Any favorite filter recipes to share? All the commercial DACs using this chip set just seem to make do with a single decoupling cap. Wonder why that is?
The CS8412 digital supply doesn't seem bothered what's on it, but it seems the TDA1543 and particularly the CS8412 analog supply are quite fussy.
I'm using a 3D layout, carbon I/V resistors, +6v TDA1543 supply, 1k0 between ICs, Sanyo SEP 330u 16v coupling caps, 470R / Wima polyester 0.47u loopfilter, and 1206 NPO 10n SPDIF coupling caps.
I started minimalist, with Meggitt 1uH/R and 0612 X7R 100n after each LM317T reg, the reg having a Panasonic FC 220u 25v directly on the output. A bit bright and thin, but sharply defined.
I've previously tried 56uH/0.3R followed by a triplet 1206 Y5V 4.7u / 0805 X7R 220n / 0612 X7R, good in a lot of ways but ultimately too soft and dry.
I retried this triplet (all X7R) with 1uH/1R on the TDA1543, adding a Panasonic FC 220u 25v. This is somewhat soggy and metallic cap, and so produces a nicely balanced sound with the triplet. Still a bit bright though, and the dynamics aren't wide enough.
Swapping the Meggitt 1uH/1R for a Murata 1210 47uH/1.3R inductor yields a cleaner sound. This inductor is incidentally a very handy size for getting in nooks and crannies.
A much better cap than the FC is the Sanyo SEP 330u 16v, a kind of polymer oscon (whatever that means) and the electrolytic to beat in my experience. This has serious weight and powerful dynamics, and is generally lush sounding. Still a little brightness though.
The best sound yet comes from the above configuration with a CS8412 analog supply of: LM317T (no output cap) > 47uH/1.3R > Sanyo SEP 330u 16v > 47uH/1.3R > ceramic triplet. Thunderous and very clean, bright but in a more neutral, clear sounding sense rather than artefaced. Note that I'm using Stax Omega II, which can be a bit fierce.
Note that non of the above combos sound particularly detailed. The best are textured and lush, but they don't have fine grain detail.
A lot of the ideas in this post comes from this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=4562
The idea for the Pansonic FC 220u 25v comes from two TNT-Audio articles, which recommed 220uF capacitance and 0.1 ESR for the proper functioning of the LM317T:
http://www.tnt-audio.com/clinica/regulators_noise2_e.html
http://www.tnt-audio.com/clinica/regulators2_impedance1_e.html
The CS8412 digital supply doesn't seem bothered what's on it, but it seems the TDA1543 and particularly the CS8412 analog supply are quite fussy.
I'm using a 3D layout, carbon I/V resistors, +6v TDA1543 supply, 1k0 between ICs, Sanyo SEP 330u 16v coupling caps, 470R / Wima polyester 0.47u loopfilter, and 1206 NPO 10n SPDIF coupling caps.
I started minimalist, with Meggitt 1uH/R and 0612 X7R 100n after each LM317T reg, the reg having a Panasonic FC 220u 25v directly on the output. A bit bright and thin, but sharply defined.
I've previously tried 56uH/0.3R followed by a triplet 1206 Y5V 4.7u / 0805 X7R 220n / 0612 X7R, good in a lot of ways but ultimately too soft and dry.
I retried this triplet (all X7R) with 1uH/1R on the TDA1543, adding a Panasonic FC 220u 25v. This is somewhat soggy and metallic cap, and so produces a nicely balanced sound with the triplet. Still a bit bright though, and the dynamics aren't wide enough.
Swapping the Meggitt 1uH/1R for a Murata 1210 47uH/1.3R inductor yields a cleaner sound. This inductor is incidentally a very handy size for getting in nooks and crannies.
A much better cap than the FC is the Sanyo SEP 330u 16v, a kind of polymer oscon (whatever that means) and the electrolytic to beat in my experience. This has serious weight and powerful dynamics, and is generally lush sounding. Still a little brightness though.
The best sound yet comes from the above configuration with a CS8412 analog supply of: LM317T (no output cap) > 47uH/1.3R > Sanyo SEP 330u 16v > 47uH/1.3R > ceramic triplet. Thunderous and very clean, bright but in a more neutral, clear sounding sense rather than artefaced. Note that I'm using Stax Omega II, which can be a bit fierce.
Note that non of the above combos sound particularly detailed. The best are textured and lush, but they don't have fine grain detail.
A lot of the ideas in this post comes from this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=4562
The idea for the Pansonic FC 220u 25v comes from two TNT-Audio articles, which recommed 220uF capacitance and 0.1 ESR for the proper functioning of the LM317T:
http://www.tnt-audio.com/clinica/regulators_noise2_e.html
http://www.tnt-audio.com/clinica/regulators2_impedance1_e.html
Hi kit:
This is a good topic, and you are very persistent with that TDA 1543/CS8412 DAC. Having followed some of its previous incarnations, it would be interesting to know the full story of its genesis.
With regard to inductors, my favorite for digital supplies is the LQH3C Murata series. Sonically, various axial and radial lead chokes did not cut it.
For a nice compact noise suppressor, effective up to 1GHz, the Murata BNX002-01 is also very nice.
The LC(R) before the regulator is normally called a card-entry filter. In a past article, Jung recommended 100uH in series with the voltage supply, and a 100uF capacitor and a 1 - 1.5 Ohm resistor in series between the V+ supply to ground. The resistor is there is to control amplitude peaking. The filter provides about 50dB noise attenuation at 1MHz, but will only filter differential noise.
Have you simulated the attenuation response for your different LC cinfigurations?
At expense of greater circuit complexity and component count, you might build a regulator with discrete and IC parts rather than use the LM 317.
Also, 100R series resistors between the CS8412 and TDA 1543 are probably a good idea to cut down on voltage spikes from capacitive loading of CS 8412. You can also accomplish the same with 100pF capacitor in series with 100 Ohm Resistor from the TDA 1543 inputs to ground.
Finally, you might want to experiment with a high-quality teflon or copper or silver foil capacitor for coupling rather than the Sanyo SEP lytic.
This is a good topic, and you are very persistent with that TDA 1543/CS8412 DAC. Having followed some of its previous incarnations, it would be interesting to know the full story of its genesis.
With regard to inductors, my favorite for digital supplies is the LQH3C Murata series. Sonically, various axial and radial lead chokes did not cut it.
For a nice compact noise suppressor, effective up to 1GHz, the Murata BNX002-01 is also very nice.
The LC(R) before the regulator is normally called a card-entry filter. In a past article, Jung recommended 100uH in series with the voltage supply, and a 100uF capacitor and a 1 - 1.5 Ohm resistor in series between the V+ supply to ground. The resistor is there is to control amplitude peaking. The filter provides about 50dB noise attenuation at 1MHz, but will only filter differential noise.
Have you simulated the attenuation response for your different LC cinfigurations?
At expense of greater circuit complexity and component count, you might build a regulator with discrete and IC parts rather than use the LM 317.
Also, 100R series resistors between the CS8412 and TDA 1543 are probably a good idea to cut down on voltage spikes from capacitive loading of CS 8412. You can also accomplish the same with 100pF capacitor in series with 100 Ohm Resistor from the TDA 1543 inputs to ground.
Finally, you might want to experiment with a high-quality teflon or copper or silver foil capacitor for coupling rather than the Sanyo SEP lytic.
Thanks for the tips slawney. Hope my own info is useful.
I use 1k0 resistors on the CS8412. Lots more slam. I could never decide if 50-100R resistors did anything, so when I read an article recommending interfacing with 47-1000R, I thought I may as well go the whole hog.
Will try the resistor filter, as I've some brightness which might be coming from the FC. I use these caps as they're better than bog standard without their ESR being too low for the LM317T and have some good qualities, but I'd like to think there's a better alternative..
Not played with a simulator, care to recommend a free one?
Where abouts ought that Murata DC filter to go?
TIA
I've not had a problem with electrolytic coupling caps. Suck it and see seems to be the case.
I use 1k0 resistors on the CS8412. Lots more slam. I could never decide if 50-100R resistors did anything, so when I read an article recommending interfacing with 47-1000R, I thought I may as well go the whole hog.
Will try the resistor filter, as I've some brightness which might be coming from the FC. I use these caps as they're better than bog standard without their ESR being too low for the LM317T and have some good qualities, but I'd like to think there's a better alternative..
Not played with a simulator, care to recommend a free one?
Where abouts ought that Murata DC filter to go?
TIA
I've not had a problem with electrolytic coupling caps. Suck it and see seems to be the case.
Kit:
>I use 1k0 resistors on the CS8412. Lots more slam. I could >never decide if 50-100R resistors did anything, so when I read >an article recommending interfacing with 47-1000R, I thought I >may as well go the whole hog.
Which article was this? Although your ground will see less voltage spikes from the CS8412 with 100R, it is true that the audible effect seems negligable. I could not get an accurate measurement of the input capacitance of the TDA 1543 pins... but the CS8412 is pretty intolerant of driving even low (~ 10 pF) capacitance loads.
>Will try the resistor filter, as I've some brightness which might >be coming from the FC.
Tell me how it works. Stick with 1R first...
>I use these caps as they're better than >bog standard without >their ESR being too low for the LM317T and have some good >qualities, but I'd like to think there's a better alternative..
If the main point is to provide your ICs with a low-noise supply, it might be better altogether to replace the LM317T with a discrete reg.
>Where abouts ought that Murata DC filter to go?
It works well before the regulator if your raw supply has very little ripple. Direct connection to pcb ground is mandatory, since the unit provides filtering for this as well (a feature which makes the filter rather unique). It can also fit between the regulator and the IC if you have the pcb real estate to spare.
Impressive thing with the Murata is the wideband LR up to 1 GHz. When I read that I was attracted immediately. I think the filter you assembled is only effective up to 2MHz with pretty strong peaking, but the math needed to be done to give an exact picture of what your LC(CCC) circuit is doing would be rather painful.
>I've not had a problem with electrolytic coupling caps. Suck it >and see seems to be the case.
Matter of taste. Am I right in assuming that you use Sanyo still for coupling?
Why did you end up with those RC values for PLL? You were experimenting with much higher C values in the past.
Best regards,
James
>I use 1k0 resistors on the CS8412. Lots more slam. I could >never decide if 50-100R resistors did anything, so when I read >an article recommending interfacing with 47-1000R, I thought I >may as well go the whole hog.
Which article was this? Although your ground will see less voltage spikes from the CS8412 with 100R, it is true that the audible effect seems negligable. I could not get an accurate measurement of the input capacitance of the TDA 1543 pins... but the CS8412 is pretty intolerant of driving even low (~ 10 pF) capacitance loads.
>Will try the resistor filter, as I've some brightness which might >be coming from the FC.
Tell me how it works. Stick with 1R first...
>I use these caps as they're better than >bog standard without >their ESR being too low for the LM317T and have some good >qualities, but I'd like to think there's a better alternative..
If the main point is to provide your ICs with a low-noise supply, it might be better altogether to replace the LM317T with a discrete reg.
>Where abouts ought that Murata DC filter to go?
It works well before the regulator if your raw supply has very little ripple. Direct connection to pcb ground is mandatory, since the unit provides filtering for this as well (a feature which makes the filter rather unique). It can also fit between the regulator and the IC if you have the pcb real estate to spare.
Impressive thing with the Murata is the wideband LR up to 1 GHz. When I read that I was attracted immediately. I think the filter you assembled is only effective up to 2MHz with pretty strong peaking, but the math needed to be done to give an exact picture of what your LC(CCC) circuit is doing would be rather painful.
>I've not had a problem with electrolytic coupling caps. Suck it >and see seems to be the case.
Matter of taste. Am I right in assuming that you use Sanyo still for coupling?
Why did you end up with those RC values for PLL? You were experimenting with much higher C values in the past.
Best regards,
James
I'll check out that filter thingy, if I can find space on the board for them 🙂
The RC values are just what I had to hand that didn't take up tooo much space. I think 470R/220nF is an objective improvement over the standard 1k0/47nF, but eveything else is quite subjective.
I think a whopping great Sanyo SEP can be a bit overpowering. I'm finding that SEP decoupling after a Rubycon ZA filter cap gives a nice balance between openess, thickness, flow and warmth.
Still using SEPs as coupling caps. Small, sound good, let lots of difference between components used elsewhere through.
I believe Guido Tent's famous digital decoupling PDF was the source for 47R - 1000R.
The RC values are just what I had to hand that didn't take up tooo much space. I think 470R/220nF is an objective improvement over the standard 1k0/47nF, but eveything else is quite subjective.
I think a whopping great Sanyo SEP can be a bit overpowering. I'm finding that SEP decoupling after a Rubycon ZA filter cap gives a nice balance between openess, thickness, flow and warmth.
Still using SEPs as coupling caps. Small, sound good, let lots of difference between components used elsewhere through.
I believe Guido Tent's famous digital decoupling PDF was the source for 47R - 1000R.
I've gone back to a single Sanyo SEP 330u 16v (+ inductor) for decoupling and binned the FC on the LM317, not for the first time 🙂 The more complex stuff is searching for the Holy Grail, but to be honest I think this cap is as good as a basic setup is going to get.
I've gone back to Sanyo SEP for the loopfilter too, 22u 25v, a small capp that doen't get in the way of the layout. It adds a bit of grease to the sound that's welcome, and doesn't increase the lock on time.
I've gone back to Sanyo SEP for the loopfilter too, 22u 25v, a small capp that doen't get in the way of the layout. It adds a bit of grease to the sound that's welcome, and doesn't increase the lock on time.
you try any Blackgates? I have them everywhere except for the SPDIF coupling caps and some Wima bypasses. I've swapped in some Nichicon Muze KZs and Panasonic FCs at times, but keep going back to Blackgates for every supply in my DAC. I'm redoing the clock supply right now following Guido Tent's suggested supply, again using Blackgates throughout.
Peter
Peter
An update for thems what might be interested.
Currently using the Murata BNX002-01 in place of a filter cap, though not the ferrite bead on the earthline. Is this wise?
It sounds much more wide open with smaller sounding individual instruments. There's less groove in the sense of the music being pushed along, but actual timing is better.
Currently using an RC filter after each LM317T, with 1R 0.5w 5% carbon resistors from Farnell. These sound much more intimate and organic than Farnell's 1% 0.5w metal films, which by comparison sound superficially cleaner and smoother but ultimately shut in and weedy.
22u 20v on the CS8412 (the smallest SEP Farnell had), as the rule seems the smaller the better with this chip. A big 330u 16v on the TDA1543, as with an analog part I imagine LF noise is the biggest issue.
It sounds quite plain but brightish, warm and punchy (drums sound like skin streched over a tube, etc) with quite a forward, close to the listener sound. There's less sparkle and lushness than with previous L and LC filtering schemes, but there's less of their unpredictable tonal "weirdness," too.
Currently using the Murata BNX002-01 in place of a filter cap, though not the ferrite bead on the earthline. Is this wise?
It sounds much more wide open with smaller sounding individual instruments. There's less groove in the sense of the music being pushed along, but actual timing is better.
Currently using an RC filter after each LM317T, with 1R 0.5w 5% carbon resistors from Farnell. These sound much more intimate and organic than Farnell's 1% 0.5w metal films, which by comparison sound superficially cleaner and smoother but ultimately shut in and weedy.
22u 20v on the CS8412 (the smallest SEP Farnell had), as the rule seems the smaller the better with this chip. A big 330u 16v on the TDA1543, as with an analog part I imagine LF noise is the biggest issue.
It sounds quite plain but brightish, warm and punchy (drums sound like skin streched over a tube, etc) with quite a forward, close to the listener sound. There's less sparkle and lushness than with previous L and LC filtering schemes, but there's less of their unpredictable tonal "weirdness," too.
Hi Kit
I am beginning to feel sorry for you. Some time ago i spent a few days playing with a similar DAC. Every single change had an immediate effect on the sound; it always seemed like a glorious sound may be hiding just around the corner - maybe just a resistor or two away.
If i listen for just five minutes it sounds impressive - detailed, great bass, good dynamics. Ultimately, it is really unmusical and unlistenble. I suspect you periodically come to the same conclusion and that's why the constant swap of part types and values.
Can nirvana be reached through a 1543 NOS?
I am beginning to feel sorry for you. Some time ago i spent a few days playing with a similar DAC. Every single change had an immediate effect on the sound; it always seemed like a glorious sound may be hiding just around the corner - maybe just a resistor or two away.
If i listen for just five minutes it sounds impressive - detailed, great bass, good dynamics. Ultimately, it is really unmusical and unlistenble. I suspect you periodically come to the same conclusion and that's why the constant swap of part types and values.
Can nirvana be reached through a 1543 NOS?
Hehe 🙂 I think the answer is no, but you can get a good sound for very little money.
The trick is just to get a sound that works without any obvious nasties (of which there are potentiall very many) and leave it at that. Learn to live with it, or you'll go mad looking for God sound.
The trick is just to get a sound that works without any obvious nasties (of which there are potentiall very many) and leave it at that. Learn to live with it, or you'll go mad looking for God sound.
Certainly not! That's heresy.
This particular incarnation does sound good though, very plain, no nonsense sound without being thin, muffled or dark.
This particular incarnation does sound good though, very plain, no nonsense sound without being thin, muffled or dark.
Hehe 🙂
I've swapped the SEPs for 33nf LCR polystyrenes, a one million percent capacitance difference! The type is called FSCEX or something similar, silver with a blue band at one end. They're meant to be very good.
The sonic difference is supringly modest, but the decoupling LCRs definitely have more warmth, space and room echo.
Taking the SEPs off the ADJ pin of the LM317T loses some of the "air", but also loses a certain tiring metallic quality. The sound is browner without them.
The sound is fuller and does have a certain undamped quality, so maybe a bit of euphonic distortion is going on.
I've swapped the SEPs for 33nf LCR polystyrenes, a one million percent capacitance difference! The type is called FSCEX or something similar, silver with a blue band at one end. They're meant to be very good.
The sonic difference is supringly modest, but the decoupling LCRs definitely have more warmth, space and room echo.
Taking the SEPs off the ADJ pin of the LM317T loses some of the "air", but also loses a certain tiring metallic quality. The sound is browner without them.
The sound is fuller and does have a certain undamped quality, so maybe a bit of euphonic distortion is going on.
- Status
- Not open for further replies.