Discussion:
Optocoupler datasheets
(too old to reply)
bitrex
2024-05-29 15:15:17 UTC
Permalink
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.

A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?

The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????

Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Miguel Gimenez
2024-05-29 15:21:42 UTC
Permalink
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
CTR reduces with age, so you must leave some margin.
--
Regards
Miguel Gimenez
Don Y
2024-05-29 15:34:33 UTC
Permalink
Any suggestions for how to approach methodically/mathematically selecting drive
current would be appreciated, thank you! ("Don't bother" a valid option)
One thing you can try (depending on what's on the other side of the device
and whether or not you want linear operation) is to drive it with a pulsed,
higher current source and filter on the isolated side. (assuming total power
is what's driving your design)
bitrex
2024-05-29 15:53:29 UTC
Permalink
Post by Don Y
Post by bitrex
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't
bother" a valid option)
One thing you can try (depending on what's on the other side of the device
and whether or not you want linear operation) is to drive it with a pulsed,
higher current source and filter on the isolated side.  (assuming total
power
is what's driving your design)
Right, that's a good idea.

Situations where they're just left on for DC most of the time and trying
to get minimal forward current by optimizing a resistor value seem
hopeless.

Could wrap a DC feedback loop around it which would only make sense in
the situations where you have a lot more power available on the
secondary side for some reason, hard to think of a use case.

Logic-output couplers seem somewhat better specified in the sense they
explicit state the relatively high minimum forward currents or all bets
are off.
Don Y
2024-05-29 16:26:15 UTC
Permalink
Situations where they're just left on for DC most of the time and trying to get
minimal forward current by optimizing a resistor value seem hopeless.
Exactly. You can "play games" with how hard you "overdrive" the emitter as
well as add current gain on the isolated side to store enough charge to carry
over long enough for the longest off period (duty cycle and frequency related)
that you want to tolerate.

We use this technique for driving outputs that need to be self-reseting
in the event that the processor "dies" (you have to make sure you design
your code so the processor can't crash in a loop that keeps pulsing the
output(s) -- I've seen more than one fool make that mistake!)
Could wrap a DC feedback loop around it which would only make sense in the
situations where you have a lot more power available on the secondary side for
some reason, hard to think of a use case.
<frown> Unless you're looking for a linear transfer function, feedback
(except as a diagnostic) is usually not appropriate.
Logic-output couplers seem somewhat better specified in the sense they explicit
state the relatively high minimum forward currents or all bets are off.
john larkin
2024-05-29 15:53:28 UTC
Permalink
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Keeps the amateurs away.
Post by bitrex
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
For on/off, use a logic coupler or an opto SSR. They have well defined
input requirements.

A dual transistor-type optocoupler makes a fun totem-pole logic
coupler, with zero static power dissipation on the isolated side. It
basically always swings rail-rail out and gets faster as LED side
current goes up. Or spike that to save driver power.
bitrex
2024-05-29 22:48:56 UTC
Permalink
Post by john larkin
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Keeps the amateurs away.
Post by bitrex
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
For on/off, use a logic coupler or an opto SSR. They have well defined
input requirements.
A dual transistor-type optocoupler makes a fun totem-pole logic
coupler, with zero static power dissipation on the isolated side. It
basically always swings rail-rail out and gets faster as LED side
current goes up. Or spike that to save driver power.
SMT optocopulers don't cost much more than discrete transistors, so they
can be useful sometimes for level shifting, current flipping, reference
shifting, and other situations that don't strictly require isolation.

It would be cool if there were some cheap implementation of an abstract
element you might call something like a "current flipper", with a
predictable CTR. Current goes through one side and comes out flipped
and/or scaled on the other.

The current mirror is sort of like this but it tends to work best placed
referred directly to ground or the supply, not "in between" stuff. Also
it usually tends to need too many parts to do anything useful, to make
it regularly worth using with discretes.
piglet
2024-05-29 15:56:18 UTC
Permalink
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don’t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
--
piglet
Phil Hobbs
2024-05-29 16:39:29 UTC
Permalink
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don’t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.

Cheers

Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com
piglet
2024-05-30 08:01:51 UTC
Permalink
Post by Phil Hobbs
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don’t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.
Cheers
Phil Hobbs
Sorry, maybe my language was sloppy. I meant keep phototransistor
collector from bottoming and reduce C-B miller effect. Not necessarily
by rationing photons. Keeping Vce constant by feeding straight into a
transistor base is brutally effective. See the post about halfway down here:

<https://electronics.stackexchange.com/questions/136928/under-what-conditions-does-an-optocoupler-work-fastest>

piglet
Phil Hobbs
2024-05-30 11:29:18 UTC
Permalink
Post by piglet
Post by Phil Hobbs
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don’t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.
Cheers
Phil Hobbs
Sorry, maybe my language was sloppy. I meant keep phototransistor
collector from bottoming and reduce C-B miller effect. Not necessarily
by rationing photons. Keeping Vce constant by feeding straight into a
<https://electronics.stackexchange.com/questions/136928/under-what-conditions-does-an-optocoupler-work-fastest>
piglet
If you have the base pinned out, you can do more stuff, true. But at the
end of the day you’re still dealing with a phototransistor.

BITD TI and HP made optos with actual specs, but these days, not so much.

Linear mode works great when there’s overall feedback, as in your typical
offline switcher, which has a TL431 to do the actual regulating.

Cheers

Phil Hobbs
--
Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC /
Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics
john larkin
2024-05-30 13:37:32 UTC
Permalink
On Thu, 30 May 2024 11:29:18 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by piglet
Post by Phil Hobbs
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don’t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.
Cheers
Phil Hobbs
Sorry, maybe my language was sloppy. I meant keep phototransistor
collector from bottoming and reduce C-B miller effect. Not necessarily
by rationing photons. Keeping Vce constant by feeding straight into a
<https://electronics.stackexchange.com/questions/136928/under-what-conditions-does-an-optocoupler-work-fastest>
piglet
If you have the base pinned out, you can do more stuff, true. But at the
end of the day you’re still dealing with a phototransistor.
BITD TI and HP made optos with actual specs, but these days, not so much.
Linear mode works great when there’s overall feedback, as in your typical
offline switcher, which has a TL431 to do the actual regulating.
Cheers
Phil Hobbs
A c-b schottky clamp would help, sort of a 74LS photocoupler.

But the really good logic couplers these days aren't optical.
Phil Hobbs
2024-05-30 18:58:36 UTC
Permalink
Post by john larkin
On Thu, 30 May 2024 11:29:18 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by piglet
Post by Phil Hobbs
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don’t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.
Cheers
Phil Hobbs
Sorry, maybe my language was sloppy. I meant keep phototransistor
collector from bottoming and reduce C-B miller effect. Not necessarily
by rationing photons. Keeping Vce constant by feeding straight into a
<https://electronics.stackexchange.com/questions/136928/under-what-conditions-does-an-optocoupler-work-fastest>
piglet
If you have the base pinned out, you can do more stuff, true. But at the
end of the day you’re still dealing with a phototransistor.
BITD TI and HP made optos with actual specs, but these days, not so much.
Linear mode works great when there’s overall feedback, as in your typical
offline switcher, which has a TL431 to do the actual regulating.
Cheers
Phil Hobbs
A c-b schottky clamp would help, sort of a 74LS photocoupler.
But the really good logic couplers these days aren't optical.
Yup. Even with a better photoreceiver, most of the usual speedup tricks
don't work with LEDs, on account of their diffusion-dominated carrier
dynamics.

Cheers

Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com
john larkin
2024-05-30 21:52:39 UTC
Permalink
On Thu, 30 May 2024 14:58:36 -0400, Phil Hobbs
Post by Phil Hobbs
Post by john larkin
On Thu, 30 May 2024 11:29:18 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by piglet
Post by Phil Hobbs
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don’t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.
Cheers
Phil Hobbs
Sorry, maybe my language was sloppy. I meant keep phototransistor
collector from bottoming and reduce C-B miller effect. Not necessarily
by rationing photons. Keeping Vce constant by feeding straight into a
<https://electronics.stackexchange.com/questions/136928/under-what-conditions-does-an-optocoupler-work-fastest>
piglet
If you have the base pinned out, you can do more stuff, true. But at the
end of the day you’re still dealing with a phototransistor.
BITD TI and HP made optos with actual specs, but these days, not so much.
Linear mode works great when there’s overall feedback, as in your typical
offline switcher, which has a TL431 to do the actual regulating.
Cheers
Phil Hobbs
A c-b schottky clamp would help, sort of a 74LS photocoupler.
But the really good logic couplers these days aren't optical.
Yup. Even with a better photoreceiver, most of the usual speedup tricks
don't work with LEDs, on account of their diffusion-dominated carrier
dynamics.
Cheers
Phil Hobbs
I did test a Cree white LED for speed. It hit my detector response of
about 7 ns, phosphor included. I was surprised.
Phil Hobbs
2024-05-31 00:12:30 UTC
Permalink
Post by john larkin
On Thu, 30 May 2024 14:58:36 -0400, Phil Hobbs
Post by Phil Hobbs
Post by john larkin
On Thu, 30 May 2024 11:29:18 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by piglet
Post by Phil Hobbs
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don’t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.
Cheers
Phil Hobbs
Sorry, maybe my language was sloppy. I meant keep phototransistor
collector from bottoming and reduce C-B miller effect. Not necessarily
by rationing photons. Keeping Vce constant by feeding straight into a
<https://electronics.stackexchange.com/questions/136928/under-what-conditions-does-an-optocoupler-work-fastest>
piglet
If you have the base pinned out, you can do more stuff, true. But at the
end of the day you’re still dealing with a phototransistor.
BITD TI and HP made optos with actual specs, but these days, not so much.
Linear mode works great when there’s overall feedback, as in your typical
offline switcher, which has a TL431 to do the actual regulating.
Cheers
Phil Hobbs
A c-b schottky clamp would help, sort of a 74LS photocoupler.
But the really good logic couplers these days aren't optical.
Yup. Even with a better photoreceiver, most of the usual speedup tricks
don't work with LEDs, on account of their diffusion-dominated carrier
dynamics.
Cheers
Phil Hobbs
I did test a Cree white LED for speed. It hit my detector response of
about 7 ns, phosphor included. I was surprised.
Yes, some LEDs are much faster than others.

We sell a LED-based pulsed light source that has <6 ns rise and fall times,
using any of three part numbers at different wavelengths.

With a fancy $20 LED, it gets down to 2 ns.

Speedup caps , reverse bias, and so on do zilch to speed it up.

Cheers

Phil Hobbs
--
Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC /
Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics
john larkin
2024-05-31 10:47:50 UTC
Permalink
On Fri, 31 May 2024 00:12:30 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by john larkin
On Thu, 30 May 2024 14:58:36 -0400, Phil Hobbs
Post by Phil Hobbs
Post by john larkin
On Thu, 30 May 2024 11:29:18 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by piglet
Post by Phil Hobbs
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don?t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.
Cheers
Phil Hobbs
Sorry, maybe my language was sloppy. I meant keep phototransistor
collector from bottoming and reduce C-B miller effect. Not necessarily
by rationing photons. Keeping Vce constant by feeding straight into a
<https://electronics.stackexchange.com/questions/136928/under-what-conditions-does-an-optocoupler-work-fastest>
piglet
If you have the base pinned out, you can do more stuff, true. But at the
end of the day you?re still dealing with a phototransistor.
BITD TI and HP made optos with actual specs, but these days, not so much.
Linear mode works great when there?s overall feedback, as in your typical
offline switcher, which has a TL431 to do the actual regulating.
Cheers
Phil Hobbs
A c-b schottky clamp would help, sort of a 74LS photocoupler.
But the really good logic couplers these days aren't optical.
Yup. Even with a better photoreceiver, most of the usual speedup tricks
don't work with LEDs, on account of their diffusion-dominated carrier
dynamics.
Cheers
Phil Hobbs
I did test a Cree white LED for speed. It hit my detector response of
about 7 ns, phosphor included. I was surprised.
Yes, some LEDs are much faster than others.
We sell a LED-based pulsed light source that has <6 ns rise and fall times,
using any of three part numbers at different wavelengths.
With a fancy $20 LED, it gets down to 2 ns.
Speedup caps , reverse bias, and so on do zilch to speed it up.
Cheers
Phil Hobbs
Why are IR LEDs so much faster? A 10 GBPS SFP transceiver module costs
$16 from Amazon (with Prime free shipping!)
Phil Hobbs
2024-05-31 11:20:53 UTC
Permalink
Post by john larkin
On Fri, 31 May 2024 00:12:30 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by john larkin
On Thu, 30 May 2024 14:58:36 -0400, Phil Hobbs
Post by Phil Hobbs
Post by john larkin
On Thu, 30 May 2024 11:29:18 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by piglet
Post by Phil Hobbs
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don?t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.
Cheers
Phil Hobbs
Sorry, maybe my language was sloppy. I meant keep phototransistor
collector from bottoming and reduce C-B miller effect. Not necessarily
by rationing photons. Keeping Vce constant by feeding straight into a
<https://electronics.stackexchange.com/questions/136928/under-what-conditions-does-an-optocoupler-work-fastest>
piglet
If you have the base pinned out, you can do more stuff, true. But at the
end of the day you?re still dealing with a phototransistor.
BITD TI and HP made optos with actual specs, but these days, not so much.
Linear mode works great when there?s overall feedback, as in your typical
offline switcher, which has a TL431 to do the actual regulating.
Cheers
Phil Hobbs
A c-b schottky clamp would help, sort of a 74LS photocoupler.
But the really good logic couplers these days aren't optical.
Yup. Even with a better photoreceiver, most of the usual speedup tricks
don't work with LEDs, on account of their diffusion-dominated carrier
dynamics.
Cheers
Phil Hobbs
I did test a Cree white LED for speed. It hit my detector response of
about 7 ns, phosphor included. I was surprised.
Yes, some LEDs are much faster than others.
We sell a LED-based pulsed light source that has <6 ns rise and fall times,
using any of three part numbers at different wavelengths.
With a fancy $20 LED, it gets down to 2 ns.
Speedup caps , reverse bias, and so on do zilch to speed it up.
Cheers
Phil Hobbs
Why are IR LEDs so much faster? A 10 GBPS SFP transceiver module costs
$16 from Amazon (with Prime free shipping!)
Those are lasers. The carrier dynamics of a laser running above threshold
are dominated by radiative recombination, which is much quicker.

Lasers are also designed to avoid the horrible diffusion delay of most
LEDs—lower doping, thinner epi, and so on.

Cheers

Phil Hobbs

Cheers

Phil Hobbs
--
Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC /
Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics
john larkin
2024-05-31 11:41:56 UTC
Permalink
On Fri, 31 May 2024 11:20:53 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by john larkin
On Fri, 31 May 2024 00:12:30 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by john larkin
On Thu, 30 May 2024 14:58:36 -0400, Phil Hobbs
Post by Phil Hobbs
Post by john larkin
On Thu, 30 May 2024 11:29:18 -0000 (UTC), Phil Hobbs
Post by Phil Hobbs
Post by piglet
Post by Phil Hobbs
Post by piglet
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Why do you want to saturate the photo transistor?
If you don?t you can get much higher speeds out of even jelly bean cheap
couplers. Even without a base connection it is possible.
Because unless there's overall feedback, running it unsaturated gives
you a beta-dependent circuit that's further dependent on the LED
efficiency, the transparency of the white snot filling the opto package,
temperature, you name it.
Cheers
Phil Hobbs
Sorry, maybe my language was sloppy. I meant keep phototransistor
collector from bottoming and reduce C-B miller effect. Not necessarily
by rationing photons. Keeping Vce constant by feeding straight into a
<https://electronics.stackexchange.com/questions/136928/under-what-conditions-does-an-optocoupler-work-fastest>
piglet
If you have the base pinned out, you can do more stuff, true. But at the
end of the day you?re still dealing with a phototransistor.
BITD TI and HP made optos with actual specs, but these days, not so much.
Linear mode works great when there?s overall feedback, as in your typical
offline switcher, which has a TL431 to do the actual regulating.
Cheers
Phil Hobbs
A c-b schottky clamp would help, sort of a 74LS photocoupler.
But the really good logic couplers these days aren't optical.
Yup. Even with a better photoreceiver, most of the usual speedup tricks
don't work with LEDs, on account of their diffusion-dominated carrier
dynamics.
Cheers
Phil Hobbs
I did test a Cree white LED for speed. It hit my detector response of
about 7 ns, phosphor included. I was surprised.
Yes, some LEDs are much faster than others.
We sell a LED-based pulsed light source that has <6 ns rise and fall times,
using any of three part numbers at different wavelengths.
With a fancy $20 LED, it gets down to 2 ns.
Speedup caps , reverse bias, and so on do zilch to speed it up.
Cheers
Phil Hobbs
Why are IR LEDs so much faster? A 10 GBPS SFP transceiver module costs
$16 from Amazon (with Prime free shipping!)
Those are lasers. The carrier dynamics of a laser running above threshold
are dominated by radiative recombination, which is much quicker.
Lasers are also designed to avoid the horrible diffusion delay of most
LEDs—lower doping, thinner epi, and so on.
I have noticed that the capacitances of LEDs vary all over the place,
like 50:1.
Edward Rawde
2024-05-31 15:57:14 UTC
Permalink
Post by john larkin
On Thu, 30 May 2024 11:29:18 -0000 (UTC), Phil Hobbs
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
...
Post by john larkin
But the really good logic couplers these days aren't optical.
Have you used anything like these?
https://www.digchip.com/datasheets/parts/datasheet/477/ISO485P.php

What other methods are available which can cope with communication between equipment on different power line phases?
Apart from the obvious transformer isolation used in computer networking?
And RF isolation such as Wireless networking and Bluetooth.
john larkin
2024-05-31 17:21:34 UTC
Permalink
On Fri, 31 May 2024 11:57:14 -0400, "Edward Rawde"
Post by Edward Rawde
Post by john larkin
On Thu, 30 May 2024 11:29:18 -0000 (UTC), Phil Hobbs
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
...
Post by john larkin
But the really good logic couplers these days aren't optical.
Have you used anything like these?
https://www.digchip.com/datasheets/parts/datasheet/477/ISO485P.php
What other methods are available which can cope with communication between equipment on different power line phases?
Apart from the obvious transformer isolation used in computer networking?
And RF isolation such as Wireless networking and Bluetooth.
We use isolated ADCs (like ADUM7703) and logic isolators (like the
ADUM1400 series.) Some newer ADCs and isolators include the isolated
dc/dc converter to power the high side, which is very cool.

There are some cheap electrical-metering multichannel ADCs that
include the power isolation.

There are also isolated mosfet gate drivers.

Clive Arthur
2024-05-29 21:03:51 UTC
Permalink
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Not really answering the question, but there's more than one way to skin
a cat...

https://www.nve.com/Isolators
--
Cheers
Clive
boB
2024-05-30 02:16:22 UTC
Permalink
On Wed, 29 May 2024 22:03:51 +0100, Clive Arthur
Post by Clive Arthur
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Not really answering the question, but there's more than one way to skin
a cat...
https://www.nve.com/Isolators
You can also turn the LED off a bit faster and reduce off-time
interference/noise by pulling the anode low, activiely.
Not sure how much better it is though.

boB
piglet
2024-05-30 07:55:19 UTC
Permalink
Post by boB
On Wed, 29 May 2024 22:03:51 +0100, Clive Arthur
Post by Clive Arthur
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Not really answering the question, but there's more than one way to skin
a cat...
https://www.nve.com/Isolators
You can also turn the LED off a bit faster and reduce off-time
interference/noise by pulling the anode low, activiely.
Not sure how much better it is though.
boB
I thought IREDs and LEDs are very fast, far faster than the
phototransistor but someone has done as you suggest, see:

<https://www.edn.com/optocoupler-speed-up-also-reduces-power-consumption/>

piglet
john larkin
2024-05-30 13:40:36 UTC
Permalink
Post by piglet
Post by boB
On Wed, 29 May 2024 22:03:51 +0100, Clive Arthur
Post by Clive Arthur
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Not really answering the question, but there's more than one way to skin
a cat...
https://www.nve.com/Isolators
You can also turn the LED off a bit faster and reduce off-time
interference/noise by pulling the anode low, activiely.
Not sure how much better it is though.
boB
I thought IREDs and LEDs are very fast, far faster than the
<https://www.edn.com/optocoupler-speed-up-also-reduces-power-consumption/>
piglet
Loading Image...
boB
2024-05-30 19:01:15 UTC
Permalink
Post by john larkin
Post by piglet
Post by boB
On Wed, 29 May 2024 22:03:51 +0100, Clive Arthur
Post by Clive Arthur
Post by bitrex
Optocoupler datasheets seem like kind of a mess, I try not to use them
too often in situations where there's any kind of power budget because
other than "shove some relatively huge current through the LED like 5-10
mA" it's hard to know what you can get away with.
A light load on the transistor side will definitely reduce the forward
current required (and of course slow the speed to a crawl) but who can
say by how much while still ensuring the thing will turn on sufficiently
to saturate the output?
The CTR varies widely from process variation, varies with temperature,
varies with collector emitter voltage, varies with forward current, and
the data sheets are full of caveats like "At I_f < 1 mA, note CTR
variation may increase" and "Graphs are representative, not indicative
of actual performance." ????
Any suggestions for how to approach methodically/mathematically
selecting drive current would be appreciated, thank you! ("Don't bother"
a valid option)
Not really answering the question, but there's more than one way to skin
a cat...
https://www.nve.com/Isolators
You can also turn the LED off a bit faster and reduce off-time
interference/noise by pulling the anode low, activiely.
Not sure how much better it is though.
boB
I thought IREDs and LEDs are very fast, far faster than the
<https://www.edn.com/optocoupler-speed-up-also-reduces-power-consumption/>
piglet
https://www.dropbox.com/scl/fi/himnbhrq97vj6eplokek5/Faster_Opto_Totem.JPG?rlkey=ste2tfj8v3zpi45wh1wepyaov&raw=1
Interesting approach, John.

Piglet, the first comment says where I first saw this which was from
Avago/HP and their IGBT driver chips.

boB
Loading...