A high-end digital amplifier for everyone
A high-end digital amplifier for everyone or everyone working on a high-end digital amplifier
(See the complete document with images here.)
We would first like to introduce ourselves to everyone. We are a small group of former and actual Electrical and Computer Engineering students that worked over the last two years on a digital audio amplification project. The reason of this post on the forum is that our work is going to its final stage and we are planning to release it to the DIY audio community. We believe that this forum (www.diyaudio.com) is simply the reference in that domain and this is why we are making our official announcement here.
What we developed is a complete digital audio amplification development platform and we invite you to read forward to understand all the benefits of our work. What we are about to release is not a conventional simple Class-D amplification project; this project won the 115th Audio Engineering Society convention student design competition last month in New York.
When we first started to work on digital audio amplification we considered the existing solutions. First, there was the conventional Class-D amplifier basically made of an analog audio input, a triangle wave signal, a comparator and a switching output. This solution while being simple and efficient did lack in performance and creativity. The input is still analog and the process itself suffers from gain-bandwidth limitations. On the second hand there was the direct PCM to 1-bitstream conversion in the digital domain. This concept really goes deep into the idea of bringing a digital audio signal to life. Considering the CD as the actual primary source of audio input, it was obvious that the idea of keeping the source in the digital domain in the process of amplification was the key. This is therefore on what we started to work.
The people around this forum who know the most about digital audio amplification probably already know what big companies are offering. While these companies were developing their digital audio amplification solutions, we were doing the same. However, now that we have a functional digital audio amplifier, instead of taking our work and putting it into a plain integrated circuit, we are offering the DIY community the chance to work with us on this cutting edge technology.
The main idea behind digital amplification is the conversion of a pulse-code-modulated (PCM) digital signal to a 1-bit stream that can be amplified by a class-D analog output stage (power switch) to a high voltage digital code. This high voltage 1-bit pulse stream is then low-pass filtered in order to produce an amplified version of the original analog signal across the loudspeakers.
Our platform splits into two entities: the hardware and our reference design firmware. The hardware we put in place has the goal of being universal and allows development of different digital amplification schemes. We separately developed a modulation platform and an output stage board so that they are independently upgradeable. The hardware and software have been designed to allow for the addition of various add-ons such as analog inputs, LCD, tone control, mp3 decoding etc.
a) Digital platform
The heart of the hardware platform is an FPGA (Field Programmable Gate Array). This device brings a lot of advantages on the table. It is re-programmable, it allows the integration of various independent blocks that have access to any of the numerous I/O pins, it operates at high clock speed, etc. Basically, with such a universal device, it is possible to implement the PCM to PDM converter, but also any digital signal processing blocks (equalizer, tone control, digital filters etc...). In addition, all the control functions can be implemented in the same device, no need for an additional micro-controller to add an LCD or an infrared receiver for example. Therefore, this means the FPGA can be used to implement just about anything you can imagine without any hardware change. In this regard, it is really a universal and evolutionary core for the project.
In order to make processing of the digital audio signal easier, two integrated circuits have been used. The first one, a digital audio decoder (Cirrus Logic, CS8416), is used to decode the S/PDIF or AES3 data stream and output the PCM audio data (e.g.: I²S format) for easy processing by the FPGA device. The second one, an asynchronous sample rate converter (TI - Burr Brown, SRC4193), makes it possible to support any audio sample rate (e.g.: 44.1 kHz, 48 kHz, 96 kHz, 192 kHz etc.) at a glance. These two ICs additionally offer practical features such as logarithmic attenuation, balance, etc. They have been selected because they are the two best devices available as of today.
As a matter of convenience, the platform offer digital input connectors (coax and optical), all the required power supply regulators and 79 I/O's from the FPGA distributed on header connectors.
Tremendous effort has been put into the design of the printed circuit boards. The modulator board requires a four layer board to allow for wide power and ground planes which drastically reduces inductance and provides well defined impedance for the high frequency noise. Particular attention was given to the FPGA decoupling and signal integrity.
b) Output stage
The main goal of the output stage is to reproduce the 1-bit code generated by the modulator at a higher voltage. A low pass filter is used to remove the high frequency content from the amplified PDM waveform before sending it to the speaker. The IC selected for the H bridge amplifier is the TAS5112 manufactured by Texas Instruments. It is one the most recent H-bridge IC designed for audio applications. It can deliver up to 50W of stereo power into 6-ohm speakers at very high efficiency (90%). It includes a logic interface (controlled by the FPGA), integrated bridge drivers, high efficiency MOSFET outputs, and all necessary protection circuitry.
As stated earlier, the output stage was built separately from the digital platform to increase flexibility. In fact, additional channel and different output power can be obtained by adding/changing the output stage, without changing the digital platform. The switching nature of the output stage circuit imposes very strict constraints on the layout and choice of external components. A two-layer board was made with proper use of ground planes to reduce EMI.
Having this hardware in place, we developed the FPGA reference design firmware. It mainly consists of a function control block, a digital audio receiver and a digital signal processing block.
The heart of the digital amplifier resides in the modulator, which converts the PCM audio data into a 1-bit stream capable of driving a power switch. The modulation scheme we have implemented is different from most of the existing PWM-based solutions; it is known as sigma-delta modulation (pulse-density modulation – PDM). Notwithstanding their high linearity and low clock frequency, sigma-delta modulators are generally not considered for digital power amplification because of their high pulse repetition frequency (PRF) and they encounter stability problems at high input amplitudes. Inspiring ourselves from work published by M.B. Sandlers in the journal of the Audio Engineering Society, we have designed a stable 7th order sigma-delta modulator with PRF reduction algorithm. Most of the design work was performed in Matlab/Simulink. We believe that the current design gives very high quality results. However, it might be interesting to compare with other modulation schemes.
In the end, what we have is a complete and functional, fully digital amplifier. What makes it really powerful is the fact that it is completely evolutionary. Anyone with this hardware could see its amplifier gain functionalities as more work is done on the firmware, a firmware in which all the efforts of the community could be invested. This is exactly why the DIY community makes the project so powerful. Releasing the firmware in open-source would allow putting all the efforts in the development of one single high-end digital amplifier. Furthermore, not only the developers could benefit from the evolution of the project, but any hobbyist interested in audio projects. Imagine having an amplifier that could sound better and have more features at every firmware update, an amplifier that any DIYer could own.
Today, we are making our official announcement to see what interest the DIY community could have in the project. We already have a few prototypes working and we think the project has reached a milestone. Unfortunately, we cannot support the project financially anymore. We must admit that while we are still very interested in adding new features and improving things, we start to lack manpower. In a near future, we will be able to release the technical manuals, but meanwhile we ask you to give feedback and visit our website for more information: www.digitalamplification.com. Feedback for us is very important in deciding whether or not we are going to release everything in the public domain. It would take us a tremendous efforts on our part to prepare this release (documentation, code organisation and commenting, CVS server, General Public License, etc.), so we want to be sure that the interest is there and that the project will live on.
Have looked around your doc. Nice work ;) But I am missing the feedback loop from the output in the documentation/pictures. Without this any pulse modulated power amp will suffer from power supply modulation and hence distortion related to this. Not to mention dead-time distortion. What about this?
well, first of all, I consider this a very interesting effort. Releasing a development into the DIY public is a strange thought, tough. I'm interested in seeing this evolving and am curious about it's potential in the hands of DIYers instead of "pros"...
Something I'd like to know clearly in the first place: Do you (DAX) consider releasing the firmware *only*? Or is it the whole project (including schematics, layouts and documentations) you want to "live on" in public?
Cause I can't imagine myself developing software for a company which earns money with selling the only hardware (as their property) that the software can work with...
Being basically a power-DAC* it wouldn't be possible to apply feedback from the output stage.
The TacT Millennium AMP (digital PWM) doesn't have NFB either, this means that such an amp is basically feasible but there are some very stringent requirements to be met by the PSU. This is also valid for the switching behaviour of the output stage. The problems are greatly reduced by the fact that those guys use a power stage that is designed with such applications in mind (the TI amps use the same working principle as the TacT amp!).
This is not the first SD-amplifier however (but the first with exactly this topology) to my knowledge) however: Sharp SM-SX100 and the Tripath amps are also SD amplifiers. The "Technische Universität" in Munich owns a patent for an SD amp as well, I don't know however if anyone makes commercial use of it:
For your project I wish you good luck.
* If one takes the meaning of "digital" serious then there will not be such a thing like a digital amp since the amplification as such is a purely analogue process.
If you're less strict then you can call power DACs (like this one) digital amps, but never ever is an ordinary class-d amp a digital amplifier.
Congratulations on winning the AES award!
I'd like to join the others and encourage you to share as much information as possible.
I'm particularly interested in your PRF reduction algorithm. I think this would be particularly beneficial for dealing with DSD bitstreams from SACD sources (completely bypassing the PCM format).
I'm curious if your signals to the output section drive the H-bridge upper sources and lower sinks in a direct complementary fashion (with dead-time between switching), or if you've used an extended technique that slightly varies the upper and lower switching moments.
The closest commercial products to your approach that I'm aware of are Sony S-Master Digital Amplifier processors with Sigma Delta modulation (i.e. Renesas M65817AFP). I don't think these have been discussed yet in this forum (I haven't played with them yet myself since I've been busy with the TI Equibit chips).
One thing that has become apparent to me is that for any of these modulation conversion technologies that use an H-Bridge power output section without feedback, it would be very easy to use an FPGA as a switch for easy measurement and listening comparisons between the various commercial modulators and self-developed algorithms.
From my own experiments, I found that I needed to begin with my focus on digital interfaces and power supplies for the amplifiers' output sections. I'm guessing that you haven't focused primarily on the power supplies. My initial conclusion has been that the power supplies are the biggest opportunity for improvement over existing commercial digital amplifier designs, more so than the modulators.
I can appreciate your wanting some interest before you spend the time and effort to document and present your work up to this point. I've achieved some excellent results with direct-digital interface and DSD-PCM conversion circuits for going between disk players and digital amps. I intend to post the details on them in the near future, but it's a lot of work to bring the information together in a presentable form. Also, I get hung up because I always know that there's some type of further improvement I could make.
As you mention, digital amps are finally at a point where we can consider making evolutionary changes and improvements. It's really good to start to be getting past the point of simply getting one to function.
Experimenting with this stuff does take a lot of time. The more people that are involved experimenting with this technology and sharing information, the better.
Very interesting. Will be excited to hear more.
I'd definately be interested in learning more.
Wow, really interesting work you've got there. The use of a 4-layer PCB as well as a FPGA will deter many DIYers unless a PCB is provided, if so it would be very interesting for anyone wanting to build a "digital amp"...
I will be watching this thread closely to see how your project develops. This could get very interesting!
After reading your release paper I have more questions than answers, but I am sure you will get to them.
The first one I have is what kind of license will you be using? Public Domain? GPL? OpenCores?
This might become the worlds first Open Source Amplifier :)
I also agree with JoeBob, that unless the PCB and FPGA were supplied (at reasonable cost) then interest from DIYers will be very limited. Most do not have the technical knowledge or equiptment to fault find or optimise a design this complex even if they do have the enthusiasm and interest (in spades). Previously, I looked at the Tripath gear but lost interest as the RF management was too difficult for me to deal with, not having my lab any more.
However, should you want a beta tester, I'll volunteer.
|All times are GMT. The time now is 05:08 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio