First post on the subject, but this has been in the invention stage for a few weeks now...
It all started when my venerable Accoustic Energy 5.1 went on the fritz last month. Thanks to kids, etc, it's the only bit of hifi I've got left and is used for everything: TV, music, blueray, dvds, ATV streaming, etc, etc. My immediate response was to start trawling the usual suppliers trying to find a replacement home cinema amplifier on an extremely tight budget. Not only did I realise I'd be getting a load of rubbish for the cash (because I'm pretty skint - refer back to statement on kids), but I'd get a load of rubbish in a big metal box with far too many flashing lights, buttons, knobs, and everything else I'll never use and don't want sitting under the TV! Hence, only one solution, design my own...
This isn't going to be a spec-beating audiophile system with THD=0.0001%, its going to be a compromise solution, based on:
- cost (I'm skint)
- complexity (for diy build)
- power enough to fill a typically british living room (i.e. pokey)
- size and functionality (i.e. invisible, but very very audiable)
- 'good' audio quality (it must be as good as the digital sources driving it)
- energy efficiency (checked your electric bill lately?)
The main requirement is to beat the pants of any £200-£300 home cinema amp you could buy. Hopefully we'll achieve something much better.
Ok, lets start with the immediate fall-over condition: proprietary licensing. Anything with a Dolby badge needs a dolby license. Hence, even though you see plenty of 1-chip dolby decoders, you can't buy them without waving the magic license. Not only is this licenses expensive, but it's also very restrictive in publication. I did a job a few years back that needed it and was told flatly "no" because my middle name isn't Pioneer or Kenwood.
So, a couple of beers later and the solution presents itself. Dolby Digital, aka AC-3 got standardised for ATSC audio transmission. As such, its not a free protocol (provided you don't call it Dolby): ATSC A/52. GNU already has an open source decoder library for doing this, as they have for DTS (but don't call it DTS, otherwise ker-ching).
This means we're going to need a reasonably decent micro with I2S ports galore. More later...
Next fall-over condition is complexity of design. HMDI inputs are right out, at the moment. We're talking gigabit LVDS SerDes signalling - this is definitely out the realm of my PCB skills and any micro I could buy. Perhaps the future evolution would be to put a cheap FPGA dev board in there, performing audio extraction. So, we'll stick with S/PDIF.
The overall system design works out to be:
- S/PDIF input processing with source selection
- central microprocessor
- 6 channel DAC output and electronic attenuators (need customisable attenuators for levelling 6 channels)
- user interface / control
- power supply
- amp x 6
I've attached an overview design showing very simplified makeup.
For the micro, I'm thinking of something very very new - actually came out in perfect time for this project: Beaglebone (google it)!! This is an embedded micro board running linux. The processor is an evolution of the TI OMAP processor which is essentially a DSP+ARM combo. Perfect for audio processing and comes with lots of I2S interfaces!!
I'll slowly upload the design over the next few days.
Got the SPDIF design: Wolfsson WM8805 - low cost, good performance and easily available
Got the micro hopefully sorted: (Beagle)
Got the DAC/attenuators: Wolfsson WM8766 - perfect fit!
Currently worrying about the amp. My immediate thought is to use Class D. Partially because they don't need massive heat sinks (breaking size requirement). Partially because they're very efficient (saving my electric), and mostly because my current amp is a class D variety (although proprietary solution) and I'm still in awe of the detail/clarify of reproduction.
I spent a long time worrying about power ratings. My current system has never seen the volume knob go higher than a quarter turn, so this is a good benchmark. Found it very difficult to get a proper power spec for it, so looked at the PSU rating instead, and boy was I suprised. The entire 5.1 system, including sub power, is running off a 12v/5A single rail supply!!! So much for "it needs to be 100w/channel"!! I'm hazarding a guess that because class D is ~85% efficient, you can effectively half the power requirements from a class A which is ~40% efficient...
Anyway, any comments/thoughts on the amp would be appreciated. Ideally needs to be single rail supply, to fit with Switch-mode requirement (got my eyes on an18V/8A from HP docking station or 15A/10A meanwell on RS). TK2050 chipset could be perfect, but thats 3x £16 just for the chipsets alone, unless I can find another supplier.
Right, I'll stop waffling. The point of posting is to get thoughts/feedback/suggestions/ideas/help/whatever??? If I'm going to the time/expense of producing my own designs, they may as well get used more than once!!
Mat
It all started when my venerable Accoustic Energy 5.1 went on the fritz last month. Thanks to kids, etc, it's the only bit of hifi I've got left and is used for everything: TV, music, blueray, dvds, ATV streaming, etc, etc. My immediate response was to start trawling the usual suppliers trying to find a replacement home cinema amplifier on an extremely tight budget. Not only did I realise I'd be getting a load of rubbish for the cash (because I'm pretty skint - refer back to statement on kids), but I'd get a load of rubbish in a big metal box with far too many flashing lights, buttons, knobs, and everything else I'll never use and don't want sitting under the TV! Hence, only one solution, design my own...
This isn't going to be a spec-beating audiophile system with THD=0.0001%, its going to be a compromise solution, based on:
- cost (I'm skint)
- complexity (for diy build)
- power enough to fill a typically british living room (i.e. pokey)
- size and functionality (i.e. invisible, but very very audiable)
- 'good' audio quality (it must be as good as the digital sources driving it)
- energy efficiency (checked your electric bill lately?)
The main requirement is to beat the pants of any £200-£300 home cinema amp you could buy. Hopefully we'll achieve something much better.
Ok, lets start with the immediate fall-over condition: proprietary licensing. Anything with a Dolby badge needs a dolby license. Hence, even though you see plenty of 1-chip dolby decoders, you can't buy them without waving the magic license. Not only is this licenses expensive, but it's also very restrictive in publication. I did a job a few years back that needed it and was told flatly "no" because my middle name isn't Pioneer or Kenwood.
So, a couple of beers later and the solution presents itself. Dolby Digital, aka AC-3 got standardised for ATSC audio transmission. As such, its not a free protocol (provided you don't call it Dolby): ATSC A/52. GNU already has an open source decoder library for doing this, as they have for DTS (but don't call it DTS, otherwise ker-ching).
This means we're going to need a reasonably decent micro with I2S ports galore. More later...
Next fall-over condition is complexity of design. HMDI inputs are right out, at the moment. We're talking gigabit LVDS SerDes signalling - this is definitely out the realm of my PCB skills and any micro I could buy. Perhaps the future evolution would be to put a cheap FPGA dev board in there, performing audio extraction. So, we'll stick with S/PDIF.
The overall system design works out to be:
- S/PDIF input processing with source selection
- central microprocessor
- 6 channel DAC output and electronic attenuators (need customisable attenuators for levelling 6 channels)
- user interface / control
- power supply
- amp x 6
I've attached an overview design showing very simplified makeup.
For the micro, I'm thinking of something very very new - actually came out in perfect time for this project: Beaglebone (google it)!! This is an embedded micro board running linux. The processor is an evolution of the TI OMAP processor which is essentially a DSP+ARM combo. Perfect for audio processing and comes with lots of I2S interfaces!!
I'll slowly upload the design over the next few days.
Got the SPDIF design: Wolfsson WM8805 - low cost, good performance and easily available
Got the micro hopefully sorted: (Beagle)
Got the DAC/attenuators: Wolfsson WM8766 - perfect fit!
Currently worrying about the amp. My immediate thought is to use Class D. Partially because they don't need massive heat sinks (breaking size requirement). Partially because they're very efficient (saving my electric), and mostly because my current amp is a class D variety (although proprietary solution) and I'm still in awe of the detail/clarify of reproduction.
I spent a long time worrying about power ratings. My current system has never seen the volume knob go higher than a quarter turn, so this is a good benchmark. Found it very difficult to get a proper power spec for it, so looked at the PSU rating instead, and boy was I suprised. The entire 5.1 system, including sub power, is running off a 12v/5A single rail supply!!! So much for "it needs to be 100w/channel"!! I'm hazarding a guess that because class D is ~85% efficient, you can effectively half the power requirements from a class A which is ~40% efficient...
Anyway, any comments/thoughts on the amp would be appreciated. Ideally needs to be single rail supply, to fit with Switch-mode requirement (got my eyes on an18V/8A from HP docking station or 15A/10A meanwell on RS). TK2050 chipset could be perfect, but thats 3x £16 just for the chipsets alone, unless I can find another supplier.
Right, I'll stop waffling. The point of posting is to get thoughts/feedback/suggestions/ideas/help/whatever??? If I'm going to the time/expense of producing my own designs, they may as well get used more than once!!
Mat
Mat,
I'm gonna put some infos toghether to build something like your amp.
Unlike yours, mine will be slightly simpler because (for me) TI processors are too complicated to program.
I'd go with a PIC18 Mcu controlling ST audio processors via I2C bus.
I just saw the S/PDIF I/F schematic and what I don't understand is why you used so many optical inputs.
Is just one not sufficient to get all the 4.1 data from eg.a DVD player?
Cheers and good work!
Frank.
I'm gonna put some infos toghether to build something like your amp.
Unlike yours, mine will be slightly simpler because (for me) TI processors are too complicated to program.
I'd go with a PIC18 Mcu controlling ST audio processors via I2C bus.
I just saw the S/PDIF I/F schematic and what I don't understand is why you used so many optical inputs.
Is just one not sufficient to get all the 4.1 data from eg.a DVD player?
Cheers and good work!
Frank.
I'm basing most of the structural design on Eurocard-format 160x100mm sized PCBs. It makes life a lot simpler when mounting, as hopefully I can use eurocard enclosures with pcb guide rails already in place (look for verotec lbx series). This will work out as a single digital board, a single analogue board, then seperately mounted computer/control interface/etc.
The digital board schematic and PCB layout is now complete and attached. *NOTE: it's not tested yet*. This board contains a 6 channel S/PDIF to I2S converter (the input), then the 6 channel I2S to Audio (the output) Control of both interfaces is via common SPI serial bus, but each I2S line is kept seperate.
I've been testing a new (free) EDA tool for this application: Designspark. Alas most of my work stuff is done on CADStar, which I can't use for a home application. Designspark isn't bad for a free application!
So, with this post you get the digital board:
- schematic (pdf)
- pcb (pdf)
- gerber files
- entire designspark project (please point designspark to the included libraries before opening the project).
Enjoy!
Mat
The digital board schematic and PCB layout is now complete and attached. *NOTE: it's not tested yet*. This board contains a 6 channel S/PDIF to I2S converter (the input), then the 6 channel I2S to Audio (the output) Control of both interfaces is via common SPI serial bus, but each I2S line is kept seperate.
I've been testing a new (free) EDA tool for this application: Designspark. Alas most of my work stuff is done on CADStar, which I can't use for a home application. Designspark isn't bad for a free application!
So, with this post you get the digital board:
- schematic (pdf)
- pcb (pdf)
- gerber files
- entire designspark project (please point designspark to the included libraries before opening the project).
Enjoy!
Mat
Frank,
I'd be interested in knowing more at your mention of ST Audio Processor?
If you go with the full TI OMAP processor, then yes - it can get complex if you start delving into the DSP side. However, I'm probably going to be using "beaglebone" - when it finally comes in stock in the UK, which is a more standard ARM Cortex processor. It means essentially I'm programming Linux and only need to worry about the hardware for input/output. Remember: if you want to decode Dolby/DTS, then you need to do it in software, because you won't be allowed to by any hardware decoders without flashing your dolby license (especially not from ST - they're very pedantic and controlling of their markets).
For the S/PDIF inputs, I've settled on 4 optical and 2 co-ax. You only need a single input per device (dvd/etc), but I've personally got a few different devices I want to swap between (and it just so happens most of them are optical). If you only need a single input, then things could get much simpler, as you can either hardware-lock the WM8805, or even better; swap to a WM8804 which is a simpler device.
A total of x8 S/PDIF intputs can be supported on the chipset. I've just gone with x6 because it fits the PCB size. If a different combination of optical/co-ax is needed, just change the schematic - I've included the entire project here.
Mat.
I'd be interested in knowing more at your mention of ST Audio Processor?
If you go with the full TI OMAP processor, then yes - it can get complex if you start delving into the DSP side. However, I'm probably going to be using "beaglebone" - when it finally comes in stock in the UK, which is a more standard ARM Cortex processor. It means essentially I'm programming Linux and only need to worry about the hardware for input/output. Remember: if you want to decode Dolby/DTS, then you need to do it in software, because you won't be allowed to by any hardware decoders without flashing your dolby license (especially not from ST - they're very pedantic and controlling of their markets).
For the S/PDIF inputs, I've settled on 4 optical and 2 co-ax. You only need a single input per device (dvd/etc), but I've personally got a few different devices I want to swap between (and it just so happens most of them are optical). If you only need a single input, then things could get much simpler, as you can either hardware-lock the WM8805, or even better; swap to a WM8804 which is a simpler device.
A total of x8 S/PDIF intputs can be supported on the chipset. I've just gone with x6 because it fits the PCB size. If a different combination of optical/co-ax is needed, just change the schematic - I've included the entire project here.
Mat.
Hi Mat,
The "Core" of my project will be a TDA7415 (497-11669-1-ND - 8,6€) controlled by a PIC18 Mcu (PIC18F47J53-I/PT-ND - 4,3€) programmed in C.
I will keep the things as simple as possible since this is the first digital amp I do, so no Dolby/DTS decoding.
Unfortunately, few days ago while searching the parts, I discovered that a lot of them were not accessible for DiYers.
Just to mention suppliers like Digikey,Mouser or Farnell, they have most of the intresting parts "not in stock" forcing you to buy a thousend or more pieces.
Also the Japanese parts are quite unavailable.
There are a lot good IC's with intresting functions made by Rhom, JRC, AsaiKasei etc....
Take a look here :
USB Audio Decoder ICs | Audio ICs | ICs | ROHM CO., LTD.
These chips are nowhere to be found plus, they don't give you the data sheet as well.
Said that, the part list will be determined by the "what's on the shelf" choiche.
I'm also getting infos on how to solder 0,5mm pitch LQFP64 without going crazy. That's gonna be funny. LOL.
The "Core" of my project will be a TDA7415 (497-11669-1-ND - 8,6€) controlled by a PIC18 Mcu (PIC18F47J53-I/PT-ND - 4,3€) programmed in C.
I will keep the things as simple as possible since this is the first digital amp I do, so no Dolby/DTS decoding.
Unfortunately, few days ago while searching the parts, I discovered that a lot of them were not accessible for DiYers.
Just to mention suppliers like Digikey,Mouser or Farnell, they have most of the intresting parts "not in stock" forcing you to buy a thousend or more pieces.
Also the Japanese parts are quite unavailable.
There are a lot good IC's with intresting functions made by Rhom, JRC, AsaiKasei etc....
Take a look here :
USB Audio Decoder ICs | Audio ICs | ICs | ROHM CO., LTD.
These chips are nowhere to be found plus, they don't give you the data sheet as well.
Said that, the part list will be determined by the "what's on the shelf" choiche.
I'm also getting infos on how to solder 0,5mm pitch LQFP64 without going crazy. That's gonna be funny. LOL.
Last edited:
For amp combine this and this to get 6 channel.
This one is even simpler & cheaper : 7.1 TK2050
Have no idea about the quality, but I'm happy with other TK2050 board.
This one is even simpler & cheaper : 7.1 TK2050
Have no idea about the quality, but I'm happy with other TK2050 board.
I'm looking at this exact board for 7.1 from Asus Xonar as the preamp , seems a very inexpensive way of making something that arcam sell for £1000's although I do have Arcam 250 AVR and thinking of stripping out video or cloning just the sound stage.
Also I came across this only today trawling at work , may be some tip and trick in the article you could use ?
http://www.kafka.elektroda.eu/podstrony/pgapreamp.php
Last edited:
Interesting links.
For parts availability, I'm sticking strictly to items available from CPC/Farnell or RS. If these guys don't do it, you can guarantee its going to be painful to buy. They're pretty straight forward to order from.
Class D is definitely the way to go. As much as I'd like to run a set of gainclones, the heat dissispation of 6 channels means massive boxes with big heatsinks, kind of defeats the object. I've finally decided on choice of amp: going to be a set of 3 Tripath TA2021b boards from DIY! HIGH END! AUDIOPHILE AMLIFIER BASED ON THE MOST ADVANCED TRIPATH 'S CLASS-T ® TA2021B CHIP! (currently running a special offer 3for2 this month). Will take a risk at that price!
Power supplies are the other worry, especially if you spec 100w/channel amps. I've settled on a SMS (Meanwell GS220-15). Might have to smooth things down a little...
And finally,
Got the first quote back from my PCB production company: £183 for a 1off digitial input board!!!!! (or £58 each if I buy 5 or more). Currently checking why it's so damn expensive - they might be silkscreening, which adds a £75 offset to the order.
Will digest the links properly tonight. Good luck with LQFP64 soldering - you're braver than I am!
Mat
For parts availability, I'm sticking strictly to items available from CPC/Farnell or RS. If these guys don't do it, you can guarantee its going to be painful to buy. They're pretty straight forward to order from.
Class D is definitely the way to go. As much as I'd like to run a set of gainclones, the heat dissispation of 6 channels means massive boxes with big heatsinks, kind of defeats the object. I've finally decided on choice of amp: going to be a set of 3 Tripath TA2021b boards from DIY! HIGH END! AUDIOPHILE AMLIFIER BASED ON THE MOST ADVANCED TRIPATH 'S CLASS-T ® TA2021B CHIP! (currently running a special offer 3for2 this month). Will take a risk at that price!
Power supplies are the other worry, especially if you spec 100w/channel amps. I've settled on a SMS (Meanwell GS220-15). Might have to smooth things down a little...
And finally,
Got the first quote back from my PCB production company: £183 for a 1off digitial input board!!!!! (or £58 each if I buy 5 or more). Currently checking why it's so damn expensive - they might be silkscreening, which adds a £75 offset to the order.
Will digest the links properly tonight. Good luck with LQFP64 soldering - you're braver than I am!
Mat
Its funny how the same projects come around at the same time of year. Dark nights, I guess.
We'll, 12months on from the original project and I've not been entirely idle, although no where near as productive as I wanted - mainly thanks to two house moves this year (thanks landlords).
For the sake of keeping my ears (and wife) happy, I decided to start with something much simpler - 2 channels only. What I've ended up with I'm now almost happy with, and has been. I posted the schematics on a different post here, but it's a toslink-in, speaker-out solution. I used baby-brother versions of the chips for this project - just a single SPDIF input and dual-channel DAC with attenuator, both Wolfson chips. System control via a little 16bit uC, which handles remote control inputs. I'm smug, mainly because it's proved some of the concepts for the "big" design.
Progressing the decoding stage, it looks like there won't be quite as much programming as planned. using a embedded Linux platform (still on beaglebone), both ALSA, GStreamer and JACK audio frameworks look like they could be configured to do exactly what I need. I'm struggling to get decent information on these (idiots guides), but all the lego pieces are definitely there.
What I can say from the 2 channel design is that those original german Tripath boards I found sound utterly fabulous for the money, powered from the big Meanwell SMP. It's breaking every law in audio design, work sounding fab at the moment.
So, in conclusion, the amps are de-risked, the DACS and SPDIF chipsets are pretty well de-risked, the pre-amp stage works, although needs refining. All thats left is to upgrade the DACS and prove the microprocessor stage (ok, thats the complex bit)...
Mat
We'll, 12months on from the original project and I've not been entirely idle, although no where near as productive as I wanted - mainly thanks to two house moves this year (thanks landlords).
For the sake of keeping my ears (and wife) happy, I decided to start with something much simpler - 2 channels only. What I've ended up with I'm now almost happy with, and has been. I posted the schematics on a different post here, but it's a toslink-in, speaker-out solution. I used baby-brother versions of the chips for this project - just a single SPDIF input and dual-channel DAC with attenuator, both Wolfson chips. System control via a little 16bit uC, which handles remote control inputs. I'm smug, mainly because it's proved some of the concepts for the "big" design.
Progressing the decoding stage, it looks like there won't be quite as much programming as planned. using a embedded Linux platform (still on beaglebone), both ALSA, GStreamer and JACK audio frameworks look like they could be configured to do exactly what I need. I'm struggling to get decent information on these (idiots guides), but all the lego pieces are definitely there.
What I can say from the 2 channel design is that those original german Tripath boards I found sound utterly fabulous for the money, powered from the big Meanwell SMP. It's breaking every law in audio design, work sounding fab at the moment.
So, in conclusion, the amps are de-risked, the DACS and SPDIF chipsets are pretty well de-risked, the pre-amp stage works, although needs refining. All thats left is to upgrade the DACS and prove the microprocessor stage (ok, thats the complex bit)...
Mat
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.