From the op amp manufacturer's perspective I can tell you that most modern op amps have internal back-to-back diodes across the input terminals to protect the input devices.
While necessary, these diodes are notorious for producing unexpected behavior during transients at the op amp input or output. We see this issue a lot in analog circuits with multiplexed inputs. Switching between channels produces a step at the op amp input, causing a diode to become forward biased and increasing the input bias current for an instant. And, of course, they ruin your plan to use that 4th op amp in the package as a comparator.
Many JFET input op amps don't have these diodes. Off the top of my head I know that these don't:
OPA140, OPA141, OPA1641/2/4, OPA827.
The OPA192 is a CMOS op amp, but was designed specifically to remove these diodes. The applications section in the datasheet discusses this.
While necessary, these diodes are notorious for producing unexpected behavior during transients at the op amp input or output. We see this issue a lot in analog circuits with multiplexed inputs. Switching between channels produces a step at the op amp input, causing a diode to become forward biased and increasing the input bias current for an instant. And, of course, they ruin your plan to use that 4th op amp in the package as a comparator.
Many JFET input op amps don't have these diodes. Off the top of my head I know that these don't:
OPA140, OPA141, OPA1641/2/4, OPA827.
The OPA192 is a CMOS op amp, but was designed specifically to remove these diodes. The applications section in the datasheet discusses this.
Ok, you made me curious so I went through a few of those datasheets. The noise graph in the OPA140 one reminded me of this old NatSemi advert:
😀😀😀
That and also leakage currents - good luck finding some discrete diodes that would not screw up the offset / drift of FET / CMOS op amps. You would be looking for "pico ampere" diodes, a legacy item that costs orders of magnitude more than the op amps 😀
If you need those diodes there, then better get an op amp that already has them inside.
If the presence of them is highly detrimental (i.e. you want to buffer/distribute a voltage reference and don't want the transients in the loads to propagate back to the reference), then get an op amp that doesn't have the diodes.
A transistor converted to diode connection makes a very good pico ampere leakage protection diode.
Now the question becomes:
which small signal transistors in sot23 make the best low pica ampere transdiodes for input protection?
Now the question becomes:
which small signal transistors in sot23 make the best low pica ampere transdiodes for input protection?
Pity the engineers don't get to write the marketing copy any longer isn't it? That one must've been from Bob Widlar.
Now you're opening a completely different can of worms 🙂
For low leakage, you'll want to use the B-E junction, which unfortunately will not hold up to more than 5-6V inverse voltage....
Alternatively you can look at JFETs, there's a nice can of worms there too.... 😀
Didn't know that. Could be a useful little gem of knowledge. 🙂
Another difference between diode junctions and base-emitter junctions is better log conformity of the Vf / If relationship, which is advantageous in things like log antilog amplifiers.
Have a look at the Pass A-20 discrete power amp from 1977 (link). It's just a big discrete transistor opamp with a high power output stage. Pass installs the input diffamp's protection diodes, a different way. Marshall Leach copied this idiosyncrasy in his early power amps. Notice that Pass's method of connecting protection diodes would work just as well with N>1 diodes in series, and the series connection would reduce both leakage current and capacitance.
The protection diodes across the inputs of an opamp and also across the input of a differential input power amplifier normally only see a very few millivolts.
Leach Lo Tim discusses this and shows what happens when a very fast signal is applied.
In abnormal operation, a higher differential can exist.
In this abnormal situation we need to protect our input stage.
We do this with a pair of parallel connected diodes one inverted relative to the other. The maximum differential voltage across the diodes is Vf and this applies for both polarities. That is the purpose of the inverse parallel diodes.
Using transdiodes invokes the same maximum differential voltage. Vf for either polarity.
Zener breakdown does not enter the equation.
I have never heard of anyone suggesting that a jFET can be used/wired as a transdiode.
Where do you get your information?
Last edited:
W.Jung mentions it in a couple of his articles.
I have seen it mentioned by a few other authors.
W.Jung specifies a particular bjt in at least one of his articles, but as is usual, I can't remember which article, nor which transistor.
IIRC Analog Devices was making some transistors with very good log conformity (like MAT02). It was documented in the datasheets.
back to back LEDs will try to block voltage/current in both directions.
One needs parallel diodes to pass the overvoltage current.
One needs parallel LEDs to pass the overvoltage current.
The 2N4117A is often used as a very low leakage diode, G to (S+D). Leakage for the -A version is 1pA max. It is obsoleted by the original manu but LIS makes them again. Expect to pay $ 4 or so each though.
Jan
Jfet transdiode to protect opamp input
Thanks Jan.
Ti has a short paper on using a jfet that way.
ti.com/lit/an/sboa058/sboa058.pdf
Future electronics will sell us 2500 sst4117 @ 43cents each
Farnell will sell us 1000 sst4118 @ 98p each
Thanks Jan.
Ti has a short paper on using a jfet that way.
ti.com/lit/an/sboa058/sboa058.pdf
Future electronics will sell us 2500 sst4117 @ 43cents each
Farnell will sell us 1000 sst4118 @ 98p each
Last edited:
Yes but I don't need 1000 ;-)
Actually you'd want the -A version; the non-suffix version has 10x max leakage.
jan
What I meant to say was yes parallel leds each connected in a different polarity instead of back to back, it gives more headroom than diodes
I don't know - aren't LEDs quite leaky in reverse?
Edit: checked some data sheets from Knightsbridge, they spec reverse current of 10uA. That's 10,000,000 times higher than the 2N4117A... ;-)
I wouldn't put this across my opamp inputs unless its not high performance required.
I don't think headroom is an issue - in normal operation, the voltage difference between the + and - input of an opamp is just a couple of mV.
jan
Last edited:
Maybe a Member in the USA could order them and sell a dozen, or so, to each Member to cover his/her costs.
Do you think we could get 200Members together to fully reimburse the volunteer?
- Status
- Not open for further replies.