Simple shunt reg

[Onvinyl]

Here is a simple shunt regulator with good isolation and low noise. One of the design goals was to use things waiting in the parts bin for too long. Another was low parts count because I plan to use some of those in a phono pre.
Not shown are output/input caps.
There nothing too special about it except it works really well.

One thing I'd like to know is: How can I test/verify for output impedance?

Rüdiger

 

[Werner]

By simulation: attach a 1A current source as load and sweep it over the frequency range of interest. The voltage measured at the output node is 1:1 equivalent to the output impedance.

http://www.tnt-audio.com/clinica/regulators2_impedance4_e.html

 

[aparatusonitus]

One thing I'd like to know is: How can I test/verify for output impedance?

Rüdiger

Since you are using LT SwCad do this: attach an 1A AC current source I1 at ouput, run AC analize, when card popoup, mark Vout, change Vout to Vout/I(I1) by presing Alt, OK, you will be able to see output impedance vs freq.

 

[Onvinyl]

Ok, I did.
How do the dB-numbers correlate to Zout? Is it mandatory to get the curve flat?
Rüdiger

 

[Werner]

plot voltage (or impedance) on a linear scale, not a log/dB scale!

 

[aparatusonitus]

Place a cursor on the Y-axis, press left or right ( I can't remember right now), mark linear scale instead dB.

Impedance shuld be as low as posible and flat as well to 100kHz to 1Mhz region, after that you will see a rise up, play a little bit with a "real" capacitor at output (or without it).

 

[Onvinyl]

Thanks! yes, I forgot to place an output cap before sim

this is with 100uF and assumed ESR of 0.1R, at the output
Rüdiger

 

[unclejed613]

right click on the db scale, you can change it to lin or log voltage scale. you can do this with the FFT output as well to easier compute %THD (log is better for that). if you find that your results are "compressed" by using a lin scale, see if you get better results with a log scale. while in the plot scaling window, you can also set the high and low limits of your trace, and you can use this feature to magnify certain portions of the trace (you can also do this by selecting a portion of the output trace by holding down the left mouse button and drawing a box around the portion of interest).

 

[sawreyrw]

Onvinyl,

How does this circuit work? Where is the voltage reference? If it's dependent to the transistor parameters, I don't think it will be very dependable.

Rick

 

[Onvinyl]

I'll do my best:

Q2/M1 and R12 form the current source against the shunt reg will work.
The differential fet (2sj109) is the error amplifier. The leds are the voltage reference, held stiff by J3. A portion of the ouput voltage (both DC and AC) is compared to the voltage reference through R2/R6. I made part of R2 a trimpot, so I can vary between 20 - 26 Volts DC.
Q1/Q3 give further gain and provide current. The current through Q3 varies against load changes, so net current through R12 is always the same. In fact, you adjust the amount of the netcurrent through this resistor.
The point to see here is that Q3 shunts current to ground, the less is needed by the load, the more it eats.
Hope that's accurate,
Rüdiger

edit: if a circuit would not depend on transistor parameters, one wouldn't need them, right?

 

[Onvinyl]

I replaced R7 with a 10mA current source (only in sim for the moment). The output impedance drops 'a bit': flat till 20kHz at 0.2mOhms, then rising at 60mOhms at 1MHz.
Sim indicates that replacing Q1 with a 2SC1775 almost halves those numbers.
I guess that extending the bandwith (e.g. falt to 1Mhz) is not that easy?
Rüdiger

 

[Onvinyl]

oops, that were 2mOhms for B C550 and 0.5 mOhms for 2sc1775...

 

[aparatusonitus]

I guess that extending the bandwith (e.g. falt to 1Mhz) is not that easy?

You should try to increase current thru CFP shunt transistor (as much as thru the load and more, if needed), size a CCS accordingly.

 

[AndrewT]

Hi,
how do you make a low voltage drop CCS?

What is the Vdrop of the post1 CCS?

A 317 configured as a CCS drops 1.25V plus the Vdrop of the regulator. This can be around 2V to 4V, depending on the CCS current.
Would an LDO reg offer significant improvement in Vdrop without any other disadvantages?
What else could be used/built up?

Onvinyl,
are you averse to adding a couple of caps?
Maybe a bypass across R6 and also across the LED Vref.
Would these interfere with the sound quality?

Edit,
I think I've got the cap locations right this time

 

[Onvinyl]

Andrew, you're a genius!

Placing 100n across R6 (and 0.1R in series for simulating (a really bad) series resistance) lowers Zout to 150µOhms in the audio band and 5mOhms at 1Mhz. I filtered the led's with 1.9k/100n as well.

A different error amp (with cfp compound pair as well) simulates shocking 3µOhms in the audioband and 150u at Mhz. Not too bad for open loop...
This thing would be a pain to get stable, I fear.

Rüdiger

 

[Onvinyl]

... 100u for both is better for noise...
Rüdiger

 

[Onvinyl]

Since no-one is asking, I post it anyway...

this circuit has not been build yet! I expect it to be very nasty to calm down possible oscillations!

It sims with 9uOhms in the audioband and 250uOhms at 1Mhz.

Noise sim is 2.4nV/Hz(1/2) from 10Hz-1MHz. Did someone ask if I think these are realistic numbers?

Rüdiger

 

[Onvinyl]

Well, I breadboarded the complicated regulator and it's like I thought: It's a pain to get it working, but if you get it, you got it. ... As long as you don't think you tuned the oscillations beyond the capability of your scope ...
Seriously, I think one needs a good board with ground plane.
Did you ever make a stable cfp-input pair with a perf board?

 

[Onvinyl]

no comments? *sigh*

 

[AndrewT]

Did you ever make a stable cfp-input pair with a perf board?

Hi,
maybe the few that have p2p this have not read the thread.