Keantoken's CFP cap multiplier

Searching for a replacement for the venerable 2SA1837/2SC4793 pair I ran across the new Toshiba TTA004B & TTC004B transistors which, at least according to the DS, seem to be a good substitute.
In any case I'm ordering them for the two 200mA K-Multipliers that I'm about to build.

Regards,
Braca
 
So far it's been a rather disappointing experience with the circuit.
I built first the positive version on the PCB, with TTA004B and BC337-40, but the voltage drop across Q2 was some 2.4 Volts, and the LED does not extinguish after the initial light-up. Suspecting that the BC337-40 was out of spec (it is not OnSemi), I tried a couple of others from the same bag, but without significant improvement. It filters, to be sure, but only about 45dB @ 100Hz.
Replacing BC337-40 with BC54xx only worsened the voltage drop by up to 0.2V.
The capacitors in the positions C1 and C2 are 100uF low-ESR Oscon from Panasonic.
I then put together the negative version on the breadboard, and tested a number of configurations with TTC004B and 2SC5171 (Toshiba, from a German distributor) as Q2, and BC327-40 (10 parts), BC55x & BC560 as Q1. The best result re. voltage drop achieved was 2.2V with the LED never fully extinguishing, and the voltage drop being dependent upon the load (the goal for the latter is 170mA at an input voltage of approx. 20V).

All BC3xx that I have come from the same manufacturer (apparently Continental Device India Ltd.), and are marked with CTBC3xxx, so now I ordered the OnSemi parts from a reputable distributor.
I'll report back after I tested the circuit with the OnSemi parts.

Regards,
Braca
 
Well, my OnSemi parts arrived, but there was almost no change in the K-Multiplier performance, i.e. the LED kept faintly glowing. So it seems that my CTBC33xx do not differ much from the OnSemi ones.
The LED in all the tests so far has been the same one - a generic, no-name, 5mm red one. Having replaced it with a Kingbright L-1034IDT (also red) the Q2 voltage drop went down from 2.3V to 2.18V - still higher than the 1.8V desired, but I thought there might be a trend appearing here.
Luckily, there was an IR LED in the junk box, and with it in place, the voltage drop hit the bull's eye at 1.84V.
With my K-Multiplier now being at a correct operating point, I did a small study of the LED parameters that I thought may be of importance in this application. The results are summarized in the table attached.
The first test was to measure the LED voltage & current at the onset of light emission as I visually perceived it, followed by the LED voltage at a current of 100uA.
The second group of numbers refer to the K-Multiplier performance with the three LEDs used. It is seen that the Q2 voltage drop reduces with practically the same increment as the LED voltage @ 100uA as the LEDs are rolled. The last column in the table shows that it is the Q1 bias voltage that determines the Q2 voltage drop, since the Q1 base currents (Iled) do not differ much in the three cases shown.
I obtained similar results with the BC337-40/2SA1930 pair at the same operating point, so it seems that the new Toshiba TTA004B is a good replacement for the non-obtainable 2SA1837 in this application.
As regards the filtering action, my input ripple was to low at 155mV,rms, and the measurement limit was reached too soon.

Now, I know that several Members built this circuit with red LEDs and reported correct results, but I still don't know why it was different in my case.

Regards,
Braca
 

Attachments

  • LED_Tests.png
    LED_Tests.png
    24.2 KB · Views: 546
Now, I know that several Members built this circuit with red LEDs and reported correct results, but I still don't know why it was different in my case.

Regards,
Braca

The version on my webpage uses 3 diodes instead of an LED, it is a bit old. Since the current boards use an LED, they seem to have a bit more voltage drop. You could replace the LED with 3 1N4148 in series, although the IR LED is apparently also a solution.
 
I'm about to build another two K-Multipliers, and have now a few more IR LEDs at my disposal, so the table from post #1130 will probably get an update.
However, in my experiments with the PCB offered by Keanu, only IR LEDs with shorter wavelengths (< 900nm) seem to have been capable of producing the target voltage drop of 1.8V across Q2.


Regards,
Braca
 
As promised, here are the voltage drop measurements for several LEDs in my K-Multiplier with BC334-40/Toshiba TTC004B (the circuit diagram from this thread, assembled on the Kean's PCB):
1XDrGvBLG6gAAAAASUVORK5CYIIA
It seems that only the 850nm (wavelength) IR LED produces the desired voltage drop of 1.8V across the pass transistor Q2.
The results for the negative polarity K-Multiplier (BC327-40/Toshiba TTA004B) are virtually the same, so I'm not quoting them.

Regards,
Braca
 

Attachments

  • IR_LED_Test.png
    IR_LED_Test.png
    15.3 KB · Views: 205
Last edited:
SMPS Attenuation

I was thinking of using a K-Multiplier to attenuate ripple in the output of a 27.6v SMPS (adjusted to 32v). The ripple is 200mv ptp and I'd run it first through a T filter (3rd order) and then the K-multiplier. After that I generate a +-15v virtual ground based on LM317/LM337 regulators.

The T filter I configured attenuates at 60db per decade as frequency rises above 328 Hz. What does the attenuation (input isolation?) curve look like for the k multiplier? From reading the specs it's somewhere around 66db for the positive version, but I don't know how frequency dependent that is. Not as good as a regulator, but it also doesn't consume power like a regulator.

Also max current looks to be 1.4A. I need just a little more than that -- about 1.55A @ 32v. Is there enough headroom to accommodate that? You say that it's linear up to 400ma. What's the behavior beyond that?

thx
eris
 
You can use the D44H11G/D45H11G and BC550C from Onsemi with 50V capacitors. Since the higher current transistor is slower, the PSRR corner frequency is around 30KHz.

Ok.

Just to clarify...

1) You're saying that the low-pass curve of the k-multiplier drops from 66db down to 3db at 30khz? What's the shape of that? Linear? Butterworth?

2) When using the higher current transistor, when does attenuation leave the linear region? With the default transistor it was about 400ma.

Thanks for all your work on this.