Today I'm officially released from the quarantine period which means I'm allowed to go outdoors 
I've spent the past few quarantined days not on DACs, rather on a digital project (which has been in my in-tray for a while) to see what's possible with ST's cheapest ARM CPU. I blogged about the STM32F030 (see here : Cheapest 32bit CPU to date... - diyAudio ) and planned to interface it to a DAC. Well its only taken 5+ years to get around to but this afternoon I got the result I was hoping for and its implementing a very rudimentary I2S pass-through.
The reason for the doubt and pondering is this device is so basic it doesn't support I2S, I had to coax its SPI interface to work in I2S mode. Perhaps I'll start another thread to talk about that though - I'll put the link up when I've opened it.
@pelopidas - glad you got your hiss sorted. I was going to suggest a test with an Audacity file which contains only dither, but as its solved now I won't describe that.
I've spent the past few quarantined days not on DACs, rather on a digital project (which has been in my in-tray for a while) to see what's possible with ST's cheapest ARM CPU. I blogged about the STM32F030 (see here : Cheapest 32bit CPU to date... - diyAudio ) and planned to interface it to a DAC. Well its only taken 5+ years to get around to but this afternoon I got the result I was hoping for and its implementing a very rudimentary I2S pass-through.
The reason for the doubt and pondering is this device is so basic it doesn't support I2S, I had to coax its SPI interface to work in I2S mode. Perhaps I'll start another thread to talk about that though - I'll put the link up when I've opened it.
@pelopidas - glad you got your hiss sorted. I was going to suggest a test with an Audacity file which contains only dither, but as its solved now I won't describe that.
Several things I have in mind for it. First is to slow down the I2S feed to my DACs, from the usual 64fs to 32fs. I made a circuit that did this a few years ago and I got the impression that the SQ improved with the slower clock. But that was on a much earlier DAC design with passive I/V, so I wonder if the apparent SQ lift will be anywhere noticeable on a PhiDAC Quad.
Second target is to create a digital volume control so that a future PhiDAC can be more 'stand-alone' in that it could work in conjunction with a CD player or SD card player.
Other ideas are to create a dual-mono and balanced DAC where one stack of chips is used to handle both phases of a single channel (either L or R) and to implement some kind of digital filtering - either to create an active XO or just to give better stop-band rejection of image frequencies.
On this last notion, I am still planning to create a much lower power implementation of the venerable old SAA7220. That chip is well known for being a major contributor to lousy SQ - not because of its filtering but because its a noisy power hog. The latest Cortex M0+ chip out of ST holds out the promise of being able to replicate the filtering of the 7220 but at perhaps just 5% of its power draw. The idea is to create a plug-in PCB which can replace the original chip for those upgrading old style CD players.
Beyond that I think the STM32F030 (and the later generation G030) could be made into digital audio 'lego' bricks to build more complex digital filters and also potentially be used to enhance the distortion performance of multibit DACs. They could also be the core of a DAC which uses not-for-audio DAC chips - those chips don't have I2S interfaces so the CPU can do the interface translation.
Any of those thoughts interesting to anyone else besides me?
Second target is to create a digital volume control so that a future PhiDAC can be more 'stand-alone' in that it could work in conjunction with a CD player or SD card player.
Other ideas are to create a dual-mono and balanced DAC where one stack of chips is used to handle both phases of a single channel (either L or R) and to implement some kind of digital filtering - either to create an active XO or just to give better stop-band rejection of image frequencies.
On this last notion, I am still planning to create a much lower power implementation of the venerable old SAA7220. That chip is well known for being a major contributor to lousy SQ - not because of its filtering but because its a noisy power hog. The latest Cortex M0+ chip out of ST holds out the promise of being able to replicate the filtering of the 7220 but at perhaps just 5% of its power draw. The idea is to create a plug-in PCB which can replace the original chip for those upgrading old style CD players.
Beyond that I think the STM32F030 (and the later generation G030) could be made into digital audio 'lego' bricks to build more complex digital filters and also potentially be used to enhance the distortion performance of multibit DACs. They could also be the core of a DAC which uses not-for-audio DAC chips - those chips don't have I2S interfaces so the CPU can do the interface translation.
Any of those thoughts interesting to anyone else besides me?
Last edited:
Richard judging by the fact that you are sticking your tongue out, and you're being stupendously productive, as usual, bodes all is well there.
Loving the 8 chip LingDac variant I have, pity that development reached a conclusion so soon, although I don't have time to tinker at all, and am a complete electronics pleb, I read as much as possible here, in the advent that one day i will have the time to at least try.
Bring it on Sir!
Loving the 8 chip LingDac variant I have, pity that development reached a conclusion so soon, although I don't have time to tinker at all, and am a complete electronics pleb, I read as much as possible here, in the advent that one day i will have the time to at least try.
Bring it on Sir!
Ok, just to show how ignorant I am of how all this works, I will ask a question that I am not even sure how to ask correctly.
If you can do dual mono, would you not also be able to do 128 chips per channel to get it to 24bit resolution? So, a modest 256 chip NOS DAC to play high bit rate files?
If you can do dual mono, would you not also be able to do 128 chips per channel to get it to 24bit resolution? So, a modest 256 chip NOS DAC to play high bit rate files?
Its an interesting question, one I've thought about myself on and off for some time. I think the answer is likely 'yes' but first I'd want to reduce the target number of bits. Even though plenty of recordings touted as '24bit' exist, they aren't in reality 24bits (in the sense they don't have 144dB dynamic range) as the bottom 4 bits (at least) will be noise. I can't see any point in working hard to get the noise level down when the DAC's internal noise will be swamped by noise on the recording. Therefore I think a more realistic target for a parallel array of 16 bit DACs is 20bits.
The typical noise spec of TDA1387 is 98dB which is 22dB away from 120dB. Noise will in general reduce 3dB for each doubling of number of chips, therefore we will need 7 doublings to reach 119dB, your modest suggestion of 256chips should get us to 122dB. Simply paralleling chips doesn't get us more resolution though, just lower noise. To improve the resolution we'll need some DSP to send the various DACs slightly different data, the same DSP could also randomize the low-order bits to effect distortion reduction. Looks like it could be possible with an array of cheap STM32 chips, not sure how many we'd need though. Definitely a worthwhile project - anyone want to contribute to it?
The typical noise spec of TDA1387 is 98dB which is 22dB away from 120dB. Noise will in general reduce 3dB for each doubling of number of chips, therefore we will need 7 doublings to reach 119dB, your modest suggestion of 256chips should get us to 122dB. Simply paralleling chips doesn't get us more resolution though, just lower noise. To improve the resolution we'll need some DSP to send the various DACs slightly different data, the same DSP could also randomize the low-order bits to effect distortion reduction. Looks like it could be possible with an array of cheap STM32 chips, not sure how many we'd need though. Definitely a worthwhile project - anyone want to contribute to it?
I like your idea of slowly down the I2S speed Richard. Very clever of you.
With regards to noise, do we really need 122db reduction ? I have experimented with my AyA dac for many years & my findings are to to minimize PS noise as much as possible first.
Here transformer types, caps values etc matters alot. The last experiment that I tested was suggested by Max on Ultimate 1541 thread, a powder coated ferrite ring surprisingly it work very very well. Caveat though is the wire used must be at least 1mm in dia. Give this a try cost peanuts. Yes Im aware that the engineer part of you will just look at component specs but I can assure you that there's really a place for so call boutique caps in audio. In fact no need boutique caps to test, every brand of caps of same value will sound different. On my dac there's more then 6 types/brand of caps that I use on different locations.
Cheers
With regards to noise, do we really need 122db reduction ? I have experimented with my AyA dac for many years & my findings are to to minimize PS noise as much as possible first.
Here transformer types, caps values etc matters alot. The last experiment that I tested was suggested by Max on Ultimate 1541 thread, a powder coated ferrite ring surprisingly it work very very well. Caveat though is the wire used must be at least 1mm in dia. Give this a try cost peanuts. Yes Im aware that the engineer part of you will just look at component specs but I can assure you that there's really a place for so call boutique caps in audio. In fact no need boutique caps to test, every brand of caps of same value will sound different. On my dac there's more then 6 types/brand of caps that I use on different locations.
Cheers
Hi @sumotan, thanks for chipping in! I've gotten the slower I2S working and while I can't put my finger on any particular changes, its more addictive to listen to. Perhaps the low frequency bloom is more pronounced, can't say yet - I need my wife's ears to examine it. Yes, agree with PS noise being the focus, but I have addressed that already on PhiDAC (even made some measurements which showed decreased LF rail noise with more caps, hence PhiDAC SE).
As for 122dB - well its a noise level which makes sense for 20bit input material. I don't think I have any as pretty much all I have is RBCD. But I like the challenge of getting there. The idea can go the other way too - I have some 'industrial' 12bit DACs sitting in my intray, I'd like to build an array of them to give 16bit performance.
For those who love optimizing their sound with fancy caps, I say great. I don't have the inclination to do that, I prefer to focus on things where I can understand the whys. If I tweak a cap to a boutique one I haven't learned anything about why it raised the SQ, to me that's not giving any satisfaction, I want to tie improvements to circuit details so I understand what's happening to some degree.
As for 122dB - well its a noise level which makes sense for 20bit input material. I don't think I have any as pretty much all I have is RBCD. But I like the challenge of getting there. The idea can go the other way too - I have some 'industrial' 12bit DACs sitting in my intray, I'd like to build an array of them to give 16bit performance.
For those who love optimizing their sound with fancy caps, I say great. I don't have the inclination to do that, I prefer to focus on things where I can understand the whys. If I tweak a cap to a boutique one I haven't learned anything about why it raised the SQ, to me that's not giving any satisfaction, I want to tie improvements to circuit details so I understand what's happening to some degree.
I respect you thoughts Richard & bravo but you'll never know the last frontier in optimizing SQ is from components used regardless of what Dac or topology. The other thing I forgot to mentioned, using multiple caps in high values to reduce noise may measure well but my own findings is when value are too high it makes the sound of the dac too stiff & tight if that's the right way to describe, it's just does not bloom out of the speakers.
I'm very happy to leave the last frontier to others - its great that @pelopidas is doing with PhiDAC what I'm not inclined to do.
I agree about 'tight' being a way to describe the effect of adding more capacitance. I recently did a side-by-side comparison of two DACs with different amount of capacitance on the rails to my wife. She picked up on the difference in SQ as 'tighter' (the one with more caps) then we discussed whether that's better or worse. But so far I haven't noticed less 'bloom out of the speakers' from this. Could be your layout is different to mine when you add more caps? Small circuit details like the path of the ground current really matter. For example when you add more caps on your circuit, does it have balanced supplies (typical for opamps) or single?
I agree about 'tight' being a way to describe the effect of adding more capacitance. I recently did a side-by-side comparison of two DACs with different amount of capacitance on the rails to my wife. She picked up on the difference in SQ as 'tighter' (the one with more caps) then we discussed whether that's better or worse. But so far I haven't noticed less 'bloom out of the speakers' from this. Could be your layout is different to mine when you add more caps? Small circuit details like the path of the ground current really matter. For example when you add more caps on your circuit, does it have balanced supplies (typical for opamps) or single?
Good that means we're on similar terms in listening. Lol
Depending on where the caps are being used, I've noticed the achieving low esr is not the holy grail. Speaking of op amps, this is what I've done & perhaps you can try it just to satisfy your curiosity, I use higher value caps on then negative then positive rails of the op amp on analog stage. It does have a pronounce impact on sound.
Depending on where the caps are being used, I've noticed the achieving low esr is not the holy grail. Speaking of op amps, this is what I've done & perhaps you can try it just to satisfy your curiosity, I use higher value caps on then negative then positive rails of the op amp on analog stage. It does have a pronounce impact on sound.
Ah, so your answer to my question about balanced supplies is you are using balanced. Got it. If you examine PhiDAC you'll see I use single-ended supplies - I abandoned balanced supplies for opamps years ago for the reason that power supplies are so critical (subjectively judged) that having two of them just multiplies problems by 2. So I suggest try your extra caps on a PhiDAC and report back.
I would expect most opamps to have worse PSRR on the -ve rail than the +ve one - meaning the -ve rail is most sensitive to rail noise. But using a single-ended rail gets around the negative rail PSRR problem completely so I only have to worry about +ve rail PSRR in my designs.
Now you've said you're using balanced supplies I have an idea why you get less bloom from your speakers when you add more caps. Your rail caps will terminate on 0V and that 0V is also your signal reference - adding more caps means coupling more LF noise into your signal 0V reference. More LF noise reduces bloom in my experience. If you adjust your circuit layout I reckon you can get a more tight sound without losing bloom.
I would expect most opamps to have worse PSRR on the -ve rail than the +ve one - meaning the -ve rail is most sensitive to rail noise. But using a single-ended rail gets around the negative rail PSRR problem completely so I only have to worry about +ve rail PSRR in my designs.
Now you've said you're using balanced supplies I have an idea why you get less bloom from your speakers when you add more caps. Your rail caps will terminate on 0V and that 0V is also your signal reference - adding more caps means coupling more LF noise into your signal 0V reference. More LF noise reduces bloom in my experience. If you adjust your circuit layout I reckon you can get a more tight sound without losing bloom.
Last edited:
Yes pls try. Btw this tweak that I use does not apply only to op amp.
I did this on my SS amp as well. I realize that I 'll be shot by many here in diy but it works. I adopt an different thinking when it comes to tweaking. Proof in the pudding is in the listening. In ways Im glad that Im not as clever as you Richard. Id rather purchase a properly design kit & apply my thoughts in tweaking to see if I can improve upon it.
I did this on my SS amp as well. I realize that I 'll be shot by many here in diy but it works. I adopt an different thinking when it comes to tweaking. Proof in the pudding is in the listening. In ways Im glad that Im not as clever as you Richard. Id rather purchase a properly design kit & apply my thoughts in tweaking to see if I can improve upon it.
I agree - I first found out about more 'tightness' from my experiments with chipamps. I went completely crazy on the caps there, always in the pursuit of more bass 'tightness'. But in those days I didn't have a DAC which delivered the huge quantities of bloom that I do now. As for buying ready-made kits, that's what I did about 10 years ago in DACs, bought a cheap board from Taobao using AD1955. If I'd not been very surprised how much PCB layout matters I'd probably not be designing my own DACs now
Bottom line - if you're talking Taobao/Aliexpress kits those guys are relatively clueless about PCB layout for correct grounding. So adding caps will only get you so far.
Bottom line - if you're talking Taobao/Aliexpress kits those guys are relatively clueless about PCB layout for correct grounding. So adding caps will only get you so far.