Anyone interested in a digital amplifier project? - Page 2 - diyAudio
Go Back   Home > Forums > Amplifiers > Class D

Class D Switching Power Amplifiers and Power D/A conversion

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 22nd November 2002, 01:31 AM   #11
diyAudio Member
 
Join Date: Sep 2002
Location: CA, USA
Default Where to get the TAS5100EVM Board

It looks like a very interesting probject. I have just done a 300B mono block amp. It sounds great. However I figured that in tube + high efficient speakers world. It's kind of difficult to tweak things around. Only thing you can change is upgrading components or try another topology. And we all know none of components, capacitors, transformers, speakers, valves, have perfect flat frequency response. It's even difficulit to put them together to end up with better result. In most cases, the system performance is determined by the weakest link.

In digital world, it will be much easier to fine-tuning things with program. As a software engineer, I am particular interested in the digital audio system. But I don't have a lot digital audio process knowledge now :-(

Therefore, I want use this as a start point. I plan to use this chip set to build a board and write progams to control them, even include some sort of digital filtering. The ultimate goal will be drive "$75 Radio Shack speakers like $10000 ones" I know this is going to be tough. But it must be fun, isn't it?

Ok, back to where I started. I have ordered TAS5100A and TAS5015. The next it build a borad. Do you guys know how to get their TAS5100EVM board? That will save a lot of time that building my own.

Thank,

Ted
  Reply With Quote
Old 22nd November 2002, 09:04 AM   #12
deandob is offline deandob  Australia
diyAudio Member
 
Join Date: Oct 2002
Location: Brisbane, Australia
I'm also looking at doing a DIY digital amp project, and tossing up between the Tripath and TI chipsets. Although the Tripath chips are well regarded, the TAS5015 has better specs and is completely digital so I'm interested in seeing how others have implemented the TAS5015, especially the volume control. Should be a fun DIY project.

There is a rumour that Tripath will release a purely digital chip, but with all the financial difficulties the company has been having (as with most small hi-tech firms at the moment), it would not surprise me if they have cut back their product development - has anyone heard any news??

I'm keen to hear if anyone has compared the modern digital amplifiers (Spectron, Tripath eg. Bel Canto, Tact, a TI 5015).

There is an interesting site www.classd.org that talks about a reference class D design from Philips DSL that the author thinks is better than most of the other class D implementations. Anyone have any info on this??

Regards,

Dean
  Reply With Quote
Old 22nd November 2002, 06:23 PM   #13
diyAudio Member
 
Join Date: Aug 2002
Location: Wisconsin
Default Chips and Eval. Boards

Perhaps we should discuss the definition of what a 'Digital Amplifier' is.

There are some companies that are refering to conventional Class D amplifiers as digital (especially some car audio products). The basic PWM switching process is either turning the output on or off, which in a way might be construed as digital, but PWM still has an analog time component. Strictly defined, a digital signal is one that can be stored, modified, and transmitted in a numeric representation. The *information* in a digital signal should be immune to amplitude and time irregularities. (Amplitude and time characteristics of the information only become critical when a conversion to or from analog is involved.) Moreover, the entire process of a conventional Class D is analog. In my view, a conventional Class D amplifier shouldn't be called a digital amplifier. (I don't mean to imply any evaluation here of the overall quality potential of Class D. The above mentioned work by Philips looks quite promising, as does Analog Devices' upcoming AD1992/AD1991.)

Equibit 'Amplifiers' (TacT, TI) are fully digital implementations of a PCM to PWM conversion, with high current output. Technically speaking, they aren't amplifiers (which actually is one of their advantages).

The Tripath devices are simalar to conventional analog Class D amplifiers in that they start with a low level analog input and produce a switched power output with analog feed back. Instead of PWM with a fixed carrier frequency and variable duty-cycle, they use variable frequency spread spectrum switching. The generation of the output timing is done digitally. To help overcome the traditional gain-bandwidth product limitations associated with analog feedback of a switched signal, Tripath uses adaptive filtering so that it can learn the behavior of the system and predict the optimum output timing. These devices can probably be accurately called 'digital amplifiers' because they are actually amplifying a signal, and using a digital process to do it. From an external 'black box' perspective, however, they are analog: low level analog in to high power analog out.

Tripath's website mentions that they plan to come out with a direct digital input version, so it's more than rumour, but they don't even offer a projected year of release. It should be promising if they can ever make it happen.

The Sharp SM-SX100 and SM-SX1 digital amplifiers take a direct DSD bitstream from an SACD and produce a high power switched output. These can probably be called 'digital amplifiers' in the purest sense. I'm not familiar with the specifics of their internal circuits.

A lot of the promise of single bit DSD is that it avoids quantisation issues associated with multibit PCM. The main problem I have with DSD is that I don't know how to process it without converting it to PCM. I've read claims from Sony and Phillips that their studio equipment can directly process DSD, but I haven't found any information at all regarding the specifics of the implementation. (I'd appreciate it if anybody could point me in a good direction.) In the meantime I've got some NPC SM5816AF six channel DSD to PCM converter chips that will be on my next board project after the digital amp.

********

Regarding DIY projects:

The Tripath chipsets are probably the easiest high-quality switching amplifier chip sets to work with. They require almost no digital experience, offer good application notes, and should have less sensitivity to power supply design than the Equibit ones (Tripath uses negative feedback which helps with PSRR, while Equibit is completely open loop.) Tripath also offers a pretty wide selection of reasonably priced ($250 to $700) eval boards, with one to six channels of amplification at various power levels. I suspect that most DIYers would be best off just using the eval boards instead of creating their own.

For those, such as myself, who are interested in digital audio processing, a Tripath based amp doesn't offer any implementation advantages over other conventional amplifiers. They still require conventional multichannel low-level DACs. (If I was to use conventional DACs, I would still strongly consider a Tripath amp. They appear to offer a real good quality/price ratio.)

IMO, TI's Equibit chips offer the best potential for a high quality direct digital system that offers flexibility and expandability for an experimental platform.

Early this year, TI announced a 'Home Theater Development Kit' featuring their new TMS320DA610 Aureus DSP, but neither have been available for general sale. The Aureus DSP is going to be released in several commercial products, so it apparently does already exist. I haven't yet tried to contact TI to see when these would be offered for general sale, or what their price will be. I expect that this will be the best experimental platform yet for audio processing.

In the meantime, the TAS5100EVM eval board is probably the best one that they have to offer ($499 direct from TI). It's a two channel board with a TAS3004 DAP (with integrated codec) 48KHz/24bit (this can be bypassed and IIS TTL directly input), and a TAS5100A Equibit modulator and two TAS5010 output bridges (30W / channel - 4 ohms). It has a National Simple Switcher to generate a 5V supply, and some TI linear regulators to generate 3.3V digital and 3.3V Equibit PLL supplies. It requires an external gate drive supply and main audio output supply. The supply power sequencing must be controlled by the user. There is a PC interface that can optionally be used for control and configuration. The board is laid out so that it can be repopulated to use higher performance Equibit chipsets such as the TAS5012 / TAS5110DAP combo. It won't support chips such as the TAS5015 or TAS5182 that use discreet output drivers. (Supposedly there will be an upcoming eval board for the TAS5182.)

Because of the TAS5100EVM's cost, and because it didn't exactly support what I wanted, it was mostly useful to me as a reference design (TI has full documentation on their website). I decided to roll my own development board. It's taken most of my spare time for the last half year, and it's still not quite done (almost!). But I felt the initial investment in time and development were worth the effort and also valuable because it will be much easier for me to crank out final designs when the time comes.

Another option for experimentors is the Panasonic SA-XR10 that I previously mentioned. It appears to be using three TAS5012 Equibit modulators (six channels total) with five discrete power output sections. The sixth (subwoofer) channel is just integrated back into a line level output. Similar to what many are doing with CD players: tapping into the IIS signals to attach an external DAC, I believe that it would be fairly easy to bypass the input ADC's of this receiver and directly feed the IIS lines with 96KHz/24bit audio.

My first digital amplifier board is a 'stereo' design using a TAS5012 Equibit modulator, and TAS5110DAD integrated output bridges. The Equibit section is roughly based on the TAS5100EVM eval board reference design, but with my own playground section for using various audio processing devices (TAS3103 DAP, AD1896 ASRC, and EPM3512 EPLD), various power supply options, and a PIC18F252 micro to run the show and provide user interface options.

Presuming the first proto works satisfactorily as a stereo amplifier unit, I'll then reprogram it to be a two way digital speaker driver and build five more. I've got six B&W DM602S3 speakers to use for experimentation.

Once that's done, then I can do a new board for the real experimentation in audio processing.

Regards,
Brian.
  Reply With Quote
Old 23rd November 2002, 04:32 AM   #14
deandob is offline deandob  Australia
diyAudio Member
 
Join Date: Oct 2002
Location: Brisbane, Australia
Informative post, thanks Brian.

It sounds like power supply design / layout will be critical to the final sound of a digital amp based on TI's chipsets and possibly as complex as the main amp itself.

Apart from a number of individuals who have tweaked a Tripath implementation, you seem to be making progress and forging the way forward through the DIY unknown of digital amps. Please continue to share your digital amplifier "voyage" as well as the end results. Are you planning to share your results with the DIY community or is all this R&D going into a commercial product (or both)?

Regarding the volume control, can you explain the sentence "it seems to me that 24bit data with 72bit processing should be pretty good by itself". Are you saying that the TI chips process up to 72 bits if you feed them a 72 bit word? If that is the case, then you certainly have lots of bits to play with (truncate) for volume control in the digital domain, and easier to implement the volume digitally than doing stuff like varying supply voltage.

Regards,
Dean
  Reply With Quote
Old 27th November 2002, 11:47 PM   #15
diyAudio Member
 
Join Date: Aug 2002
Location: Wisconsin
Default Equibit Power Supplies

Quote:
Originally posted by deandob

It sounds like power supply design / layout will be critical to the final sound of a digital amp based on TI's chipsets and possibly as complex as the main amp itself.
The role of the power supply is one of my main curiousities.

The quality of the supply will certainly be critical, but probably in different ways than is usually a concern in audio.

For example, because Equibit is completely open loop, the stability and stiffness of the supply in the audio frequency band will have a direct impact on sound quality. Any noise or variance at audio frequencies will be connected directly to the speakers (at least it won't be amplified). If two or more channels are being powered from the same supply, there will be crosstalk unless the supply is completely stiff.

The real question I have is the impact that ultrasonic noise above the audio band will have. I think the Equibit design may actually prove to have real good immunity to this. All of the timing used for signal generation is based on a PLL running off of a dedicated linear supply that is completely separate from the high power supply. So any ultrasonic noise on the high power supply shouldn't cause any jitter or other undesirable timing based audio effects. Any ultrasonic noise that passes from the high power supply towards the speaker should be caught by the same low pass filter that's used to get rid of the 384KHz PWM switching stuff.

Presuming the above analysis is correct, I'm thinking that an off-line switcher followed by a linear regulator would be ideal in terms of quality, size, cost, and efficiency.

I'm going to start with an LT1083 or LT1084. These are the best low noise integrated regulators that I'm aware of that can work up to 30V at higher current levels. I can put an LRC filter after the output of the switcher (gets rid of noise, but drastically diminishes the response bandwidth of the supply). The linear regulator would then be used to hold the supply rock steady in the 0 - 40KHz audio range.

I'm using a single fixed voltage switcher but with a separate linear regulator for each channel.

(Just as a point of reference, the TAS5110 output driver costs $5.45 in single quantity. An LT1084 regulator costs $7.75.)

For later higher-powered versions, I'll probably use a separate switcher for each channel, with the feedback set up to maintain about 1.5V across the linear regulator.

In order to get an idea of how well the supply is actually performing, I plan doing measurements and listening to the following variations:

To see if any ultra-sonic switcher trash getting through the linear regulator is affecting sound, I'll replace the switcher with a stack of gell cell batteries in series (with and without the linear regulators). This should be about as clean of a high current supply as possible, and it will make a good reference. (I actually considered Ni-MH batteries because they can deliver even better high current pulses, but I reconsidered after calculating the cost to be over $1000 for the system.)

My guess is that the batteries won't be an appreciable improvement. But if they are, an Equibit amp is efficient enough to make them a practical consideration. (I calculated a little over one cubic foot of gell cells for 4-6 hours of listening between recharges. A class A amp of comparable output would probably need almost a closet full!)

I'm also going to try running a switching supply without the LCR filter or linear regulator. (TI actually runs a switching power brick for their eval board.) I expect that this will still sound pretty good, if not as good as the above approaches.

Probably the really ideal set up would be to have a separate switching supply for each channel, with no post filter (to keep the response bandwidth above the audio range), no post linear regulator (wouldn't be necessary if the switcher's response bandwidth is high enough), and having the switcher run synchronous to the amp's PWM switching (synchronous, but out of phase). This way the energy output by every PWM pulse of the amp would be precisely replaced by one pulse from the switching supply.

Even though this last approach has the potential of being ideal, it requires that the response and balance of the entire system be precisely tuned and optimized. I won't even consider trying at the same time I try my first digital amp.

Quote:
Apart from a number of individuals who have tweaked a Tripath implementation, you seem to be making progress and forging the way forward through the DIY unknown of digital amps. Please continue to share your digital amplifier "voyage" as well as the end results. Are you planning to share your results with the DIY community or is all this R&D going into a commercial product (or both)?
You know, I was actually hoping to find someone else out there who had already tried this. Especially to talk about the power supplies.

Sure I'll discuss it some more. It's nice to have a place where people are interested. Most of the audio and music friends I have aren't into electronic design, and most of the people I know in electronics aren't into audio. I can only share my enthusiasm with them for just so long before they start getting a certain glazed look in their eyes.

I'm doing this project for myself, not to create a product. (Who knows, it might make a nice side business. But with my luck, you'll probably be able to buy these things at Walmart before I'm done.)

Actually, one of the things that I've found especially enjoyable about this project is that I can do whatever I want without having to justify it to anyone. For better or worse I get to make all of the decisions without playing to the whims of marketing, management, or other designers. It's been quite a few years since I did anything electronic as a hobby. I forgot how much fun electronics can be!

Quote:
Regarding the volume control, can you explain the sentence "it seems to me that 24bit data with 72bit processing should be pretty good by itself". Are you saying that the TI chips process up to 72 bits if you feed them a 72 bit word? If that is the case, then you certainly have lots of bits to play with (truncate) for volume control in the digital domain, and easier to implement the volume digitally than doing stuff like varying supply voltage.
To recap for others, there's two ways to control the volume with an Equibit amp: Varying the power supply voltage, or numerically scaling the digital signal being fed into the amp.

TacT uses the approach of varying the supply voltage. The reason being that this preserves the full 24bit dynamic range when listening at lower levels. This requires a variable power supply. For some power supply approaches, this can add significant cost. For some of the switching power supplies I'm considering, this wouldn't be much additional work.

My general thinking on this is that 24bits is sufficient to cover the entire dynamic range of human hearing and then some. If you use the approach of digital scaling for volume to reduce the level, sure, you reduce the resolution of the original least significant bits, but these will now be *way* below the threashold of hearing.

The quoted comment above was refering to the TAS3103 Digital Audio Processor that I'm using. This part can handle 32bit audio data (not to mention 24 or 16bit). It has 48bit internal resolution with 72bit accumulators for multiplication. With reasonable programming diligence, this is more than enough to prevent rounding errors, especially for simple volume adjustment.

Even though I'm pre-biased on this issue, my prototype board will accomodate both methods to control volume. I'll just have to give it a listen and see.

Regards,
Brian.
  Reply With Quote
Old 28th November 2002, 09:59 AM   #16
Rookie is offline Rookie  Serbia
diyAudio Member
 
Rookie's Avatar
 
Join Date: Nov 2002
Location: Serbia
Brian,

Thank you for participating on this thread. Your posts are invaluable for me and others interested in digital amps.

You said that Panasonic SA-XR 10 uses TAS5012 modulators and discrete output bridges. I know that TAS5012 has diferential PWM outputs, so I would like to how it is conected to the gate drivers. I couldn't find any gate driver with diferential input. Does this mean that only + outputs of TAS5012 are used?

Best regards,
Dejan
  Reply With Quote
Old 28th November 2002, 11:47 AM   #17
Rarkov is offline Rarkov  England
diyAudio Member
 
Rarkov's Avatar
 
Join Date: Aug 2002
Location: North East, UK
Send a message via ICQ to Rarkov Send a message via Skype™ to Rarkov
Last year I built an Class D amp around a Zetex chip. I had to build my own 8A 20uH coils since the ones they quoted didn't exist!!! After that I bought the last two mid-power Tripath chips in the UK (according to Profusion!)...I haven't gotten around to using them yet.

I have a large interest in Digital amps...but I would like to get something sorted. You can feed PWM into a speaker - but don't. Various theories say that speakers will act as a low pass filter for them selves...you decide!

At some point - you have to convert to analogue...That's what music is! Fair enough - feed in digital - but from my experience - the low pass output filters (for PWM) are not easy to design in mind of low distortion and easily sourced parts.

My point is that you can't knock Analogue totally. It is what we hear - and indeed - what is played into a mic at a studio...It's also far easier to work on an analogue signal to change basic parameters (such as Volume, Balance, Tone controls etc) than it is with DSP. The advantage to DSP is a quickly customised product and the ability for far more clever tricks! At the end of the day - is more tricks HiFi?

Anyway - I post my interest!

Gaz
  Reply With Quote
Old 29th November 2002, 01:40 PM   #18
diyAudio Member
 
Join Date: Aug 2002
Location: Wisconsin
Default Equibit Gate Drive

Quote:
Originally posted by Rookie

You said that Panasonic SA-XR 10 uses TAS5012 modulators and discrete output bridges. I know that TAS5012 has diferential PWM outputs, so I would like to how it is conected to the gate drivers. I couldn't find any gate driver with diferential input. Does this mean that only + outputs of TAS5012 are used?
These outputs aren't actually differential. The signals themselves are just single ended 3.3V logic.

I think I know what your concern is:

An Equibit output section for one channel can have one of two forms:

1.) A half-bridge with a series cap (to block DC) driving one terminal of the speaker. The other speaker terminal is connected to a common. (This is a lower cost, lower performance configuration that also saves space.)

2.) A full H-bridge configuration with a separate half-bridge on each speaker terminal - no series cap. (Called BTL - Bridge Tied Load.) This configuration has almost double the power output, lower distortion, and is easier to design the output filter. IMO this is the only configuration worth considering for DIY.

The TAS5012 has eight gate drive outputs, four for each channel.

The four outputs each control one gate driver:
AP: the upper left FET of the H-bridge.
AM: the lower left FET of the H-bridge.
BP: the upper right FET of the H-bridge.
BM: the lower right FET of the H-bridge.

So to send current in one direction through the speaker, AP and BM will be on, AM and BP will be off.

To send current in the other direction through the speaker, AP and BM will be off, AM and BP will be on.

Some of the other lower performance Equibit modulators (i.e. TAS5000) only had control outputs for the upper FETs. These could still be used to control a full H-Bridge. The missing AM control signal would instead be taken from BP, and BM would be taken from AP. This works because in general the transistors are always turns on in diagonal pairs of the H-Bridge.

There are two reasons the higher performance modulators (i.e. TAS5012) use four separate control signals:

1.) The power FETs have switching time delays. This requires an 'underlap' time where neither diagonal pair is on. This is to prevent the possibility of current 'shoot-through' from an upper half bridge FET directly into the lower FET (without going through the speaker) that might occur if the turn-off time of one FET overlapped with the turn-on of the other.

Having separate gate drive control signals for the upper and lower gives better control, which allows the necessary underlap time to be reduced. In turn, this reduces distortion.

2.) Having separate control for upper and lower FETs allows the possibility of slightly staggering the turn-on or turn-off of a diagonal pair. This allows finer resolution for a given carrier frequency and improves dynamic range.

So back to your question, the SA-XR10 uses all of the control signals. Each one drives a (single-ended) gate drive circuit.

Regards,
Brian.
  Reply With Quote
Old 29th November 2002, 02:03 PM   #19
diyAudio Member
 
Join Date: Aug 2002
Location: Wisconsin
Quote:
Originally posted by Rarkov

I have a large interest in Digital amps...but I would like to get something sorted. You can feed PWM into a speaker - but don't. Various theories say that speakers will act as a low pass filter for them selves...you decide!
Most speakers don't have high enough inductance to block the switching frequency. You could still feed the PWM straight into the speaker, but then the speaker would have to disipate all of this extra power that wasn't being used for music. In most cases this power would exceed the power for the music. Another issue is that the efficiency of the amp would be reduced.

At one point I thought that this approach could at least be made to work for woofers. But the ones I measured still didn't have enough inductance to prevent the need for an external inductor.

Quote:
At some point - you have to convert to analogue...That's what music is! Fair enough - feed in digital - but from my experience - the low pass output filters (for PWM) are not easy to design in mind of low distortion and easily sourced parts.
TI has has some good application notes regarding the design of output reconstruction filters. Their 384KHz carrier frequency is higher than most class D, so the filter doesn't have to be as steep. (TI recommends a two pole 12db/oct design.) This is similar in concept to an oversampling DAC that doesn't require an analog brick wall filter.

Quote:
My point is that you can't knock Analogue totally. It is what we hear - and indeed - what is played into a mic at a studio...It's also far easier to work on an analogue signal to change basic parameters (such as Volume, Balance, Tone controls etc) than it is with DSP. The advantage to DSP is a quickly customised product and the ability for far more clever tricks! At the end of the day - is more tricks HiFi?
I'm not indending to knock analog.
I'll agree that digital lends itself to easy implementation of 'tricks' that I don't find usefull or desirable. But I do think that digital offers many possibilities from an 'audio purist' perspective. TacT and Meridian are two great examples of this, but they're expensive. One of the main reasons I'm playing around with this stuff myself is that I wanted a non-gimmicked digital system at a cost I could afford.

Regards,
Brian.
  Reply With Quote
Old 29th November 2002, 02:47 PM   #20
Rarkov is offline Rarkov  England
diyAudio Member
 
Rarkov's Avatar
 
Join Date: Aug 2002
Location: North East, UK
Send a message via ICQ to Rarkov Send a message via Skype™ to Rarkov
Hi,
I was reading your reply...My point about the output filter is that for a high power amp (100W in the case of the zetex), it needed about 6 or 8 amps current handling through a 20uH inductor (I'm trying to remember these figures off the top of my head though). These don't exist. Upon searching the internet for them - I came accross a 40A (!) Switched Mode power supply with these exact ratings. I got the cores and wound them myself...It required about ten wires in parallel to handle the current!

I said it was difficult to design with low distortion and easily sourced parts...That wasn't something I considered to be either. If the TI does have a far higher smapling rate, a smaller inductor will be used - and therefore - higher current handling...Smiles all round. So maybe the morel of the story is not to go off Class D - but to avoid Zetex...As I said earlier...

This is turning out to be quite interesting! Have you tried looking at the DSP Guide. It is a free online book of DSP. Very useful!

My other problem is implementing DSP functions...Microchip are starting to bring things out - but they are not upto audiophile grades yet...More instrumentation etc...

Gaz
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Noise measurement amplifier - interested? jackinnj Solid State 15 27th March 2008 09:36 PM
New digital XO project borges Multi-Way 2 8th January 2008 06:49 PM


New To Site? Need Help?

All times are GMT. The time now is 03:35 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2