Keantoken's CFP cap multiplier

I keep waffling, in simulation all the gain groups act the same so I use the -40 gain group when convenient. The models are from NXP/Phillips and always seemed to be very good quality, so I would not be surprised to find the -40 gain group working as expected in reality.
 
the SMPS issue brought up here could have another interesting angle/aspect

I have a box loads of those standard wall plug thingies
tried to use one with my guitar effects
but even with this a simple regulated DIY supply sounds better

now Im thinking if the K-multiplier might help to filter the SMPS
but should be easy to check :scratch2:
 
If one wishes to focus on input rejection, performance can be improved without increasing Vf dropout by splitting the 1kohm resistors R2 and R4 into two 499 ohm resistors each, and connecting a capacitor from the R2a/R2b junction to ground (component identifiers refer to schematic on KeanToken's webpage).

KR and Happy New Year!
 
Your suggestion doesn't increase Vf dropout, it turns the reference filter into a second-order one. Increasing Vf may improve PSRR, but I've tried it and it's not much.

Webpage for reference:

The K Multiplier

As far as a better reference filter (R2/C1) I have considered this option before. Input RF leaking through the RF filter depends on the inductance of C1 to ground. Let's say its 22nH, one inch.

For the PSRR to be worse than the Kmultiplier at 100KHz, which is at max say 66db, the inductive reactance has to be 500mX. An inductor with 500mX reactance at 100KHz would be 796nH. 22nH has 14mX, giving a max rejection of over 96db. The Kmultiplier itself is already passing 10 times this much.

ESR of the cap is more of a concern. Using a cap with an ESR of 100mR gives a filter rejection of 80db at max, still swamped by the 66db rejection of the Kmultiplier. However with 500mR, the total Kmultiplier rejection is reduced by half. Most caps at the usual size of C1 have much less ESR than this, but at high voltages you may need caps with higher ESR. At this point current load tends to be reduced so I increase R2 to 10k at the expense of an extra .5V drop.

A much more effective way to improve PSRR is to put an RC filter in front of the Kmultiplier. Unfortunately this increases voltage drop and DC output impedance, unless you take the reference input to before the RC, in which case you probably will need jcarr's suggestion so the the input filter doesn't ruin the fun.

If anyone wants to try it, I'll be interested to hear your results. The Kmultiplier's main draw is its PSRR/Vf ratio, but if you have Vf to spare it's a good idea.
 
🙂
 

Attachments

  • IMG_8143.JPG
    IMG_8143.JPG
    180.6 KB · Views: 541
I use a PCB covered in masking tape. I cut and lift away the tape in spots around the circuit for ground connections and to solder wires to with which I clip or anchor components to the board. This gives me an effective ground plane prototype.
 
How about a lc for rf before your K-multiplier . Did you try the bd140 on the negative rail as a k-multiplier ?
To me this looks like a very good stage isolator with little overhead . Well done . Using this for each stage and your low current for the front end would provide freedom in circuit chooses not able to be used because of very poor PSRR other wise. Interesting.