[design log] Neat 2x170W I2S in, I2C controlled, integrated DSP amp (TAS3251)

The purpose is to design the "amplification brick" of digital centered audio systems, like systems based on Raspberry Pi, or PC music servers relying on digitalized music files. Some versatile and cheap micro-controllers (uC) have also rich audio features very suitable for audio projects like multi I2S or SPDIF outputs and DSP functions.


The sound quality shall be High. No a "No compromize" quality, but sort of 95% of that quality for a fraction of the price (adjusted pareto rule :)


Target applications are classic stereo systems, but more specifically multi channel active speakers, for which there are less easy to implement options for the hobbyist.


A key idea is modularity. The amplification module shall easily interface with generic, open (manufacturer independant) and well spread protocols. This ensures the possibility to be used in various projects, and on the long term:

  • I2S for the audio streams (reference interface since the beginning)
  • I2C for the control of the device (propocol that we can find in any uC


The DSP in feature opens possibilities of removing analog filters, easy performance of tests and experiments on filters, easy implementation of active speakers with oly one digital in wire....

As much as possible the design of the "brick" shall not enforce usage options. I should, as far as possible, let the end user choose how he wants to use the amplification chip. This is a key difference with the ready made amps, where the cheap price comes with an heavy lack of flexibility. Some flexibility for the tuning of some key components would be an advantage (like output selfs and caps).


The final system should be cost competive with the latest Hypex fusion plates DIYclassd.com. 300€ for 2x125W + DSP + Power supply. Is it achievable to have a similar thing for 100€ ?


Last, It would be possible to build from PCB by the hobbyist, with achievable SMD soldering. This is to ensure that even if no manufacturing, the design could be useful to hobbyists.
 
Considered options:
Since two years, I control LX-minis with a STM32 uC and 2x FX-802 amps feeded with SPDIF. The STM32 is connected to the server with Async USB, performs the DSP based filtering and generates the 4 audio feeds over the 2 SPDIF lines. Not bad for a generic 30€ dev board.


The FX-802 amps are based on the proven STA326. I like the sound a lot, but they are not powerfull enough for the woofer of my speakers. I also liked a lot the Full Digital Amp idea. However it is more that I liked the idea (less is more unverified belief) than I did extended comparisons with other technologies.


One path is to follow ST path with a 2 chip options: STA311B + STA510. I have not seen developments in this range since years. May not be an area of strong focus from ST anymore.


Another path is to jump to TI similar set-up like the one implemented by IVX in its Tiny TAS5558 + TAS5624 design.



Last one is to investigate the TI TAS3251; which is basically a TPA3251 fitted with a PCM5242 (miniDSP+DAC) on the same chip.


The TPA32xx familly seems to have proved its performance here, especially when well implemented. Looking at the TI roadmap for >50W amps, it is their reference for high performance designs (Webinar - Class-D Amplifiers Above 50W | TI Training)


TI sells that in their usage range, those amps have a lower THD than their class AB amplifiers... So it seems that they cranked their current best engeeniering in those chips. I know that amps that measure well don't always sound well and reverse. But good measures and positive feedback and interest here on DIYaudio show that this horse has significant potential.


TPA3251 has balanced inputs. A question is wether going to a separate balanced DAC option + TPA3251 or to integrated TAS3251. Balanced DACs are not so well spread. Existing boards are not so cheap (especially when you need 2 or 3 of them). Building a PCM5242 board doesn't looks obvious to me as that chip seems difficult to hand solder. However, it looks like a perfect fit for the TPA3251.


Bottom line:
The TAS3251 is pretty interesting, integrating the DSP; balanced DAC and Amp in one chip, for 10€ (unit price in EU).


The TAS3251 has extensive implementation advice. There is an EVM that demonstrate TI proposed implementation, including gerber files. This should avoid (big) mistakes.


Different threads discuss the implementation of TPA32xx, providing useful advices.


Soldering looks achievable for the hobbyist.


=> will go for the TAS3251
 
Datasheet: TAS3251 175W-Stereo, 350W-Mono, Ultra-HD Digital-Input Class-D Amplifier | TI.com


Evalutation board: TAS3251EVM TAS3251 175W Stereo / 350W Mono Ultra-HD Digital-In Class-D Evaluation Module | TI.com


Attached the simplified schematics from the project.


In my final set-up a Stm32 nucleo board will contrrol several amp boards:
- configure the miniDSPs,
- receive music from USB (async),
- manage the volume,

- send the I2S audio streams, based on the board clock,

- control the few I/O (like mute) and acquire the amps outputs (fault...)



NUCLEO-F746ZG - STM32 Nucleo-144 development board with STM32F746ZG MCU, supports Arduino, ST Zio and morpho connectivity - STMicroelectronics


JMF
 

Attachments

  • simplified schema.JPG
    simplified schema.JPG
    65.3 KB · Views: 4,442
Part of the price equation (compared to Hypex fusion plates) is the power supply. We need 36 V. My application requires:
- 2x80W peak in 8R,
- 2x25W peak in 4R.


This would add up to 210W (I suspect that my math here are not fully correct, but hope the order of magnitude is).



The identified candidate is Mean Well LRS-350-36 (MEAN WELL LRS-350-36 Module d'Alimentation a Decoupage SMPS 36V 9.7A 350W - Audiophonics).


Any advice about pro/cons and alternatives (DIY or not) are warmly welcomed :)


JMF
 
Hi,


Requesting the files on the TI community, I have now the TAS3251EVM Gerber and CAD files.


As the used CAD is Altium, and I don't have this professional product. I used a web service Altium2KiCAD to translate the files. It does not seems perfect, but looks like a starting point. At least it allows to start study the TI design.


Unfortunately, it is a 4 layers board. Top and bottom are the key layers. One internal is almost plain copper. The second internal one has some traces to the TAS3251 and for PFFB part.


I have to study if I can go to 2 layers PCB, which would be much cheaper. However as it is my second PCB attempt, I will be very interested by the advice of knowledgeable ones to avoid big mistakes.


I also have to install the PPC3 TI tool that allows to generate the TAS3251 config headers, to ensure that it can be achieved without buying the real TAS3251EVM. I'm a bit cautious about those policies where you are blocked in a proprietary system designed for another customer target (big consumer electronics companies).


Last, I wonder if implementing the psot filter feedback part is useful or not, if looking for the best sonic performance and not the best measurements. Anybody compared ?


Best regards,


JMF
 
As of today:


I just get the access to the TAS3251 module in TI configuration tool PPC3. Good to have. I will now study it and confirm the possibility to configure the chip without having the TAS3251EVM board.


Studying the different reference designs, it happens that the TPA3255EVM has only 2 layers. Fine ! This confirms the acceptability ! I will do a mix between the TPA EVM. concepts and the TAS EVM ones.


I'm currently learning about KiCAD basics. I will try to submit a schematic for the project soon.


JMF
 
JMF11, by the way, I considering the design hardware sharing in case if it will get no commercial interest.


Hi, and thanks for the proposal. The priority for me is success of the design. If it could be usefull to the hobbyist communnity: all good !


I must admit that I asked myself what could be the market size for such a project, sold as "powerfull and audiophile RPi hat". Do not know. What I learned is that it looks possible today to manufacture small batches of electronic boards at reasonable price (which was not possible before). It looks to me not so easy to find the right place between the cheap cheap Taboa / Aliexpress boards and the renowned brands.


However, I'm convinced that the TAS3251 is a "no possible failure" project that will deliver great value :)


Status:
I have now the import of the altium project in KiCAD, which was a bit tricky. Next phase is to remove unused parts of the EVM board to focus on the needed configuration.


JMF
 
My initial intention was to have the MCU out of the amp board, to keep the system modular, and possible to evolve to a multi-amp system. Maybe a small functional analysis would be helpful to list the needed functions.


However, as of today, I try to learn enough to draw the PCB. The import from Altium inside KiCAD proves to have some limitations that make the process bumpy and the result questionable. I may have to port "manually"the design in KiCAD.


Humm those beginners :)....


JMF
 
I have experience with hardware and software design for the TAS5754M which uses PurePath Console 2. It is a lower power (2x50W) brother from a previous generation. If you need any help you can send me a message. I think a lot of aspects of the design will be similar.

Maybe it has changed, but a big downside for development with the HybridFlow DSP devices is that realtime adjustment is only possible with the motherboard (PPCMB) and the EVM. For the final product you need to dump the DSP registers into a .h (header) file from within the console GUI, which contains all registers and the value to be written to them. If the PCB boots, the companion microcontroller needs to write this header file to the TAS device, which takes about 100mS via I2C.

This means that end users can not adjust the DSP because the EVM and motherboard are needed. Maybe someone can reverse engineer the register values that correspond with the biquads, but TI keeps it a secret.

Besides that, it is a great device that combines high end features in a small and cheap package.
 
Member
Joined 2015
Paid Member
KiCAD

My initial intention was to have the MCU out of the amp board, to keep the system modular, and possible to evolve to a multi-amp system. Maybe a small functional analysis would be helpful to list the needed functions.


However, as of today, I try to learn enough to draw the PCB. The import from Altium inside KiCAD proves to have some limitations that make the process bumpy and the result questionable. I may have to port "manually"the design in KiCAD.


Humm those beginners :)....


JMF

Hi JMF11,

Here are two documents I found very useful starting then progressing with Kicad.
First one is a ham radio club project, it’s very basic but it allows me, step by step, to understand the various Kicad module and the whole design process then to be able to start at new project from scratch.
Second one, from Dave Jones founder of EEVBlog, give you very practical and useful insights on the way to design a pcb.
Those two docs are not recent but, with the help of Kicad documentation and some practice, I’m know working on 4 layers projects...and i’m still progressing!
Hope you’ll find those docs useful, I follow your threads with attention as I’m also using STM32 in my projects.
Chris

Link to First doc : https://www.robertponge.com/telechargements/ebooks/kicad-2.pdf
 

Attachments

  • PCBDesignTutorialRevA.pdf
    384.9 KB · Views: 342
PCB drawing skills it's like bicycle riding, once you get it, never forget. However, without understanding some physics, it is a cheap ability.


@IVX: True :) That's why I want to mimic the EVM boards layout. And hopefully the datasheet discuss the topic with good details.


@Lutkeveld & IVX: this is OK for my use case. However having information about the registers layout would help being independent from the proprietary tools (which you can only access by requesting explicitly for them). I don't know if IVX has more information on that, to develop an application.


@AIM65: thanks for the documents. Will read them. And I like the Stm32 and nucleo boards for audio applications, with their digital audio dedicated SAI modules (plus I2C features, USB capable)....


JMF