Confused: Best Approach to Cap Replacement?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I have searched here and across the web but come up with conflicting answers to the question of whether to increase cap value when upgrading caps.

Encouraged by the positive effect of upgrading the DC blocking caps on the balanced outs of my TEAC UD-501 DAC, I'm thinking about changing out the generic 47uF SMD electrolytics around the opamps and DAC chips. For the DAC chips, thinking about Oscons on the 3.3 digital supply and Silmics on the analog 5 volt Vcc pins. And Silmics to replace the two electros at each of the Op Amps.

Should I replace the existing 47uF caps with 47uF caps or increase the values? Reports from my fellow tweakers assure us that audio ecstasy requires nothing more than sticking in the largest cap that will fit. On the other hand, manufacturers and engineers in data sheets, application notes, test results and scope traces warn these coupling caps are part of a filter network that has been specifically tuned to filter out the frequency of noise actually found in that circuit: change the cap value without tuning the rest of the network and you'll probably end up getting more noise and worse sound.

So which is it? Max out the capacitance or keep it the same? Also any other feedback on choice of cap brands, other mods to consider on the UD-501, what to watch out for, etc etc, is welcome! Thanks much.
 
47uF does seem very stingey for decoupling. That said there is some truth in the claim that bigger isn't always better - not so much because the values are specifically chosen to filter out specific frequencies but because in most commercial circuits there's no separation between signal 0V and power 0V. Putting in bigger caps tends to work to couple more noise from power supplies onto signal 0V. My own way around this is to decouple +ve to -ve supply - i.e. across the opamp power pins with the additional cap leaving the 47uFs in place.

Recommended brands are Nichicon and Nippon Chemi. Go for their low ESR ranges. Oh and avoid SMT caps - they tend to suck compared to the leaded ones.
 
Thanks to those that have replied so far. I'm sensing a consensus around leaving the values (more or less) alone, possibly going up to 100uF.

RE: brands of caps. I guess I need to further mull over which brands to use, given your comments on Nichicon, Nippon Chemi and using Oscons in analog.

I'm trying to move the sound a bit warmer with more body and I think I've noticed a nice warming/added body affect from Silmics and this is also consistent with comments on the web. Thats why I was thinking to use those on the analog side. What do y'all see as the advantage of the alternative brands vs Silmics?

I do have a bunch of 100uF (non SMD) Oscons on hand, but there seems to be a lot of folks saying they just don't work well on the analog side of things.

Abaxalito: your saying simply a cap from +V to -V pins on the opamps? Large value cap here? I've got a bunch of 1000 and 2200uF Panasonic FMs I might use. Is there anything similar you would recommend for the DAC chips?
 
Yes I'm suggesting directly across the opamps between pin4 and pin7 (if singles) or pin8 (if duals). I don't pay much attention to brands myself - I mentioned Nichicon and NCC because another member here with a lot of experience in re-capping (stephensank) says they're the most reliable ones. I've not been in the business of substituting caps long enough to know that. Recently I've found a Korean manufacturer whose specs for ESR beat the Japanese - Samyoung. However they might not be widely available. If you can identify what the DAC chips are maybe I'll have some suggestions. Some DAC chips contain opamps - in which case definitely those opamps are going to be the most critical ones for sound quality because they tend to have very poor PSRR. But if the DAC hasn't any internal opamps its decoupling is far less critical.
 
Pics of the PCM 1795 DAC Chip

Here, these might help: pics of the block diagram and the pinout for the 1795.
 

Attachments

  • Block Diagram of DAC.gif
    Block Diagram of DAC.gif
    41.1 KB · Views: 213
  • Pinout of DAC.gif
    Pinout of DAC.gif
    14.7 KB · Views: 213
Right - PCM1795 is a current-out DAC hence no internal opamps. What does benefit this line of DACs is attention to the Iref pin. Some guys on another thread - maybe the DDDAC one - found that the bass can be improved by modifying the suggested circuit feeding this pin. I think they used a JFET CCS but can't be sure, I'll check out the details. The opamps that need the most attention will be the ones being fed from the Iout pins of the DAC - they should have extreme decoupling - including ceramics - because they're slewing at extreme frequencies. Also if the usual 5534s have been fitted you'll probably get SQ improvement from going to LM6171 (5532s -> LM6172s).
 
OK, let me try to summarize what I think is the way forward. Please correct me as needed:

1. Adding a JFET on the Iref pin might be beyond where I want to go, although I am interested in hearing what's involved.

2. Replacing the op amps. The UD-501 uses a pair of Muses 8920's beyond each DAC chip that then feed the balanced outputs. From my limited electronic understanding, these are pretty good op amps. I will leave them in place and try to enhance the decoupling.

3. Decoupling the opamps. Per your suggestion I'll solder a 1000 or 2200uf Panasonic FM across pins 4 and 8 of each opamp. I will add a ceramic in parallel across pins 4 and 8. Or would a the lower ESR 100uF OSCON be a better choice than the higher value Panasonic here? I can see SMD (probably ceramic) caps and some resistors already in circuit around the Muses. So, without deconstructing the current circuit or negating what the TEAC engineers have already done I'm not sure where else I could add anything?

4. Upgrading 18 generic 47uF SMD electros. Will replace the pair accompanying each opamp with Silmic caps. Will replace the four on the analog side of each DAC (Vcc1; VcomL; VcomR; Vcc2L and Vcc2R) with Silmics. Will replace the one on Vdd with an OSCON. I have Silmics and OSCONS in both 47 and 100uF. The current electros are in circuit with SMD ceramics and resistors. Is it advisable to modestly increase the electros to 100uF or should I leave well enough alone and replace like for like value?

4. Replacing Caps on the outputs. In addition to what we've discussed, I will be replacing the current DC blocking caps feeding the balanced outputs with teflons. I did the calculation based on the my downstream amp's input impedance and found I can lower the size of this cap all the way down to .47uF from 100uF.

OK before I start making smoke: Am I missing anything? Am I about to screw anything up? Any alternative suggestions?

I appreciate everyone's help. Thanks again!
 
Hmmm... not sure what you mean re: voltage? My caps are rated at 16 and 25v which should be fine for 3.3 and 5v supplies.

Are you saying the plan looks good? If so, my only questions were whether to stick with the 47uF or increase to 100uF when replacing the current SMD caps at the DAC and Op Amps. And whether its better to use a low value OSCON (for the low ESR) or use the high value Panasonic cap across the Op Amps.
 
Last edited:
Your Muses opamps are running on 5V or +/-5V? For 3.3V you'll get a more compact size if you go for 6.3V working caps. 16V would be a good choice if the rails are +5/-5V. I have no experience with using SMD caps other than measuring some on a PCB I bought a couple of years ago and finding they sucked big-time. So if you want suggestions about what to do with your SMD caps I'd say ditch them and replace with normal leaded ones. Also about OSCONs vs Panasonics I also don't know which will be best - normally I avoid OSCONs because they're so much more expensive than normal 'lytics. So if it were me I'd go for the bigger value but I can't swear hand on heart it'll give you better SQ than OSCONs.
 
If the pcb layout is made for smd caps, you risk to do more harm than good by changing them for leaded caps. There are extremely good smd caps these days. Considering how low ESR is achieved these days, the quality of soldering and the pcb/parts layout play a big role in the effectiveness of decoupling.

Btw, it's good to keep in mind that, above 100khz, a 100uf polymer will be more effective than a 1000uf low esr electro (more and more so as you go higher in frequency).

For the DAC pins, there are polymer caps aplenty in smd package, have a look at mouser and digikey. They'll be fine for both the digital and analog power pins. They're not required for the vcom pins though, a decent electro will do fine. However, if you change the electrolytics, you probably also have to change the ceramic in // to avoid resonances. As a rule of thumb, I'd start at least with 2.2uf ceramics in // with oscons.

But before anything, I'd try at least to determine what are those "generic" smd caps and grab their datasheet. Maybe aren't they that bad to start with.
 
Thanks Ben. The more I get into this the more my thinking is in line with what you're suggesting. I've added Oscons between the pos and neg on the opamps as discussed with Abraxalito and not sure I'm hearing any difference. In looking at what a professional modder did with the UD-501, I noticed they left all the SMD electros alone but did change out some of the leaded electros on the board--probably for like values but higher quality. I may take a similar approach. I also have a quad of teflon caps I just bought to replace the DC blocking caps on the XLR outputs. I expect a notable improvement there, as I had earlier replaced the factory caps with Silmics and got quite an improvement.
 
Thanks. I checked the Pin 1s and they're grounded, so all is OK there.

Maybe I AM hearing a difference with the Oscons across the op amps. Right after installing them I thought the system sounded thin with little body, like maybe the lower mids had gone missing. Later the thinness was gone, but I thought it was sounding light in the bass. At that point I figured it was all in my mind.

But then last night and again this morning I had some Sinatra/Riddle playing. I've never heard it like this before. Everything was very smooth or creamy sounding but most notably Sinatra's voice sounded recessed, so recessed that on one song I swear it sounded like it was out of phase!

Either I'm losing my mind, or maybe that thing John mentioned above, about Oscons burning in for 3-400 hours, maybe that's what I'm hearing?
 
Don't know about OSCONs but caps do take time to 'bed in' most certainly. When new the sound is often not very satisfying, 'thin' is a good way to describe it. Pin1 being grounded isn't the issue - rather how its grounded. Does it go onto the PCB or has it a wire direct to chassis without a PCB connection? My bet is the former.
 
Well you're right my expat friend. Pin 1 does go into the board. But it goes into a multi cap/resistor network and then to what looks like an isolation transformer. One x-former for each pin, actually. The complexity gets me thinking the Teac double Es had something thought out in mind here and maybe I shouldn't mess with it. And when I said grounded earlier I prob should have said "showing continuity to the chassis".

BTW I just noticed those big dual Muses op amps for each channel seem to feed into a little single N5532A op amp before heading off to the balanced out XLR connectors. hmmmmm....
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.