Hi All,
I'm trying to understand filtering in an unregulated linear PSU. I understand the purpose of it but have become confused about the objective.
If we take a CLC filter for example used after a FWB rectifier, is the objective to reduce Vripple's amplitude, Vripple's frequency or both? It seems to me that you can approach pure DC by doing either.
I have attempted to calculate the values for a filter to reduce Vripple's frequency below 1Hz and end up with C=~50,000uf per pole and L=~1H. C doesn't seem out of place compared to other PSU designs I've seen, but L seems rather large.
When searching through this forum and the interweb the values of the chokes used are much lower, but with comparable C, 10kuf to 100kuf.
Any help is greatly appreciated.
Cheers!
I'm trying to understand filtering in an unregulated linear PSU. I understand the purpose of it but have become confused about the objective.
If we take a CLC filter for example used after a FWB rectifier, is the objective to reduce Vripple's amplitude, Vripple's frequency or both? It seems to me that you can approach pure DC by doing either.
I have attempted to calculate the values for a filter to reduce Vripple's frequency below 1Hz and end up with C=~50,000uf per pole and L=~1H. C doesn't seem out of place compared to other PSU designs I've seen, but L seems rather large.
When searching through this forum and the interweb the values of the chokes used are much lower, but with comparable C, 10kuf to 100kuf.
Any help is greatly appreciated.
Cheers!
The goal is to reduce the ripple's amplitude. The ripple frequency stays constant, regardless.
There's a charging current pulse per half cycle that increases with larger capacitance, and
this can only be taken so far. Also, larger value capacitors are more expensive.
For a given voltage and case size, the capacitor's uF value is limited, since the case volume
is related to the product of uF and voltage. So, a second stage of filtering is often used, even
in higher current circuits, with the exception of transistor output stages where feedback is
relied on instead to reduce the ripple at the output, because of the high load currents needed.
Last edited:
As Rayma says, ripple is always 100/120Hz.
We sometimes work the math as the high-cut frequency of the filter.
Yes, 1H and 50,000uFd is 0.7Hz. 140 times lower than 100Hz ripple. Since it is 2-pole, we expect 140*140 or 20,000:1 ripple reduction.
We rarely need this much reduction. Certainly not for "big" power. There are better (more economic) ways to clean up for small power (preamps).
We never get this much reduction in a single filter because of parasitics. 20,000:1 requires very low cap ESR and choke low core-loss.
> confused about the objective. ... you can approach pure DC
Yes, if you use infinite choke and cap the output is dead-steady DC forever. But infinite parts are not in stock and probably cost infinite bucks. The objective is to make "good enough" DC, for your purposes, with available parts/budget.
We sometimes work the math as the high-cut frequency of the filter.
Yes, 1H and 50,000uFd is 0.7Hz. 140 times lower than 100Hz ripple. Since it is 2-pole, we expect 140*140 or 20,000:1 ripple reduction.
We rarely need this much reduction. Certainly not for "big" power. There are better (more economic) ways to clean up for small power (preamps).
We never get this much reduction in a single filter because of parasitics. 20,000:1 requires very low cap ESR and choke low core-loss.
> confused about the objective. ... you can approach pure DC
Yes, if you use infinite choke and cap the output is dead-steady DC forever. But infinite parts are not in stock and probably cost infinite bucks. The objective is to make "good enough" DC, for your purposes, with available parts/budget.
So if I understand this correctly, filtering is not being used in a power supply in the same fashion as say a speaker Xover? Generally, we're not trying for a low pass filter? Instead, we are trying to reduce the amplitude of Vripple as a primary function. Any frequency rejection is secondary, perhaps affecting any HF noise on the line.
If my understanding is correct, how is the value of L and C selected in a CLC filter for this purpose? The formulas I've managed to dig up on this topic are aimed at high or low pass frequency filters, not amplitude.
Thanks.
If my understanding is correct, how is the value of L and C selected in a CLC filter for this purpose? The formulas I've managed to dig up on this topic are aimed at high or low pass frequency filters, not amplitude.
Thanks.
The lopass behavior is very much wanted, because our ears are more sensitive to noise in the mid than bass range. But a first-order filter is enuff in order to look for that sensitivity curve. In a certain power amplifier I use a CRC filter with corner frequency around 250Hz, reducing noise and inrush currents.
Component values are also selected for their power behavior such as voltage drop over a capacitor, when it is becoming discharged. So you have to look at the supply load, which is the amplifier(s).
Component values are also selected for their power behavior such as voltage drop over a capacitor, when it is becoming discharged. So you have to look at the supply load, which is the amplifier(s).
Use PSUD2 to analyse your PSU.
It is fairly easy to use and shows pics of the ripple in the filter and after the filter.
Start with a simple C only PSU.
This is really an rC filter where the r is quite small due to the resistances in the cables and components before the capacitor.
Then you can add on an extra RC to create an rCRC PSU.
You will see the reduction in ripple gained with the extra stage.
Change the R to an L+R to see the effect of the L @ 100/120Hz. The effect is quite small for all small inductors.
It's the effect at High Frequencies (interference) that is massive. But PSUD2 does not show the HF effect. A scope shows it by comparing the spiky waveform of an rC with the rounded ripple from an rC(L+R)C
The rounding of the ripple tells you that the proportion of the HF component is much reduced.
It is fairly easy to use and shows pics of the ripple in the filter and after the filter.
Start with a simple C only PSU.
This is really an rC filter where the r is quite small due to the resistances in the cables and components before the capacitor.
Then you can add on an extra RC to create an rCRC PSU.
You will see the reduction in ripple gained with the extra stage.
Change the R to an L+R to see the effect of the L @ 100/120Hz. The effect is quite small for all small inductors.
It's the effect at High Frequencies (interference) that is massive. But PSUD2 does not show the HF effect. A scope shows it by comparing the spiky waveform of an rC with the rounded ripple from an rC(L+R)C
The rounding of the ripple tells you that the proportion of the HF component is much reduced.
No, the very opposite. PSU filtering is being used as a low pass filter because that is what reduces ripple, by rejecting higher frequencies (higher than the filter corner frequency). This is exactly the same as a speaker crossover.ChesterBelle said:
Amplitude (of frequency components) is exactly what filters are intended to modify. They also modify phase, but that is of less interest in a PSU.
You need to do some more reading on signals and filters. You cannot design a filter until you have begun to understand what it does and how it does it.
Dont forget that a choke to be effective must be magetised ... so it will have an effec only if there is a load at the ouput of power supply ... in the old day there was often a bleeder resistor to have a minimum load ... with a choke the more current you have the best filtering you have..
With a capacitor the less current you have the less ripple you will get...
With a capacitor the less current you have the less ripple you will get...
> a simple C only PSU. This is really an rC filter where the r is quite small
Actually, there are two "R": the charging path (low R) and the discharge path (high R). There is no analogous situation in loudspeaker crossovers, because it distorts grossly.
Taking cheap values: 1,000uFd loaded in a 50 Ohm amp gives discharge "frequency corner" of 3.3Hz. Well below 100Hz ripple frequency. I remember many amps playing fine with about this much filtering. At high power there is considerable ripple, so the amplifier must have power supply rejection (emitter followers do OK). In clipping the 100/120Hz cross-modulates the audio peaks for a harsh sound, but such low-price amps got harsh even before clipping. With today's low-low price for soup-can caps, nobody uses this little uFd.
But another lesson from crossovers. A 6dB/Oct filter at 120hz may pass a relative bunch of ~~1KHz into a woofer and that may be audible despite squawker output. An 18dB/Oct crossover gives a LOT more midrange reduction in the high-cut. Once you can prove that a single C stage is not enough (yet it often is!), you get the most from your cap-dollar with multi-stage filtering. In pure R-C filters, 3 to 5 stages is economic. R-C filters suck (or sag) on power amps, and I have not thought-through L-C filter economics. (Perhaps because I think a skilled designer cna work with one hand tied behind his back, and at low frequencies chokes give the least bang for the buck of any modern component, so that's the hand I tend to tie.)
Actually, there are two "R": the charging path (low R) and the discharge path (high R). There is no analogous situation in loudspeaker crossovers, because it distorts grossly.
Taking cheap values: 1,000uFd loaded in a 50 Ohm amp gives discharge "frequency corner" of 3.3Hz. Well below 100Hz ripple frequency. I remember many amps playing fine with about this much filtering. At high power there is considerable ripple, so the amplifier must have power supply rejection (emitter followers do OK). In clipping the 100/120Hz cross-modulates the audio peaks for a harsh sound, but such low-price amps got harsh even before clipping. With today's low-low price for soup-can caps, nobody uses this little uFd.
But another lesson from crossovers. A 6dB/Oct filter at 120hz may pass a relative bunch of ~~1KHz into a woofer and that may be audible despite squawker output. An 18dB/Oct crossover gives a LOT more midrange reduction in the high-cut. Once you can prove that a single C stage is not enough (yet it often is!), you get the most from your cap-dollar with multi-stage filtering. In pure R-C filters, 3 to 5 stages is economic. R-C filters suck (or sag) on power amps, and I have not thought-through L-C filter economics. (Perhaps because I think a skilled designer cna work with one hand tied behind his back, and at low frequencies chokes give the least bang for the buck of any modern component, so that's the hand I tend to tie.)
> a choke to be effective must be magetised
No. Chokes are used without DC in many places in audio. Notably the loudspeaker crossovers the OP has mentioned. Also old tone controls.
> in the old day there was often a bleeder resistor to have a minimum load
That is a different thing. Aside from technician safety, a Choke INPUT DC power supply outputs 0.9*VAC at high current but 1.414*VAC at low- or low-load. There is a minimum size choke for current. Zero current needs infinite inductance. The 1.57X rise of voltage at turn-on (before tubes heat and suck; or if a tube fails) can do damage. The bleeder can be sized to present a minimum load for a reasonable size choke.
> with a choke the more current you have the best filtering you have..
Not generally true. For an ideal choke, heavy load means less filtering. Real choke inductance tends to sag with load, so even less filtering. (A "swinging choke" varies inductance very widely, allowing the high I needed for choke-input use with low bleeder current then dropping to a lower I at high current so it does not cost too much.)
No. Chokes are used without DC in many places in audio. Notably the loudspeaker crossovers the OP has mentioned. Also old tone controls.
> in the old day there was often a bleeder resistor to have a minimum load
That is a different thing. Aside from technician safety, a Choke INPUT DC power supply outputs 0.9*VAC at high current but 1.414*VAC at low- or low-load. There is a minimum size choke for current. Zero current needs infinite inductance. The 1.57X rise of voltage at turn-on (before tubes heat and suck; or if a tube fails) can do damage. The bleeder can be sized to present a minimum load for a reasonable size choke.
> with a choke the more current you have the best filtering you have..
Not generally true. For an ideal choke, heavy load means less filtering. Real choke inductance tends to sag with load, so even less filtering. (A "swinging choke" varies inductance very widely, allowing the high I needed for choke-input use with low bleeder current then dropping to a lower I at high current so it does not cost too much.)
Ok, I think I have now got a handle on this. Correct me if I’m wrong.
Generally speaking, in a speaker xover we use a filter to attenuate a signal so that it is no longer in the audible range from the speaker being crossed over, i.e. we don't want to use that signal, so we make it move closer to the noise floor. In a PSU we still want to use the signal, but we want to smooth it out, i.e. we want to get closer to the ideal, being DC.
The mains supplies a frequency (in my case 50Hz, ergo 100Hz for a FWB) and that frequency cannot be changed, as pointed out to me. A filter is simply attenuating a signal so that at the cut off frequency the signal will be -3db. Referencing the -3db point may be useful in speaker xover’s, but is not necessarily useful in PSU design. So just looking at the mechanics of how this process works: if, for example, -12db, -18db or -24db was the level we wanted our 100Hz ripple to drop by, we can design a filter with a suitable Fcut that by the time the roll off gets to 100Hz, the amplitude of Vripple has dropped by the desired amount.The desired amount is determined by the requirements of the load.
Am I on the right track now?
Generally speaking, in a speaker xover we use a filter to attenuate a signal so that it is no longer in the audible range from the speaker being crossed over, i.e. we don't want to use that signal, so we make it move closer to the noise floor. In a PSU we still want to use the signal, but we want to smooth it out, i.e. we want to get closer to the ideal, being DC.
The mains supplies a frequency (in my case 50Hz, ergo 100Hz for a FWB) and that frequency cannot be changed, as pointed out to me. A filter is simply attenuating a signal so that at the cut off frequency the signal will be -3db. Referencing the -3db point may be useful in speaker xover’s, but is not necessarily useful in PSU design. So just looking at the mechanics of how this process works: if, for example, -12db, -18db or -24db was the level we wanted our 100Hz ripple to drop by, we can design a filter with a suitable Fcut that by the time the roll off gets to 100Hz, the amplitude of Vripple has dropped by the desired amount.The desired amount is determined by the requirements of the load.
Am I on the right track now?
Think in terms of voltage, current and load when designing a PSU.
DC Power Supply design
http://www.skillbank.co.uk/psu/smoothing.htm
DC Power Supply design
http://www.skillbank.co.uk/psu/smoothing.htm
Last edited:
Chester, the ripple is not a pure 100 Hz tone but contains harmonics. Amplifiers tend to reject power supply ripple the most completely at lo frequencies such as 100 Hz and belo but less completely above. Also due to ear sensitivity curve discovered by persons Fletcher and Munson, a lopass with corner f=250 Hz is usually already all, which is needed for noise reduction improvement of the most basic supply. At first I calculate the capacitance for a ripple of one tenth of supply voltage. Then I split it in two and connect both halves with a resistor, which gives corner f=250 Hz. Sure the resistor must be able to produce some heat, calculate its power for maximum load. Exchanging the resistor for an inductor may improve behaviour but complicates design.
Hi Grasso789,
I’m confused about the filtering at 250Hz. As I understand it, your ripple voltage would remain unchanged and the only effect would be to attenuate harmonics 250Hz and higher. All signals below 250Hz would remain non-attenuated. For example, if we have a 2Vpk ripple at 100Hz, a passive component filter with an Fc of 250Hz will do nothing to attenuate that 2Vpk ripple; it will only take a 250Hz signal and attenuate it by -3dB, the 100Hz signal would remain at 2Vpk. If the goal is to reduce ripple then a passive component filter needs to have an Fc that will attenuate 100Hz by the desired amount, i.e. an Fc somewhere around 100Hz or less.
Am I missing something here?
I appreciate that I may be approaching this from a non-standard way of thinking, but electronics is a new discipline to me and so I have a limited electronics ‘vocabulary’ from which to draw from to understand the new concepts. I feel that it is important to understand the science and engineering principals (at least at a basic level) that underpin these topics so that when I use programs like PSUD or LTSPICE I have an idea of what is happening and why.
Thanks for your help.
I’m confused about the filtering at 250Hz. As I understand it, your ripple voltage would remain unchanged and the only effect would be to attenuate harmonics 250Hz and higher. All signals below 250Hz would remain non-attenuated. For example, if we have a 2Vpk ripple at 100Hz, a passive component filter with an Fc of 250Hz will do nothing to attenuate that 2Vpk ripple; it will only take a 250Hz signal and attenuate it by -3dB, the 100Hz signal would remain at 2Vpk. If the goal is to reduce ripple then a passive component filter needs to have an Fc that will attenuate 100Hz by the desired amount, i.e. an Fc somewhere around 100Hz or less.
Am I missing something here?
I appreciate that I may be approaching this from a non-standard way of thinking, but electronics is a new discipline to me and so I have a limited electronics ‘vocabulary’ from which to draw from to understand the new concepts. I feel that it is important to understand the science and engineering principals (at least at a basic level) that underpin these topics so that when I use programs like PSUD or LTSPICE I have an idea of what is happening and why.
Thanks for your help.
Rite, with most transistor power amplifiers, the supply does not need to be pure DC, some ripple is allowed. Tho our ears are sensitive, and any nite noise can drive us mad, especially if we are unable to find the source, we should be glad, that they are not as sensitive at 100Hz as they are at 3KHz. Just when you go exotic, with feedback-less class-A or electron valves, you have to take measures.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.