I've had lots of fun modding one of the DACs from China.
This cute little board:
TDA1541
Turned into...
http://s3t.it/data/uploads/disaster.jpg
I know it looks bad. This isn't reference-level at all. It's playground.
I've tried many options on many points of this circuit - some ideas gave me a better sound, some didn't - and i had to revert the changes.
I've tried to stay "on budget", with classic Phillips I/V+LPF from the 80s.
The last and best of all decoupling options for 7220 was the board cut. This beast drives the 7805 reg HOT with that largish heatsink. Just imagine this kind of load runs on 6MHz.
In addition to board cut,
- shielding with a piece of dynamat i had around
- 20 Ohm resistor between DF and DAC PSU grounds - the lowest impedance point between DAC and DF.
- 100 Ohm in series to I2S
- 10-20pF cap to ground on each I2S wire, near the DAC
Don't forget to take the GND for the (+) input of OpAmps in I/V from the DAC ground.
4.7uF X7R caps are pretty nice for the bits decoupling - and they do fit under the DAC.
Kinda loose power supplies i've got. Nice noise performance due to CRCs, but creepy load regulation on the low frequencies.
This one should be reference grade:
http://overture.org.ua/wp-content/uploads/2010/03/Overture2_3.gif
Unfortunately to English speakers, everything is in Russian
Google translator may help...
http://overture.org.ua/?tag=tda1541
But proper layout is 50% of the DAC.
ThorstenL,
I really like your posts And agree with you. That "reference DAC" always raised the "WTF" question in me
I'll bookmark this thread for the next time i'll play with 1541s. Lots of useful info you have!
This cute little board:
TDA1541
Turned into...
http://s3t.it/data/uploads/disaster.jpg
I know it looks bad. This isn't reference-level at all. It's playground.
I've tried many options on many points of this circuit - some ideas gave me a better sound, some didn't - and i had to revert the changes.
I've tried to stay "on budget", with classic Phillips I/V+LPF from the 80s.
The last and best of all decoupling options for 7220 was the board cut. This beast drives the 7805 reg HOT with that largish heatsink. Just imagine this kind of load runs on 6MHz.
In addition to board cut,
- shielding with a piece of dynamat i had around
- 20 Ohm resistor between DF and DAC PSU grounds - the lowest impedance point between DAC and DF.
- 100 Ohm in series to I2S
- 10-20pF cap to ground on each I2S wire, near the DAC
Don't forget to take the GND for the (+) input of OpAmps in I/V from the DAC ground.
4.7uF X7R caps are pretty nice for the bits decoupling - and they do fit under the DAC.
Kinda loose power supplies i've got. Nice noise performance due to CRCs, but creepy load regulation on the low frequencies.
This one should be reference grade:
http://overture.org.ua/wp-content/uploads/2010/03/Overture2_3.gif
Unfortunately to English speakers, everything is in Russian
Google translator may help...
http://overture.org.ua/?tag=tda1541
But proper layout is 50% of the DAC.
ThorstenL,
I really like your posts
And agree with you. That "reference DAC" always raised the "WTF" question in me I'll bookmark this thread for the next time i'll play with 1541s. Lots of useful info you have!
Last edited:
Hey Thorsten,
I'm not sure if you noticed but the second SPDIF transceiver on Joshua's list uses the Wolfson WM8804
The S/PDIF Transceiver Module
Cheers,
Jeff
I'm not sure if you noticed but the second SPDIF transceiver on Joshua's list uses the Wolfson WM8804
The S/PDIF Transceiver Module
Cheers,
Jeff
Hi,
TBH, I'd rather have a very high impedance across most of the audio range and then use a choke to make up the rest, then to lower the in band impedance to make it flat in the audio range.
In the end there are so many options for doing this, parts availability and ease of implementation need to be considered together with performance.
Hence no direct recommendation on precise designs from me. Too many variables and cooks.
Ciao T
TBH, I'd rather have a very high impedance across most of the audio range and then use a choke to make up the rest, then to lower the in band impedance to make it flat in the audio range.
In the end there are so many options for doing this, parts availability and ease of implementation need to be considered together with performance.
Hence no direct recommendation on precise designs from me. Too many variables and cooks.
Ciao T
Hi,
I can see you had fun!
Yes, it's a bit of a gas guzzler. It also is not necessarily the best filter. I re-deployed it's space gainfully in my DAC.
Okay, if you show me yours...
Here some pics of my "hacked up, experimental china DAC", quite a few years old now, but still serves as "old faithful".
This is the input section, it is active using a pair of PNP Transistors, the circuit is a micture of Jocko Homo and Jim Hagerman, specific for the Cirrus Logic receiver you can see near the top of the pciture.
The little green Cap's are Sanyo 2,200uF/6.3V ultra-low Z types, you can also see an Os-Con and a TL431 Shunt Reg right next to it.
You can also see the PSU for one of the two supply voltages of the Cirrus Logic Receiver, 470uF Os-Con and TL431 Shunt. You can also see the local decoupling at the Cirrus Chip, the 0603 Cap is 0.1uF and the others are 1uF and 10uF respectively in 1206.
The DIP chips in the far right are a 74HC4040 and a 74AHC04 for DEM Reclocking. As can be seen, each chip has it's own Sanyo 2,200uF Cap, Os-Con and SMD decoupling is below the PCB (suboptimal, SMD chip would be better).
Another shot of the receiver section. This also shows the chokes in the power supply lines (including ground!). The chokes are 2,2mH each.
On the other side of the chokes are LM317 CCS connected, one as current sink in the ground and another as AC Current Source but DC regulator in the supply line, so there is no "ground" in the traditional sense, the whole supply floats.
In the corner on the right you can see the space where the SAA7220 used to be, it is now occupied by the I2S attenuators and you can also see the wires carrying the DEM reclock signal.
The DAC is below the Tube stage PCB. The Tube stage is modified from the Universal Tube stage I did for Diyhifisupply, with a purely passive supply using off-board chokes and Film Caps, the rest is similar to my published "final" Tube stage but uses 6072 and not 6922.
The PSU has fewer voltages from the transformer than I would like and the case is already cramped, so there is only around 30V (raw) supplying the DAC, using 4 pcs 3,300uF/50V Elna Silmic II Capacitors in a CRCRCRC filter, before the same AC Current Source, DC current Sink arrangement using LM317 as before.
The CS8414 and rest of the digital section has to make do with only one more voltage. It is split into three lines, one for the input section, one feeding only the CS8414 PLL Supply pins and the other the Digital section and DEM reclocking circuitry.
They use 6,800uF/25V Nichicon caps, four each with a CRCRCRC arrangement per supply before we get to the various CCS's...
Yes, despite many distinctly "non-audiophile" features many would turn their noses up on this is very good. I think Nazar should update this using his later "Nazar Reg", though just having the 1R in series with the 3X7 regs and the final decoupling helps a lot as well...
Ciao T
I can see you had fun!
Yes, it's a bit of a gas guzzler. It also is not necessarily the best filter. I re-deployed it's space gainfully in my DAC.
Okay, if you show me yours...
Here some pics of my "hacked up, experimental china DAC", quite a few years old now, but still serves as "old faithful".
This is the input section, it is active using a pair of PNP Transistors, the circuit is a micture of Jocko Homo and Jim Hagerman, specific for the Cirrus Logic receiver you can see near the top of the pciture.
The little green Cap's are Sanyo 2,200uF/6.3V ultra-low Z types, you can also see an Os-Con and a TL431 Shunt Reg right next to it.
You can also see the PSU for one of the two supply voltages of the Cirrus Logic Receiver, 470uF Os-Con and TL431 Shunt. You can also see the local decoupling at the Cirrus Chip, the 0603 Cap is 0.1uF and the others are 1uF and 10uF respectively in 1206.
The DIP chips in the far right are a 74HC4040 and a 74AHC04 for DEM Reclocking. As can be seen, each chip has it's own Sanyo 2,200uF Cap, Os-Con and SMD decoupling is below the PCB (suboptimal, SMD chip would be better).
Another shot of the receiver section. This also shows the chokes in the power supply lines (including ground!). The chokes are 2,2mH each.
On the other side of the chokes are LM317 CCS connected, one as current sink in the ground and another as AC Current Source but DC regulator in the supply line, so there is no "ground" in the traditional sense, the whole supply floats.
In the corner on the right you can see the space where the SAA7220 used to be, it is now occupied by the I2S attenuators and you can also see the wires carrying the DEM reclock signal.
The DAC is below the Tube stage PCB. The Tube stage is modified from the Universal Tube stage I did for Diyhifisupply, with a purely passive supply using off-board chokes and Film Caps, the rest is similar to my published "final" Tube stage but uses 6072 and not 6922.
The PSU has fewer voltages from the transformer than I would like and the case is already cramped, so there is only around 30V (raw) supplying the DAC, using 4 pcs 3,300uF/50V Elna Silmic II Capacitors in a CRCRCRC filter, before the same AC Current Source, DC current Sink arrangement using LM317 as before.
The CS8414 and rest of the digital section has to make do with only one more voltage. It is split into three lines, one for the input section, one feeding only the CS8414 PLL Supply pins and the other the Digital section and DEM reclocking circuitry.
They use 6,800uF/25V Nichicon caps, four each with a CRCRCRC arrangement per supply before we get to the various CCS's...
This one should be reference grade:
http://overture.org.ua/wp-content/uploads/2010/03/Overture2_3.gif
Unfortunately to English speakers, everything is in Russian
Google translator may help...
Post tagged TDA1541 -
Yes, despite many distinctly "non-audiophile" features many would turn their noses up on this is very good. I think Nazar should update this using his later "Nazar Reg", though just having the 1R in series with the 3X7 regs and the final decoupling helps a lot as well...
Ciao T
The entire DAC you recommend building depends on the synch reclocking obviously for there to be quality improvement over a typical board.
Yet there are no means to synch reclock a standalone DAC without a complex PLL/VXCO. I hate to call you on this as it seemed you had good intentions but I am not sure. You obviously understand that all asych USB audio is 64 bit frame rate I2s and incompatible with the TDA1541 unless you don't care about the 16th bit?
As far as using a Wm8805 are you saying use its external PLL clock to synch reclock the I2s? Surely not?
What people are saying is simply that a separate TDA1541 DAC needs synch reclocking to be a significant improvement over the generic epay boards.
Yet there is no means to do this synch reclocking with a standalone DAC that has proven performance. So I still think the TDA1541 as great as it is is too sensitive to jitter to be separated from the transport, that means today as it always has been is a disc player only chip.
We don't have the technology available to make a top quality standalone TDA1541 DAC. I mean ECDesigns gives up in his first few attempts in the "reference thread" and absolutely refuses to use an asynch usb implementation. The hundreds of pages of that thread are him building a TDA1541 "disk player" not a DAC, somehow folks have completely missed that.
Last edited:
Hi,
Yes I notice. It uses Hardware mode.
This means it is not possible to use a number of optimisations the WM880X makes available, plus there is no 176.4 KHz support, which Sux as a lot of excellent reference recordings stuff comes in 176.4/24...
The layout and decoupling is by far more competent than what we see on the boards Raindrop_hui sells. I have one here - it will be a major job to hack it into acceptable operation - I intended to fit it to my TDA1541 DAC, but it uses software mode.
Ciao T
Yes I notice. It uses Hardware mode.
This means it is not possible to use a number of optimisations the WM880X makes available, plus there is no 176.4 KHz support, which Sux as a lot of excellent reference recordings stuff comes in 176.4/24...
The layout and decoupling is by far more competent than what we see on the boards Raindrop_hui sells. I have one here - it will be a major job to hack it into acceptable operation - I intended to fit it to my TDA1541 DAC, but it uses software mode.
Ciao T
Hi,
To a point, but yes we need a decnet clock from somewhere.
Hence my point that a suitable source delivering I2S and a low jitter MCK must be added.
The ones that output correct I2S are compatible directly with TDA1541 data format even is 64Fs is used as bitclock. However the TDA1541 requires "16-Bit Mode" with 32Fs bitclock for quad speed. There are ways around this, I was deliberately leaving the source open to any options anyone would wish to employ.
Depends on the precise implementation of the WM880X. The quality of the clock can be quite good. It is certainly capable of deliver lower phase noise than what can be measured on the output of a TDA1541 due to the TDA's self noise and 16 Bit wordlength limitations. This is not possible with Cirrus logic receivers that John Brown worked with exclusively.
Actually, to be a major improvement it only needs decent layout, adding I2S attenuators and DEM reclocking. In order to make sure we get the best possible results from the best clock we have available in the whole system (which may not be great, but usable) we apply the reclocker, which also works (implicitly) as level shifter and offers balanced output I2S attenuation.
Note, this is not an attempt to equal ecdesigns board or any of my commercial designs, it is an attempt to come up with something that may be assembled by almost anyone on a veroboard at no more cost than the parts and some of their time, with better performance than the e-bay kits.
Yes, there is, but not available to DIY at this time. The WM880X implemented with care offers a performance that quite good and good enough to not materially degrade that of the TDA1541.
Speak for yourself. I have...
Just because he refuses to consider certain approaches and technologies does not mean that it is not possible to get good results using them. The AMR CD-77 has had an USB Input since it's launch in 2006 and it is not even asynchronous. Yet the results are pretty much comparable with using the clock slaved CD-Drive. So it is possible to get excellent results if sufficient effort is employed. Having seen John's approach to handling USB I am unsurprised that he was not impressed by the results.
Hence John has elected to direct his efforts into other directions and got excellent results that way. I have no issues with that, everyone is entitled to their own approaches. To me "SD Transport" is simply not acceptable, as to me it would be a huge retrograde step (by 2006 I was using PC as main digital music source) so I applied my efforts into that direction and as you may find, I got excellent results.
Ciao T
To a point, but yes we need a decnet clock from somewhere.
Hence my point that a suitable source delivering I2S and a low jitter MCK must be added.
The ones that output correct I2S are compatible directly with TDA1541 data format even is 64Fs is used as bitclock. However the TDA1541 requires "16-Bit Mode" with 32Fs bitclock for quad speed. There are ways around this, I was deliberately leaving the source open to any options anyone would wish to employ.
Depends on the precise implementation of the WM880X. The quality of the clock can be quite good. It is certainly capable of deliver lower phase noise than what can be measured on the output of a TDA1541 due to the TDA's self noise and 16 Bit wordlength limitations. This is not possible with Cirrus logic receivers that John Brown worked with exclusively.
Actually, to be a major improvement it only needs decent layout, adding I2S attenuators and DEM reclocking. In order to make sure we get the best possible results from the best clock we have available in the whole system (which may not be great, but usable) we apply the reclocker, which also works (implicitly) as level shifter and offers balanced output I2S attenuation.
Note, this is not an attempt to equal ecdesigns board or any of my commercial designs, it is an attempt to come up with something that may be assembled by almost anyone on a veroboard at no more cost than the parts and some of their time, with better performance than the e-bay kits.
Yes, there is, but not available to DIY at this time. The WM880X implemented with care offers a performance that quite good and good enough to not materially degrade that of the TDA1541.
Speak for yourself. I have...
Just because he refuses to consider certain approaches and technologies does not mean that it is not possible to get good results using them. The AMR CD-77 has had an USB Input since it's launch in 2006 and it is not even asynchronous. Yet the results are pretty much comparable with using the clock slaved CD-Drive. So it is possible to get excellent results if sufficient effort is employed. Having seen John's approach to handling USB I am unsurprised that he was not impressed by the results.
Hence John has elected to direct his efforts into other directions and got excellent results that way. I have no issues with that, everyone is entitled to their own approaches. To me "SD Transport" is simply not acceptable, as to me it would be a huge retrograde step (by 2006 I was using PC as main digital music source) so I applied my efforts into that direction and as you may find, I got excellent results.
Ciao T
Thats a clever use of a inductor as long as the current source is simple enough to give a first order rolloff.
I was thinking with floating supplies we have a three wire system do you put any effort att common mode rejection? Maybe some creative use of optocouplers could make the basis of a nice common mode current source.
But to make it much simpler maybe just a CMC choke with suitable leakage inductance could possible be used to get both effects with a single component.
Hi,
LM317 (as used in my TDA1541 "Xperimenters" DAC - this means some times I will have an idea, will pull it apart and try it...) does have this rolloff at around 10KHz and it is 2nd order.
The 2.2mH inductors do not rally compensate that (they are only around 1.4k Ohm each at 100KHz), so they help more to block RF noise (common and differential mode) out of the circuit.
The LM317 is around 300K at 10KHz and then looses impedance, at 100KHz it will be around 3K. As I effectively use two 317's in series (one as sink in the negative line and one as CVCCS) I have around 500K CCS impedance below 10KHz in total and still around 5K at 100KHz, where the 431 shunt impedance starts to rise...
So from around 3Hz (1K resistors & 100uF bypass cap) to around 20KHz the whole combo has over 120dB ripple rejection and still over 90dB ripple rejection at 100KHz (including the chokes) and around 15uV (RMS @ 20KHz BW) unweighed noise and 0.2 Ohm output impedance up to 100KHz. Nothing outstanding compared to more complex circuits, but comparing complexity for return it is quite efficient.
The most sensitive supply for the TDA1541 is -15V, there we have around 60dB PSRR, so the 15uV noise translate into around 15nV noise contribution to the output of the TDA1541, or around -160dB... That is 50dB below the TDA1541 self noise, so I refuse to loose sleep over using the "noisy TL431" in this application (I would not use it for a clock supply though).
I use a 2-wire system and with my design common mode filtering is basically only 12dB worse than differential mode filtering as I have more or less identical differential mode impedances in either line. The first layer of common mode filtering happens in RC filter chain at the PSU input and the final one with the chokes that connect the powersupply to main PCB...
So for the DAC for example it is:
Rectifier ->
3,300uF ->
2.2 ohm in each line ->
3,300uF ->
2.2 ohm in each line ->
3,300uF ->
2.2 ohm in each line ->
3,300uF ->
CCS in each line ->
choke in each line ->
Tapped shuntreg ->
Local decoupling (470uF/16V Elna Silmic II & Film)
Each RC cell adds 20dB attenuation at 100Hz (so 60dB compared to a single 3,300uF cap) and continues to a maximum of 140dB (limited by the capacitors ESR). Add to that 120dB from the CCS/Shunt combo and any audio band noise from the mains is squashed down 180dB (in theory at least, I have no way of measuring this low, the supplies self noise gets in the way)...
I have used "combined" filtering for common and differential mode for ages BTW... You can see it most in my tube designs. Commercial implementations usually do not pick this part up (AMR is an exception).
Ciao T
LM317 (as used in my TDA1541 "Xperimenters" DAC - this means some times I will have an idea, will pull it apart and try it...) does have this rolloff at around 10KHz and it is 2nd order.
The 2.2mH inductors do not rally compensate that (they are only around 1.4k Ohm each at 100KHz), so they help more to block RF noise (common and differential mode) out of the circuit.
The LM317 is around 300K at 10KHz and then looses impedance, at 100KHz it will be around 3K. As I effectively use two 317's in series (one as sink in the negative line and one as CVCCS) I have around 500K CCS impedance below 10KHz in total and still around 5K at 100KHz, where the 431 shunt impedance starts to rise...
So from around 3Hz (1K resistors & 100uF bypass cap) to around 20KHz the whole combo has over 120dB ripple rejection and still over 90dB ripple rejection at 100KHz (including the chokes) and around 15uV (RMS @ 20KHz BW) unweighed noise and 0.2 Ohm output impedance up to 100KHz. Nothing outstanding compared to more complex circuits, but comparing complexity for return it is quite efficient.
The most sensitive supply for the TDA1541 is -15V, there we have around 60dB PSRR, so the 15uV noise translate into around 15nV noise contribution to the output of the TDA1541, or around -160dB... That is 50dB below the TDA1541 self noise, so I refuse to loose sleep over using the "noisy TL431" in this application (I would not use it for a clock supply though).
I use a 2-wire system and with my design common mode filtering is basically only 12dB worse than differential mode filtering as I have more or less identical differential mode impedances in either line. The first layer of common mode filtering happens in RC filter chain at the PSU input and the final one with the chokes that connect the powersupply to main PCB...
So for the DAC for example it is:
Rectifier ->
3,300uF ->
2.2 ohm in each line ->
3,300uF ->
2.2 ohm in each line ->
3,300uF ->
2.2 ohm in each line ->
3,300uF ->
CCS in each line ->
choke in each line ->
Tapped shuntreg ->
Local decoupling (470uF/16V Elna Silmic II & Film)
Each RC cell adds 20dB attenuation at 100Hz (so 60dB compared to a single 3,300uF cap) and continues to a maximum of 140dB (limited by the capacitors ESR). Add to that 120dB from the CCS/Shunt combo and any audio band noise from the mains is squashed down 180dB (in theory at least, I have no way of measuring this low, the supplies self noise gets in the way)...
I have used "combined" filtering for common and differential mode for ages BTW... You can see it most in my tube designs. Commercial implementations usually do not pick this part up (AMR is an exception).
Ciao T
So... now your digital signal needs to go trough a ground impedance to get to DAC? What does that to the signal shape at the DAC input and how much jitter induces that impedance in series with that signal? All your "close decopling" now is worth zero... because you just added that impedance in series with the load.
There is absolute no reason to server the grounds, since anyway they are tied together INSIDE the DAC (unless your DAC has a transformer or optical isolator INSIDE of it).
Hi,
Why would I do something as stupid as that?
The DAC has a complete groundfill and all grounds are linked. However the POWEWRSUPPLIES are fully seperate and include chokes in both + & - lines. In effect the - line only becomes "ground" when it connects to the shuntregulators negative pin
I wouldn't know, as only a complete ejit would think up something so harebrained in the first place.
Ciao T
Why would I do something as stupid as that?
The DAC has a complete groundfill and all grounds are linked. However the POWEWRSUPPLIES are fully seperate and include chokes in both + & - lines. In effect the - line only becomes "ground" when it connects to the shuntregulators negative pin
I wouldn't know, as only a complete ejit would think up something so harebrained in the first place.
Ciao T
Nazar in his TDA1541 3.0 version probably has gone for his cap-less PSU.
But he didn't published the schematics for it.
Joshua_G,
Nazar's 1541s are reference-grade in my opinion, but they are from Russia (actually Ukraine), and could be pricey and troublesome to get.
I don't really see a point in tubes in the DACs. They can't be reference-graded - just some "cool sweet-sounding thing with heaters shining in the dark".
But he didn't published the schematics for it.
Joshua_G,
Nazar's 1541s are reference-grade in my opinion, but they are from Russia (actually Ukraine), and could be pricey and troublesome to get.
I don't really see a point in tubes in the DACs. They can't be reference-graded - just some "cool sweet-sounding thing with heaters shining in the dark".
Have you heard it? What did you compare it with?
Any link?
My view is just the opposite of yours.
Joshua_G, i haven't heard his designs, but have followed some of his suggestions on the 1541 design and liked the sound of how it turned out.
He is well regarded DAC designer on russian forums, more techie (really techie, quite mad about proper audio design) than esoteric forums. I think know only 3 guys who really know what the DACs are about, and he is one of them
His website is - In developing audio devices we can always find creative non standart solutions that will ensure the best sound
Ehhm, not 3 guys, probably more like 5, but you got the idea
He is well regarded DAC designer on russian forums, more techie (really techie, quite mad about proper audio design) than esoteric forums. I think know only 3 guys who really know what the DACs are about, and he is one of them
His website is - In developing audio devices we can always find creative non standart solutions that will ensure the best sound
Ehhm, not 3 guys, probably more like 5, but you got the idea
Hi Thorsten, how about this diversion, and I've been only doing it since 2003
peter daniel 1543 / Loesch 1541 - Google Fight : make this fight with googlefight !
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.