# Exponentiator: Creating lower limit for op amp output and other problems

Discussion in 'The Projects Forum' started by JingleJoe, Sep 5, 2013.

1. ### JingleJoe Thread Starter Member

Jul 23, 2011
185
10
Okay I know the drill here, you're going to ask me for lots of circuits and facts and figures So here you go:

• Supply: single rail 12V
• Circuit: (see attachments)
• Problem: random oscillation and lower limit too low.
• Input: from a pot wired as voltage divider connected to same power rails and a triangle wave generator on the same power rails, my problems still occur when niether are used.
• Output: just goes to the o'scope via ground sheilded cable.

So I've got a little circuit working great, it's a diode based exponentiator of my own design, though probably invented before by someone else.

Explanation of the circuit:

D1 and the feedback diode are the components responsible for making the input signal exponential but the circuit is linear below the forward voltage of D1. To get it to operate in the exponential region I simply bias the input up to the forward voltage (Vf) of D1. However, the input is variable and adding or taking away input voltages results in changing any additional offset added to the input voltage divider.
My solution to that was to have the bottom resistor of the input voltage divider connected to the same kind of diode used for D1, biased into conducting, thus offsetting the input by a few mV below the forward voltage of D1 & D2.
It varies by about 100mV as I connect/disconnect the inputs to ground, but that had little effect on the functionality of the circuit; It was still biased into the exponential region enough for good operation.
This means the output is also biased up by about 0.6V (Vf) however this is desireable for my circuit, I acctually want the output biased up by 1.25 V.
To acheive this I simply replaced D1 and D2 with IR LEDs with a Vf of 1.2V, however that's where my problems began... Also I would rather avoid using 2 or 3 diodes in series, I'm keeping the parts count down.

Now, the problems:
I've got it working really well with ordinary diodes but when I try to use other diodes with a higher forward voltage, it oscillates when the input voltage is at, or just below, the diode forward voltage!

I've isolated the oscillation to the diode/resistor feedback network of the op amp. It still occurs with input straight from a potentiometer voltage divider, not the input circuit in the diagram.

One solution to this is a 100pF cap in the feeedback of the op amp, but I want to avoid that because I feel like that's the equivalent of putting a sticking plaster over a massive wound. It solves it but only in that it is covering it up, take the cap away and the problem still exists. Besides, I want to keep my parts count down.

I have no idea why it's oscillating, I assume it's the diode going in and out of it's conducting state because it's right on the cusp of the "avalanche voltage" as they call it. However, why does it work well with a 1N914 diode and poorly with an LED? If I can get some answer to that part of the riddle then I can properly solve the problem at the root cause.

I've tried reducing the value of R3 but that only works to a small degree: as I connect more inputs the voltage at the non-inverting input lowers and the exponentiaor oscillates. Exchanging R3 for a much lower value results in too high a voltage accross D2, which is strange I would have thought it would stay at the diode forward voltage? I suppose it's got something to do with the input resistor network it's connected to. The voltage accross D2 also changes too greatly with a low value of R3.

so, what can I do to overcome this oscillation and/or keep my minimum output voltage at approx 1.25V?

A solution I came up with to the low biasing with ordinary diodes was to bias the output up with a pull up resistor (like a voltage divider referenced to +VS, not ground) that works really well and I can tune it to whatever I want, however it also squashes the exponential signal a bit and requires an additional buffer- I'm trying to keep parts down and my op amps down to one op amp.

• ###### exponentiator.png
File size:
94 KB
Views:
47
Last edited: Sep 5, 2013
2. ### wayneh Expert

Sep 9, 2010
13,435
4,272
Bypass cap on the power supply lines to the amp? Your circuit feeds supply voltage to the amp, so any noise in the power affects the result.

Just a wild guess.

3. ### JingleJoe Thread Starter Member

Jul 23, 2011
185
10
I'm way ahead of you, 2200uF ahead of you but the oscillation I'm seeing is ≈ 1 volt p-p. The main distortion on my power supply is significantly less and I know what it looks like

4. ### wayneh Expert

Sep 9, 2010
13,435
4,272
I know you want to keep the parts count down, but for the sake of sorting this out, have you tried 2 or 3 regular diodes in series instead of the LEDs? I'm just thinking the LEDs have a wacky behavior we're not expecting. If the circuit oscillates with regular diodes, then the problem must be with the bias voltage itself, not how it gets there.

5. ### JingleJoe Thread Starter Member

Jul 23, 2011
185
10
I'll try that now and get back to you with some results

I was hoping someone would give me an alternate solution to limiting the minimum output voltage and/or biasing the input to approx 1.2V aswell. I know my method is a bit ramshackle and bodged but it works for regular diodes.

by the way I tried a buffered D2 & R3 for biasing the input up but had almost exactly the same results as the above circuit.

6. ### JingleJoe Thread Starter Member

Jul 23, 2011
185
10
With two diodes in place of D1 and D2 the oscillation is considerably less and occurs at a much lower voltage. This is good and would be workable in practice however the input biasing gets too low when additonal inputs are connected. (grounded inputs, the idea bieng that the output stays at about 1.25V with 0 input voltage) it's barely 1.2 with nothing connected and as additional inputs are connected it drags the bias voltage down to 1 volt. 100mV below Vf (1.15V) is really the lowest I can work with.

Lowering the value of R2 brings the voltage up a bit, however, not enough for my applications and this also lowers the input range; I can't max out the exponentiator with two inputs connected. Previously I could hit the upper rail with about 6 volts, which is what I'd prefer. I need to keep the input resistors >100k for impedence reasons as the output impedence of some of the circuits this will be receiving it's input from, can be about 10k. Do I need to go into more detail regarding that?

Apr 5, 2008
16,967
2,961
Hello,

What are you trying to do?
You might want to have a look at the attached PDF on log / anti log amplifiers.

Bertus

File size:
25.5 KB
Views:
35
8. ### JingleJoe Thread Starter Member

Jul 23, 2011
185
10
In simple terms I'm trying to turn a linear voltage into an exponential voltage which is biased up about 1.25V from ground. I was thinking the easiest way to do that would be to use a diode in my circuit with a Vf of 1.25V and bias the input up to that.

Now I'm considering all kinds of weird things, like making a little circuit to make a -0.6V rail to which the circuit it is driving can be referenced, this will lower the required offset by 0.6V allowing me to use regular diodes (which worked okay). But that'll use so many more components!

There must be a simple way to bias up the input or limit the minimum output voltage of the exponentiator op amp, right?

P.S. I've seen those circuits in the pdf before but they seem to use more op amps than I'd like and even in the first part of the description it says "the transistor is basically acting like a diode" which makes me think, why bother with transistors? why not use diodes?
My circuit exponentiates really well, it functions fine I just need to bias things right really and I could have done that if the LEDs didn't want to oscillate!

9. ### joeyd999 AAC Fanatic!

Jun 6, 2011
2,964
3,788
It's all a matter of desired accuracy and temperature compensation.

10. ### crutschow Expert

Mar 14, 2008
16,206
4,330
There is a reason to use transistors instead of diodes. Transistors, when used as diodes (collector connected to base), generally have a more ideal exponential behavior with less variation between units, then diodes.

11. ### JingleJoe Thread Starter Member

Jul 23, 2011
185
10
This isn't for anything super critial so don't worry about accuraccy- that's not part of the specification here.
What about that biasing? Any suggestions for ways to raise up the minimum/low output of an op amp?

I had a thought last night; my previous set up was: diode exponentiator > biasing pull-up > buffer, which used two op amps (one for the buffer and one for the expo). However, I could buffer the initial input summing network, then put in some biasing with simple resistor voltage divider stuff after the buffer which means additional inputs won't mess up the biasing offset!
the downside is that this does use another op amp but I'm considering common collector transistor buffers... but perhaps this invalidates my exponentiator, maybe a transistor variant doesn't need that offset voltage to get it out of the linear range? I suspect not but maybe that's why I've never seen anyone using diodes in an exponentiator

12. ### JingleJoe Thread Starter Member

Jul 23, 2011
185
10
just a little update:
the idea posted above works great, my circuit is basically:

input summing > transistor buffer > biasing to Vf of expo diode > exponentiator.

the downside is that the buffer transistor requires an offset voltage of 0.6 to start and additional/fewer inputs to the input summing resistor network change this offset... I'm going to fiddle with the resistor values and biasing a bit more before I write this one off as unsuitable, for now it's a solution which saves an op amp which is what I wanted really.

Aug 21, 2008
3,820
1,038
14. ### JingleJoe Thread Starter Member

Jul 23, 2011
185
10
That link was really informative, thankyou
However I didn't see much about phase shift... maybe I missed something but it does seem like it could have been the cause of my oscillation, however, there was no amplification in my feedback loop, as is suggested is required for unwanted ringing in the linked article.

Also I'm still mystified about biasing the input and by the way,
my original biasing, with the conducting diode, had temperature compensation! as the temperature of the feedback diode rose, voltage went up but as the biasing diode temperature increased, voltage went down! So they canceled each other out (maybe only to a small degree, one may have more effect than the other).

What I'm thinking of now for a more complex and accurate circuit with all op amps is in 3 parts:
CV buffer
biasing diode buffer
exponentiator

The cv buffer negative or ground rail is referenced to the output of the biasing diode buffer. That way it should allways be biased up by the biasing diode without a lot of effect from additional inputs.
You may want to also reference the input summing stage to the biasing diode buffer.... acctually didn't I mention this already? whatever it's an idea which could work if I could spare more op amps.