Keantoken's CFP cap multiplier

IIRC, the IRFP part needs 4V gate bias just to turn on. This leaves us with a high dropout voltage. The driver needs high Gm, and Jfets don't give good PSRR.

To be honest, the MJE15032/33 aren't terrible. According to the models they are about 3 times faster than MJL0281/0302 outputs. There seems to be a big speed gap between these and the MJE210/200 though and it really bugs me.
 
Yes, the IRFPs are pretty hungry units and I got some 2sk3497/j618s from Patrick's F5X group buy that about halves this, but haven't tried them out yet.

Okay, will get some of both the MJE 15032/33s and the 210/200s and give them a trial - don't know what will happen with these "faster" supply components to the sound - it'll be interesting.

How critical is that 820uF cap?

Thanks again.
 
Here is a poweramp version of the Kmultipliers. I use this on the class AB BJT amps I design. FET amps don't seem to improve that much with it.

As you can see, going over a few amps gets messy. This is the version I will try out next; I haven't tested it yet but it should be better than my previous version. Consider it experimental.

My supply is a beefy mains filter feeding an E-core, with a secondary snubbed at 1.5u+7R, parallel with 10nF, to the rectifier, then 1u+1R snubber, then 12mF, and then the power Kmultipliers.

I would say the line filter, trafo snubbing, and Kmultiplier are what made the biggest positive changes. The Kmultiplier improves bass, treble, imaging and general realism. The snubbing makes things clearer and more intelligible.
 

Attachments

  • Power_Kmultipliers.png
    Power_Kmultipliers.png
    121.5 KB · Views: 573
Okay, cap can be anything reasonable, the old 40MHz Schlumberger CRO should be able to see the ripple down to there, I think - it's not really up to seeing the "glitches" in the diode transfer function but still quite usable.

I also added an isolation transformer ( an old industrial one that allows the split secondary +/- 120V ac out here (balanced primary supply lines, or something like that!) to clean up the power - just generally quieter and nothing really dramatic - expected more - bloody great 2kVA EI lump of metal in an old brown ugly case!

I haven't used any "R+C, in // C" on the secondaries - must try this out - Hagermans, yes? I usually stuff around with tuning the damn secondary windings - a right real PIA that takes a lot of time and patience.

I was going to try those MSR diodes, the super quiet ones (MSR1520, or similar?) and something similar to your diode snubbers (4.7uF + 0R47).

I see you didn't need to use the C-R-C but just straight into the Kmx - that certainly makes it simpler - will play it myself -

I remember 'juma' saying the simpler single cap gave superior performance than the C-R-C, with his simple Cmx for his smaller triple fets F5 version - another variation to try out.

Is that extra transistor to increase the base drive on the Q2 o/p transistor, and the C1, R5 for extra stability? Increased bias to 5 volts? 56 volt rails, a big power unit?

Interesting things, these Cmultipliers - don't know why they've become "unfashionable" these days and it's hard to find people that have ever added a resistor to a cap and hung it across the diodes - never could figure out why it's not taken off as it's so simple to try.

It's a bit the same with trying to get people to knock up a simple Schroeder defuser and lean them against the wall in the room - people look at you funny when mention these things (or more than usual, anyway!)

Yes, I 'recon' a half way decent power supply is "money for nothing" in audio terms, and a good one .....
 
The power Kmultiplier must push the same currents as the output stage it's powering. The third transistor is needed, if not for drive, then for increasing current gain so that R6/R13 can be very high. Furthermore, at these currents the output and driver will have bad Early voltage so we need a TO92 BJT with great quasi-sat specs to apply feedback and fix it up.

C1/R5 are for stability.

There seems to be a split in philosophies for designing audio power supplies. Personally I've found that filtering noise and getting rid of resonances enhances the sound. They cause a sort of blaring sound to me. At the same time I know a lot of other people have tried snubbers and filtering but say the music loses something.

I think it is partly the person's preferences and partly the circuit. Amps which don't have much innate realism I think tend to become "livelier" when agitated by noise sources. And the result of this can be very pleasing to the ear. But since noise tends to be anomalous and unpredictable, there often is little rhyme or reason to making these tweaks. It makes you look like a fool and it's a moving target, which can disappear at any time.

In my amplifiers I use a more modern type of stability compensation that is more resistant to EMI and gives greater PSRR. This increased imaging and realism quite a bit, and after this I found that most of the time more inert, snubbed rails sounded better.

FWIW
 
Member
Joined 2009
Paid Member
Thanks HiFi, a lot of knowledge passes in private conversations, much of it worth sharing...

The schematic I sent you contains some possibly misleading features:

1: Output devices are 2SD669A/B649A. These will work to 100mA, but the 2SC5171/A1930 work to 300mA, are just as fast (if not faster) and are easier to get for some people.
2: Drivers are BC547C/557C. The BC550C/560C have slightly better noise specs, not sure if it matters but I prefer to buy the superior part anyways if it's of little consequence to the wallet.

Here is a cleaned up, nice schematic representing the ideas I described to HiFi.

- keantoken

Can I use the cfp multiplier after SSHVSR?

You should be able to, as long as you use an appropriately rated capacitor. However I don't think that output impedance will be improved (HF impedance might).

- keantoken

Can I use this schematic for HV 150-350VDC only with caps appropriate rated voltage capacitor? I want to use before the Salas Shunt Regulator HV2.
 
Hi,

Great work - I'm hoping to try this out for my DAC op amps that use a Kubota reg. BTw, are you familiar with Semelab BUP40 / BUP41 ?

I have a question about the circuit describing the Darlington multiplier on your linked page. You have drawn Q2 as a NPN but labelled it as BC560C ? Is this correct ? I'm no expert ...

Anyway, thanks very much for posting this circuit. Great stuff !
 
Last edited:
Yes, you can use it before the HV2. Tell me what your input ripple is and the current draw, since I will need to change the circuit a bit. If you increase R2 (website schematic) to 10k, you can use 82u instead of 820u. You may even be able to use 100k and 8.2u; it depends on loading.

Thanks KlipshKid. I haven't messed with the BUP40/41. My focus has always been on parts that are easy to get and that can be bought from authorized vendors with the certainty they are not counterfeit. This has just been my preference so far, as I don't want unexpected problems while I am testing and developing my circuits. If they are still easy to find not counerfeit, I may consider making designs with them.

If BC550C is the NPN part then that's what it should be. I mix them up all the time! Must fix that...
 
Member
Joined 2009
Paid Member
I don't know the input ripple: I will use CRCRC and CLCLC between the tx & the cap multiplier, for sure I will use a hybrid SS+silicon rectifier Graetz type with tubes AZ1 & 6AXGT, load will be #26 tube preamp or other tube preamps, SSHV2 spends 20mA to operate for #26 or other tube preamp I will set the SSHV to 30mA.
 
If you have 30mA load current and 470uF capacitance after the rectifier you should have .5V ripple. Input ripple for the Kmultiplier needs to be lower than 1.8V. For 1.8V ripple you need at least 140uF, if the Kmultiplier were right after the rectifier. Tell me if your supply is not beefier than this.
 
Member
Joined 2009
Paid Member
Well AZ1 now have connected only 2 x 40uF capacitance ASC386X MKP in oil = 80uF. 6AX4GT still not installed waiting to arrive. I have now the Valve Itch phono tube preamp that I have installed 4 UF4007 as bridge rectifiers followed with CRCR 220uF/220R/220uF/220K bleeder resistor, I think can I use here?
 
So if I understand correctly 240VAC->340VDC. 30mA*220R=6.6V. So I have 13.4V to work with.

Merlin, what do you think about this? The 220R/220u RC is from your supply. It took some determination to make it safe from overvoltage and inrush current, but I like how it turned out.

Don't put more than 330uF between the Kmultiplier and the HV2. Capacitance after the HV2 is okay because it will be limited to 30mA inrush.
 

Attachments

  • HV-Kmultiplier.png
    HV-Kmultiplier.png
    103.9 KB · Views: 636
Last edited:
Thanks KlipshKid. I haven't messed with the BUP40/41. My focus has always been on parts that are easy to get and that can be bought from authorized vendors with the certainty they are not counterfeit. This has just been my preference so far, as I don't want unexpected problems while I am testing and developing my circuits. If they are still easy to find not counerfeit, I may consider making designs with them.

If BC550C is the NPN part then that's what it should be. I mix them up all the time! Must fix that...

Hi,

I'm in the same situation but I can't get 2SA1930 / 2SC5171; Q2 and Q3. Could I use Q2 BC560C and Q3 BC550C instead (for applications under 50mA) and still get good psrr and stability ? I'll still use Q4 BC327-25 and Q5 BC337-25 as your page recommends.

Is anyone intrested in doing a PCB and group buy ? I can see this becoming very popular if it can be made suitably small.

Thanks for your help Kean !
 
If you use smaller outputs you need to mind inrush. At 50mA TO92 outputs may work but may blow up when trying to fill a 470uF output cap. This was my problem with the HV Kmultiplier. A TO92 output would work great for this but would be destroyed the first time it was turned on.

The BC327/337 are more suited as outputs than the BC5x0C, I would use them instead.

An SMD version could be made, NXP has some ideal BISS transistors for this purpose I think. Scratch that, the only way you can get them is through the sample kit!
 
BD139 will do I think though I haven't tested it myself. You have to get it from the right manufacturer though, or it'll be something aweful. The BD433 is one of the slowest transistors I've ever seen, 3MHz at 250mA! That is a new one, for me. It's slower than an MJL21194!

BC639/640 from NXP/Phillips are a good bet if you can get them anymore. IIRC BD139/639 from ONSemi are no good.

Dunno if any of this helps. The BUP40/41 will probably work great if you have them.