Urgen question about de-coupling capacitor for PCM1704

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
hi all
The decoupling capactiors for PCM1704 in datasheet:
"Aluminum electrolytic capacitors are recommended for larger values."
REF DC = 47uf
SERVO DC = 47uf
BPO DC = 100uf
Is there ESR critical ? May I use the ultra-low ESR capacitors (e.g. X5R MLCC) to replace? (Sound quality is not matter, will it be unstable?)
If anyone knows the answer, please let me know, very thanks.
 
hi all
The decoupling capactiors for PCM1704 in datasheet:
"Aluminum electrolytic capacitors are recommended for larger values."
REF DC = 47uf
SERVO DC = 47uf
BPO DC = 100uf
Is there ESR critical ? May I use the ultra-low ESR capacitors (e.g. X5R MLCC) to replace? (Sound quality is not matter, will it be unstable?)
If anyone knows the answer, please let me know, very thanks.

Why do you want to use ultra-low ESR caps for decoupling? Typically you wouldn't want the ESR too low, for decoupling, since the ESR might help to add damping for resonances involving the decoupling capacitance and the inductance of the supply traces or wires.
 
PCM1704 is a digital circuit building analog waveform. Global sound of your DAC will depends of sonic signature of those capacitors. The choice is not free.

What makes you think that?

You have not presented any evidence linking the quality of the bypass capacitors to the sound quality of the output.

I am NOT saying that you are wrong. I am saying that you have not yet given enough information to support your position. But maybe it's my fault, since I am not familiar with that IC.

I could still postulate that if the digital part of the chip is what depends on the bypass capacitors, then the analog output would almost certainly be unaffected by their quality, unless they were so bad that they caused a whole bit to be misconstrued, which seems extremely unlikely. That's the true beauty of going digital.

Bypass capacitors act as small point-of-load power supplies so that when the IC has sudden/high-rate demands for current, then the caps can supply it much more easily and better than the inductive power supply rail traces or wires, which, if made to try, would then also induce voltage spikes on the rails (v = L dI/dt; ouch).

Are there analog active filters or analog output amplifiers or impedance converters (buffer amplifiers) built into that IC's output stages? In that case, you could easily be correct.

You might already know all of this, but, the "standard" way to bypass, if you don't want to analyze thoroughly and it isn't too critical, is a non-"low ESR" electrolytic in parallel with a small-value lower-quality ceramic cap (i.e. not NPO or C0G type). But if I were bypassing something where it might matter for the sound quality, I might use some combination of high-quality film capacitors, and probably even also some electrolytic. BUT, because of the low ESRs, then, UNLESS I was doing a full massive analysis regimen that took into account ALL of the parasitic effects in both the actual components and in the traces and layout, etc etc etc, I would almost-certainly add a small-value (typically 3 to 30 Ohms) high-quality series resistance just upstream in series with each set of bypass capacitors. Otherwise, it's just like ASKING for problems, especially with digital stuff involved. It's even possible that you might think that you're hearing the effects of different capacitors' "sonic signatures" when it might actually be the different resonances that different caps allow to be excited, causing different-sounding problems on the power supply rails. Maybe you should even try putting a larger electrolytic capacitor on the upstream side of that small-value resistor, to better-supply the film caps. You could even split the resistor and put one on each side of the electrolytic, to provide low-pass filtering in both directions. Also, you would typically want to make sure that you have at least large-ish value and very small value bypass caps, with values depending on what the chip might try to pull, which depends on the amplitudes and the time-rates-of-change and the time-durations of the possible current requirements. You might also want to try fine-tuning the individual bypass cap values that you use. But always put the smallest ones as close to the IC pins as possible.

By the way, you should be able to use an oscilloscope and SEE whether or not anything like audio signals are being handled by the bypass capacitors.

Cheers,

Tom
 
Last edited:
Inside PCM1704

First of all, it might be useful to look inside PCM1704 to understand the utility of each capacitors.
 

Attachments

  • PCM1704 internal.jpg
    PCM1704 internal.jpg
    82.9 KB · Views: 484
The test that I made
In a prototype of a Jundac Two DAC, I made some change in a single channel (right channel for example). I change only one of those capacitor at a time.
I burn a monophonic CD and start listening comparison. Left channel is the reference; right channel is the modified one. Since the CD is monophonic, I listen to only one channel at a time on my two speakers.

I try various capacitors with various values (OS-CON, ELNA, PANSONIC, MKT, MKP...); various values, because 47µ with MKP capacitor is not possible.
Each comparison gives different results. This means that capacitors change sonic signature coming from DAC chip.
 
The test that I made
In a prototype of a Jundac Two DAC, I made some change in a single channel (right channel for example). I change only one of those capacitor at a time.
I burn a monophonic CD and start listening comparison. Left channel is the reference; right channel is the modified one. Since the CD is monophonic, I listen to only one channel at a time on my two speakers.

I try various capacitors with various values (OS-CON, ELNA, PANSONIC, MKT, MKP...); various values, because 47µ with MKP capacitor is not possible.
Each comparison gives different results. This means that capacitors change sonic signature coming from DAC chip.

Your technique of using a monophonic input, and having one channel as a reference for comparison with the modified channel, is a very good testing technique. The only improvement would be to also use measurement equipment, in addition to your ears.

I am sure that your results are essentially valid and correct, especially after looking at the internal schematic of that IC.

Actually, 47 uF polypropylene capacitors do exist and are fairly common, but are relatively expensive, and large. I just purchased two Solen 50uF 400V 5% PB-MKP-FC from madisound.com for $16 USD each, for example, to upgrade the electrolytic caps in the bass filters in a pair of speaker crossovers. Places like parts-express.com also have large-value MKP caps. The ones I just received are about 60 mm long x 44 mm diameter, with axial leads. And printed on them is "Made in France". :)

The largest-capacitance MKP cap that I saw at parts-express.com was a Solen 200uF 250V for $61.43 USD.

Some people also try the industrial-type "motor run" polypropylene capacitors (not the "motor start" types), which come in quite-large values, often at very-reasonable prices.

http://www.parts-express.com/crossover-capacitor-index.cfm
http://www.parts-express.com/wizard...AT&srchCat=163&CFID=31374676&CFTOKEN=87284457

https://www.madisound.com/store/index.php?cPath=404_5

Cheers,

Tom
 
Last edited:
I am not...

Your technique of using a monophonic input, and having one channel as a reference for comparison with the modified channel, is a very good testing technique. The only improvement would be to also use measurement equipment, in addition to your ears.

I am sure that your results are essentially valid and correct, especially after looking at the internal schematic of that IC.

Actually, 47 uF polypropylene capacitors do exist and are fairly common, but are relatively expensive, and large. I just purchased two Solen 50uF 400V 5% PB-MKP-FC from madisound.com for $16 USD each, for example, to upgrade the electrolytic caps in the bass filters in a pair of speaker crossovers. Places like parts-express.com also have large-value MKP caps. The ones I just received are about 60 mm long x 44 mm diameter, with axial leads. And printed on them is "Made in France". :)

The largest-capacitance MKP cap that I saw at parts-express.com was a Solen 200uF 250V for $61.43 USD.

Some people also try the industrial-type "motor run" polypropylene capacitors (not the "motor start" types), which come in quite-large values, often at very-reasonable prices.

Parts-Express.com - Crossover Capacitor Index: Polypropylene, Metallized Film, Metallized Polypropylene, Non-Polarized Electrolytic
POLYPROPYLENE CAPACITORS from Parts Express ship same day and come with 45 day money back guarantee. Free Shipping Available. Order free 10,000 product catalog.

https://www.madisound.com/store/index.php?cPath=404_5

Cheers,

Tom


An expert on PCB design, but my general understanding is that large capacitors like these, with long lead spacing, are going to create big problems with loop inductances for use as decoupling caps in a relatively high speed circuit environment. Real experts please confirm; I think small lead spacing, and short PCB traces from (DAC chip) supply pins to the decoupling cap are critical for best performance.
 
The test that I made
In a prototype of a Jundac Two DAC, I made some change in a single channel (right channel for example). I change only one of those capacitor at a time.
I burn a monophonic CD and start listening comparison. Left channel is the reference; right channel is the modified one. Since the CD is monophonic, I listen to only one channel at a time on my two speakers.

I try various capacitors with various values (OS-CON, ELNA, PANSONIC, MKT, MKP...); various values, because 47µ with MKP capacitor is not possible.
Each comparison gives different results. This means that capacitors change sonic signature coming from DAC chip.


So which one did you prefer most in your circuit? Any PIO caps in the bunch?

--G
 
Inductance is the killer with decoupling capacitors, small SMD devices next to the pins are best. Decoupling capacitors wont effect the sound, but bad decoupling will cause problems. Most IC manufacturers sites and captes (AVX) have copious notes on decoupling capacitors and how to do it.
MLCC are excellent for decoupling as they are a low inductance design, X7R are suitable and can be got in smaller size packages, again reducing inductance, COG I would use for crytals etc, in sensitive places. I would not reccomend a dielectric below X7R.
For the high frequency decoupling you require the smallest value caps next to the pins, and it is high frequency noise you are trying to get rid off.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.