Is it recommended to use 1000uf/35v silmic2 x 10 pieces in parallel rather than one 10000uf Mundorf AG Mlytic? The reasons are paralleling the caps will reduce the esr by 10 times and the ripple current of each silmic 2 caps is 1.6 Amps and x 10 will be 16 amperes. If I take just Mundorf it gives only 4.8 amps. But i dont have the esr for the silmic2 i believe using this will reduce the esr substantially resulting strong psu.
Does anybody knows the esr of the silmic 2 I couldnt find it in the datasheet.
I found this kind of psu system in Aussieamplifiers ..
http://www.aussieamplifiers.com/nxv101_2_1.jpg
whats your opinion. Which is better in this case?
price wise both come to close but ripple current will be pretty high and esr will be pretty low. Im only concerned with the signature of these two caps.
If the signature of elna silmic 2 is better then its better to use these as psu caps. The ripple current using in 10 parallel caps is higher than the Mundorf Mlytic HC+ which is phenomenal... but if anybody says to use Silmic 2 as psu it would be great.. On bulk purchase we can get better pricing on silmic2 caps..
Does anybody knows the esr of the silmic 2 I couldnt find it in the datasheet.
I found this kind of psu system in Aussieamplifiers ..
http://www.aussieamplifiers.com/nxv101_2_1.jpg
whats your opinion. Which is better in this case?
price wise both come to close but ripple current will be pretty high and esr will be pretty low. Im only concerned with the signature of these two caps.
If the signature of elna silmic 2 is better then its better to use these as psu caps. The ripple current using in 10 parallel caps is higher than the Mundorf Mlytic HC+ which is phenomenal... but if anybody says to use Silmic 2 as psu it would be great.. On bulk purchase we can get better pricing on silmic2 caps..
Last edited:
You are worrying needlessly. Its a power supply!
Just use good quality caps and stop reading the rubbish published about power supply capacitors.
Frank.
These are my thoughts developed over a few years:
The local decoupling supply the fast changing current demands of the load.
The smoothing capacitors see the mains voltage variations and the resulting current variations, but have to supply the slower changing load current demands.
A PSU that is working in conjunction with an adequate local decoupling implementation does not have much impact on the medium frequency and no impact on the high frequency performance of an amplifier.
The smoothing capacitors that meet ordinary commercial and industrial standards can adequately meet all the demands of a current varying load when the local decoupling is well implemented.
However, a badly implemented local decoupling scheme, or a load that has no local decoupling will put quite different demands on the PSU. Do it wrong and the performance will be adversely affected.
The local decoupling supply the fast changing current demands of the load.
The smoothing capacitors see the mains voltage variations and the resulting current variations, but have to supply the slower changing load current demands.
A PSU that is working in conjunction with an adequate local decoupling implementation does not have much impact on the medium frequency and no impact on the high frequency performance of an amplifier.
The smoothing capacitors that meet ordinary commercial and industrial standards can adequately meet all the demands of a current varying load when the local decoupling is well implemented.
However, a badly implemented local decoupling scheme, or a load that has no local decoupling will put quite different demands on the PSU. Do it wrong and the performance will be adversely affected.
Agreed andrew I was in the similar thought I tried silmic2 i didnt understand why people are behind it. I tried using it everywhere like decoupling near the pins of both the transistor pins VAS decoupling, at feedback cap, tried even directly at psu. Overall it doesnt have transparent signature and it sounds as if there is some phase issue on the entire spectrum. I have tried all together even as input coupling caps. Finally a big No to silmics as it was smearing all the details and layers in the music. Instead I would like to get something like smoother by increasing the details but this does opposite.. but some like that signature of bland sound...
That is my plan with the F4 and F5. I will use 69.000uf 80V. Four of them. One of them will be 68 000 uf because I can't find a similar one.
I wonder how 68-69 000uf affects the choice of other components. I will still use the Universal Power Supply from the DIYaudio store. Is it just to replace the 10 000 uf caps? I doubt.
Last edited:
Decoupling capacitance?
Hi Andrew
Do you know how the value of the decoupling capacitance is worked out - is say, a 10:1 ratio between smoothing and decoupling capcitance a good starting guideline?
Are there any benefits to maximising the decoupling capacitance providing the quality of the cap (ESR, ESL) between different valued caps is the same?
Hi Andrew
Do you know how the value of the decoupling capacitance is worked out - is say, a 10:1 ratio between smoothing and decoupling capcitance a good starting guideline?
Are there any benefits to maximising the decoupling capacitance providing the quality of the cap (ESR, ESL) between different valued caps is the same?
Last edited:
http://www.diyaudio.com/forums/chip...g-revisited-applies-other-chip-amps-well.html
Marcus, start there. Should give a good primer.
Marcus, start there. Should give a good primer.
Thanks. Bit too technical for me
but Tom did state that he thought there was little benefit going from 470uf to 1000uf:
"I saw very little benefit going from 470 uF to 1000 uF. I wouldn't bother with more than that on the PCB. It depends on the inductance and resistance of the wiring from the main supply cap to the board, however.
The main supply reservoir caps need to be sized for the biggest acceptable ripple voltage under worst-case load conditions"
but Tom did state that he thought there was little benefit going from 470uf to 1000uf:"I saw very little benefit going from 470 uF to 1000 uF. I wouldn't bother with more than that on the PCB. It depends on the inductance and resistance of the wiring from the main supply cap to the board, however.
The main supply reservoir caps need to be sized for the biggest acceptable ripple voltage under worst-case load conditions"
Last edited:
Ah, okay. Let me see if I can reach you!
A PSU will do well to effectively supply energy (have low impedance) over a very wide frequency range. At high frequencies, the long wires from the main power supply capacitors to the amp (chip in this case) are going to have pretty high impedance. That's bad, so we need to get some amount of energy storage as humanly close to the chip as we can.
Different sized capacitors of different technologies will have different impedance-vs-frequency plots. Our goal is to have enough capacitors in parallel that their low impedance regions overlap to give the chips a very broadband power supply source.
Biggest moral of the story you should take away is to get a few capacitors up and very close between V+ and Gnd and V- and Gnd so as to minimize the amount of wire/lead inductance we can (which rolls off the highs). The highest frequency-tuned capacitor closest (smallest), working down the range.
Those capacitors should be staggered in size/material/technology so as provide bypass across a wide frequency spectrum. Look at the impedance vs frequency plots in the first post of that link. We want that as low and as flat as sensible.
E.g. the 4.7 uF X7R is there to effectively bypass high frequencies (100 kHz-1 MHz), but isn't big enough to have low impedance where the 20 uF OSCON is most effective(10k-100k), and the OSCON is nowhere big enough to help with the low frequency stuff (where the 470 uF vs 1000 uF is). All 3 of these caps (6 per chip, to go to both rails) are local to the chip and are functionally the power supply the chip "sees".
The main capacitor bank is charged with the task of smoothing out the ripple voltage from rectifying mains (or switching/whatever). Tom leaves up to the reader to determine (use a PSU calculator here, it'll be good enough) how much capacitance you need for your system at the highest current draw you expect. You need to account for the psrr of the chip to boot.
Or just go nuts here.
A PSU will do well to effectively supply energy (have low impedance) over a very wide frequency range. At high frequencies, the long wires from the main power supply capacitors to the amp (chip in this case) are going to have pretty high impedance. That's bad, so we need to get some amount of energy storage as humanly close to the chip as we can.
Different sized capacitors of different technologies will have different impedance-vs-frequency plots. Our goal is to have enough capacitors in parallel that their low impedance regions overlap to give the chips a very broadband power supply source.
Biggest moral of the story you should take away is to get a few capacitors up and very close between V+ and Gnd and V- and Gnd so as to minimize the amount of wire/lead inductance we can (which rolls off the highs). The highest frequency-tuned capacitor closest (smallest), working down the range.
Those capacitors should be staggered in size/material/technology so as provide bypass across a wide frequency spectrum. Look at the impedance vs frequency plots in the first post of that link. We want that as low and as flat as sensible.
E.g. the 4.7 uF X7R is there to effectively bypass high frequencies (100 kHz-1 MHz), but isn't big enough to have low impedance where the 20 uF OSCON is most effective(10k-100k), and the OSCON is nowhere big enough to help with the low frequency stuff (where the 470 uF vs 1000 uF is). All 3 of these caps (6 per chip, to go to both rails) are local to the chip and are functionally the power supply the chip "sees".
The main capacitor bank is charged with the task of smoothing out the ripple voltage from rectifying mains (or switching/whatever). Tom leaves up to the reader to determine (use a PSU calculator here, it'll be good enough) how much capacitance you need for your system at the highest current draw you expect. You need to account for the psrr of the chip to boot.
Or just go nuts here.
Stereo and a linear supply on a 50/60 hz line? I can't imagine needing more than 20,000 uF per rail for a +/- 28V supply with an absolute max steady state power around 70 W into a 4 ohm load (and an optimistic 85 dB PSRR), so we'll call that my cut-off for nuts vs not nuts. Whatever middle-of-the road 35V electrolytics that you can get from your supplier for cheap are all that are needed here.
Then again I haven't personally done the math, as that's a project much further down the line (it's on the to-do list) and I'm inclined to go SMPS with 10,000 uF per rail. The higher frequency of the SMPS greatly reduces the capacitance needed (higher frequency ripple).
Best to grab a calculator, spec your desired xformer, and see how the output changes with load.
Then again I haven't personally done the math, as that's a project much further down the line (it's on the to-do list) and I'm inclined to go SMPS with 10,000 uF per rail. The higher frequency of the SMPS greatly reduces the capacitance needed (higher frequency ripple).
Best to grab a calculator, spec your desired xformer, and see how the output changes with load.
use the formula linking Farads of capacitance, Charging/discharging rate of current flow, Voltage change and a chosen time period: 1Farad (F) changes by 1Volt (V) when 1Ampere (A) flows in for 1second (s).
i.e. V = A * s / F
A 100nF decoupling capacitor can supply 0.5A continuously for 0.1us and the voltage on the capacitor will change by 0.5V
or can be stated as .5V in 0.1us = 5V/us of voltage change for 500mA into or out of a 100nF capacitor.
Use this to determine the current and timing for a current pulse into the circuit.
Do the same for the medium frequency pulses and the recharging of the HF decoupling.
Nothing to do with ratios of sizes of capacitors.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.