What is the rail voltage level for your amplifier buffer? Is it (+) V only or symmetrical?
Make sure you count for the peak current draw. 200mA plus is very good spare.
M1 needs be PMOS, M2 needs be NMOS. Reverse in a negative reg. So no single type would do.
What's the better reg to use from those three depends on the application. Tell us more about that buffer.
It is a calvin buffer out buffer. It run on +-15v and use about 100ma for two channels. I have an transformer that can out +-25vdc after filter capacitors. I did have a typo. I have both pmos and nmos version of the mosfet.
The Calvin Buffer is very capable of driving cable capacitance.
You must allow for peak transients much higher than what the typical 10k to 100k loads would draw.
Your 300mA less 20mA as the Shunt minimum leaves 280mApk for the two Calvins.
140mApk compared to 50mA quiescent seems like a big enough margin (90mApk) for transients.
Can you see that the Iq is almost completely irrelevant to sizing a CCS/Shunt regulator?
You must allow for peak transients much higher than what the typical 10k to 100k loads would draw.
Your 300mA less 20mA as the Shunt minimum leaves 280mApk for the two Calvins.
140mApk compared to 50mA quiescent seems like a big enough margin (90mApk) for transients.
Can you see that the Iq is almost completely irrelevant to sizing a CCS/Shunt regulator?
You may use a couple of Reflektor-D with R1=2 Ohm, two LM329 instead of LEDS (cathodes in reverse than LEDS) and a JFET (G+S shorted) instead of R6 with your MOSFETS. You can stack them for symmetry and virtual center point ground as shown at the bottom of the Ref-D pdf guide. Will do 15.22V symmetrical @ 300mA current limit per rail that way. Give them 22V raw DC in at least. Calculate dissipation* and get proper sinks. They should give you less than 0.001 Ohm extended output impedance and good stability. You may use J113 if difficult to get surely genuine 2SK117GR, with a small noise penalty, but well passable for an analogue line level buffer's rail noise demand.
*
WM1=ICCS*(VinDC-VoutDC)
WM2=VoutDC*(ICCS-ILoad)
@Andrew: I got what you mean. Iq will only applied when there is no input signal. And we need to take care of maximum peak current when we use Salas Shunt regulator.
@Salas: Thank you for your circuit recommend and extra info! I will try to follow it and will come up later with a schematic to confirm. BTW, how 2SK246GR or 2SK170BL compare to 2SK117GR in R6 position?
@Salas: Thank you for your circuit recommend and extra info! I will try to follow it and will come up later with a schematic to confirm. BTW, how 2SK246GR or 2SK170BL compare to 2SK117GR in R6 position?
Last edited:
K246GR at 3-4mA Idss OK for Vgs(off) vs M2 VGS(on). K170 avoid due to TC. J113 may be no worse than K246GR for noise and stays in Vgs range for all samples. BF245A is good there also.
Looks drawn alright. Being based on my 2011 linked example from post#2 using IRF640 it has C1,R5 heavier values. Better use 1R & 0.1uF for those lighter gm and Ciss FQP MOSFET types.
As for 2SK117GR, if you are not 100% sure they are genuine, better use J113. In a very recent example for SSHV2 in the HV regs thread an Australian user had drift and noise problems because of fakes.
As for 2SK117GR, if you are not 100% sure they are genuine, better use J113. In a very recent example for SSHV2 in the HV regs thread an Australian user had drift and noise problems because of fakes.
P.S. Also why you designate 8dB noise figure PNP's when 2N4403s as in the original are much better for noise?
2SK2145
edit: the 2145 is essentially two SK117
Although Toshiba has discontinued all of the leaded parts afaik, a dual in SM exists, the 2SK2145, which might be useable with some sort of adapter socket. They also tied the two channels together instead of bringing both ends of them out and using a sixth pin, which I found remarkably short-sighted.
edit: the 2145 is essentially two SK117
2SK880GR is safe and 100% compatible to K117GR if one cares to use an SMT adapter. Must only notice its a 100mW device.
I didn't know that 2N4403s was low noise, I checked again and I think in same condition with 2N5401, 2N4403s has NF = 1dB. I choice 2N5401 because I have a lot of them.
I have some Toshiba A1015GR (original - get from Japan) which NF = 1dB. I also have some OnSemi BC556B, Fairchild KSA992. Do you think it is ok to use one of those? If not, I will try to get some 2N4403s.
I think I can use J113 in J1 and 2SK880 in J2. PJ2 = 4ma x 14V = 56mW seem still ok.
I have some Toshiba A1015GR (original - get from Japan) which NF = 1dB. I also have some OnSemi BC556B, Fairchild KSA992. Do you think it is ok to use one of those? If not, I will try to get some 2N4403s.
I think I can use J113 in J1 and 2SK880 in J2. PJ2 = 4ma x 14V = 56mW seem still ok.
Last edited:
Yes, rbb' for 4403 and 1015 is about the same, with the 1015GR having a somewhat higher beta than the typical 4403 if memory serves.
It's interesting that the 2N4403 was designed for switching applications, and it was regarded by some as a trade secret that it had low noise with low source impedances. One friend made something of a name for himself at TRW by designing a preamp for very-low-resistance detectors using the 4403, back in the 70's, that was better than that of the resident low-noise-preamp specialist.
J1 will see about 19V across, also being degenerated to 2mA IDSS by RR (test on bench for RR and J1 sample). That is 38mW.
JC was using those for low noise audio back then. Something he learned in Ampex if I remember well by reading one of his reminiscing posts.
Hi Salas
Can I use one CCS to feed two different voltage regs ?
Imagine a BiB with two different output mosfets and corresponding error voltage circuitry sharing the same input CCS..... is it possible ?
Can I use one CCS to feed two different voltage regs ?
Imagine a BiB with two different output mosfets and corresponding error voltage circuitry sharing the same input CCS..... is it possible ?
- Status
- Not open for further replies.