Keantoken's CFP cap multiplier

Thanks, Kean and Salas.

I will replace the 2N5551 and 2N5401 with BC550C and BC560C. I have run out of them and all shops are closed Sunday today. I will order them as soon as I can.

1: Remove R1 and see if performance improves.- keantoken

Without R1, some big teeths (a few mV) with shapes like pyramid were found.

Wait, voltage at R1 is 17.2 and at the base of Q1 is 16.2? This means 1V across D7, which means something is shorted or broken.

- keantoken

What can be broken? Diodes are 2N4002 and all parts can stand much higher than the 18V input voltage. Also, the measured 1V accross the 1k resistor is consistent on both rails and with two boards built. Also note that on the negative rail, the output appears almost flat, even with something "broken". Opps! I forgot to mention that I have 9,100uF after the K-Multiplier before the Marantz reg which would smooth the ripples out. Without those, the ripples must be higher. So you are right, something must be broken, or it can work even better (as I works well for my purpose now).

Last edited:


2006-08-10 12:32 am
Iko, for some reason I kept building things with the 5088/5087, but the datasheets show 50 and 40MHz Ft for both, which is WAY too slow for me. I used to think they were fast... Not sure why I thought that... The MPSA18 is still pretty good though. Unfortunately it has no good PNP counterpart.

The BC560/550 are the best I know of.

HiFi, something is apparently broken. Can you check to see if C1 is not shorted or something? You might add another diode to the string just to make sure, but I don't think this is right...

- keantoken
If you have 2n5088/2n5087 they might give you exactly what you need. mpsa18 has even higher gain.


Thanks a million. I have just looked at the datasheet and yes, the 2n5088/2n5087 are even better!

I think I have quite a few 2n5088 left from the Iko reg projects. I don't have 2n5087 though. The 2n5088 is used in the positive rail, so I can put it in and see if it fixes the problem.

The 1V across the 1k still worries me though.


We typed at the same time. In that case I may wait until Monday to get the BC550C/BC560C. I found the 5088 has a higher Hfe.

The C1 were brand new capacitors. For one board they were 1000uF. For the second one they were 820uF (As I ran out of the 1000uF).

If it was due to a particular C1, I would imagine that the problem would appear on one board only, not both. What do you think?



diyAudio Moderator Emeritus
2008-05-29 2:41 am
Bill, IMHO you got bigger problems than worrying about the 2n5088 being slower than bc550. This filter isn't working as it should. You could use the 2n5088 to troubleshoot the circuit.

kt, I know what you're saying, but it's just a filter. I wouldn't worry about speed in this application. He can use a common mode ferrite choke for higher frequency garbage that comes down the line. The filter won't help with that much anyway.

Edit: of course, it's just an opinion.
Iko, you are totally right. My reasoning is that the AC characteristics of the master will affect the circuit's stability. The slave's characteristics are more important for filtering at AC, with lower Cob being better.

Anyways, there is actually very little speed difference between the:


Except what is imagined due to the nonstandard way of measuring Ft. However the BC550/560 are best for this circuit because of their low noise, and the possibility of using the higher Hfe grades. One might say they are the perfect pair...

- keantoken
I swapped the 2N5551 with 2N5088.

Measurements for the positive rail are as follows:

Voltage at Input: 15.70V
Ripples at Input: 400mV
Voltage before 1k resistor after two diodes: 14.95V
Voltage after 1k resistor at top of C1: 14.95V
Voltage at the base of the pass transsitor: 14.97
Voltage at output: 14.30V.
Ripples at output: 4mV.

I could not get detailed readings for the negative rail because this is the real circuit board on the Marantz PSU board mounted on the side of the chassis that blocks the access to the solder joints. I only managed to measure input ripples 400mV and output ripples less than 0.5mV. Input and output voltages are the same as the positive rail.

While the voltages seem to be reasonable, the output ripples suggest that the negative rail may (or may not, since I have 9.100uF after it) work well as simulated while the positive rail fails short of the simulated result.
Okay, this is confusing me. Let's get it all in the open.

For both regulators, please post:

Input voltage
Input ripple
Measured current draw
Output voltage
Output ripple

Then proceed with whichever measurements you have made, with positive and negative reg measurements side by side, so we can see what's different. Also post capacitor values and diode part numbers and what we need to know in order to simulate what you have. Also scope shots of any interesting/choppy looking traces would help.

It is good that there is little/no voltage across the 1k resistor, this means the transistor is operating where it's supposed to.

- keantoken
I have just checked with my latest ISP, bigpond / Telstra, the largest in Australia, and they told me that I need to subscribe their web hosting package in order to have any amount of web space. Free web space of 10MB has always come free with any ISP account with anybody. I will definitely boycoot their web hosting packages. This is rather rediculous, as I am paying $100 per month just to have a rarely used phone line and 25G bandwidth of cable internet access. The government made this "monopoly" so this is the only ISP that can supply cable internet to me, while the rest are of ADSL or mobile. I guess any of you living anywhere else can laugh at Australia.

Which means, I won't be able to post any photos in the forum any more.

Kean, I will send you some photos when done. For now, I am leaving home...

The above given numbers in post #93 should cover what you asked for. I am pretty sure that the negative rail has very simular numbers.

I measured it again after removing some obstacles in the Marantz player, poking the probes trying to reach some component legs. I almost got a full set of numbers again which show both the positive and negative rails share simular numbers, as follows:

Test 1:

Input: +15.91, -15.91V
Output: +14.51V, -14.32V
After two diodes: +15.15, -15.05
Pass transistor base: +15.17, -15.27

Test 2 with same board but 1 hour later. I conducted this because in the previous test I forgot to measure the voltage at the base of the driver transistor. (Note the K-Multiplier is before the regulator, the readings can be different each time because of the mains voltage fluctuations.)

Input: +15.65V, -15.64V
Output: +14.19V, -14.01V
After two diodes: +14.86, -14.75

At this point, I was to measure the voltages at the bases of the transistors, due to the difficulty of reaching the component legs, the probe somehow shorted or earthed a component - a small cracking sound - the measurement indicated that on the negative rail some of the components must have been destroyed and the circuit no longer working. I will take out the board and repair it.

A couple of days ago I measured accurately using a series resistor which showed that the load draws very close to 300mA currents on both rails. I can trust this number.

The input ripples have been measured many times, including this time, to be 400mV on both rails.

The positive rail output has about 4mV peak to peak, while the negative rail output has less than 0.5mV ripples.

From 400mV to 4mV (also assisted with 9,100uF after the K-Multiplier), this is a -40dB rejection, comparing to the theoretical, simulated -65dB rejction (with 20uF at the output?), so perhaps something can be done.

For me, the sound is already a huge improvement that I am very happy with. This is because the original Marantz -60dB rejection seems to be insufficient, but combined with the K-Multiplier's -40dB, I have now -100dB line rejection, which is excellent.

Sorry for no photos taken.

Even so, I am more than happy to continue the effort to perhaps gain even more and help with testing this circuit to perfect it.

Last edited:
The components I used to build the K-Multipliers:

Driver: Board 1 uses 2N5088, 2N5401. Board 2 uses 2N5551, 2N5401.
Pass: 2SA1930, 2SC5171
Diodes: All 1N4002
Capacitor (C1): Board 2 uses Rubycon ZL 25V 1000uF || MKP 1.1uF. Board 1 uses Rubycon ZL 25V 820uF || MKP 1.1uF
Input Capacitors: 2 x Rubycon ZL 25V 2,200uF || MKP 1.1uF
Output Capacitors: 2 x Rubycon ZL 25V 2,200uF || 1 x Elna 4,700uF (Marantz)