How much capacitance is proper?
I'm a rank amateur to electronics. I'm collecting all of the parts to assemble an SET amp with a separate PS. I'm using an LC filter after the rectifier tubes & from there providing separate LC's to each channel (a la Lundahl).
I've read everything I can find about sizing of the LC components & have wound up with two ranges of cap sizes. The RCA handbook has led me to use a 4uf cap after the first choke of 4.5H & 25uf caps after the two channel's separate 4.5H chokes. The power output is 350vdc & 250mA.
The other indicated sizing is for 110uf after the first choke & 230uf after the other two. The attitude with these higher caps is "all you can afford".
I'm sure that too much capacitance must have bad effects such as ringing.
I enjoy reading DIY Audio & learn a great deal from it. Mainly, that there is more than one way to skin a cat.
I'm a rank amateur to electronics. I'm collecting all of the parts to assemble an SET amp with a separate PS. I'm using an LC filter after the rectifier tubes & from there providing separate LC's to each channel (a la Lundahl).
I've read everything I can find about sizing of the LC components & have wound up with two ranges of cap sizes. The RCA handbook has led me to use a 4uf cap after the first choke of 4.5H & 25uf caps after the two channel's separate 4.5H chokes. The power output is 350vdc & 250mA.
The other indicated sizing is for 110uf after the first choke & 230uf after the other two. The attitude with these higher caps is "all you can afford".
I'm sure that too much capacitance must have bad effects such as ringing.
I enjoy reading DIY Audio & learn a great deal from it. Mainly, that there is more than one way to skin a cat.
If the first (reservoir) capacitor is too large then the charging pulses become large and narrow. The result can be buzz due to induction and ground currents, and excessive transformer heating. A valve rectifier may have a maximum cap specified in the data sheet.
After that the smoothing caps do not matter too much. One proviso: the caps can resonate with the smoothing choke to create subsonic ringing.
Note that old designs use small caps and big chokes, because that was the best fit with the technology available in the 1950/60s. Modern practice uses bigger caps and smaller chokes, as big caps are now available in reasonable sizes and prices.
After that the smoothing caps do not matter too much. One proviso: the caps can resonate with the smoothing choke to create subsonic ringing.
Note that old designs use small caps and big chokes, because that was the best fit with the technology available in the 1950/60s. Modern practice uses bigger caps and smaller chokes, as big caps are now available in reasonable sizes and prices.
Thanks for your response.
Yes, the info for the 5R4GYA rectifiers shows 4uf max for the first cap, but that is for a CLC filter. I think I used the RCA handbook's instructions & charts correctly for sizing the 4uf & 25uf caps after the chokes. Yes, that was the old RCA. Anything new online scoffs at LC filters except for Lundahl where I gather that "more is better". I already had the three 4.5H chokes (the one after the rectifier is huge) & that's what I wanted to include.
Yes, the info for the 5R4GYA rectifiers shows 4uf max for the first cap, but that is for a CLC filter. I think I used the RCA handbook's instructions & charts correctly for sizing the 4uf & 25uf caps after the chokes. Yes, that was the old RCA. Anything new online scoffs at LC filters except for Lundahl where I gather that "more is better". I already had the three 4.5H chokes (the one after the rectifier is huge) & that's what I wanted to include.
Last edited:
LC filters in linear power supplies are kinda rare these days. Any reason why you don't just use the CLC filter suggested by the SR4GYA tube data? Anyway, I believe the capacitance formula is C = 1/2pi1400fc where C is farads and fc is your cutoff freq (60 or 120?). The 1400 is your load resistance.
Nothing wrong with LC filters if the load is fairly constant as an SET amplifier would be. Industrial 3 phase power supplies mainly use LC or just L because they give good power factor and transformer utilisation as well as working well with phase controlled rectifiers. Consider a voltage regulator as power supply rejection is very poor with single ended designs
Everything I've read has convinced me to go with choke input. The advance of cheap capacitors has made the CLC filter the popular way to go. The expense of the larger chokes in LC filters make them unpopular.
Unless I can get verification that larger caps are beneficial, I'll have to stay with the modest sizing I worked out using the old RCA handbooks. Even the formulas found on the internet lean heavily towards the more popular CLC & ignore the LC types. Lundahl Transformers appears to have decided that the LCLC is superior.
Unless I can get verification that larger caps are beneficial, I'll have to stay with the modest sizing I worked out using the old RCA handbooks. Even the formulas found on the internet lean heavily towards the more popular CLC & ignore the LC types. Lundahl Transformers appears to have decided that the LCLC is superior.
Are you sure? L does not appear in it. The formula you gave was for a CR filter.sofaspud said:
With choke input the problems with peak charging current created by a big first cap no longer apply. Bigger caps are then better, from a ripple point of view. Subsonic resonance still needs thinking about.dobias said:
Well, no one has offered a correction, and I dug up a reference that agrees with me (Illustrated Encyclopedic Dictionary of Electronics by John Douglas-Young 1981 p434). Maybe I should have given it as C = 1/2pi x square root of L/C. I "cheated" and used figures given previously.
The does seem to be a lack info on this. I don't have a really good book on passive filters, but figured the speaker-builders would be all over it. I guess people use tables and software these days.
The does seem to be a lack info on this. I don't have a really good book on passive filters, but figured the speaker-builders would be all over it. I guess people use tables and software these days.
OK. sqrt(L/C) would be the characteristic impedance of an LC section, not load resistance, although I guess it could be equal to the load resistance which gives a particular damping characteristic. In that case f would have to be the corner frequency, not the line frequency. Quoting a formula with values already half filled-in can create confusion.
Yes, many people use tables or simulations these days. This can avoid the need to think, which in turn avoids the need to understand. Or you can use sims as a tool to aid understanding, in conjunction with formulae and where they come from.
Yes, many people use tables or simulations these days. This can avoid the need to think, which in turn avoids the need to understand. Or you can use sims as a tool to aid understanding, in conjunction with formulae and where they come from.
DF96,
"With choke input the problems with peak charging current created by a big first cap no longer apply. Bigger caps are then better, from a ripple point of view. Subsonic resonance still needs thinking about. "
That's exactly what I was trying to ask about.. how big (cap) is too big?
Now, how do I figure the max. cap size & still avoid subsonic resonance?
"With choke input the problems with peak charging current created by a big first cap no longer apply. Bigger caps are then better, from a ripple point of view. Subsonic resonance still needs thinking about. "
That's exactly what I was trying to ask about.. how big (cap) is too big?
Now, how do I figure the max. cap size & still avoid subsonic resonance?
There will be a subsonic resonance at f=1/(2pi sqrt(LC) ), where L is the choke and C is the cap. Typically f will be well below the audio range (otherwise you get no smoothing), so then the issue is how f relates to record warps (0.5-2Hz?), arm-cartridge resonance (5-15Hz?) and mains voltage variations (0.01-0.2Hz?). Obviously the first two are not an issue if you use CD.
The resonance will be damped by the load on the PSU. I can't remember the exact formula, but it is something like Q=R/sqrt(L/C), where R is the load resistance. You want Q to be lowish, but not too low. A expert on PSUs may be able to help.
The resonance will be damped by the load on the PSU. I can't remember the exact formula, but it is something like Q=R/sqrt(L/C), where R is the load resistance. You want Q to be lowish, but not too low. A expert on PSUs may be able to help.
If what is right is not known, it's difficult to say what is wrong. I can only offer what I found for the inverted-L low-pass LC filter:
Values of C and L are calculated from the following formulas:
C = 1/6.28fcR
L = R/6.28fc
where:
C = farads
L = henries
R = nominal terminating resistance [=sqrt of L/C]
fc = cutoff frequency
If a CLC pi filter is going to be used instead, this is all rather mute anyway.
Values of C and L are calculated from the following formulas:
C = 1/6.28fcR
L = R/6.28fc
where:
C = farads
L = henries
R = nominal terminating resistance [=sqrt of L/C]
fc = cutoff frequency
If a CLC pi filter is going to be used instead, this is all rather mute anyway.
I would use an oil filled 4uf then the choke, the output capacitor is often "voiced" to improve the low end sound.
When the reactance of the choke (= 2 x Pi x F x L) is equal to the reactance of the filter capacitor (= 1/( 2 x pi x F x C) that is where you have a resonance. F is in Farads and L is in Henries. If you set this frequency to where your output transformer is around 3 to 6 db down on the low end you may find you prefer the sound of the finished amplifier. However for the most neutral amplifier this power supply resonance frequency should be 1/10 of the frequency at which the output transformer is down by 3 db from midband.
The amount of ripple that is acceptable depends on your loudspeakers sensitivity and the noise floor in your listening space.
So if you begin to understand there are trade-offs between ripple and resonance that will provide some insight into one of the causes of different amplifier sound.
As you wish to use a specific choke you should calculate how loud the ripple voltage would be in your loudspeakers. 2.83V into 8 ohms gives the reference volume number. 20 log(ripple/2.83) gives you the amount you subtract from your sensitivity reference number. If drops below 20db you probably have just enough filtering.
ES
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.