# Is there a simpler way to create a variable voltage drop than a differential amplifier?

I am working on a preregulator circuit design (it's derived from the one in LTC AN32), and am trying to design the test fixture for it. Two of the three components are clear:

• I can use an AC-output wall transformer to provide a safely isolated AC supply for testing
• A small optoisolated DC-output SSR (like the ones Ixys/Clare make for telecom work) can be used to add pulses wherever I wish with the aid of an isolated supply voltage (which can be obtained easily from a standard DC-output wall transformer). Pulse generation itself is also a solved problem.

However, I'm trying to generate the DC input voltages for testing purposes, and while generating a baseline is easy (I can build a variable reference around a TL431, a LM334, and a pot), how can I go about generating a voltage that is X volts below it, for user-variable (via a pot) values of X (I can accept a range for X from say 1 to 10 V), in a simple way? Output currents can be on the order of a couple of milliamps, as all this will be hooked to is an op-amp circuit's inputs, and the voltage accuracy doesn't need to be great either (if I can get within 5% or so at room temperature, I'll be fine). See the behavioral schematic below.

simulate this circuit – Schematic created using CircuitLab

The easiest way to handle this I can come up with uses an op-amp differential amplifier to subtract the output of a second variable reference from the first, but it seems to me like there should be a simpler way to do this...

• @next-hack I was actually thinking about that, but was not at all sure if it'd work -- guess I'll have to try it and report back :) Sep 5, 2017 at 22:53
• Do you want digital control, analog electrical control, or mechanical control of the voltage? Sep 5, 2017 at 23:07
• @ThePhoton -- edited in a note on that -- having both voltage and differential, if you will, controlled via pots is plenty adequate for my purposes Sep 5, 2017 at 23:33
• In ICs, I often see a BJT mirror current source that goes through a resistor to create a drop. That gets driven (the BJT collector doesn't care and you don't seem to be worried about the Early Effect.) It's a pretty simple arrangement. There are actually three different single-BJT things that come to mind. But I'm not fully visualizing where you are at right now. So I'm not sure what to suggest without seeing more of where you are and want to go (behavioral schematic?)
– jonk
Sep 5, 2017 at 23:45
• Would a negative linear regulator work? It'd be limited in the range of drops it could produce, but easy to make. Sep 5, 2017 at 23:52

The TL431 you mention can be used to create a variable positive or negative voltage from 2.5 to 37V up to 100mA limited of I*V=Pd drop rating of the package and temperature rise.

To make it a zener for negative voltages the series R goes on the cathode instead and anode is your fixed reference point while cathode is the output voltage biased by a more negative input series R.

Using a Triac primary control with optoisolation for pre-regulating DC output into a high power LDO was a basic method used by Lambda in the 70's in lab supplies. Their front end was much more sophisticated than I am suggesting in order to regulate heat, surge current and minimum Dc with ripple for the linear regulato.

It worked with better pre-regulated voltage to prevent needed higher no-load voltage rated capacitors and had oodles of LC filtering, but it offered a smaller package size per 100W than anyone else. Still nowhere near as good as a SMPS.

Ah, heck. I'll just draw the behavioral thing for you, below. Worst case you tell me all kinds of reasons why it won't help you. But either way, perhaps it helps you think of your own solution?

simulate this circuit – Schematic created using CircuitLab

I tacked a current mirror underneath your reference. I'm suggesting BCV61 here ($V_{BE}$ matching), though if you have more money you can also use BCM61 instead (both $V_{BE}$ and $\beta$ matching.)

If you are cheap or just want to hack it then you can just use PN2222A or 2N3904 devices and add emitter degeneration (to improve $V_{BE}$ matching) and a $\beta$-compensation resistor (to improve $\beta$ matching, of course.) Or play with measuring a few until you get a semi-match.

$Q_3$ just lightens the load on your OUTHI and hauls up and down on the emitter. (You could go "Darlington" on this, if you want to seriously lighten the load.) $R_1$ drops some fixed voltage based upon your resistance setting for it and the mirrored $10\:\textrm{mA}$, so the voltage at OUTLO will be one $V_{BE}$, plus this resistor drop, below OUTHI. If you want the range of say $10\:\textrm{V}$ to play with, then $R_1=1\:\textrm{k}\Omega$.

Be aware of where the ground reference is located. If that's annoying for some reason then it's not hard to move things around and add a part to this.

At least this gets one idea across about level shifting that isn't too involved and isn't difficult to follow. I also think you can play with the basic idea on your own to make it fit better if you want.