Talking about the lytics at the power pins of the DAC, Servo, filter and Opamps in CD players. I don't see the need to go higher than 220uF (audio grade lytics for analogue rails, Oscons for digital). However, since I'm going to order 100 caps at once, going 100uF would be quite a bit less expensive for all the CD players I want to recap. 47uF would be even better for my wallet 
So compare 47uF vs 100uF vs 220uF. What difference will it make in sound if you're using high quality caps? Do higher values give better lows or better clarity?
So compare 47uF vs 100uF vs 220uF. What difference will it make in sound if you're using high quality caps? Do higher values give better lows or better clarity?
Actually in all my Philips CD players I see the values are 33-47uF. Using audio grade caps of the same values would already be a noticable improvement. I'm just wondering what difference higher values would make on top of that.
Most people choose 220uF or 470uF for the lytics and tell to use no less than 220uF for Oscons on the digital rails but I have no idea why? Why would 100uF be worse than 220uF for these caps?
Last edited:
Signal capacitors are best replaced with same value but higher quality or with different type (bipolar electrolytic, MKP, MKS, etc). Increasing capacity to attain lower bottom end is a vain job. Use this formula to calculate how the capacitor size affects low-end frequency response at circuit input:
F=1/(2*pi*R*C) (note that F is at -3dB roloff)
With PSU and smoothing capacitors, higher capacitor means lower ripple and clean DC is very important for digital circuits, hence the popular belief that increasing a rail bypassing cap in a digital circuit will improve performance. Engineers would not save cost there but rather in large, more costly capacitors in PSU. Speaking of cost, there is negligible difference in price between 100 and 220uF caps.
F=1/(2*pi*R*C) (note that F is at -3dB roloff)
With PSU and smoothing capacitors, higher capacitor means lower ripple and clean DC is very important for digital circuits, hence the popular belief that increasing a rail bypassing cap in a digital circuit will improve performance. Engineers would not save cost there but rather in large, more costly capacitors in PSU. Speaking of cost, there is negligible difference in price between 100 and 220uF caps.
Thanks for the clarification. What resistor should I be looking at for R in your formula?
I've settled on Nichicon FG 25V which is €16 at Mouser for 100 pieces of 100uF and €48 for 100 of 220uF, so the difference is pretty big. Same for Silmics which are €36 for 100 of 100uF and €64 for 100 of 220uF.
Also, what about voltage rating? I've read that starting at 35V these audio grade caps sound best. Going 35V for Nichicon FG 100uF is only a €5 price difference.
I've settled on Nichicon FG 25V which is €16 at Mouser for 100 pieces of 100uF and €48 for 100 of 220uF, so the difference is pretty big. Same for Silmics which are €36 for 100 of 100uF and €64 for 100 of 220uF.
Also, what about voltage rating? I've read that starting at 35V these audio grade caps sound best. Going 35V for Nichicon FG 100uF is only a €5 price difference.
Resistor in series with the signal cap, which is usually at the input of a circuit.
Where do you want to use FG? In signal path or for smoothing? I do agree that capacitors make a difference but the sound of different voltages is already masturbation.
BTW, if you decide to go with higher capacity or voltage, remember that audio-grade caps usually have bigger physical dimensions, so make sure they will fit where intended.
Where do you want to use FG? In signal path or for smoothing? I do agree that capacitors make a difference but the sound of different voltages is already masturbation.
BTW, if you decide to go with higher capacity or voltage, remember that audio-grade caps usually have bigger physical dimensions, so make sure they will fit where intended.
In addition, read this first: "What is proper decoupling and why" from Analog Devices:
http://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf
I have a friend who tweaks high end audio for a living, oftentimes he will reduce the possibility of vibration of ceramic bypass caps.
Ceramic decoupling caps and vibration is not going to be a problem for DACS, some cheep laptop supplies have been known to have singing caps in the SMPS due to certain frequencies when burst modes etc are on.
But for general decoupling the MLCCs vibrating wont be a problem in domestic audio, I isn't a problem for high reliability designs that do suffer some extreme vibration, but then this is for SMD devices which are what you need for next to the pins on a DAC device.
The devices that do suffer from vibration are electors especially SMD electrolytic caps, these do have a bad habit of breaking during vibration tests.
But for general decoupling the MLCCs vibrating wont be a problem in domestic audio, I isn't a problem for high reliability designs that do suffer some extreme vibration, but then this is for SMD devices which are what you need for next to the pins on a DAC device.
The devices that do suffer from vibration are electors especially SMD electrolytic caps, these do have a bad habit of breaking during vibration tests.
For general smoothing of ripple and recharging of the MF & HF local decoupling, bigger is better.
At the points of current demand, where there are fast changes in current demand, one must insert HF decoupling of appropriate values.
These need to be recharged very quickly, in time for the next change in demand.
This requires MF decoupling, again fairly close (to allow the fast recharge) to the the demand locations.
In the interests of cost saving many will combine the LF recharging with the MF decoupling.
I think there is something to be gained by separating these two, so that we end up with
LF, MF & HF local decoupling on the PCB.
I see nothing wrong with 1mF and more on the PCB, with a few 100uF to 220uF local to the main circuits and very many 1uF and smaller at every demand location.
At the points of current demand, where there are fast changes in current demand, one must insert HF decoupling of appropriate values.
These need to be recharged very quickly, in time for the next change in demand.
This requires MF decoupling, again fairly close (to allow the fast recharge) to the the demand locations.
In the interests of cost saving many will combine the LF recharging with the MF decoupling.
I think there is something to be gained by separating these two, so that we end up with
LF, MF & HF local decoupling on the PCB.
I see nothing wrong with 1mF and more on the PCB, with a few 100uF to 220uF local to the main circuits and very many 1uF and smaller at every demand location.
For digital the capacitors next to the pins should be in the smallest package going to reduce parasitic inductance as this is the killer for decoupling efficiency...
Most digital (and analogue to a lesser extent) decoupling schemes are pretty ad hoc, very few designs are simulated due to the cost of simulation, but luckily digital is pretty robust so can handle it. Having a closely coupled power and ground plane is a must as this provides some planar capacitance (critical over 20MHz) 0402 caps next to the pins 0.001 -0.100uF, local reservoir caps (MF) 100uf-400uF scattered around near the devices, then larger caps further out. How many of each depends on the design, but every power pin should have at least one small device next to the pins, the MF and LF caps depends on board size, number of devices etc.
When you get FPGAs or complex processors on the board the initial switching current requirements can be in the 100A range.... For simple DIY DAC designs with a few devices getting it right will decrease the power supply noise and minimise simultaneous switching noise helping keep everything nice and clean...
AVX, KEMET, Panasonic and other cap manufacturers have loads of info on decoupling as do the chip manufacturers...
Some search hints:
AN0035.pdf
an1325.pdf
AN295.pdf
gl-cyintro.pdf
guide05.pdf
JDI_X2Y_STXII.pdf
Most digital (and analogue to a lesser extent) decoupling schemes are pretty ad hoc, very few designs are simulated due to the cost of simulation, but luckily digital is pretty robust so can handle it. Having a closely coupled power and ground plane is a must as this provides some planar capacitance (critical over 20MHz) 0402 caps next to the pins 0.001 -0.100uF, local reservoir caps (MF) 100uf-400uF scattered around near the devices, then larger caps further out. How many of each depends on the design, but every power pin should have at least one small device next to the pins, the MF and LF caps depends on board size, number of devices etc.
When you get FPGAs or complex processors on the board the initial switching current requirements can be in the 100A range.... For simple DIY DAC designs with a few devices getting it right will decrease the power supply noise and minimise simultaneous switching noise helping keep everything nice and clean...
AVX, KEMET, Panasonic and other cap manufacturers have loads of info on decoupling as do the chip manufacturers...
Some search hints:
AN0035.pdf
an1325.pdf
AN295.pdf
gl-cyintro.pdf
guide05.pdf
JDI_X2Y_STXII.pdf
For digital it depends on the frequency range and that is determined by the rise time of the signal, so you may end up with different values on different devices... The board I am doing now has small decoupler's (next to the device pins) ranging from 1nF up to 680nF depending on the device, most pins having two caps a small two digit nF and a 3 digit nF...
MLCC X7R or X8R are a good choice due to the very small package sizes available.... and are probably best for caps next to device pins in SMD packages.
What are audio grade capacitors... I would choose a decent commercial brand personally.
MLCC X7R or X8R are a good choice due to the very small package sizes available.... and are probably best for caps next to device pins in SMD packages.
What are audio grade capacitors... I would choose a decent commercial brand personally.
Nichicon FG are the capacitors I'll be using, pretty well priced at €0.20 a piece for 100uF. I want to use them for the TDA1541 DAC. SAA7220 filter. 470uF is recommended by modders, the original ones are 33-47uF. Not sure how much the original ceramics are but its definitely lower than 100nF, modders use 100nF though. I'm still don't know what the sound difference would be with higher values than 100uF and 100nF.
I think I need to update myself on the latest thoughts on capacitor types and bypassing.
The general consensus (many many years ago) used to be in order of ascending quality:
Industrial grade electrolytics
Audio grade electrolytics ( Os-cons were considered the best apart from the rare slit foil DNM caps)
Ceramics
Metallised film
Polypropylene
polystyrene
silver Mica
Bypassing was generally in the ratio of 100:1 (e.g 1000uF audo grade electrolytic bypassed with 10uF polypropylene and .1uF polystyrene in ascending order of quality)
Brands of capacitors was a different discussion!
Is this still largely correct?
The general consensus (many many years ago) used to be in order of ascending quality:
Industrial grade electrolytics
Audio grade electrolytics ( Os-cons were considered the best apart from the rare slit foil DNM caps)
Ceramics
Metallised film
Polypropylene
polystyrene
silver Mica
Bypassing was generally in the ratio of 100:1 (e.g 1000uF audo grade electrolytic bypassed with 10uF polypropylene and .1uF polystyrene in ascending order of quality)
Brands of capacitors was a different discussion!
Is this still largely correct?
From what I've gathered from all the Marantz CD player mod threads, Oscons are no good for analogue rails postreg (low ESR should be avoided postreg) but ideal for digital rails. Same applies for X7R ceramic caps which should be used prereg only (PPS caps are more suitable postreg).
Also it seems that lytic cap choice is pretty important with Nichicon FG being the best for my personal taste (slight bump in lower mids). Nichicon KZ is supposedly very neutral. I'm not sure about Silmic II caps, for some it has a warm, soft sound but I've seen people complaining that they sound harsh. I do want to give the Silmics a shot for one of my amp recaps though.
Anjump, thanks for pointing out the ratio. 100nF might be a bit too low if I use 100uF caps? Most people use 470uF + 100nF in the Marantz players though.
Also it seems that lytic cap choice is pretty important with Nichicon FG being the best for my personal taste (slight bump in lower mids). Nichicon KZ is supposedly very neutral. I'm not sure about Silmic II caps, for some it has a warm, soft sound but I've seen people complaining that they sound harsh. I do want to give the Silmics a shot for one of my amp recaps though.
Anjump, thanks for pointing out the ratio. 100nF might be a bit too low if I use 100uF caps? Most people use 470uF + 100nF in the Marantz players though.
Last edited:
I'm pretty sure that the ratios are very variable. For instance a 1000uF cap bypassed with a 10uf wouldn't make much difference (??) whereas a 100uF or more may.
I'm sure there's way of mathematically complicating things but I'm not sure that the results will be better than suck it and see if you like the difference.
I, like you, read differing opinions on this brand or that brand and get confused
The Os-cons I liked were the ofc wired ones (SA or SG?) for analogue power supply rails, but as I said that was many years ago and I need to update my knowledge as to the most recent technologies.
I'm sure there's way of mathematically complicating things but I'm not sure that the results will be better than suck it and see if you like the difference.
I, like you, read differing opinions on this brand or that brand and get confused
The Os-cons I liked were the ofc wired ones (SA or SG?) for analogue power supply rails, but as I said that was many years ago and I need to update my knowledge as to the most recent technologies.
If you are decoupling digital or analogue circuits I would suggest you do some proper research, there is a wealth of information out there gleamed from years of research, solid engineering information regarding the hows and whys of decoupling... Proper decoupling is critical to circuit operation and noise reduction...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.