The simplistic Salas low voltage shunt regulator

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Then where's your hum needle at 50Hz on second pic? Maybe you had the probe at a better angle to the toroid?

Probably "drowned" in noise b/c we are at the limits of the equipment. It is visible on my "audio range" picture.
 

Attachments

  • audio noise.png
    audio noise.png
    37.5 KB · Views: 365
Should be a Mastech or clone. ELFA lists one such linear PSU.

Its from ELFA :) It was a bargain of about €100 a couple of years back. My previous lab supply currently serves as psu for my friends B1...
Yes it has taken us a couple of years to get back to this diy project - he's very happy with his B1 (prototype, with black gates, tantalum resistors and all) :)
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Its from ELFA :) It was a bargain of about €100 a couple of years back. My previous lab supply currently serves as psu for my friends B1...
Yes it has taken us a couple of years to get back to this diy project - he's very happy with his B1 (prototype, with black gates, tantalum resistors and all) :)

Very practical boxes for prototyping work non the less. A TTi (Thurlby Thandar) with remote sensing, digital control & keypad may wait for better days.:)

Just keep your DC cables to B1 short and your toroid away in the end installation. Have fun.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Not in my old tests with resistor divider bias, one Vbe control and bootstrap cap as far as I remember, but that has to do with the voltage level and BJT bias too. Maybe it got a bit hot as you were going up with 4k7 only, or has other Hfe. And it wasn't your Mosfet type or CCS level either, correct.
 
Sorry, I'm not sure I understand.

My FET is IRF9610.
The BJT didn't get hot.

I just saw that the current decreased when input voltage got above some 21-22V and was slightly puzzled.
Only thing I could think of was that the FET (9610) needed less Vgs for the lower I (100mA) and different type (not IRFP9240), then the BJT needs to drive more current through the 2x 4k7. I changed them to 10k and it improved a bit.
Changed them to 47k and it worked, atleast up to 35V :)
 
Using two 47k for R2 & R3 allows about 0.22mA for Q1 Ic, when Vs=25Vdc.
This seems very low.
With R4 passing ~100mA, then the base current of Q1 is ~0.25mA if hFE=400 @ Ic = 0.22mA.

I would suggest you aim to run Q1 such that Ic>>Ib preferably a ratio of 10 or more.
That seems to indicate that your first choice of 4k7 was about the maximum for these resistors.

I think you need to look at other reasons for the non constant CCS current.
Replacing R3 with a 3mA to 4mA jFET may help fix the Ic of Q1 at a more suitable value, but I suspect that instability is at the root of the problem. 431 plus 9610 for Q6 and the eight separate connections for the sense circuit could all give stability symptoms.
Forget Remote Sensing if it does not suit your needs. BUT YOU MUST properly collect the sense connections together and attach just TWO sense measurement points into the +ve & -ve rails. You have shown 5 connections spread along the top +ve rail and 3 connections spread along the bottom -ve rail.

And why did you adopt a pair of 9610s when a 510+(520 or 530) combination would almost certainly perform better and guaranteed to not perform worse.
 
Last edited:
Hi Andrew, thanks for your comments :)

Why is 0,22mA for Ic for the BJT very low?
How do you get the value 0,25mA for Ib? Given hFE of 400 it would be more like 0,55uA?
Maybe you could explain to me how this CCS works?

It looks obvious: BJT will regulate Ic/Vce to keep Vbe about 0,6V. If Id (FET) increases, so will Vbe(=R4*Id), increasing Ic. Now V drop over R2+R3 will increase and Vgs(=Vce) will decrease and Id will decrease and we're back, constant Id (In-/decrease as numerical values, ie. |Vgs| etc).

Then what happens when Vin increases (assuming we keep Ic and Id constant)?
Vce and Vds will increase. For the BJT we assume constant Ic so only Vbe must decrease. With increase in Vds AND Vgs (=Vce) FET will see 2 factors giving increase in Id => Vbe(=Vr4) will increase. But it has to decrease to keep Ic konstant at higher Vce.
Vce then has to decrease. If Ic is given by Vin-Vce/(R2+R3) then Ic will increase...
Eventually Ic, Vce(=Vgs) and Vbe(=R4*Id) will find a new "steady state" and it is not obvious to me if it is with higher or lower Id than before Vin increase.

BTW: |Vgs| in my prototype for the CCS FET is 4,27V.

I don't see any stability issues with my prototype. All connections from D of CCS FET to S of shunt FET and output are connected to the same breadboard copper trace (shown with black marker on top).

I use IRF9610s because they were in my "stash" :) and are more than adequate dimensioned. One of the purposes of this prototype was exactly to test if they can be used in this circuit.

I have a number of things to try in a new version of the prototype. I do not like the "BJT controlled" CCS as I cannot easily comprehend its functionality as described above, so it has to go. Also the TL431 for a number of reasons.

But please keep comment, I'll learn some (analog-) electronics which is obviously too far away in my little head :)
 
diyAudio Chief Moderator
Joined 2002
Paid Member
I run a simulation, no problem at say 25V in with 4k7,4k7, to the contrary Vbe naturally is a bit higher for more Ic so it gains 10mA for the Vbe/R1 CCS governing rule (R4 in yours). Only thing I saw is you better put 10R in series on your bootstrap capacitor (C4) between now nodes, especially if its low ESR it gives a PSRR curve kink.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
P.S. From 1.2 and 1.2R experience tight layout matters in such CCS. And/or an interaction with the lab PSU? Maybe a 0.1uF decoupling cap when test feeding from bench PSU long leads. A good thing in this CCS type is it gives you a steady parameter (Vbe) to choose setting resistors against, when LEDs and Vgs solution give play. Especially in hot rod settings the one Vbe solution is signifiicantly cooler for Rset too.
 
I run a simulation, no problem at say 25V in with 4k7,4k7, to the contrary Vbe naturally is a bit higher for more Ic so it gains 10mA for the Vbe/R1 CCS governing rule (R4 in yours). Only thing I saw is you better put 10R in series on your bootstrap capacitor (C4) between now nodes, especially if its low ESR it gives a PSRR curve kink.

Hi, being and old-school engineer, I usually try to understand the circuit before simulating :D

But thanks for the input - could I ask you to post the schematic + result from the simulation?

The R in series with the bootstrap C is noted :)
 
...A good thing in this CCS type is it gives you a steady parameter (Vbe) to choose setting resistors against, when LEDs and Vgs solution give play. Especially in hot rod settings the one Vbe solution is signifiicantly cooler for Rset too.

Yes, I like(d) this CCS too because of the tight Vbe setting, it gives low loss and you can expect Id to be close to 0,6V/Rset (R4 in my version).

But as said before, I'm not sure I can understand exactly what happens when Vin changes, which gives me headache...

I might choose a 3xLED driven by a CCS FET (K170) as "Vref" for the CCS, experiment with a proper Rset value and accept the higher loss.

Regards Jesper
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.