DEM options
I add one more confirmation to the record. I too have followed this path. First teh 200pf cap on 16 & 17, followed later by the CD 9009 mod. The initial cap was a big improvement in sound quality, but left a nagging sense that there was still room to improve. The second mod that syncs back to WS was a large step forward, and after a few days listening there is not the same need to correct something.
This is not to say the dac is perfect. Changes elsewhere can make a big difference in resolution. Changing a USB cable, a change in parameters in the PC based source and its power supply can have lots of impact on resolution, but for some reason they are all different shades of a satisfying sound. Does this mean that the jitter induced by intermodulation within the 1541 chip is the most irritating form of jitter? perhaps
I add one more confirmation to the record. I too have followed this path. First teh 200pf cap on 16 & 17, followed later by the CD 9009 mod. The initial cap was a big improvement in sound quality, but left a nagging sense that there was still room to improve. The second mod that syncs back to WS was a large step forward, and after a few days listening there is not the same need to correct something.
This is not to say the dac is perfect. Changes elsewhere can make a big difference in resolution. Changing a USB cable, a change in parameters in the PC based source and its power supply can have lots of impact on resolution, but for some reason they are all different shades of a satisfying sound. Does this mean that the jitter induced by intermodulation within the 1541 chip is the most irritating form of jitter? perhaps
Don't know about measuring the freq across the pins 16, 17 - I think it's done elsewhere.
About that 180pF cap on the DEM pins - Suggest you get the capacitance across pins 16, 17 up to about 230 <-> 240pF - at 180pF you're too high freq and can't get close to a "lock" so leading edge emphasized and tops give that ssss - do it by ear - quick and straightforward, particularly if adding the trimmer cap in parallel - suggest propylene/styrene/teflon - really good foils here, not the ceramic types.
If you're using 1uF decoupling caps like B32529s for example, or the more recent SMDs, I suggest try a 4.3uF cap on pins 13, and 18. - a good quality film cap - Sonicaps are fairly uniform here, Jantzen SuperiorZ are quite forward and a bit toppy, Mundorf Supremes on the other hand are heavier bass and mellow mids - you get the idea from these examples.
Fascinating how this old chip still bedazzles us!
About that 180pF cap on the DEM pins - Suggest you get the capacitance across pins 16, 17 up to about 230 <-> 240pF - at 180pF you're too high freq and can't get close to a "lock" so leading edge emphasized and tops give that ssss - do it by ear - quick and straightforward, particularly if adding the trimmer cap in parallel - suggest propylene/styrene/teflon - really good foils here, not the ceramic types.
If you're using 1uF decoupling caps like B32529s for example, or the more recent SMDs, I suggest try a 4.3uF cap on pins 13, and 18. - a good quality film cap - Sonicaps are fairly uniform here, Jantzen SuperiorZ are quite forward and a bit toppy, Mundorf Supremes on the other hand are heavier bass and mellow mids - you get the idea from these examples.
Fascinating how this old chip still bedazzles us!
Having tried the DEM reclocking from http://www.diyaudio.com/forums/digital-line-level/11949-tda1541-dem-reclocking.html and I think even another circuit I don't care to remember, I was never satisfied with the sound and quickly returned to datasheet solution (470 pF capacitor). BUT the Grunding solution works like a charm! I never thought the TDA1541 is capable of such dynamics and clarity. Thank you guys for sharing this tweak! btw, who designed the CD9009 player? He must have know something about TDA1541. (I only know Grundig usually outsourced its designs)
you surely mean an dynamic matching instead of static caps. No, its not such a bad idea
Grundig CD9009 mod
Hi,
i also tried the Grundig 9009 mod in OS player, a very large step forward over full frequency range.
Don't know why Philips did this 25years before.
In the past,I did some good mods, but with this mod the cheap old Sony CDP750 let me hear how the musicians breath flow and the digital chips work,hahahaha, ;-( ...some klick noise during playing.
I added a small 100mH coil in series direct from -15Vcap pin.
This solved the problem completely, background is deep black.
A suitable master-clock circuit for the TDA1541A would be
helpfull to go further.
Don't want to go NOS, so for OS-player, if DEM is reclocked with BCK,you need freq-divider /2 to BCK and /4 to DEM also flipflop /1 to filter and flipflop /1 to decoder,true or mistaken?
I made a forum search but cannot find suitable post.
Please, could someone give me some advice where to find, read and study those informations? Thanks!
Regards Harry
PS: Hallo Andreas, danke für Deine email über zukünftige Angebote auf Deiner nanocamp-HP
Hi,
i also tried the Grundig 9009 mod in OS player, a very large step forward over full frequency range.
Don't know why Philips did this 25years before.
In the past,I did some good mods, but with this mod the cheap old Sony CDP750 let me hear how the musicians breath flow and the digital chips work,hahahaha, ;-( ...some klick noise during playing.
I added a small 100mH coil in series direct from -15Vcap pin.
This solved the problem completely, background is deep black.
A suitable master-clock circuit for the TDA1541A would be
helpfull to go further.
Don't want to go NOS, so for OS-player, if DEM is reclocked with BCK,you need freq-divider /2 to BCK and /4 to DEM also flipflop /1 to filter and flipflop /1 to decoder,true or mistaken?
I made a forum search but cannot find suitable post.
Please, could someone give me some advice where to find, read and study those informations? Thanks!
Regards Harry
PS: Hallo Andreas, danke für Deine email über zukünftige Angebote auf Deiner nanocamp-HP
I just tried the Grundig solution, where the DEM oscillator is driven from WS (4x OS), that is one full cycle per sample. I haven't listened yet, just checked the waveform on pin 16, looks nice. Pin 17 is AC filtered to GND. On the Grundig diagram the MSB filter capacitors at pins 13 and 18 are 1uF, all the rest are .1uF. Shall I increase all capacitors to 4x their value if I go for NOS? Why would be 44.1 kHz DEM frequency too low?
I'm using the NOS and wasn't so happy with the 'Grundig' technique so I went back to the 470pF styrene cap across pin16, 17 but reduced this to 300pF via a variable air cap and then reduced it again to 156pF (120pFstyrene + variable air cap) across the chip pins to give a dem freq about 700kHz - much better than using the WS system.
Surprising improvement in the "smoothness" of the sound - poor description but that's how it seems - the adjustments of the air cap are pretty touchy but quite easy to find the "sweet spot".
It seems to indicate that higher dem freq leads to smoother sound - heard about this before, yes!
[Also using the 1uF Epcos on the 6 single dacs and a 4uF Solen FC on the pins 13 & 18.]
Surprising improvement in the "smoothness" of the sound - poor description but that's how it seems - the adjustments of the air cap are pretty touchy but quite easy to find the "sweet spot".
It seems to indicate that higher dem freq leads to smoother sound - heard about this before, yes!
[Also using the 1uF Epcos on the 6 single dacs and a 4uF Solen FC on the pins 13 & 18.]
I think 44.1kHz is too slow for fDEM. For the matching to be fully effective, the DEM shift register should spend time in each of its 4 states during every CD sample period. The lowest fDEM which achieves this is 176.4kHz. So it's OK to clock the DEM using WS of 176.4kHz in 4xOS mode as in the CD9009, but probably not such a good idea at a WS frequency of 44.1kHz in NOS mode.
The working of the 4-state DEM used in the TDA1540/TDA1541 is described in this paper
http://www.diyaudio.org.ua/download/file.php?id=2148
If fDEM is 44.1kHz, the DAC will still sound good, but there may be more distortion above 11kHz. Another problem with using fDEM of 44.1kHz is that some of the DEM switching ripple is at fDEM/2 i.e. 22.05kHz, and this affects filtering.
I believe the "Alternative External Control Circuit" for the DEM as used in the Grundig CD9009 was published by Philips in their ADC/DAC applications book for manufacturers, but it was never used in production Philips or Marantz machines. This may be because it did not improve the performance of the DAC as they chose to measure it.
The working of the 4-state DEM used in the TDA1540/TDA1541 is described in this paper
http://www.diyaudio.org.ua/download/file.php?id=2148
If fDEM is 44.1kHz, the DAC will still sound good, but there may be more distortion above 11kHz. Another problem with using fDEM of 44.1kHz is that some of the DEM switching ripple is at fDEM/2 i.e. 22.05kHz, and this affects filtering.
I believe the "Alternative External Control Circuit" for the DEM as used in the Grundig CD9009 was published by Philips in their ADC/DAC applications book for manufacturers, but it was never used in production Philips or Marantz machines. This may be because it did not improve the performance of the DAC as they chose to measure it.
Ok, I can't and want not dispute this. But philosofically, tell me, why theoretically wrong solution sounds so much better than datasheet solution? And all that to save 1 capacitor on which they had large volume discount anyway (used at DEM pins)? (and one resistor and one zener).
Speaking of distortion, the non-OS approach gives much more precise and better defined, though little bit muted, treble than oversampled dacs. Probably this more than offsets the potential distortions you mention.
Speaking of distortion, the non-OS approach gives much more precise and better defined, though little bit muted, treble than oversampled dacs. Probably this more than offsets the potential distortions you mention.
4x OS means we have new samples calculated by the digital filter at 176.4 kHz. According to your logic, the DEM frequency should be 4x the sample frequency, that is 705.6 kHz.
Also not clear why should be more distortion above 11 kHz if fDEM is 44.1 kHz (NOS case), and where the 22.05 kHz switching ripple comes from.
I am just trying to understand the way it works, and find the optimal solution (if such exists). Thanks