# My power supply project-filters questions.

#### prashanthb

Joined May 17, 2015
34
Tx >> Bridge rectifier >> capacitor bank >> Voltage reg >> current limiter >> output.
This is my idea, yet to be implemented.

My desire is to get DC as straight line as possible(and devoid of noise). I am just a hobbyist and it is out of my capability to design circuits to achieve it. But still the desire remains! I want to check if the following idea is workable or a completely/fundamentally wrong idea. Please do tell me.

I have come across this filter simulation page here:
http://sim.okawa-denshi.jp/en/RLCtool.php
I plugged in some values to play around till I came up with this RESULT :
RLC Filter: R=1 ohm,L=0.005H,C=0.1F

Cut-off frequency, fc = 7.11762543417[Hz]
Quality factor, Q = 0.22360679775
Damping ratio, ζ = 2.2360679775
The system does not oscillate.

I do not understand what exactly is Damping ratio, I guess it is synonymous with the mechanical damping definition. I guess the higher the value the greater the high frequencies will be suppressed. Hence I think I will be able to attenuate noise/fluctuations in my DC. I may be wrong in my understanding here. Please do tell me.
I dont know what quality factor is.

Please see attached plot images, and tell me if it is wrong and to drop this idea till I learn more or go ahead and implement it.

#### Attachments

• 1.8 KB Views: 9
• 1.2 KB Views: 6
• 1.1 KB Views: 5

#### #12

Joined Nov 30, 2010
18,217
You have gotten involved in minutia but failed to say anything about power. Do you have a schematic? How much voltage? How much current? How many outputs? How, "perfect" do you need the DC?

Customary methods may prove far more cost effective.

#### prashanthb

Joined May 17, 2015
34
You have gotten involved in minutia but failed to say anything about power. Do you have a schematic? How much voltage? How much current? How many outputs? How, "perfect" do you need the DC?

Customary methods may prove far more cost effective.
30~35VDC, @ 3A final usable output, single +& - output.
I can spend some 50 to 60 hours working on the filter part of the project. Within that whatever perfection is achievable, I will be satisfied with it.

#### #12

Joined Nov 30, 2010
18,217
I can spend some 5 or 6 minutes working on the filter and achieve .003% RMS output noise with 60 db power supply rejection ratio. How much better do you need it?

#### Attachments

• 279.6 KB Views: 24
Last edited:

#### prashanthb

Joined May 17, 2015
34
Thats excellent for most applications. I am planning on making LED lights next. But I do also have interests in audio.
But for audio circuits I have heard/read that the better quality DC, the better is the sound. People say this but add no scientific evidence. Even if they provide evidence I wouldnt understand, but thats another matter. Based on their experiences I want to add the filter.

EDIT: That noise reduction is for frequencies <10KHz. I want attenuation for a larger frequency spectrum.

#### #12

Joined Nov 30, 2010
18,217
The true test of DC quality is signal to noise ratio. You can add all the filters you want, but there is a point where enough is enough. The way you talk, you want to reduce hum and noise by infinity...millionths of a billionth of a volt of noise. There is a practical limit.

Why do you think there are frequencies over 10KHz in the power line to your home?
If they aren't there, why filter them?
Because you want to?
Why do you think the first filter capacitor in the power supply will work at 50Hz or 60Hz, but not at 10KHz and above?
Why do you think the filter capacitors after the voltage regulator will not work at frequencies above 10KHz?
The usual way to do this filters frequencies up to several MHz.

Last edited:

#### prashanthb

Joined May 17, 2015
34

#### R!f@@

Joined Apr 2, 2009
9,747
You should be more worried about the Amp generating the noise when it is idling

#12

#### #12

Joined Nov 30, 2010
18,217
http://www.tnt-audio.com/clinica/regulators_noise2_e.html

Kindly go thru the above link.
I dont have an oscilloscope or the skills to do such analysis. But I can assemble a filter to attach at the output of my power supply and hope that the noise is cleansed.
Thank you for that link. It demonstrates that the usual capacitors associated with a 3 terminal regulator reduce noise above 10KHz by 84db. You are still allowed to design more filtering if you want to. Some of the educated people on this site can even do the math for undetectably low noise levels.

#### R!f@@

Joined Apr 2, 2009
9,747
Still I believe the Amp will produce more noise than the PSU.
OP do not have a scope to see the noise, so how can he be sure of something he cannot hear ?

#### #12

Joined Nov 30, 2010
18,217
Still I believe the Amp will produce more noise than the PSU.
If the first stage of amplification is an operational amplifier, that sort of chip typically has a power supply rejection ratio of about 60 db. That would place power supply noise down by 144db. Here's a nice op-amp rated at 1.1 nv/root Hz.

#### Attachments

• 1.2 MB Views: 2

#### ronv

Joined Nov 12, 2008
3,770
Thats excellent for most applications. I am planning on making LED lights next. But I do also have interests in audio.
But for audio circuits I have heard/read that the better quality DC, the better is the sound. People say this but add no scientific evidence. Even if they provide evidence I wouldnt understand, but thats another matter. Based on their experiences I want to add the filter.

EDIT: That noise reduction is for frequencies <10KHz. I want attenuation for a larger frequency spectrum.
Your filter doesn't help the noise at low frequencies. (low pass filter) Because of there size the inductor and capacitor will be pretty expensive. $50 or so. If you really think you need better supplies (I seriously doubt you could hear the difference if the amplifier design is good.) I would make the regulator better. Like this one at the end: http://tangentsoft.net/elec/opamp-linreg.html #### ian field Joined Oct 27, 2012 6,539 Tx >> Bridge rectifier >> capacitor bank >> Voltage reg >> current limiter >> output. This is my idea, yet to be implemented. My desire is to get DC as straight line as possible(and devoid of noise). I am just a hobbyist and it is out of my capability to design circuits to achieve it. But still the desire remains! I want to check if the following idea is workable or a completely/fundamentally wrong idea. Please do tell me. I have come across this filter simulation page here: http://sim.okawa-denshi.jp/en/RLCtool.php I plugged in some values to play around till I came up with this RESULT : RLC Filter: R=1 ohm,L=0.005H,C=0.1F Cut-off frequency, fc = 7.11762543417[Hz] Quality factor, Q = 0.22360679775 Damping ratio, ζ = 2.2360679775 The system does not oscillate. I do not understand what exactly is Damping ratio, I guess it is synonymous with the mechanical damping definition. I guess the higher the value the greater the high frequencies will be suppressed. Hence I think I will be able to attenuate noise/fluctuations in my DC. I may be wrong in my understanding here. Please do tell me. I dont know what quality factor is. Please see attached plot images, and tell me if it is wrong and to drop this idea till I learn more or go ahead and implement it. One option is the so called "capacitor multiplier" - basically just an emitter follower that doesn't actually regulate voltage, it has a bias resistor from base to collector to make it pass the required current, but instead of a voltage stabilising zener - it has a ripple damping capacitor. In most cases, a few hundred uF is a reasonable starting point. Thread Starter #### prashanthb Joined May 17, 2015 34 One information I would like to add : I will wind my own coils. That will bring down the costs down to quite an extent. #12 & R!f@@ : I totally understand what you are trying to say. But audiofile people are weird people. They are always aiming towards heavens. Unfortunate, but hard to get detached from the desire of experiencing heavenly bliss. The debate to how worthy an effort noise reduction is, is a lengthy and unending one. I know amplifier idle noise is a more relevant challenge, that is why I am stating, if this filter works, I will implement it and be done with it. Or else I will forget it and move to the next task. Thread Starter #### prashanthb Joined May 17, 2015 34 Your filter doesn't help the noise at low frequencies. (low pass filter) Because of there size the inductor and capacitor will be pretty expensive.$50 or so.
If you really think you need better supplies (I seriously doubt you could hear the difference if the amplifier design is good.) I would make the regulator better. Like this one at the end:
http://tangentsoft.net/elec/opamp-linreg.html
Thank you for the link. I have heard that better regulation works equally good.
One option is the so called "capacitor multiplier" - basically just an emitter follower that doesn't actually regulate voltage, it has a bias resistor from base to collector to make it pass the required current, but instead of a voltage stabilising zener - it has a ripple damping capacitor.

In most cases, a few hundred uF is a reasonable starting point.
Thank you for the idea.
I will try this out in the coming days.If its simple, why not both the filter and capacitor multiplier ....just kidding.

Going by the responses so far I think there is no fundamental flaw in the filter idea. Only the cutoff frequency has to be changed.

Last edited:

#### Lestraveled

Joined May 19, 2014
1,946
I have known a few fanatical audiophiles. There is generally no auguring with them. They are as bad as a southern Baptist on a Sunday morning in Georgia. But I will give it a try.

Your power supply, supplies voltage. Somewhere below that voltage potiential is the noise of the power supply. The power supply voltage and its noise is used to power a good audio op-amp in your super amplifier. That op-amp has a "power supply rejection" spec of about 60 dbV. This spec means the output of the op-amp will not be effected by noise on the power supply by 60 dbV. In other words a 1 volt rms ripple on the power supply will cause a 0.001 volt rms ripple on the out put of the opamp. This is a thousand to one ratio. But the ripple on your power supply is not 1 volt rms, it is .0001 Volts rms (-84dbV at 10Khz), per #12 power supply filter model in post #9. The op-amp will still reject the noise at a thousand to one ratio, which means the op-amp output will have an rms ripple of .0000001 Volts or 100 nano-volts. Good luck hearing that.

Do you know that everything produces electrical noise? Your beard, your eardrums, the sweat on your forehead, and the resistors, capacitors, etc, etc in your amplifier generate noise. A great guy by the name of Ludwid Boltzmann discovered how much energy matter gives off due to thermal heat. Without getting into the math, a 10K resistor, the normal input resistance of an audio amplifier, will produce about 63 nano-volts rms of energy at 10Khz bandwidth at room temperature. This is not far from the 100 nano-volts that the power supple ripple rejection yields.

In other words, your eardrums produce about the same noise as the noise output of the first amplifier, from the combined effect of power supply noise at 10 KHz.

Edit: an audio amplifier manufacturer, I think Crown, published an article about this. If I find it I will post it.

Last edited:

#### ian field

Joined Oct 27, 2012
6,539
Thank you for the link. I have heard that better regulation works equally good.

Thank you for the idea.
I will try this out in the coming days.If its simple, why not both the filter and capacitor multiplier ....just kidding.

.
In general - you fit the biggest reservoir capacitors you can afford/fit in the space - but you have to watch the turn on surge where the mains transformer secondary sees the discharged capacitors as a dead short for a brief instant. There are various ways of limiting turn on surge, which you can get advice on here. The turn on surge usually causes the rectifier to fail short circuit - if there's no fuse, the fault current usually lets the magic smoke out of the transformer.

The capacitor multiplier is one strategy you can use to get away with slightly less reservoir capacitance - and less turn on surge.

#### prashanthb

Joined May 17, 2015
34
I have known a few fanatical audiophiles. There is generally no auguring with them. They are as bad as a southern Baptist on a Sunday morning in Georgia. But I will give it a try.
Totally agree with you Sir. But..................................................................
Sir, I am a person who uses 78xx ic for calibrating multimeters, not the sophisticated reference voltage ICs. You and me are on the same page 99% of the times. There's no conflict at all.

Thank you.

#### ronv

Joined Nov 12, 2008
3,770
@prashanthb
I simulated 4 circuits.
single 317
two 317's (one following the other)
The capacitor multiplier.
I'll post some pictures in a little bit- they are on another computer.
Both your circuit and the emitter follower don't regulate as well as the 2 317's of course since they are open loop, so the transient response is a little funny.
From the 100 Hz ripple standpoint the dual regulator is best, followed closely by the emitter follower. The emitter follower could be equal I think with a larger capacitor.
Your filter is very good at higher frequencies.
From the standpoint of problems.
The very large capacitor in the filter holds the regulator in current limit for a long time on power up. I don't know what your input voltage to the regulator is but it may be high enough to blow the regulator. (>37 volts)
The regulator needs a larger filter cap at it's output to deal with the RC circuit of the emitter follower or it oscillates. (10 Ufd. worked)
My personal favorite is the dual regulator. The you could market it as double regulated.

#### Attachments

• 95.2 KB Views: 4
• 133.6 KB Views: 4

#### ian field

Joined Oct 27, 2012
6,539
I have known a few fanatical audiophiles. There is generally no auguring with them. They are as bad as a southern Baptist on a Sunday morning in Georgia. But I will give it a try.

.

Its amazing the bogus advertising they fall for - I once saw advertised, square loudspeakers for digital audio.