Optocoupler suggestions

What are some optocouplers you like that would be good for designs where they're used in unusual ways, like that "HV Opamp" thing JL likes where they're used as the "output stage" of an amplifier?

Are any characteristics that I should look for in the datasheet that makes a particular coupler a better performer in an application where linearity is more important than high-speed switching as in e.g. digital signaling? Not all datasheets seem to provide good charts of CTR over all the relevant independent variables.

Reply to
bitrex
Loading thread data ...

I don't know of any that are particularly well specified, except for the dual-photodiode ones, which are still a bit sloppy (enough that you need trimmers for a modestly precise analog coupler circuit).

CTR isn't reliable due to emitter and detector nonlinearity, and emitter aging, AFAIK. The nonlinearity is modest over current (+/- 25%?), but the aging is... something like max to min spec over lifetime (decades)? Unsure.

It's rare enough to find one specified for B-E resistance (for speeding up digital applications -- not much use for analog, AFAIK*).

*This uses the (photoactive) B-E junction as a summing node, which is interesting; speed (a few MHz) is about as fast as the opto can go, but the response still looks first-order, suggesting there is yet more speed to be won.
formatting link

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

I'm experimenting with using optocopulers as the gain-control element of a Wien bridge-style oscillator; the LED nonlinearity provides the corrective action for small deviations, the phototransistors feed a lil bias network/current source that controls the gain of an OTA in the main feedback loop for coarse corrective action vis a vis temperature and component variation etc.

In sim at least it produces lower distortion sines than diodes alone and tracks frequency changes faster than that JFET + integrator...thing...

Reply to
bitrex

You can buy 40 GHz photodiodes. Phototransistor optocouplers are slow because they use large-area transistors. Phototransistor couplers are a lot faster when used in photodiode mode, in a proper Hobbsonian circuit.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

e
d
r

but

d

How low a distortion do you need? A gain stage like this, just clips the tops off.

+->|---+ +-|
Reply to
George Herold

Eehhh...i've used it before, it's a very fiddly circuit. Usually shown with trimpots in the feedback loop which is ugly and problematic.

My goal with this thing is to have the positive feedback loop gain auto-level itself into a slightly over-unity state where the diode clipper in the negative loop can work most effectively, without grossly murdering the peaks due to an out-of-whack trimpot or out-of-tolerance resistor. It's not too hard to set up if the diodes are in optocouplers driving photodiodes/phototransistors. If the phototransistor/photodiode is pushing large currents on the peaks that means the positive feedback loop gain is too high, filter and mirror it around into an OTA in the positive loop and turn down the gain. Vice versa if they're not putting out any current at all.

It should respond faster than the JFET + integrator gain control thing because the OTA gain is being controlled by charge balancing the current coming out of the photodiodes into a small cap with a current source pulling current out to ground, you don't have to wait for a DC voltage integrator with a large cap in the integrator feedback loop to wind up and wind down.

Reply to
bitrex

Here's the general idea:

Reply to
bitrex

the

eed

ter

), but

)?

ng

ut

eed

of

l

in

nd

I use the above in production, 1% R's.. no trimming.

Hey I also made one with a light bulb..the bulb costs ~$1. (I think) Way back when I had issues with it motorboating sometimes. Turned out I needed better caps in the R/C part of the circuit. (The light bulb transient response is the pits. :^)

I assume you read J. William's long piece on the subject. I tried the Jfet thing, but could never get it to balance nicely.

George H.

Reply to
George Herold

the

eed

ter

), but

)?

ng

ut

eed

of

l

in

nd

Hmm OK, I'm mostly confused. I've never used an OTA...

George H.

Reply to
George Herold

On the input side OTAs like e.g. the LM13700 are just a differential pair with the "other side" of the current mirror feeding the tail brought out to a pin so you can put a varying current in to control the gain of the pair.

The collectors are connected to a structure of current mirrors arranged such that when one side of the diff pair is drawing more current than the other the IC output acts as a current source that "blows" the difference current out, vice versa the output acts as a current sink that "sucks" it in.

It's an V-I converter with an adjustable transconductance basically. The V-I curve is only approximately linear in the area where the tanh function is approximately linear so if you want a linear current response to control voltage you need to divide large signals down before putting them onto the inputs.

The response of the tanh function itself is a pretty good way to massage a trinagle wave into a sine, too, though that's not what I'm doing here

Reply to
bitrex

Older ones fail because the cruddy filler goes opaque over time when you drive the LED too hard, which you usually have to do because of the poor CTR of even a phototransistor. IIRC that happens faster than LED degradation.

Shining a bright LED on a reasonably-sized photodiode can get you 10% CTR and will easily go as fast as the LED can manage. (*) This isn't that fast, a few megahertz on a good day for a display LED. You can speed it up a bit by reverse biasing, but recovery isn't that fast even so.

LEDs are much slower than diode lasers because they're limited by spontaneous recombination, whereas in a laser stimulated emission causes much faster depletion of the upper state when power is removed. UV LEDs are a lot faster than visible or IR ones, because the transition rate wants to go like 1/lambda**2 (Fermi's golden rule, for physics fans).

High speed IR LEDs are specified at super high current densities, where high-level injection causes the carrier lifetime to drop by a lot--they're not nearly as fast at reasonable current densities.

There are LEDs intended for optical communications that go as high as 1 GHz because their minority carrier lifetime is intentionally trashed. IIUC most of the additional recombination is nonradiative, which hurts their efficiency.

Cheers

Phil Hobbs

(*) I did this with JL and Co. some years back in a nanoamp photoreceiver, which their marketing department called "a unique photon-coupled architecture." ;) It used photocurrent feedback in a TIA, which isn't usually worthwhile because you get 3 dB worse shot noise and worse capacitance.

The new wrinkle in that one was using two photodiodes _in series_, with feedback applied to keep them from fighting each other. Somewhat amusingly, that gives you half the capacitance and _half the shot noise power_. The noise penalty from the photocurrent feedback was thus only

10 log(1.5) = 1.6 dB rather than 3 dB, so it really was shot noise limited. It was a pain to build, by all accounts, and didn't sell well, so they discontinued it. However, the QL01 introduced by HEO last year is cheaper and quieter anyway. :)
--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

I briefly played with some Cree white LEDs. I didn't use a super fast detector, but the combination seemed to be in the 7 ns ballpark, which surprised me because I'd assumed that the phosphor would be slow.

That was a technical success but not successful commercially. The market seemed to be one-off researcher types, not volume users, and it's hard to make money off researchers. Let them buy PMTs. We did learn a lot, which is always useful.

We don't, for some reason, seem to have luck selling free-space detector stuff. We do sell lots of fiber things and fast diode laser drivers. The ideal customer is an optics-heavy OEM founded by physicists that doesn't have a big internal EE staff.

In an electronics-centric company, engineers are the superheroes. In a company founded by scientists, engineers are 3rd class citizens, and the best ones tend to move on.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

e
d
r

but

o.

huh, OK help me here. I thought spontaneous emission went as the frequency^3, 1/lamda^3

formatting link
sion

George H.

Reply to
George Herold

he

ed

er

, but

?

so.

Maybe Thor labs could resell/market it for you?

George H.

Reply to
George Herold

White LEDs are actually pretty fast on account of the blue LED and the inorganic fluor, which I think is some salt of europium. Wideband fluors are pretty fast in general.

Well round here all we have are physicists. ;)

We're in the process of licensing a customized APD version of the QL01 to a large biomed company, assuming that nothing goes south before the contract is signed.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Hmm, I seem to be missing a factor of 1/lambda, but I don't know where it comes from.

I was thinking of Fermi's golden rule, which says that the transition rate is

Gamma = |

Reply to
Phil Hobbs

Sounds like Phil has a better and cheaper design now. I'll leave the free-space stuff to him.

We did have a product in the Thorlabs catalog once. They want a huge markup.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

OK good, I looked at Fermi's golden rule, and I thought fine. The square of the matrix element, which is some energy, so energy squared... But then I remembered this nmr talk which started by reminding us of the cube dependence.. which if you scale down to ~10 MHz nmr frequencies... is like forever, proton magnetic moments (and such) in low magnetic fields hardly ever do a spontaneous emission, it's all the environment. T1 in nmr speak.

George H.

Reply to
George Herold

r the

need

itter

%?), but

es)?

you

oor

en so.

ses

EDs

re

1

s

Factor of 2-3? TBH I look at Thor labs and then figure I can find someone to sell me the gizmo for 1/2 to 1/3.. (if I buy a bunch) at least that's my limited experience. I think I'm soon going to need a 405 nm laser diode, last time I looked Thor labs wants ~$100 for one, so $30-40 from sanyo, maybe less?

George H.

Reply to
George Herold

for the

ou need

emitter

25%?), but

ades)?

n you

poor

0%

't

n

even so.

auses

LEDs

te

).

t

ch

here

as 1

d.

rts

a

Oh we buy mirror mounts and other optical stuff from Thor labs by the bucket full....

GH

Reply to
George Herold

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.