> 1. System must be independent from PC and OS.
> These thing become obsolete in a few months, so I'd avoid micro-atx board and similar.
Yes and no ; open source software doesn't die because the manufacturer drops support.
> 2. An old dedicated PIII based PC has no cost, but is enough for audio
> purposes.
Certainly (if you don't ask it to play H264 video or do digital room correction)... and these old chips need less power, which means fanless supplies, which is good.
Just like my RME card. This is neat. Another benefit of pro boards is that (unlike Soundblasters) they will not do funky stuff to your data unless you specifically ask for it.
Yes you could, but you shouldn't.
PC and ground should be completely isolated so that the PC noise does not go into the DAC. Therefore, you must have isolated transmission of
- clock from DAC to PC (I assume the master clock is in your DAC which is the way to go)
- audio data from PC to DAC
Now, a clock signal and SPDIF have no DC component so you can use cheap pulse transformers, providing a very good isolation level without compromises. For your clock, rip an isolation transformer out of an old Ethernet card, for instance.
Audio data can be transmitted either as spdif (transformer isolated) or optical.
However if you want to transmit I2S you'll need high speed optocouplers because these signals have a DC components ; these chips are expensive.
Therefore, I advise you to simply keep using your toslink for audio data and use a transformer or optical for slaving the soundcard to the dac clock.
Since your master clock is in your DAC the jitter added by the toslink is not a problem (provided you have proper reclocking).
Enjoy !
> These thing become obsolete in a few months, so I'd avoid micro-atx board and similar.
Yes and no ; open source software doesn't die because the manufacturer drops support.
> 2. An old dedicated PIII based PC has no cost, but is enough for audio
> purposes.
Certainly (if you don't ask it to play H264 video or do digital room correction)... and these old chips need less power, which means fanless supplies, which is good.
Just like my RME card. This is neat. Another benefit of pro boards is that (unlike Soundblasters) they will not do funky stuff to your data unless you specifically ask for it.
Yes you could, but you shouldn't.
PC and ground should be completely isolated so that the PC noise does not go into the DAC. Therefore, you must have isolated transmission of
- clock from DAC to PC (I assume the master clock is in your DAC which is the way to go)
- audio data from PC to DAC
Now, a clock signal and SPDIF have no DC component so you can use cheap pulse transformers, providing a very good isolation level without compromises. For your clock, rip an isolation transformer out of an old Ethernet card, for instance.
Audio data can be transmitted either as spdif (transformer isolated) or optical.
However if you want to transmit I2S you'll need high speed optocouplers because these signals have a DC components ; these chips are expensive.
Therefore, I advise you to simply keep using your toslink for audio data and use a transformer or optical for slaving the soundcard to the dac clock.
Since your master clock is in your DAC the jitter added by the toslink is not a problem (provided you have proper reclocking).
Enjoy !
Hi Gary,
Yes I've a Wolfson 8740 EV board. It is quite cheap and sounds much better than E-mu DAC (Cirrus CS4398) that really is not bad.
This board has not a master clock. I have a master clock, but it is a second daughter card of Emu 1212M. I don't use it yet because I still don't know how connect cables. The Wolfson dac has just pins to receave a clock and external serial data with unpopulated stands for caps and resistors to adapt cables. If it senses an external clock it adapts to it. Otherwise extracts clock from spdif.
Both, Wolfson and Emu1212M would act as slave of an external master clock. Not necessarily the Emu one. But I have it, and would try if I only knew how to do. Tomorrow I could buy a Tent's master clock and put it into tha DAC case.
Hy peufeu,
What I mean is that new software often requires new hardware. For istance Emu wants XP, so you can't use old Pentium I (that for Foobar is enough and needs only a passive heatsink). Another example are SATA hard disks: best HD are now SATA and in few months IDE HD will became obsolete. So PIII and older P4 motherboards are out and all micro ATX you can find today. I just wait my office P4 become obsolete with Windows Vista.
With P4 dual core too, it is better to make execute specific tasks to custom cards. Room correction is made by Emu DSP, without using CPU resources. The same for video: since years video cards have far better processors than PC CPU and use very little CPU resources.
Why not? Pro cards have an isolating transformer on board. Emu is very quiet, even if I play with its internal dac (it has balanced outputs too).
If I could use SDIN SCLK and LRCK signals from Emu, with a clean external clock, I would have the perfect transport. No need to design from scratch new hardware.
Spdif is obsolete. I have some pro recordings on a DAT at 24bit 192Khz and can't hear to them because spidif has only 24bit 96Khz. The same for DVD-A. Wolfson DAC can handle 192Khz and can be controlled via software (parallel port) to switch to different sample frequencies and bit rates.
If you need more channels, Emu has can handle 32 ASIO channels at 48Hz and four or six, if I remember, at 24bit 192khz. To each channel you can apply all sort of filters using Emu DSP.
Please can you help me with connecting SDIN SCLK and LRCK to Wolfson DAC?
Ciao
Thomas
Yes I've a Wolfson 8740 EV board. It is quite cheap and sounds much better than E-mu DAC (Cirrus CS4398) that really is not bad.
This board has not a master clock. I have a master clock, but it is a second daughter card of Emu 1212M. I don't use it yet because I still don't know how connect cables. The Wolfson dac has just pins to receave a clock and external serial data with unpopulated stands for caps and resistors to adapt cables. If it senses an external clock it adapts to it. Otherwise extracts clock from spdif.
Both, Wolfson and Emu1212M would act as slave of an external master clock. Not necessarily the Emu one. But I have it, and would try if I only knew how to do. Tomorrow I could buy a Tent's master clock and put it into tha DAC case.
peufeu said:
Hy peufeu,
What I mean is that new software often requires new hardware. For istance Emu wants XP, so you can't use old Pentium I (that for Foobar is enough and needs only a passive heatsink). Another example are SATA hard disks: best HD are now SATA and in few months IDE HD will became obsolete. So PIII and older P4 motherboards are out and all micro ATX you can find today. I just wait my office P4 become obsolete with Windows Vista.
With P4 dual core too, it is better to make execute specific tasks to custom cards. Room correction is made by Emu DSP, without using CPU resources. The same for video: since years video cards have far better processors than PC CPU and use very little CPU resources.
Why not? Pro cards have an isolating transformer on board. Emu is very quiet, even if I play with its internal dac (it has balanced outputs too).
If I could use SDIN SCLK and LRCK signals from Emu, with a clean external clock, I would have the perfect transport. No need to design from scratch new hardware.
Spdif is obsolete. I have some pro recordings on a DAT at 24bit 192Khz and can't hear to them because spidif has only 24bit 96Khz. The same for DVD-A. Wolfson DAC can handle 192Khz and can be controlled via software (parallel port) to switch to different sample frequencies and bit rates.
If you need more channels, Emu has can handle 32 ASIO channels at 48Hz and four or six, if I remember, at 24bit 192khz. To each channel you can apply all sort of filters using Emu DSP.
Please can you help me with connecting SDIN SCLK and LRCK to Wolfson DAC?
Ciao
Thomas
Thanks. I wondered wether I should evaluate CS4398 or not for my next DAC... It's the flagship from Cirrus (but it still uses switched capacitor filters). I guess I won't try it
I thought your master clock was more DIY... hum, can you post more information ?
What I mean is that new software often requires new hardware. For istance Emu wants XP, so you can't use old Pentium I (that for Foobar is enough and needs only a passive heatsink). Another example are SATA hard disks: best HD are now SATA and in few months IDE HD will became obsolete. So PIII and older P4 motherboards are out and all micro ATX you can find today. I just wait my office P4 become obsolete with Windows Vista.
That's why I prefer Linux. Besides :
- An audio PC does not need a harddrive, only an ethernet port and a net-boot, or a flash drive
If the DSP does its thing with enough bits of precision, that's cool.
What I meant is that you should avoid to connect your audio gear ground to your PC ground because the PC ground potential contains wideband noise that you really don't want to get into your audio system.
Yes, absolutely ; this is the only way to get 24/192 so you can experiment with ribbon cables but for the final version I still recomment using some form of isolation like optocouplers, or differential digital signalling.
So now I suggest you look at what signals the EMU sends, especially the timings, and look closely at the Wolfson datasheet to see if your DAC can understand them and how to tell it which format to expect. This should be quite simple, but I hope you have a scope...
Yeah, but first make sure you read the datasheets and know what signals your soundcard sends. This should not be too hard.
You'll need a master clock to put in your DAC. In fact you'll need two because 44.1k and 48k based sample rates are not compatible... maybe you can hack your E-Mu master clock, or just use a pair of Tent Clocks, having only one of the two powered at any given time (that's what I plan to do)
peufeu said:
Wolfson DAC has no clock on board. It has a receiver to get spdif, I2s and other formats. But it has pins to get directly digital signals and clock, bypassing the receiver.
I've two seagate HD 300Gb each, almost full. Expensive to hire so much space on a remote server... No scope
.Emu has no datasheets. And they didn't want to answer when I asked these informations. Emu, before they were bought by Creative, had the best DSP for pro samplers they produced in stand alone sampler machines (these were indeed custom PC with custom OS, ram and hard disk). Their technology is a secret.
So, with empyrical approch, i followed the traces that arrive from the cable that connects Emu board to the CS4398 chip on the daughter board. This way I know which wires from Emu board carry SDIN SCLK and LRCK signals, that are signals that CS4398 accepts. The cable is the same flat one used for floppy disks.
These three wires would be connected to Wolfson DAC instead of Cirrus chip.
The problem is compatibility. In datasheet from Cirrus and Wolfson are well described signal formats. But I can't understand
Here are the datasheets:
Wolfson 8740 EV board
Wolfson 8740 chip
CirrusCS4398 chip
Emu second daughter board is a complete masterclock. It has 44.1k. 48k, 96k and 192k. It uses a Xilinx fpga to provide these clocks (the same used by Digidesign). The selection of clocks is via software. It has two BNC, one for output clock, one for input clock.
This clock is not connected to PCI bus, only to Emu card and powered by it. I could place it in the dac board, if I want, and power it with external supply.
So, all I'd need is to know if signals are compatible, and wich caps and resistors to use for coupling cables on Wolfson board.
If you are looking for a new DAC, Wolfson is the best oversampling one avalaible. To have a better one the only way is to program an AD Sharc DSP.
Ciao
Thomas
OK, I understand...
If you want to get serious in DIY you'll need a scope ; I bought mine on ebay (a very good analog Tektronix, 100 MHz x2) for about $200. Add $100 to get a professional to gut it, replace the old capacitors and calibrate it. It's a very good investment.
There are some formats that the Crystal chip uses that the Wolfson can't decode (the right-justified modes are identical, but in the left-justified modes, the Wolfson needs a blank bit before the data whereas the Crystal does not).
If I were you, I'll buy a scope. But you can also be rock'n'roll and just connect the wolfson to the e-mu board, if you figured out the pinout correctly nothing should burn. Just check with a voltmeter if the pins are not +/- 15V supplies...
Then once it's connected, the DAC will probably put out some garbled audio ; you can then use the parallel port tu put it in the right format (try them all).
If you want to get serious in DIY you'll need a scope ; I bought mine on ebay (a very good analog Tektronix, 100 MHz x2) for about $200. Add $100 to get a professional to gut it, replace the old capacitors and calibrate it. It's a very good investment.
There are some formats that the Crystal chip uses that the Wolfson can't decode (the right-justified modes are identical, but in the left-justified modes, the Wolfson needs a blank bit before the data whereas the Crystal does not).
If I were you, I'll buy a scope. But you can also be rock'n'roll and just connect the wolfson to the e-mu board, if you figured out the pinout correctly nothing should burn. Just check with a voltmeter if the pins are not +/- 15V supplies...
Then once it's connected, the DAC will probably put out some garbled audio ; you can then use the parallel port tu put it in the right format (try them all).
Currently in development.
The thing now has a website :
http://audio.peufeu.com/
I believe you can even add comments if I managed to configure this thing correctly...
The thing now has a website :
http://audio.peufeu.com/
I believe you can even add comments if I managed to configure this thing correctly...
In the internal EMU cards the problem is not the CS4398. CS4398's sound very nice but it needs CLEAN supplies. PC ground is a mess... I know someone that has an 1820m using an external 12V/-12V power supply he made - completely isolated from the PC ground (there is a proprietary protocol similar to Ethernet between the PCI and the external box - Creative uses XILINX spartan chips for routing) - and it sounds fantastic.
Peufeu,
I believe you have though out this solution many times but what are the reasons you want to use a special board with custom-made outputs instead of a single-board computer (pref. miniITX) with PCI sound card with its i2s channels taken out and fed to external DACs? I have seen this solution discussed here several times http://www.diyaudio.com/forums/showthread.php?threadid=22019
http://www.diyaudio.com/forums/showthread.php?threadid=22556
I really enjoy your thread and webpage and for a while have been thinking of trying a DRC/3-way crossover based on linux PC - kind of what you are working on now. Thanks for your comments.
I believe you have though out this solution many times but what are the reasons you want to use a special board with custom-made outputs instead of a single-board computer (pref. miniITX) with PCI sound card with its i2s channels taken out and fed to external DACs? I have seen this solution discussed here several times http://www.diyaudio.com/forums/showthread.php?threadid=22019
http://www.diyaudio.com/forums/showthread.php?threadid=22556
I really enjoy your thread and webpage and for a while have been thinking of trying a DRC/3-way crossover based on linux PC - kind of what you are working on now. Thanks for your comments.
Hm, why ?
- For the fun & challenge
- For learning new stuff (FGPAs...)
- Because isolation is easier
- Possibility of oversampling filters in the FPGA
- As many channels as I want
- More flexibility, embedded volume control etc
The Ethernet part is done. I'm coding I2S transmitters now, then the only remaining thing will be software.
http://audio.peufeu.com/node/46
- For the fun & challenge
- For learning new stuff (FGPAs...)
- Because isolation is easier
- Possibility of oversampling filters in the FPGA
- As many channels as I want
- More flexibility, embedded volume control etc
The Ethernet part is done. I'm coding I2S transmitters now, then the only remaining thing will be software.
http://audio.peufeu.com/node/46
Good question.
The TAS5518 operates as a slave only: input are I2S signals and master clock.
PWM Processing
• 32-Bit Processing PWM Architecture With 40 Bits of Precision
• 8x Oversampling With 5th Order Noise Shaping at 32 – 48 kHz, 4x Oversampling at 88.2 kHz, and 96 kHz
and 2x Oversampling at 176.4 kHz and 192 kHz
http://www-s.ti.com/sc/techlit/SLES115
BWRX has a design for the TAS5261 power stage. We are trying to build it:
http://www.diyaudio.com/forums/showthread.php?postid=1028657#post1028657
The TAS5518 operates as a slave only: input are I2S signals and master clock.
PWM Processing
• 32-Bit Processing PWM Architecture With 40 Bits of Precision
• 8x Oversampling With 5th Order Noise Shaping at 32 – 48 kHz, 4x Oversampling at 88.2 kHz, and 96 kHz
and 2x Oversampling at 176.4 kHz and 192 kHz
http://www-s.ti.com/sc/techlit/SLES115
BWRX has a design for the TAS5261 power stage. We are trying to build it:
http://www.diyaudio.com/forums/showthread.php?postid=1028657#post1028657
I just finished coding the I2S encoder for mono chips like PCM1704, or chips with two data lines like TDA1541. A stereo version takes about 45 slices.
I have implemented word clock delaying and bit clock gating for TDA1541, along with an option to send the data bits interleaved with their inverted version to reduce digital feedthrough.
These options are controlled from software, of course.
I plan on doing a test dac board with all the interesting dac chips I can lay my hands on...
I have implemented word clock delaying and bit clock gating for TDA1541, along with an option to send the data bits interleaved with their inverted version to reduce digital feedthrough.
These options are controlled from software, of course.
I plan on doing a test dac board with all the interesting dac chips I can lay my hands on...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.