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.
diyAudio Chief Moderator
Joined 2002
Paid Member
ikoflexer said:
I got oscillation with a 2200uF on the output. I don't trust the simulator that much. Probably best to try some values in practice and see what's best. I've been thinking, and I'm not sure that for a power amp extended bandwidth on output impedance is something that one would want. Perhaps best would be a flat output impedance and phase in the audible frequency range?

He can always revert to post#825 schematic if it does not go well with amp and V2. No worries.
 

iko

Ex-Moderator
Joined 2008
BTW, salas, have a look at my modified biasing of the CCS mosfet. It seems to be less sensitive to minute changes of the current setting resistor (R1 in your case), and the resistor is no longer in the "signal" path so to say. What do you think?
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Good, but 2 points. Recommended Rset value depends on certain IDSS, so the user must be able to calculate, or get certain IDSS? We must always make sure that the LEDS are running over 5mA and the Rset isn't big enough so to contribute more Johnson than a LED?

P.S. Can it be a trimmer?
 

iko

Ex-Moderator
Joined 2008
Yes, but with a known power resistor as load, not with zero load. I never start with the real load. Once the desired current is set, then I connect the real load.

Also, one can use one of the larger Idss 2sk170 below the "current set" resistor, so that the value of the resistor will be even lower for a given desired shunt current. I would not worry about the noise of such small value resistors. I would say it's definitely below the LED noise.

Edit: a 175 ohm resistor, at 120Hz, 25 Celsius, will have about 18.6 nV RMS noise. Would such value be of concern?

http://www.sengpielaudio.com/calculator-noise.htm
 
diyAudio Chief Moderator
Joined 2002
Paid Member
ikoflexer said:
Edit: a 175 ohm resistor, at 120Hz, 25 Celsius, will have about 18.6 nV RMS noise. Would such value be of concern?

It is broad spectrum noise we are considered with, not only at 120Hz, but it is very low if the resistor is not into kOhm range. It calculates at 0.26uV for 200R at 35C and 20k BW. As much as a LED. What Rset value you need for a typical 200mA CCS and 10mA IDSS?
 

iko

Ex-Moderator
Joined 2008
Salas said:
10mA IDSS in a 100pcs BL bag are a dime a dozen and seem adequate for our purposes. A bit more complex for a DIYer to set, but hey, this is V2. Takes a bit of more involvement.:cool:

Come on, it's not so bad :D

Rset = (Vbias - VLED1 - VLED2) / Idss

For a desired bias voltage for the CCS mosfet of 4 volts, if the voltage drop across each LED is about 1.8 volts, and if Idss is 10mA

Rset = (4 - 1.8 - 1.8) / 0.01 = 40 ohm

Simple :angel:
 

iko

Ex-Moderator
Joined 2008
Salas said:
So a 100R or 220R trimmer will do. The only prediction variable will be individual Mosfet part Vgs if with fixed Jfet IDSS, I guess.

Yes, and perhaps people use slightly different LEDs, so that could be a variable too. But as long as the LEDs have a voltage drop of less than 2 volts, the procedure would be to set the trimmer in advance to a low value, such as 10 ohm, and then increase its value in the circuit, while measuring the current. Very easy, really.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
JLH69 or any high bias class A like Zen etc. are the best customers for a high current shunt by the way. They already burn the juice on their outputs on big sinks all the time, and the shunt needs only a comparatively small sink for idling just a bit over that, and the bulk of its heat is only on the CCS Mosfet, which can be held at 5-6V higher only, if the input ripple is under 1Vpp by using a not so big filter cap, or two with 0.33R in between. Can make do with a common between IRFPs 1W/C sink in other words if the load is always connected.
 

iko

Ex-Moderator
Joined 2008
Tham, sorry about the moving target here. If you really wanted to try something with a wider bandwidth than v1 for your t-amp, I have this suggestion:

reg-tamp-v1.5-2a.png


This is the type of application where the loop gain is best provided by high frequency transistors. Q2 is BFT92, nicely working up to 2GHz, and Q1 is its complement, BFR92P. They are also cheap, and most online places will carry them. Q1 can also be 2N2222 (not so good, but OK). Using these high freq transistors in simulation makes a big difference in terms of stability and step response. It might be a good idea to use them for the DAC shunt as well. Then it is proper broadband treatment.
 
Hi Iko,
use the schematic in post823 for all your future versions.
It shows the measuring bridge more like it should be implemented in a practical PCB.

I am not sure that Q3 & J4 should be included in the bridge. They pass a correction current (although near constant), the same as the output FET does.

Finally 823 would be even better if the top and bottom tie in points, together with the whole bridge, were brought over to the other side of the output FET and shown connected to the output terminals.
 

iko

Ex-Moderator
Joined 2008
AndrewT said:
Hi Iko,
use the schematic in post823 for all your future versions.
It shows the measuring bridge more like it should be implemented in a practical PCB.

I am not sure that Q3 & J4 should be included in the bridge. They pass a correction current (although near constant), the same as the output FET does.

Finally 823 would be even better if the top and bottom tie in points, together with the whole bridge, were brought over to the other side of the output FET and shown connected to the output terminals.

Hi Andrew, if you could please have another look at post 857. Is this more like it?

Honestly, caught in the design process, running many tests for each change, I completely ignored layout. But you're right, many people see the schematic and automatically think that also shows the layout/connection points, which may often not be the case.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.