CRCRC linear power supply - selecting the right size caps.

[mikesnowdon]

Hi.

If replacing a single cap PSU with a 3 cap CRCRC how do we determine the correct value of caps to use in the new supply?

I have a standard linear PSU with a single 4700uF smoothing cap. If I replace that with a CRCRC do I just use 3 x 4700uF caps or do I calculate the total added value of the 3 caps to aproximate the single 4700uF?

Is it the case that in a CRCRC supply the powered circuit effectively only 'sees' the last cap, meaning my selected value for all caps should be the same as the original I'm replacing - hence, 4700uF?

Also, does anyone have any info on how to choose the value of resistors to use?

[DougL]

PSUD2 will answer most of your questions.
For output Z of the supply, the last cap is the one that matters. Make it a low ESR type. You could reasonably use a 4700uF as a last cap. The resistor values and other 2 caps will determine the output voltage and Q of the supply.

[Nazar_lv]

If replacing a single cap PSU with a 3 cap CRCRC how do we determine the correct value of caps to use in the new supply?

Much better is to use filter that begins with R, such as RC or RCRC
after filter you use regulator?

[mikesnowdon]

Originally Posted by DougL

PSUD2 will answer most of your questions.
For output Z of the supply, the last cap is the one that matters. Make it a low ESR type. You could reasonably use a 4700uF as a last cap. The resistor values and other 2 caps will determine the output voltage and Q of the supply.

Cool. Where can I get this?

I was thinking as the circuit only sees the last cap then it should match the value of the sigle cap its replacing, 4700uF. So what about the other 2 caps then. Is it conventional wisdom to use 3 caps of the same value? I ask because this is what we normally see. 3 different size caps is uncommon I think.

[mikesnowdon]

I downloaded PSUD2 and entered a CRCRC based on 4700uF BHC caps and 0.47r resistors. Trouble is I'm not sure how to interpret the data. can anyone help?

[AndrewT]

BTW,
that reference to using a first R in the RCRC configuration is always met.
The transformer secondary resistance and the resistance of the wiring to the rectifier and to the smoothing caps gives the necessary r or R depending on transformer and wire gauge and wiring length.

the ripple current on the first cap of a cascaded filter is always high. Choose a very high ripple current capacitor for this duty or parallel a few lower ripple capacity caps to allow them to run cooler. This can be especially bad in ClassA amplifiers or other load that has a high continuous current demand.

Zoom in on the ripple. Measure it.
Compare after changing ONE component value.

Cascaded filters are very good at attenuating ripple. But they are very bad at achieving low output impedance. That's why the last C becomes critical for sound quality.

[mikesnowdon]

Hi Andrew.

I was hoping you would comment

Thanks.

[mikesnowdon]

I think Ive got to grips with this software now.

Interestingly removing the first R shows reduced ripple. Matching the value of the R after the first C, and again after the last C shows further improvement. This is based on the software's approximated transformer impedance of 1.7ohm. I wonder if this could be a general rule - that the resistors value should match the transformer impedance? It simulates beautifully.

So far I have compared BHC 4700uF/40v caps with Panasonic FC4700/25v caps. The BHC are marginally smoother.

Now I want to do this more accurately so I need help with regard to measuring the necessary transformer specs. I'd also like to add the correct diodes which are Schottky DQ10. Also, how do I determine the load on the power supply? (to add further complication the load on the negative supply will be less than the positive rail. So should I tailor each bank of caps accordingly?)

Can anyone help with this please?

Some screenshots...

Panny FC looks like this:



BHC looks like this:



Marginal isnt it

[AndrewT]

Hi,
looking at your screenshots.
you have plotted for 1000ms at start of power up.
you have plotted the current through C3
You have a load of 5k at a supply voltage ~25Vdc.
you have a load current of ~5mA.

Change the reporting delay to 1s or 2s
plot V(R3) and V(C3). They should be the same, exactly.
Change the load from 5k to 1k to 100r and see the effect on output voltage and on voltage ripple.
Worth looking at voltage ripple on C1 & C2.
Definitely worth looking at current ripple through C1.

The reason for one cap looking smoother is the high ESR (10times higher) This reduces the peak currents in each of the caps and the cascade attenuates the V/us even more.

[mikesnowdon]

Thanks Andrew