I'm building a portable media player, and am looking for a DAC muich like the Wolfson WM8758, but better. You'll note the figures here show 100/92dB SNR on the DAC and ADC respectively.
So, battery life is important, and PSU requirements would preferrably be 3.3V compatible, and I wondered if I could have my cake and eat it with something like the 8758 but with better SNR. It's on some ipods as far as I can tell, and I wanted to make something a bit better audio quality wise.
Would like:
Headphone driver + line out
3.3V supply if possible
I2S audio input
Nice to have:
spdif in / out
effects / enhancements like the wm8758 eq / 3d etc
Suggestions welcome
So, battery life is important, and PSU requirements would preferrably be 3.3V compatible, and I wondered if I could have my cake and eat it with something like the 8758 but with better SNR. It's on some ipods as far as I can tell, and I wanted to make something a bit better audio quality wise.
Would like:
Headphone driver + line out
3.3V supply if possible
I2S audio input
Nice to have:
spdif in / out
effects / enhancements like the wm8758 eq / 3d etc
Suggestions welcome
With 16bit source material the noise floor on the recording is highly unlikely to be better than -93dB if the appropriate dither's been used. So are you wanting better SNR because your player is working with 24bit files?
Do you take the view that audio quality is in the main a matter of SNR figure?
Do you take the view that audio quality is in the main a matter of SNR figure?
Interesting points / thoughts.
Yes, I'm looking for something that will handle 24bit files, FLAC being the most important / likely format.
So far, I've built a WAV file player on the STM32F407 discovery board with the code up on github here. Plays 44.1kHz wav files from an attached SD card in SPI mode and runs through the onboard CS43L22 codec. Use EM::Blocks for the project, the pins on which to attach the SD card and navigation buttons are in the hardware.c file iirc.
Now my plan is to use the stm32f429 discovery board to stream through another codec board using I2S. So the codec would preferrably have the features above such as HP output, supply voltage etc.
I've looked into the PCM1739, which has the basics of what I want, although SPDIF and some filtering options would be good. But, the basics are solid.
I'm fine with mixed signal PCB design, but will have to delve into the notes for high quality PCB layout for this obviously, including the differential to single ended drivers. I'd like to use the LME49600/49610 as HP output drivers, and perhaps LME4972X as differential buffer for the 496XX output.
So, the LME op-amps are going to require +/- 2.5V which means the 3.3V supply requirement isn't going to be firm, if I have to use 5V.
Any other suggestions would be good / helpful, looks like something on the Burr-Brown line at TI might fill the specs...
Yes, I'm looking for something that will handle 24bit files, FLAC being the most important / likely format.
So far, I've built a WAV file player on the STM32F407 discovery board with the code up on github here. Plays 44.1kHz wav files from an attached SD card in SPI mode and runs through the onboard CS43L22 codec. Use EM::Blocks for the project, the pins on which to attach the SD card and navigation buttons are in the hardware.c file iirc.
Now my plan is to use the stm32f429 discovery board to stream through another codec board using I2S. So the codec would preferrably have the features above such as HP output, supply voltage etc.
I've looked into the PCM1739, which has the basics of what I want, although SPDIF and some filtering options would be good. But, the basics are solid.
I'm fine with mixed signal PCB design, but will have to delve into the notes for high quality PCB layout for this obviously, including the differential to single ended drivers. I'd like to use the LME49600/49610 as HP output drivers, and perhaps LME4972X as differential buffer for the 496XX output.
So, the LME op-amps are going to require +/- 2.5V which means the 3.3V supply requirement isn't going to be firm, if I have to use 5V.
Any other suggestions would be good / helpful, looks like something on the Burr-Brown line at TI might fill the specs...
To answer your second question, SNR in itself isn't the only measure of quality, but it would be a good starting point, along with good THD figures.
What's on my design list is: (in no particular order)
* decent clock, low jitter
* good analogue routing / layout / power supply
* good DAC specs
These are applicable here without going into the whole build.
The CS4207 looks to be a good choice - more channels than I need, just lacking some of the WM8758 features
What's on my design list is: (in no particular order)
* decent clock, low jitter
* good analogue routing / layout / power supply
* good DAC specs
These are applicable here without going into the whole build.
The CS4207 looks to be a good choice - more channels than I need, just lacking some of the WM8758 features
I suppose I'm aiming to achieve listening satisfaction from a starting point of having good numbers... I don't mind the probing questions, they're good food for thought.
Feel free to try my project I linked to, the stm32f407 disco board uses an ST cortex M4F @ 168MHz, plays wav files @ 44.1kHz through the onboard CS43L22 and sounds ok using a pair of sennheiser HD202 headphones. I listened to a range of music ripped from CD, and the dynamic range seemed ok, so I started on a FLAC codec implementation. I've got down to the rice-goulomb encoding and no further at present. (C# FLAC file read and header decode for testing)
So, my thoughts at present are to make the audio codec part detachable from the main board, so I can try either the '4207 or the '8758. I've also got a homebuilt "gainclone" LM3886 amp for listening comparison, which sounds pretty good.
I'll post back any progress and results here unless someone comes up with a better alternative than the 4207 Cirrus dac. However, maybe Cirrus's acquisition of Wolfson will yield something good in the future.
I did also find a range of really good looking DACs from AKM including a lot of 32bit options, AK4375 looks great, but is CSP, AK4452 looks like it might fit the bill though.
Feel free to try my project I linked to, the stm32f407 disco board uses an ST cortex M4F @ 168MHz, plays wav files @ 44.1kHz through the onboard CS43L22 and sounds ok using a pair of sennheiser HD202 headphones. I listened to a range of music ripped from CD, and the dynamic range seemed ok, so I started on a FLAC codec implementation. I've got down to the rice-goulomb encoding and no further at present. (C# FLAC file read and header decode for testing)
So, my thoughts at present are to make the audio codec part detachable from the main board, so I can try either the '4207 or the '8758. I've also got a homebuilt "gainclone" LM3886 amp for listening comparison, which sounds pretty good.
I'll post back any progress and results here unless someone comes up with a better alternative than the 4207 Cirrus dac. However, maybe Cirrus's acquisition of Wolfson will yield something good in the future.
I did also find a range of really good looking DACs from AKM including a lot of 32bit options, AK4375 looks great, but is CSP, AK4452 looks like it might fit the bill though.
I started out my DAC design journey with a pretty similar starting point, that the best numbers would be a first stab. But then I didn't much care for the timbre of the S-D DACs I tried when compared to multibit ones. In particular they didn't do brass very well but I only noticed this fact when I played a brass heavy track (Tchaik 4th symph, 1st movement) through a TDA1543 for comparison.
Trouble is, there's no reasonably economical route to 24bit performance with a low chip-count. PCM1704 is way too expensive and NRND, PCM1702 is cheaper but obsolete, so its roll-your-own or accept the corruption of instrumental timbres. Which was why I was veering towards using the M4, to allow a home-brewed multibit DAC which accepted more than 16bits.
I did take a peek at your github but my C skills never were particularly competent and they've further atrophied through non use. I shall though keep an interest in your progress.
Trouble is, there's no reasonably economical route to 24bit performance with a low chip-count. PCM1704 is way too expensive and NRND, PCM1702 is cheaper but obsolete, so its roll-your-own or accept the corruption of instrumental timbres. Which was why I was veering towards using the M4, to allow a home-brewed multibit DAC which accepted more than 16bits.
I did take a peek at your github but my C skills never were particularly competent and they've further atrophied through non use. I shall though keep an interest in your progress.
I'm thinking in terms of using the M4 to marshall either an array of (relatively slower) DACs like TDA1387 or alternatively to pre-process data for a single much faster DAC (like an ADI TxDAC).
In the first case, the individual DACs are 16bits but with suitable pre-processing, something beyond 16bit performance probably can be achieved. Paralleling DACs reduces the noise, by summing the outputs of (say) 16 DACs, another 4 bits might be achievable. Having a local CPU means some degree of self-calibiration might be feasible.
In the second example, the TxDACs are all lower than 16bits (they max out at 14 but go down to 8) so need interpolation/upsampling to deliver appropriate audio quality, something that seems fairly well suited to the M4.
In the first case, the individual DACs are 16bits but with suitable pre-processing, something beyond 16bit performance probably can be achieved. Paralleling DACs reduces the noise, by summing the outputs of (say) 16 DACs, another 4 bits might be achievable. Having a local CPU means some degree of self-calibiration might be feasible.
In the second example, the TxDACs are all lower than 16bits (they max out at 14 but go down to 8) so need interpolation/upsampling to deliver appropriate audio quality, something that seems fairly well suited to the M4.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.