TL;DR: A friend and I are developing a DSP amp and are seeking feedback on our proof of concept.
Note to admins: This post is not intended for commercial purposes. I have deliberately avoided advertising any brand or product names. The sole purpose is to gather feedback for our feasibility study on a project that may potentially become commercial in the future.
Dear fellow engineers and enthusiasts,
This is a prototype of our audio amplifier, which is based on the ADAU1452 DSP and two TAS3251 chips in the output stage. This setup provides a total output of 4x175W (or 2x350W) at 36VDC.
After several months of development, we now face some critical decisions, the most important being the viability of this device for mass production and the potential features to add or remove since this proof of concept.
This version is not the final iteration; we are still optimizing certain parameters. Most notably, the THD+N currently measures at 0.08% at 100W into 4Ohm and 0.008% at 10W into 4Ohm, with an A-weighted noise floor around 105uV. While these numbers are quite good, they do not fully match the TAS3251 datasheet. We are continuing to optimize the PCB layout, LC filter components, and PFFB.
We aim to provide plug-and-play functionality for this device, meaning users will not need to tweak the ADAU DSP and can select one of the pre-made use cases that are easily configurable and controlled via a user interface. For more advanced users, full access to the ADAU DSP will be available using SigmaStudio.
The main firmware, based on FreeRTOS, runs on an STM32F4 MCU. This MCU initializes and controls the DSP and amp chips, among other functionalities on the board. The audio signal path is also managed through the MCU, allowing for various measurements such as level metering and FFT. Additionally, the MCU handles USB audio input and the user interface.
We are also considering integrating 3rd party module to significantly enhance connectivity and streaming capabilities.
Our current focus is to prepare the hardware fully, so all new features can be added or enabled simply by updating the firmware in the future.
Please share your thoughts on how appealing this project is to you and what features you would like to see added or removed. Your feedback will be invaluable in guiding our next decisions. Below is a detailed list of features.
Thank you!
Notable features:
Inputs:
User interface:
Powering:
Connectivity:
Upgradability:
Note to admins: This post is not intended for commercial purposes. I have deliberately avoided advertising any brand or product names. The sole purpose is to gather feedback for our feasibility study on a project that may potentially become commercial in the future.
Dear fellow engineers and enthusiasts,
This is a prototype of our audio amplifier, which is based on the ADAU1452 DSP and two TAS3251 chips in the output stage. This setup provides a total output of 4x175W (or 2x350W) at 36VDC.
After several months of development, we now face some critical decisions, the most important being the viability of this device for mass production and the potential features to add or remove since this proof of concept.
This version is not the final iteration; we are still optimizing certain parameters. Most notably, the THD+N currently measures at 0.08% at 100W into 4Ohm and 0.008% at 10W into 4Ohm, with an A-weighted noise floor around 105uV. While these numbers are quite good, they do not fully match the TAS3251 datasheet. We are continuing to optimize the PCB layout, LC filter components, and PFFB.
We aim to provide plug-and-play functionality for this device, meaning users will not need to tweak the ADAU DSP and can select one of the pre-made use cases that are easily configurable and controlled via a user interface. For more advanced users, full access to the ADAU DSP will be available using SigmaStudio.
The main firmware, based on FreeRTOS, runs on an STM32F4 MCU. This MCU initializes and controls the DSP and amp chips, among other functionalities on the board. The audio signal path is also managed through the MCU, allowing for various measurements such as level metering and FFT. Additionally, the MCU handles USB audio input and the user interface.
We are also considering integrating 3rd party module to significantly enhance connectivity and streaming capabilities.
Our current focus is to prepare the hardware fully, so all new features can be added or enabled simply by updating the firmware in the future.
Please share your thoughts on how appealing this project is to you and what features you would like to see added or removed. Your feedback will be invaluable in guiding our next decisions. Below is a detailed list of features.
Thank you!
Notable features:
Inputs:
- I2S
- SPDIF
- Bluetooth with aptX
- USB audio class
User interface:
- GPIO and Rotary encoders
- All features controllable over I2C registers and through SCPI interface on UART
- Few RGB LEDs on board
Powering:
- 36VDC up to max power
- USB-PD up to 100W (up to 240W in the future once the technology becomes available)
Connectivity:
- Point-to-point stereo link with another module using CC8520
- Integration with third-party module for streaming services such as Spotify, Apple Music, Amazon Music, Tidal, Deezer, Qobuz, TuneIn, and more...
Upgradability:
- Main firmware upgradable over USB
- USBi interface emulated by main firmware, allowing real-time DSP tuning and debugging without a dedicated hardware programmer
Attachments
Hi. Really excited about this project.
It would be good if you can get the THD+N to <0.03% at 100W.
How are you managing the connection between ADAU DSP and SigmaStudio.
I'll take a few of the first batch for sure.
Chris
It would be good if you can get the THD+N to <0.03% at 100W.
How are you managing the connection between ADAU DSP and SigmaStudio.
I'll take a few of the first batch for sure.
Chris
Hi Chris,
we would like to get close to datasheet values, meaning THD+N <0.01% at 100W. With proper utilization of PFFB, it could be even less, but I can't promise that. There is still lot of work to do.
Regarding the ADAU DSP and SigmaStudio connection, for the prototype and perhaps for the first batches, there will be the standard programming connector compatible with USBi adaptor. However, the ADAU I2C is also routed to STM32 and later we would like to emulate the USBi interface completely in the STM32, so the onboard USB could be used without the need for external USBi programmer. On all released boards, this will be enabled by FW update.
Thank you very much for your interest. We aim to launch first batch at the end of this year.
Jan
we would like to get close to datasheet values, meaning THD+N <0.01% at 100W. With proper utilization of PFFB, it could be even less, but I can't promise that. There is still lot of work to do.
Regarding the ADAU DSP and SigmaStudio connection, for the prototype and perhaps for the first batches, there will be the standard programming connector compatible with USBi adaptor. However, the ADAU I2C is also routed to STM32 and later we would like to emulate the USBi interface completely in the STM32, so the onboard USB could be used without the need for external USBi programmer. On all released boards, this will be enabled by FW update.
Thank you very much for your interest. We aim to launch first batch at the end of this year.
Jan
Last edited:
This looks very interesting, I will be keeping track of what happens here. Thanks for the work, hopefully the end result is to your liking.
Hello everyone!
I haven't been posting much, but the project is alive! We are working on it two people and mainly in our free time, so the progress has been slow, but steady.
There are most notable updates so far:
1) 90% features listed above have been implemented in the software.
2) We have tuned the components in the analog signal path and reduced the THD+N significantly. With proper utilization of PFFB and quality components in LC filters, we can now achieve THD+N as low as 0.008% at 100W. With the next PCB revision, that is currently in manufacturing, we expect this value to go even slightly lower.
3) We did extended stress testing to verify passive cooling and the results are satisfying.
4) We have implemented some nice-to-have features such as spectral analyser with configurable number of bands.
5) We have the first product candidate in manufacturing now. The PCB has been significantly improved and miniaturised. This will be the first small batch of boards that will be distributed to the first users for testing.
6) We have expanded the concept to have 3 variants of the amplifier board. Same features but different power stage on each - from a small credit card size amp to a large board equipped with TPA3255. The current board with TAS3251 will be in middle. Multiple amp modules will be chainable so various multi-channel solutions will be possible. We are also working on 3 variants of control modules that will provide user interface to our amps. From a small board with few knobs to a large HMI panel with TFT display. Each control module will be compatible with every amp variant and everything will be user-configurable without the need of coding.
Stay tuned! I will keep posting!
I haven't been posting much, but the project is alive! We are working on it two people and mainly in our free time, so the progress has been slow, but steady.
There are most notable updates so far:
1) 90% features listed above have been implemented in the software.
2) We have tuned the components in the analog signal path and reduced the THD+N significantly. With proper utilization of PFFB and quality components in LC filters, we can now achieve THD+N as low as 0.008% at 100W. With the next PCB revision, that is currently in manufacturing, we expect this value to go even slightly lower.
3) We did extended stress testing to verify passive cooling and the results are satisfying.
4) We have implemented some nice-to-have features such as spectral analyser with configurable number of bands.
5) We have the first product candidate in manufacturing now. The PCB has been significantly improved and miniaturised. This will be the first small batch of boards that will be distributed to the first users for testing.
6) We have expanded the concept to have 3 variants of the amplifier board. Same features but different power stage on each - from a small credit card size amp to a large board equipped with TPA3255. The current board with TAS3251 will be in middle. Multiple amp modules will be chainable so various multi-channel solutions will be possible. We are also working on 3 variants of control modules that will provide user interface to our amps. From a small board with few knobs to a large HMI panel with TFT display. Each control module will be compatible with every amp variant and everything will be user-configurable without the need of coding.
Stay tuned! I will keep posting!
Wonderful for the speaker designers of medium to large 3-4 ways:
@hifijim @vineethkumar01 @ernperkins@cowanaudio @perrymarshall
Hi @Gordon001
Great project! A lot of DSP amps appear recently, I also have one and looking for answers to same questions 🙂
Do you have any work done on the sw interface side?
Great project! A lot of DSP amps appear recently, I also have one and looking for answers to same questions 🙂
Do you have any work done on the sw interface side?
Hello DIYAudio community,
Couple of years ago I’ve designed a DSP Class-D amplifier called Centure 21. It has got ADAU1452 DSP, TAS5548 as I2S->PWM modulator and 4 output channels, with 4 more available on the "PWM" header.
and photo of the prototype:
on the left there is Amanero header, which can be used for amanero module to have just USB 2 ch input, or my York interface. There was also header on the left side with S/PDIF and I2S which are connected to ADAU1452. The purpose is to connect daugter boards with different input interfaces: ADC, AES/EBU, TOSLINK etc. The...
Couple of years ago I’ve designed a DSP Class-D amplifier called Centure 21. It has got ADAU1452 DSP, TAS5548 as I2S->PWM modulator and 4 output channels, with 4 more available on the "PWM" header.
and photo of the prototype:
on the left there is Amanero header, which can be used for amanero module to have just USB 2 ch input, or my York interface. There was also header on the left side with S/PDIF and I2S which are connected to ADAU1452. The purpose is to connect daugter boards with different input interfaces: ADC, AES/EBU, TOSLINK etc. The...
Have a look at this 🙂
Why not use the DSP already available withing the TAS3251?
I'll assume it's TI's grave mistake to not make programming of that "available" to a general consumer, as AD does with their Sigma Studio. Apparently, only OEMs get to have that and are not allowed to include its functionality as part of a product offering, encouraging flexible useage.
Interesting how one company encourages use of the software part of their audio DSP offerings, another restricts it. I see how you got around that.
I'll assume it's TI's grave mistake to not make programming of that "available" to a general consumer, as AD does with their Sigma Studio. Apparently, only OEMs get to have that and are not allowed to include its functionality as part of a product offering, encouraging flexible useage.
Interesting how one company encourages use of the software part of their audio DSP offerings, another restricts it. I see how you got around that.
Yes, the board has 2 independent power stages and each can be configured separately.
We have considered to use the DSP of TAS3251 but there are multiple reasons why we decided to go with ADAU1452.
- As you assumed, the DSP of TAS3251 is somewhat difficult to use. We actually did some experimenting with it and we also implemented SW libraries to control it but it would be hard to make it configurable by the end user.
- We would like to produce 3 variants of this amp board, each with different power stage (besides TAS3251, we also consider TPA3255 and one much smaller variant where the amp chip is not yet selected) and these other amp chips does not have the DSP capabilities so it makes sense to have this independent on the power stage and also to maintain one firmware and one feature set for all 3 variants.
To be honest, there are still couple of months of work for us. Besides the amp boards and control modules, we are also working on a website where we will sell our products and provide tutorials. And in two people it is advancing slowly.
Ask lutkeveld, creator of ZOUDIO AIO4CH: 4-channel amplifier with DSP and Bluetooth, how that goes...
Thanks for pointing this out. I have been watching this project for some time. Perhaps I could talk to the author.
Hi @Gordon001,
using PFFB in combination with the TAS3251 is interesting.
Which components did you choose for the PFFB and how did you dimension the necessary components? Have you compared different component values? Have you made measurements in comparison to the output without PFFB?
It was an experiment which turned out good. We used the components recommended by TI in their app note for TPA3251. We also tried different values of the components but the default ones yielded the best results. Compared to not using PFFB, the difference in THD+N is not so critical, few dBs I would say, but it is worth the low cost of PFFB components.