How much capacitance....??

[MiiB]

Hi.. I'am currently working an an amplifier design...I use a dual supply..one for the voltage gain stages and one for the output stages...

For the front end supply i use a high speed shunt at 70 V and a 10 uF..film/foil type capacitance.
I then use LED base current sources... and I know its common to use some sort of capacitance over the led to kill any noise generated there...But i am much in doubt how big these caps need to be...??
Browsing schematics gives me all kinds of suggestions from small and fast to rather big eleco's...

so what is your take on this.. how big should the caps be.. and also should they be from the rails to the bias point or from the bias-point to the GND...??

thanks

[AndrewT]

the cap across the Vref does not kill the noise.
It creates a filter and the effect of that filter is to attenuate the noise.

You can insert a real capacitor model (not pure capacitance) and evaluate the attenuation over the audio frequency range and in the sub audio range and in the VHF range.

Then you decide what attenuation you require in any or all of these ranges and choose a capacitor.
If you can't do that then you have to guess and measure the result. Guessing is just about the slowest way to get anything to work to specification.

[MiiB]

I do know it creates a filter....the reason for me asking is the use of the high speed shunt..and the 10 uf...film foil cap...So to me putting like 680 uF..from the rail to the V-ref only degrades the supply to the active devices...better then use the large caps to decouple V-ref to ground...or to use small film types caps to the rail...

this is what my sense tell me.. but i have no actual experience to back my assumptions

[MiiB]

ohh... one more thing...The filtering of LED's..V-ref.. must also depend on the noise and frequencies of this noise...so if the noise is at high-frequencies..I suppose the caps must be small an fast...film..types like stack-foil...or..

[MiiB]

Done a little testing.....and I can see to stabilize the V-ref you need a bit of capacitance...and it seems that it works the best when tied to GND...so now I have settled on 470 uF tied to GND...then the V-ref is totally flat and not modulated by the signal...!!

Then there's only the LED noise question...how high the led noise freq is and will the Elco be fast enough to filter it to GND...or does one need additional decoupling...?

[Bigun]

Some people say that they don't like the 'sound' of a CCS - whatever that means. But you have pointed out that the Vref if modulated by the signal and a capacitor stabilizes this - which means that the capacitor has some influence on the performance of the amplifier and it's quality is important.

[MiiB]

I agree on the CCS... when ever i use it in the first stage it seems to compress and veil the sound..

This is uesd for the setting of the cascode voltage for the bipolars shielding the input Jfets and also for a led voltage reference used to to set the current through the second stage...

[Nico Ras]

Originally Posted by MiiB

Done a little testing.....and I can see to stabilise the V-ref you need a bit of capacitance...and it seems that it works the best when tied to GND...so now I have settled on 470 uF tied to GND...then the V-ref is totally flat and not modulated by the signal...!!

Then there's only the LED noise question...how high the led noise freq is and will the Elco be fast enough to filter it to GND...or does one need additional decoupling...?

Hi Michael,

Capacitors has impedance related to frequency that can be calculated roughly as 1/(2PIxF) which is good enough for use in filters. Problem is with large capacitors the length of the foil and other factors causes them to have an increased impedance at higher frequencies and one places a smaller capacitor in parallel with the larger one to counter this effect.

One must also keep in the back of ones mind that a capacitor causes a time delay (integrate), in other words the higher the capacitance value across the reference the slower it will respond to change.

The trade off is what is most important, reacting to change or getting rid of noise.

Instead of an LED as voltage reference (which is not all that good anyway) you can use a low noise voltage reference . You can also use the junction of a transistor as reference giving more choice.

All of these references generate noise, regardless of how good they are, there is no set rule for the size of capacitor to reduce/filter this noise, it is a case of measurement and what you regard as good enough for the application.

[MiiB]

The reference can also be set by two resistors...Then you'll only have to juggle with resistor noise...and thus at the same time avoid any extra complexity...

[Nico Ras]

Originally Posted by MiiB

The reference can also be set by two resistors...Then you'll only have to juggle with resistor noise...and thus at the same time avoid any extra complexity...

Thermal noise is approximated by KxTxBxR, where K = Boltzmann Constant, B = bandwidth, T = temperature and R is resistance, other noises are added, such as shot noise, electron noise and the like. Then there are noise which is related to the mains, inductive noise, all kinds of stuff.

First have a look at what your problem is then solve that, to make a statement of I don't want noise is too much of a broad term. If you do not have a problem, then don't try to solve it.

Too many folk will throw components at a problem which may not exist because it may be trendy or they assume they are avoiding something that could be present but is not or it is purely for theoretical reasons. Just by doing this may create a problem that you did not have before.

My crude test for a good power supply is to take a set of headphones, decouple it with a capacitor and stick it across your power supply, if there is hum or noise audible start looking for a solution. If not chances are very good that you will not hear any power supply related noise on the speakers.