Problem with Ideal Diodes

françois · Posted: Tue Dec 16, 2008 10:59 am Post subject: Problem with Ideal Diodes

Hello all,

I thought I understood thoroughly Jim Patchell's "Ideal Diode Tutorial", and in fact I've been using ideal diodes in many places. But always with a zero volt threshold.

Now I need something a bit more elaborate. I have a bipolar control signal (typ. +/-10V, allowed to swing to the +/-15V rails). But I can only guarantee accurate operation of the controlled circuit when the control signal is between +/-5V.

So I need a three-segment piecewise linear response curve : Vout = 0V when Vin in +/-5V, Vout = 5V - Vin when Vin below -5V, Vout = Vin - 5V when Vin above + 5V (I wouldn't care if the signs were reversed). Seemingly simple, no? This should even be obvious according to Jim's paper.

But I just can't get it to work, although I tried many combinations, sometimes after careful thought, sometimes just out of anger and boredom.

Can anyone tell me what I'm getting wrong?

Thank you,

-- françois

JovianPyx · Posted: Tue Dec 16, 2008 2:29 pm Post subject:

A schematic of your current circuit would be immensely helpful to this end.

françois · Posted: Wed Dec 17, 2008 4:26 am Post subject:

Hello and thank you, Scott!

I couldn't join a schematic yesterday because I had been testing so many configurations... (moreover there was a typo in my post). After much thought I came to a working circuit, but it required 7 (!) opamps. Obviously there was something wrong. Tracking every signal from every opamp I managed to get to an almost satisfying circuit with only 3 opamps (see attached schematic).

Opamp U1 extracts the negative (V1) and positive (V2) parts of the input signal Ve (and of course inverts them). Then U2 computes Vx = V2 - V1 + 5V, which is an inverted full-wave rectified version of Ve, shifted up by +5V -- then Vx is positive only when Ve is between -5V and +5V. Finally U3 cuts off this positive part of Vx and inverts the negative ones, thus achieving the desired effect.

Now what is the purpose? I could certainly just clamp Ve to -5V/+5V with just two diodes (provided the diode voltage drop is taken into account). But it is sometimes interesting to allow some overdrive in the control signal, even if it is at the cost of accuracy. If I observe some abnormal behaviour I do not want to track which modules (there are many of them) are being overdriven. Just a LED indicating this circumstance will be of great help. Of course, I could use a window comparator that goes high when Ve exceeds +/-5V, but it would be nice also to have an indication of how much overdrive is applied. In fact I don't know if this is actually useful, but it is worth testing. And anyway, that was a good exercise.

What worries me a bit is, why 3 opamps? Couldn't it be done with just two? I tried to incorporate the last diode in U2's feedback path, to no avail. I think I have an idea why it doesn't work (at least not in the obvious manner) but I have to keep testing. Anyone who could help me getting rid of this "extra" opamp will be blessed!

Yours,

-- françois

EDIT: Oops. I forgot half of the output stage (U3). Attached is a correct version. I also forgot to mention that all "anonymous" resistors are equal, of any reasonable value (I used 10k).

Visio-Ideal Diode Clamp.pdf

Description:

Download

Filename:

Visio-Ideal Diode Clamp.pdf

Filesize:

12.46 KB

Downloaded:

366 Time(s)

JovianPyx · Posted: Mon Dec 22, 2008 12:03 pm Post subject:

Well, while reducing to 2 opamps would be great, if your 3 opamp design works well and does what you want, then I would leave it. Opamps are pretty inexpensive, so one more or less ...

I can't see a way to combine functions to reduce it to 2 opamps, but then doing those mental gymnastics isn't my forte. Perhaps someone else?

blue hell · Posted: Mon Dec 22, 2008 12:33 pm Post subject:

I'd expect :

Posted Image, might have been reduced in size. Click Image to view fullscreen.

but probably you tried that variation ... Laughing

BananaPlug · Posted: Mon Dec 22, 2008 12:54 pm Post subject:

françois · Posted: Tue Dec 23, 2008 3:45 am Post subject:

@Blue Hell: Sorry for the PDF, it is just the "natural" output format for Microsoft Visio... but you need a PDF viewer (try Foxit Reader, freeware). Attached is a JPEG file, hopefullly easier to handle. I don't know how to include an image in the message body, though.
EDIT: OK, the PDF was not the point. I just missed your image the first time. I did try this variation, but it doesn't work any better... Sad

@BananaPlug : It is not just a scaling issue. Have a look at the response curve I need. There is some scaling involved, of course, but also some thresholding, and I can't see how to do that without diodes. But full rectification is probably too much.

@ScottG: Well, opamps usually come in packs of four, so I usually have one or two unused. And, while not being a millionnaire, I still can afford a dedicated TL084 if needed. But I love these gymnastics, trying to use as few components as possible (or even as impossible in some cases).

Ideal Diode Clamp.jpg

Description:

Filesize:

134.21 KB

Viewed:

370 Time(s)

This image has been reduced to fit the page. Click on it to enlarge.

BananaPlug · Posted: Tue Dec 23, 2008 9:31 am Post subject:

françois · Posted: Tue Dec 23, 2008 12:28 pm Post subject:

BananaPlug · Posted: Tue Dec 23, 2008 1:15 pm Post subject:

Actually, your non-native English is great but I read what I expected to see rather than what you actually wrote. Wink