Speeding optoisolators

You are correct that it's VERY unlikely to have a problem. But, in my world, I NEVER take chances.

...Jim Thompson

Then Microchip would keep you busy for a while.

Well I'm tempted to draw a fuse in there, but I guess I have no choice but to bow to the sage. By popular demand, just in case 20:1 isn't enough margin:

--' '-- | \\ / ---> | | / 4v p-p | -+- |>. | \\ R23 | | | | / 4.7k '---|----------|-' | - >-------------' | | === === GND GND

Cheers, James Arthur

Classic Larkin: manhandling electrons. He's a classy brute.

Cheers, James Arthur

If my push-pull opto thing is done right, it will never smoke. And if you're really paranoid, change the LED driver circuit. It ain't rocket science.

I've used total-pole optos in linear mode too, as the output stage of a 400 volt p-p swing opamp; I posted that a while back.

Of course you do. It's just a matter of how many sigmas you're willing to pay for.

We often take chances, particularly in the performance-critical heart of a product. If a phemt is rated at 9 volts max, but tests ok to 25, we'll use it at 12. And often we test parts for behaviors that just aren't specified... we often know stuff about parts that the makers don't.

Extreme performance sometimes requires taking calculated risks. But there's no point in taking significant chances on the routine parts of a product.

John

Even simpler: don't float the input.

If the input comes from a CMOS gate, and it's floating, it's broke.

John

"Brute"?? Heck, I've met the man in person. He's really cute. ;-) ;-) ;-)

Cheers! Rich

I actually sat at a bench and measured some opto's CTR, I think one of the 4N series; anyway, I made a graph from about 0-20 mA, and was astonished when my graph turned out to be a straight line within a few percent. At a previous job, when they needed a linear transfer function, they used either TWO optoisolators, one for reference to linearize it, or they used a V-F converter, opto, and F-V converter.

I had intended to use it in a feedback loop for a design we were working on, but that project died on the vine - we couldn't afford it. I almost bankrupted the guy. =:-O

Thanks, Rich

There are optos where one LED lights two detectors (e.g. Infineon's IL388). By closing a loop on one detector (on the isolated side) you assure linearity. The 2nd, matched detector provides the signal on the signal side.

Pretty good for phone couplers, but not much in the way of absolute accuracy.

Cheers, James Arthur

The led tempcos are rotten, and they degrade over time (integrated current) too.

John

Yep, so the one-LED-lighting-two-detectors strategy works well to fix that (aging, linearity, and LED drift). Matching the detectors, then, is the weak link. Works well for audio though--you often see the series I mentioned replacing transformers in modems.

I guess you could calibrate everything out--if you had a multiplexer and a reference on the isolated side--by cycling between signal, zero, and Vref, but with that much hardware you might as well digitize and be done with it.

James

View in a fixed-width font such as Courier.

. . . . +5V +5v . FIGURE 3. -+- -+- . ========== | | . \\ | . R22 / | . 10k \\ .' . R21 | |< Q21 . 1.5k 1/2 +--------| 2n3906 . + >----/\\/\\/----. MOCD217 | |\\ . | OPTO | \\ . .--. .---|---------/--. +------>

. | | | --- |/ | | . --' '-- | \\ / ---> | | / . 4v p-p | -+- |>. | \\ R23 . | | | | / 3,9k . '---|----------|-' | . - >-------------' | | . '-----------+ . | . / . \\ R33 . / 1k0 . | . === . GND . .

At first glance I thought that might work, but I think the negative feedback will add Miller thru the opto NPN and keep it slow.

...Jim Thompson

Won't that slow it down?

John

I don't have the MOCD217, but SPICE simulation on the MOC223 shows tr/tf as about equal at 550ns. It easily handles 100KHz and he can back that input drive off to 0.5mA or so, 2mA is way too much, and there is very little danger of exceeding 100% CTR at that drive level. Yep, so NFB does make the transitions slower.

What are you asking me for?

Inspired by Fred's use of simulation, I plugged an MOC207 into LTspice and found it surprisingly faithful in modelling my original Fig. 2 circuit.

So inspired, I explored some more.

Attractive as John's totem-pole idea was, I couldn't adjust things to jerk the opto output transistors 'off.' Maybe I missed something, but once they're saturated, they're stubborn.

Fred's idea of reducing LED current is a good one; I'd originally set the LED current a bit high for CTR and aging allowance, but I was too conservative. Lower LED current is better.

Here's a faster two-transistor version that minimizes the voltage swing across the opto's output transistor, and thus manages to pass

FIGURE 4. ========== +5v -+- R41 | 2.7k | + >----/\\/\\/----. MOC207 +---------. | OPTO | | .--. .---|---------/--. | | | | --- |/ | |

--' '-- | \\ / ---> | | \\ R45 4v p-p | -+- |>. | / 4.7k | | | | \\ '---|----------|-' | --. .-- - >-------------' | | | | | | '--' R42 | | tf=90nS, tr=125nS 4.7k |

Wonder what might be done with those devices where the opto-transistor base is brought out?

...Jim Thompson

How would these circuits handle 3.0V power rails ??

What changes would need to be made ??

Thanks

don