at 24V the board will heat up less.
Regarding Sigma Studio, read this topic, it will be a little clearer.
https://www.diyaudio.com/community/...ac-with-dsp-adau1452-100w-per-channel.410804/
This is an interesting module. Especially the direct I2S to PWM conversion.
How did you do that? Did you implement that on your owm or is that a preprogrammed FPGA?
Just wondering how that is possible for the low selling price?
Which PWM resolution do you achieve?
I suspect the limited THD+N and SNR performance (in comparison to really good class-d amps a few dB lower) are caused by the TI amp chip. Did you evaluate what the bootleneck is?
How did you do that? Did you implement that on your owm or is that a preprogrammed FPGA?
Just wondering how that is possible for the low selling price?
Which PWM resolution do you achieve?
I suspect the limited THD+N and SNR performance (in comparison to really good class-d amps a few dB lower) are caused by the TI amp chip. Did you evaluate what the bootleneck is?
With great difficulty. But it was very exciting.How did you do that?
This is a program of our own development that configures the FPGA to the required configuration. Two people worked on this power DAC, I was responsible for the hardware, and the second person was responsible for the mathematics. Unfortunately, the second developer suddenly died of covid.Did you implement that on your owm or is that a preprogrammed FPGA?
The result that we were able to achieve is quite acceptable, so as not to hide it in a distant drawer.
It depends on what you mean by PWM resolution, to be honest I didn't measure PWM resolution, nonlinear distortions that I get mostly depend on LPF distortions, since it is not covered by feedback. I have a stereo amplifier implemented according to this principle, at its output I measure 0.003% at its maximum power of 50W on a 4 Ohm load. The PWM frequency in this DAC is 580 kHz.Which PWM resolution do you archieve?
There are pictures here. https://www.ebay.com/itm/395385503196
Let's say that these are the best indicators that I was able to record through measurements.
No, it's much more complicated there. Getting SNR 107 dB(A) in this approach is not a trivial task at all.I suspect the limited THD+N and SNR performance (in comparison of really good class-d amps a few dB lower) are caused by the TI amp chip.
Yes, you can.Hmmmmm would it be possible to use this one as amp for a 2.1 loudspeaker combination ?
Each channel supports the entire frequency band, so there are no problems with organizing a 2+1 system.
In post number one, a folder with projects in Sigma Studio is attached, in this folder there is a DSP setting for stereo plus LOW, the file is called I2S 2+1 - 48kHz . By default, each board is configured for 2+1. And only by agreement can I flash another project from the previously specified folder.
https://www.diyaudio.com/community/threads/3-way-dsp-amp.415065/#post-7734938

One way or another, it should be understood that this is a basic 2+1 project, and to get a positive result, you will have to independently adjust the gain and cutoff frequency in the low-frequency channel. That is, for a positive result, you need at least a USBi programmer, and it is also desirable to have a microphone, although in the low-frequency range you can use a microphone in smartphone, but this will be more difficult than using REW.
The number of steps which could be resolved by the pwm duty cycle.It depends on what you mean by PWM resolution, to be honest I didn't measure PWM resolution
If you have for example a quite common pwm frequency of 384 kHz (8x oversampling at 48 kHz) and want to get a resolution of 16 bit (65536 steps), you would need a base clock of 384000 * 65536 = 25.17 GHz for the pwm module.
If you want to do 24 Bit, you would get in the terra herz range.
That's why I'm asking how you implemented that. It's the most interesting part of the amp module.
It's definitely not trivial.No, it's much more complicated there. Getting SNR 107 dB(A) in this approach is not a trivial task at all.
This value is quite difficult to determine in my case. I did not have a direct goal to know the bit depth of the resulting reference PWM. My goal was to get the lowest possible noise level and the lowest possible level of nonlinear distortion at the output of the power DAC. At the same time, I knew that the distortions introduced by the PCM to PWM conversion algorithm are a couple of orders of magnitude lower than the distortions introduced by the power part of the DAC, since I had the results of simulating the mathematical model of the conversion.The number of steps which could be resolved by the pwm duty cycle.
A 24-bit stream was chosen for a sure margin when adjusting the volume in the algorithm, plus ASRC itself allowed this to be done without any special costs, so a 24-bit word was chosen to transfer data to the algorithm. Moreover, since the volume is not regulated by the method of cutting off bits, there was no urgent need for 24 bits, ASRC could simply issue 24 bits. In this board, the volume is not adjustable, but in the original version, the volume was adjusted in the algorithm using a valcoder.
As you yourself understand, there is no point in me describing the method by which the problem of converting PCM to PWM was solved.
At first glance, this is true, but when you start implementing such an idea in hardware, you learn that the stability of the power supply voltage and the nonlinearity of the PWM amplifiers negate all the original beauty of the reference PWM. The second, no less significant breakthrough was achieved when we understood how to organize feedback on PWM. And only when these two conditions are met can we obtain the values that I received. Reference PWM without the ability to implement feedback is practically useless.That's why I'm asking how you implemented that. It's the most interesting part of the amp module.
Last edited:
Squeezelite-ESP32 fan here, and I was wondering if you could shoot down this idea. I have WROVER's and am familiar with configuring I2S on the chips. Instead of using your input boards, could one use one external ESP32, running Squeezelite-ESP32, and connect the I2S output pins to two of your amp boards? I've never tried to connect one GPIO to two separate input pins.
Seems like this would eliminate the latency issues when syncing two WROVER chips using LMS???
I could also configure the ESP32's digital output as S/PDIF and do the same thing???
Seems like this would eliminate the latency issues when syncing two WROVER chips using LMS???
I could also configure the ESP32's digital output as S/PDIF and do the same thing???