@esl63:
Quote:
I have found TDK 0306 4,7uF caps that i have managed to solder directly on the DAC chip.
@gentlevoice
... Have you by any chance tried to measure distortion on these capacitors? If I remember correctly X5R dielectrics may distort quite some, i.e. generate out-of-pulse noise.
This capacitor that is placed so extremely close to the current consumer only has to supply current extremely fast, di/dt. Distortion in the analog world is not relevant at the supply pins. There is no Analogsignals to handle. We are talking "keeping the voltage at 5,00000 volts" whatever current the DAC chip needs. Since the bandwith is theoretically infinite in the digital switching domain I have really found a way to maximize the powersupply performance on DAC chip. The distortion and background noise is kept on the smallest possible levels.
Quote:
I have found TDK 0306 4,7uF caps that i have managed to solder directly on the DAC chip.
@gentlevoice
... Have you by any chance tried to measure distortion on these capacitors? If I remember correctly X5R dielectrics may distort quite some, i.e. generate out-of-pulse noise.
This capacitor that is placed so extremely close to the current consumer only has to supply current extremely fast, di/dt. Distortion in the analog world is not relevant at the supply pins. There is no Analogsignals to handle. We are talking "keeping the voltage at 5,00000 volts" whatever current the DAC chip needs. Since the bandwith is theoretically infinite in the digital switching domain I have really found a way to maximize the powersupply performance on DAC chip. The distortion and background noise is kept on the smallest possible levels.
Last edited:
My little PCM5122 Dac is also running from LIFEPO4 battery. It sounds much better than with a talema transformator and nichicon caps. Since i am using the battery i removed the bypass caps. It sounds even better now. Of course there are people who know better, even without trying by themselves. I can only say one thing : TRY IT.
The filter at the output in my dac is 2200pF Wima FKP and Vishay resistors.
I also tried with two LIFEPO4s and a TPS7A4700 ldo, but no chance against a not regulated LIFEPO4.
With my PCM1794a the difference was not so big. The Talema/TPS7A4700 combo for my ears was as good as the batteries.
No idea why, but the voltage out PCM5122 grew to new heights with unregulated, capless battery power.
The filter at the output in my dac is 2200pF Wima FKP and Vishay resistors.
I also tried with two LIFEPO4s and a TPS7A4700 ldo, but no chance against a not regulated LIFEPO4.
With my PCM1794a the difference was not so big. The Talema/TPS7A4700 combo for my ears was as good as the batteries.
No idea why, but the voltage out PCM5122 grew to new heights with unregulated, capless battery power.
@LinuxGeek: Hi ... Interesting to hear that others also have positive experiences with using LiFePO4 batteries for a DAC PSU. Your listening experiences are much in line with what I've observed myself using only batteries.
However, with respect to removing the decoupling capacitors I'm still "cautious here" as I've had the opposite experience, i.e. adding decoupling capacitors (on a suitably laid out PCB) clearly cleaned up the sound so when you (and bjd) say that it sounds better without the decoupling I'm admittedly still wondering if it may be because the decoupling capacitors that were removed were of "less than good" quality ... ? No offense intended but in my experience these capacitors most often do not sound "good" (I have listened to a good selection). To this end: Can I ask you which decoupling capacitors you removed (i.e. value, type, manufacturer and size)? And also which sampling frequency you mainly play back at?
Cheers,
Jesper
However, with respect to removing the decoupling capacitors I'm still "cautious here" as I've had the opposite experience, i.e. adding decoupling capacitors (on a suitably laid out PCB) clearly cleaned up the sound so when you (and bjd) say that it sounds better without the decoupling I'm admittedly still wondering if it may be because the decoupling capacitors that were removed were of "less than good" quality ... ? No offense intended but in my experience these capacitors most often do not sound "good" (I have listened to a good selection). To this end: Can I ask you which decoupling capacitors you removed (i.e. value, type, manufacturer and size)? And also which sampling frequency you mainly play back at?
Cheers,
Jesper
@gentlevoice :
I dont know why it sounds better without the caps. My DAC IC is very close to the battery and the power traces are rather wide. My bypass caps were nichicon kz and fg, tried both, kz were nicer. Dont know the exact values, will look tommorrow when i am back home.
I mainly play 44.1khz 16bit flacs and about 10% 192khz 24bit flacs.
I dont know why it sounds better without the caps. My DAC IC is very close to the battery and the power traces are rather wide. My bypass caps were nichicon kz and fg, tried both, kz were nicer. Dont know the exact values, will look tommorrow when i am back home.
I mainly play 44.1khz 16bit flacs and about 10% 192khz 24bit flacs.
Last edited:
Hi LinuxGeek ... just a brief comment ...
From the perspective of HF decoupling neither the FG nor the KZ are very effective at high frequencies (high inductance types). To decouple high frequencies typically small size SMD (low inductance - C0G) capacitors are used - often by default 100 nF. Some also use polyester types - to my memory the DDDAC in Doede's version uses Wima MKS types (also 100 nF), yet these have somewhat higher inductance than the C0G types, and thus will be less effective in decoupling HF. The question then is if this theory matters in practice and I cannot answer relative to pure KZ or FG decoupling but I've used a PCM1794 based DAC for some time only decoupling it with electrolytics (680 uF rubycon ZL I think) and adding 100 nF C0Gs to my ears made quite a positive difference.
However, if you mainly play 44.1 kHz/16 bits the clock going to the DAC most likely is 1.411 MHz in which case normal electrolytics like the KZ or FG might still be reasonably effective for decoupling. Why the battery sounds better than either of the electrolytics, well, that's a good question
In case you read this before checking the values of your capacitors then you don't need to - it's fine that I know which types they were.
Cheers,
Jesper
From the perspective of HF decoupling neither the FG nor the KZ are very effective at high frequencies (high inductance types). To decouple high frequencies typically small size SMD (low inductance - C0G) capacitors are used - often by default 100 nF. Some also use polyester types - to my memory the DDDAC in Doede's version uses Wima MKS types (also 100 nF), yet these have somewhat higher inductance than the C0G types, and thus will be less effective in decoupling HF. The question then is if this theory matters in practice and I cannot answer relative to pure KZ or FG decoupling but I've used a PCM1794 based DAC for some time only decoupling it with electrolytics (680 uF rubycon ZL I think) and adding 100 nF C0Gs to my ears made quite a positive difference.
However, if you mainly play 44.1 kHz/16 bits the clock going to the DAC most likely is 1.411 MHz in which case normal electrolytics like the KZ or FG might still be reasonably effective for decoupling. Why the battery sounds better than either of the electrolytics, well, that's a good question
In case you read this before checking the values of your capacitors then you don't need to - it's fine that I know which types they were.
Cheers,
Jesper
Hi again - just a hint on C0Gs: Samuel Groner (diyaudio member) has made measurements on C0G capacitors. Those with the lowest distortion were >= 100 VDC types - he found that many different makes are close to each other in terms of distortion. Personally, I've preferred Kemet (25 VDC types - 100 nF 1206 doesn't exist), then TDK.
Good luck with your endeavours ;-)
Jesper
Good luck with your endeavours ;-)
Jesper
Unless you are have a very wideband amplifier, this output stage is not needed either and will gain you a slight improvement in sound.
@gentlevoice :
Im not sure, but i think you might be right, the high frequencies sound a bit more detailed with the caps. I use wima pps caps now, the c0g were not better than without. But Im not really sure, i invited some guests for this weekend who work in studios and we will try blind tests. I will tell you on monday what came out of this.
Im not sure, but i think you might be right, the high frequencies sound a bit more detailed with the caps. I use wima pps caps now, the c0g were not better than without. But Im not really sure, i invited some guests for this weekend who work in studios and we will try blind tests. I will tell you on monday what came out of this.
Blind testmay fool you... If you introduce distortion like modulation, the music gets stronger. You believe its better. Louder level is interpreted by the brain as better... So when you disconnect bypasscaps you may well end up in the situation where you actually fool yourself.. If you make a tweak and you hear less. Thats a good sign! If you get calm, and the heartrate is slowing down, thats good! If it sounds dull maybe its less distortion. If you like it in long term thats good, if you can listen for hours, well done! I introduced approx 12 diffrent main filter transformers 5kVA, 2x3kVA, 2x2,5kVA, 2kVA... New fuse central, new feeding 3phase kable, filters, emi supression filters. All in all stuff for >15000 USD. The result is a "duller" maybe darker sound. Since this is my proffession I have a strong idea about the mechanisms. Its simply more music and less intermodulation distortion components. And electrostatic speakers has quite low distortion compared to dynamic speaker elements, so the difference is clear even for my wife. If you are not sure about the result of your mod... Listen for a longer time. A week at least.
The difference is there, but as esl 63 mentioned, sometimes you think it sounds better, but the measures are worse. After many hours of listening and a/b testing i ended up with pps wima caps. They sound very close to no caps, but the output has lower distortion.
Next week i will try to listen to different cap values and also measure the output. I wonder if there is a noticeable difference between different cap values. Is lower value better ? Whats the sweetspot, where a cap does help remove distortion, but is lowest possible value ?
Next week i will try to listen to different cap values and also measure the output. I wonder if there is a noticeable difference between different cap values. Is lower value better ? Whats the sweetspot, where a cap does help remove distortion, but is lowest possible value ?
Hi LinuxGeek,
A couple of comments to your post above ... Basically the optimum decoupling depends a.o.t. on the sampling frequency (possibly the clock frequency), the PSU "bleed through" frequency, the PCB layout, and the IC/DAC used. Different IC packages have different pin inductances (inductance is what mainly matters for digital PSU decoupling) and so this inductance may somewhat shift the optimum value of the decoupling capacitor.
Also different capacitor values have different resonance frequencies - which may be high or low Q - just to complicate matters further. PPS capacitors measure very well up to about 1 MHz (to my memory this is general to this type of dielectric but not entirely sure) at which point the dissipation factor increases very much causing distortion above this frequency.
I've personally been looking at this for quite some time - in order to find an optimum solution - and it may not be that simple. FYI I've attached two LTSpice simulations (DAC pins to the left 1.1 nH each), one showing that e.g. a 100 nF 1206 size capacitor (C0G) in a "real-world setting" has a non-linear impedance vs. frequency curve, thus decoupling to a very different level at different frequencies. It's the 100 nF capacitor (the combination of low capacitor impedance & low inductance & relatively low capacitance) that causes the quite noticable wrinkle at ~6 - 20 MHz with a max decoupling around 16 MHz of ~42 dBs.
On the other hand a 1uF capacitor with a 100 mohm impedance shows a much smoother impedance course - however - just for comparison - the decoupling around 16 MHz is reduced to 22 dBs.
One approach to this is to find out which frequency is mainly present on the DACs supply pins (what I call the "bleed through" frequency) and then select a capacitor that has an optimum damping here. To this end I've found that Kemet's K-Sim may be useful:
KEMET K-SIM
Here it's possible to select a wide range of different capacitors and see what the impedance/ESR plots look like and also change e.g. bias voltage and external resistance.
And so what does this means from a sound-quality point of view? IMHO not least that there are indeed many options, yet since I personally believe in low distortion (i.e. low noise) also on supply pins I've chosen to use C0Gs where I can in the smallest size that effectively filters the supply pin noise main frequency at the main sampling frequency I listen to. I.e. I optimize my DAC e.g. for 384 kHz knowing that it may then not be as good at e.g. 44.1 kHz or vice versa.
Hope this may be of inspiration.
And so now I'm off for vacation so won't be back here for the next weeks.
Cheers,
Jesper
A couple of comments to your post above ... Basically the optimum decoupling depends a.o.t. on the sampling frequency (possibly the clock frequency), the PSU "bleed through" frequency, the PCB layout, and the IC/DAC used. Different IC packages have different pin inductances (inductance is what mainly matters for digital PSU decoupling) and so this inductance may somewhat shift the optimum value of the decoupling capacitor.
Also different capacitor values have different resonance frequencies - which may be high or low Q - just to complicate matters further. PPS capacitors measure very well up to about 1 MHz (to my memory this is general to this type of dielectric but not entirely sure) at which point the dissipation factor increases very much causing distortion above this frequency.
I've personally been looking at this for quite some time - in order to find an optimum solution - and it may not be that simple. FYI I've attached two LTSpice simulations (DAC pins to the left 1.1 nH each), one showing that e.g. a 100 nF 1206 size capacitor (C0G) in a "real-world setting" has a non-linear impedance vs. frequency curve, thus decoupling to a very different level at different frequencies. It's the 100 nF capacitor (the combination of low capacitor impedance & low inductance & relatively low capacitance) that causes the quite noticable wrinkle at ~6 - 20 MHz with a max decoupling around 16 MHz of ~42 dBs.
On the other hand a 1uF capacitor with a 100 mohm impedance shows a much smoother impedance course - however - just for comparison - the decoupling around 16 MHz is reduced to 22 dBs.
One approach to this is to find out which frequency is mainly present on the DACs supply pins (what I call the "bleed through" frequency) and then select a capacitor that has an optimum damping here. To this end I've found that Kemet's K-Sim may be useful:
KEMET K-SIM
Here it's possible to select a wide range of different capacitors and see what the impedance/ESR plots look like and also change e.g. bias voltage and external resistance.
And so what does this means from a sound-quality point of view? IMHO not least that there are indeed many options, yet since I personally believe in low distortion (i.e. low noise) also on supply pins I've chosen to use C0Gs where I can in the smallest size that effectively filters the supply pin noise main frequency at the main sampling frequency I listen to. I.e. I optimize my DAC e.g. for 384 kHz knowing that it may then not be as good at e.g. 44.1 kHz or vice versa.
Hope this may be of inspiration.
And so now I'm off for vacation so won't be back here for the next weeks.
Cheers,
Jesper
@Ian:
Thanks, any input is appreciated. If you have some suggestions plz share your ideas.
@gentlevoice:
I build a stage to upsample to 192khz. I use the SRC4392 because i had some of them flying around here. Then I tried to optimize decoupling for this frequency. Thanks for your input and the KSIM link, didnt know it before.
It did not really help with the PCM5102. Still a limited soundstage compared to DACs with I/V stage. Not sure, but all of my voltageout DACS lack the ability to place instruments in the room. Compared to the I/V Dacs and my R2R of course. Is this a voltageout DAC limitation ?
But then i hooked up the SRC stage to my PCM1794a. Damn, i never thought that this one could become even better. I use two LIFEPO4s to power it and TPS7A4700s for regulation to 5V and 3.3V. The I/V stage now has its own batteries because of higher voltage and the needed negative voltage. All decoupling is optimized for 192khz and the SRC4392 is used for upsampling.
This ROCKS. It sounds beautiful. LIFEPO4s extended the bass for this DAC, but the PPS wimas now help to also make the upper frequencies beautiful. The midtones were astonishing even before the batteries when i used talema transformers, but the hights now are more "enjoyable". Not sure how to describe it, they are still very detailed, but now the details do not make them sound "sterile". Hard to discribe, english is not my main language and i cant find the right words...
I never thought that my DIY PCM1794a will destroy my new Linn streamer, but it seems so. I will make some more measurements at work tomorrow, if i find some time, but my oscilloscope here at home and the microphone i use for measurements show very promising numbers.
Thanks for all the help. For me LIFEPO4 and wima pps did the job.
@Ian :
I thought about using the IL715 to isolate my PCM5102. Is the added jitter very high ? I know you made some tests with isolation. Do you think its worth a try to synchronize I2S with flipflops only if there is only one frequency in use or is this rather a dream and i will have to buffer ? My problem is, if i buffer the signal, i will loose lipsync on my diy beamer I am thinking about flipflops that output when they get a "trigger" signal. What would happen if i trigger the flipflop with the bitclock ? Will the switching be wrong because of signals drifting apart (because of jitter). Ever tried something like this ?
Thanks, any input is appreciated. If you have some suggestions plz share your ideas.
@gentlevoice:
I build a stage to upsample to 192khz. I use the SRC4392 because i had some of them flying around here. Then I tried to optimize decoupling for this frequency. Thanks for your input and the KSIM link, didnt know it before.
It did not really help with the PCM5102. Still a limited soundstage compared to DACs with I/V stage. Not sure, but all of my voltageout DACS lack the ability to place instruments in the room. Compared to the I/V Dacs and my R2R of course. Is this a voltageout DAC limitation ?
But then i hooked up the SRC stage to my PCM1794a. Damn, i never thought that this one could become even better. I use two LIFEPO4s to power it and TPS7A4700s for regulation to 5V and 3.3V. The I/V stage now has its own batteries because of higher voltage and the needed negative voltage. All decoupling is optimized for 192khz and the SRC4392 is used for upsampling.
This ROCKS. It sounds beautiful. LIFEPO4s extended the bass for this DAC, but the PPS wimas now help to also make the upper frequencies beautiful. The midtones were astonishing even before the batteries when i used talema transformers, but the hights now are more "enjoyable". Not sure how to describe it, they are still very detailed, but now the details do not make them sound "sterile". Hard to discribe, english is not my main language and i cant find the right words...
I never thought that my DIY PCM1794a will destroy my new Linn streamer, but it seems so. I will make some more measurements at work tomorrow, if i find some time, but my oscilloscope here at home and the microphone i use for measurements show very promising numbers.
Thanks for all the help. For me LIFEPO4 and wima pps did the job.
@Ian :
I thought about using the IL715 to isolate my PCM5102. Is the added jitter very high ? I know you made some tests with isolation. Do you think its worth a try to synchronize I2S with flipflops only if there is only one frequency in use or is this rather a dream and i will have to buffer ? My problem is, if i buffer the signal, i will loose lipsync on my diy beamer
I am thinking about flipflops that output when they get a "trigger" signal. What would happen if i trigger the flipflop with the bitclock ? Will the switching be wrong because of signals drifting apart (because of jitter). Ever tried something like this ?
Last edited:
@Ian:
Thanks, any input is appreciated. If you have some suggestions plz share your ideas.
@gentlevoice:
I build a stage to upsample to 192khz. I use the SRC4392 because i had some of them flying around here. Then I tried to optimize decoupling for this frequency. Thanks for your input and the KSIM link, didnt know it before.
It did not really help with the PCM5102. Still a limited soundstage compared to DACs with I/V stage. Not sure, but all of my voltageout DACS lack the ability to place instruments in the room. Compared to the I/V Dacs and my R2R of course. Is this a voltageout DAC limitation ?
But then i hooked up the SRC stage to my PCM1794a. Damn, i never thought that this one could become even better. I use two LIFEPO4s to power it and TPS7A4700s for regulation to 5V and 3.3V. The I/V stage now has its own batteries because of higher voltage and the needed negative voltage. All decoupling is optimized for 192khz and the SRC4392 is used for upsampling.
This ROCKS. It sounds beautiful. LIFEPO4s extended the bass for this DAC, but the PPS wimas now help to also make the upper frequencies beautiful. The midtones were astonishing even before the batteries when i used talema transformers, but the hights now are more "enjoyable". Not sure how to describe it, they are still very detailed, but now the details do not make them sound "sterile". Hard to discribe, english is not my main language and i cant find the right words...
I never thought that my DIY PCM1794a will destroy my new Linn streamer, but it seems so. I will make some more measurements at work tomorrow, if i find some time, but my oscilloscope here at home and the microphone i use for measurements show very promising numbers.
Thanks for all the help. For me LIFEPO4 and wima pps did the job.
@Ian :
I thought about using the IL715 to isolate my PCM5102. Is the added jitter very high ? I know you made some tests with isolation. Do you think its worth a try to synchronize I2S with flipflops only if there is only one frequency in use or is this rather a dream and i will have to buffer ? My problem is, if i buffer the signal, i will loose lipsync on my diy beamer
Thanks for sharing your learnings LinuxGeek!
I am wondering which Linn Streamer were you comparing to?
Any new improvements you had over time with your dac/streamer project?
@Ian:
Thanks, any input is appreciated. If you have some suggestions plz share your ideas.
@gentlevoice:
I build a stage to upsample to 192khz. I use the SRC4392 because i had some of them flying around here. Then I tried to optimize decoupling for this frequency. Thanks for your input and the KSIM link, didnt know it before.
It did not really help with the PCM5102. Still a limited soundstage compared to DACs with I/V stage. Not sure, but all of my voltageout DACS lack the ability to place instruments in the room. Compared to the I/V Dacs and my R2R of course. Is this a voltageout DAC limitation ?
But then i hooked up the SRC stage to my PCM1794a. Damn, i never thought that this one could become even better. I use two LIFEPO4s to power it and TPS7A4700s for regulation to 5V and 3.3V. The I/V stage now has its own batteries because of higher voltage and the needed negative voltage. All decoupling is optimized for 192khz and the SRC4392 is used for upsampling.
This ROCKS. It sounds beautiful. LIFEPO4s extended the bass for this DAC, but the PPS wimas now help to also make the upper frequencies beautiful. The midtones were astonishing even before the batteries when i used talema transformers, but the hights now are more "enjoyable". Not sure how to describe it, they are still very detailed, but now the details do not make them sound "sterile". Hard to discribe, english is not my main language and i cant find the right words...
I never thought that my DIY PCM1794a will destroy my new Linn streamer, but it seems so. I will make some more measurements at work tomorrow, if i find some time, but my oscilloscope here at home and the microphone i use for measurements show very promising numbers.
Thanks for all the help. For me LIFEPO4 and wima pps did the job.
@Ian :
I thought about using the IL715 to isolate my PCM5102. Is the added jitter very high ? I know you made some tests with isolation. Do you think its worth a try to synchronize I2S with flipflops only if there is only one frequency in use or is this rather a dream and i will have to buffer ? My problem is, if i buffer the signal, i will loose lipsync on my diy beamer
@LinuxGeek
Isolator is doing great job to cut the ground loop thus stop the transmission of common mode EMI noise. Isolator is the only efficient way to eliminate this kind of noise. However, isolator itself introduces jitter, no matter what kind of, optical, capacitance or magnetic all of them. To use the isolator, you need to re-clock the signals, but have to use the local XO clock. Because it would be useless using the isolated clock to re-clock other signals. That's why FIFO + Isolator + Dual XO/re-clock is the most suitable solution so far. Or, you can also run you DAC at master clock mode after isolator.
Regards,
Ian
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.