Speeding optoisolators

[James Arthur]

Sometimes you need a low-power, moderately fast optoisolator.
Here's a simple-minded little rig I hooked up on my bench.
It's not much, but the circuits and measurements might hold
interest for some.

Here's the plain, standard circuit (view in fixed font):


FIGURE 1. +5V
========== -+-
|
R1 |
1.5k 1/2 |
+ >----/\/\/----. MOCD217 |
| OPTO |
.--. .- .---|---------/--.
| | | | --- |/ |
--' '--' | \ / ---> | |
4v p-p | -+- |>. | .--. .-
'---|----------|-' | | |
- >-------------' | --' '--'
+--------> 4.8v p-p
| 3.5KHz max
\ R2
/ 2.2k
\
|
===
GND


Above about 3.5KHz the output no longer swings rail-to-rail,
quickly becoming unusable at higher frequencies.

The opto's slow output transistor limits response time, but
that's easily improved:


+5V +5v
FIGURE 2. -+- -+-
========== | |
\ |
R22 / |
1k \ .'
R21 | |< Q21
1.5k 1/2 +--------| 2n3906 .--.
+ >----/\/\/----. MOCD217 | |\ | |
| OPTO | \ --' '--
.--. .---|---------/--. +------> 4.8v p-p
| | | --- |/ | | >100KHz
--' '-- | \ / ---> | | / tr ~300nS
4v p-p | -+- |>. | \ R23 tf ~1uS
| | | | / 4.7k
'---|----------|-' |
- >-------------' | |
=== ===
GND GND


Q21 prevents the opto's output transistor from saturating
and limits its voltage swing.

The same optoisolator, so equipped, now passes a 110KHz
squarewave, or 1.0uS pulses at 320KHz in my unit.

The speed limitation is still storage time in the opto's output
transistor, so anti-sat tricks on Q21 aren't helpful or needed.

Running the LED at low power ensures a very long lifetime.

Cheers,
James Arthur

 

[D from BC]

Sometimes you need a low-power, moderately fast optoisolator.
Here's a simple-minded little rig I hooked up on my bench.
It's not much, but the circuits and measurements might hold
interest for some.

Here's the plain, standard circuit (view in fixed font):


FIGURE 1. +5V
========== -+-
|
R1 |
1.5k 1/2 |
+ >----/\/\/----. MOCD217 |
| OPTO |
.--. .- .---|---------/--.
| | | | --- |/ |
--' '--' | \ / ---> | |
4v p-p | -+- |>. | .--. .-
'---|----------|-' | | |
- >-------------' | --' '--'
+--------> 4.8v p-p
| 3.5KHz max
\ R2
/ 2.2k
\
|
===
GND


Above about 3.5KHz the output no longer swings rail-to-rail,
quickly becoming unusable at higher frequencies.

The opto's slow output transistor limits response time, but
that's easily improved:


+5V +5v
FIGURE 2. -+- -+-
========== | |
\ |
R22 / |
1k \ .'
R21 | |< Q21
1.5k 1/2 +--------| 2n3906 .--.
+ >----/\/\/----. MOCD217 | |\ | |
| OPTO | \ --' '--
.--. .---|---------/--. +------> 4.8v p-p
| | | --- |/ | | >100KHz
--' '-- | \ / ---> | | / tr ~300nS
4v p-p | -+- |>. | \ R23 tf ~1uS
| | | | / 4.7k
'---|----------|-' |
- >-------------' | |
=== ===
GND GND


Q21 prevents the opto's output transistor from saturating
and limits its voltage swing.

The same optoisolator, so equipped, now passes a 110KHz
squarewave, or 1.0uS pulses at 320KHz in my unit.

The speed limitation is still storage time in the opto's output
transistor, so anti-sat tricks on Q21 aren't helpful or needed.

Running the LED at low power ensures a very long lifetime.

Cheers,
James Arthur

Yawnnnn... :O
Now..2.1Ghz RF coupling digital isolators are interesting..
Tprop < 10nS
150Mbps
2500VRMS isolation
http://www.silabs.com/tgwWebApp/public/web_content/products/Digital_Power/Isolators/en/Si844x.htm


D from BC

 

[John Larkin]

Sometimes you need a low-power, moderately fast optoisolator.
Here's a simple-minded little rig I hooked up on my bench.
It's not much, but the circuits and measurements might hold
interest for some.

Here's the plain, standard circuit (view in fixed font):


FIGURE 1. +5V
========== -+-
|
R1 |
1.5k 1/2 |
+ >----/\/\/----. MOCD217 |
| OPTO |
.--. .- .---|---------/--.
| | | | --- |/ |
--' '--' | \ / ---> | |
4v p-p | -+- |>. | .--. .-
'---|----------|-' | | |
- >-------------' | --' '--'
+--------> 4.8v p-p
| 3.5KHz max
\ R2
/ 2.2k
\
|
===
GND


Above about 3.5KHz the output no longer swings rail-to-rail,
quickly becoming unusable at higher frequencies.

The opto's slow output transistor limits response time, but
that's easily improved:


+5V +5v
FIGURE 2. -+- -+-
========== | |
\ |
R22 / |
1k \ .'
R21 | |< Q21
1.5k 1/2 +--------| 2n3906 .--.
+ >----/\/\/----. MOCD217 | |\ | |
| OPTO | \ --' '--
.--. .---|---------/--. +------> 4.8v p-p
| | | --- |/ | | >100KHz
--' '-- | \ / ---> | | / tr ~300nS
4v p-p | -+- |>. | \ R23 tf ~1uS
| | | | / 4.7k
'---|----------|-' |
- >-------------' | |
=== ===
GND GND


Q21 prevents the opto's output transistor from saturating
and limits its voltage swing.

The same optoisolator, so equipped, now passes a 110KHz
squarewave, or 1.0uS pulses at 320KHz in my unit.

The speed limitation is still storage time in the opto's output
transistor, so anti-sat tricks on Q21 aren't helpful or needed.

Running the LED at low power ensures a very long lifetime.

Cheers,
James Arthur

How about this?

http://img240.imageshack.us/my.php?image=optoppdr1.jpg

If the transient led current is large but the dc current is low, the
transistors will switch fast but not saturate too hard. Of course, b-e
resistors always help, too.

John

 

[James Arthur]

Yawnnnn... :O
Now..2.1Ghz RF coupling digital isolators are interesting..
Tprop < 10nS
150Mbps
2500VRMS isolationhttp://www.silabs.com/tgwWebApp/public/web_content/products/Digital_P...

D from BC

Hey, 100kHz, isolated, for pennies and 2 parts is a decent upgrade for
a dogmeat junk-box opto.

All I needed was 20kHz, so 'good enough' is good enough, right?

Besides, using transformers is cheating.

Cheers,
James Arthur

 

[James Arthur]

How about this?

http://img240.imageshack.us/my.php?image=optoppdr1.jpg

If the transient led current is large but the dc current is low, the
transistors will switch fast but not saturate too hard. Of course, b-e
resistors always help, too.

John

Cute. I might rig one up, just for fun.

I s'pose the next step would be doing clever anti-sat stuff to the
phototransistor, if we had access to its base (not available on the
MOCD217 (which I have a reel of (my junk box is well-stocked))).

Cheers,
James Arthur

 

[D from BC]

Hey, 100kHz, isolated, for pennies and 2 parts is a decent upgrade for
a dogmeat junk-box opto.

All I needed was 20kHz, so 'good enough' is good enough, right?

Besides, using transformers is cheating.

Cheers,
James Arthur

I'm wondering if LED capacitance is significant...


D from BC

 

[Tim Williams]

I once made an inverter like that, using 2N4401/03. Nice and sharp edges
(in the 40ns range IIRC), gotta be careful not to leave the input
disconnected, in an ambiguous logic state, though! ;-)

Tim

 

[John Larkin]

Cute. I might rig one up, just for fun.

I s'pose the next step would be doing clever anti-sat stuff to the
phototransistor, if we had access to its base (not available on the
MOCD217 (which I have a reel of (my junk box is well-stocked))).

Cheers,
James Arthur

Well, to go really fast you'd want a phototransistor (or better yet
photodiode) to work in pure current mode, into a TIA.

http://www.highlandtechnology.com/DSS/J730DS.html

This uses a silicon (for 850 nm) or InGaAs (for longer stuff) pin
diode detector driving a current-mode opamp as the tia. Analog bw is
about 200 MHz, which would be faster if I'd bothered to put more
voltage across the pin detector. I should have done that.

All non-avalanche photodiodes tend to have a sensitivity in the
ballpark of 0.5 amps/watt. If you illuminate them from an led, they
have to be big and slow to capture the fuzzy light. If you use a
laser, through fiber maybe, they can be a lot smaller and faster.


John

 

[John Larkin]

I'm wondering if LED capacitance is significant...

Sure, especially c-b (Miller) capacitance. And there's nothing to
discharge the base here except base current itself.


John

 

[James Arthur]

I'm wondering if LED capacitance is significant...

D from BC

Hmmm. Let's see: Fairchild's datasheet says 18pF, x 1.5k is about a
30nS time-constant, so I'd have to say "no," not a factor.

Storage time in the phototransistor is the main limitation, so
optimizing its drive & load is a good first step. Next would be
preventing all transistors (photo, especially, and Q1) from
saturating.

John's totem-pole actively jerks the phototransistors out of
saturation, so that'd be a nice speed-up that's not too critical about
exact photocurrent and load--just rip the things 'off'. It takes two
opto channels and bipolar drive for their LEDs, but it's brutally
elegant for speed.

All of this is pretty simple and low-power, which is nice.

Cheers,
James Arthur

 

[Jan Panteltje]

Well, to go really fast you'd want a phototransistor (or better yet
photodiode) to work in pure current mode, into a TIA.

http://www.highlandtechnology.com/DSS/J730DS.html

'The J730 is compatible with the Highland J720 E/O Conveter,'

My spell sjeker says it is ConveRter.

 

[John Larkin]

'The J730 is compatible with the Highland J720 E/O Conveter,'

My spell sjeker says it is ConveRter.

It's amazing how hard it is to get every last technical and
spelling/punctuation error out of datasheets and web pages. I sure
hope this one hasn't cost us a lot in sales.

John

 

[Vladimir Vassilevsky]

This will smoke eventually. The opto current is not limited.

How about this?

http://img240.imageshack.us/my.php?image=optoppdr1.jpg

Much better idea however it takes two optocouplers instead of one and
will smoke if the input is unconnected. BTW I wouldn't be concerned
about the speed of the LEDs compared to that of the photo transistor.

If the transient led current is large but the dc current is low, the
transistors will switch fast but not saturate too hard. Of course, b-e
resistors always help, too.

Once we had a problem with decoupling of RS-232 via H11L1. There was a
big difference in the speed (several times) depending on the particular
make and the particular part. The board was already made; the trivial
solution was just pick the fastest parts from the box


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[James Arthur]

Well, to go really fast you'd want a phototransistor (or better yet
photodiode) to work in pure current mode, into a TIA.

Yep. That's the reason for Q21, which serves as a kind of fast TIA on
the cheap.

Photodiode-mode would be much faster, eliminating phototransistor
storage time, but my opto doesn't offer up the base. If it did,
substituting the back-biased photojunction and adjusting R23 should
speed things up a lot.

http://www.highlandtechnology.com/DSS/J730DS.html

Nice, as usual for HTI gear.

James

 

[James Arthur]

This will smoke eventually. The opto current is not limited.

Not so. The opto's current transfer ratio (CTR) limits the
phototransistor's current. Since the LED current is about 2mA, the
output current is about the same, and the opto's phototransistor will
dissipate 10mW, typical. The detector in the specified opto can
safely dissipate 150mW, which would be quite impossible given the
combination of limited LED current, CTR of roughly 100%, and 5v
supply.

Much better idea however it takes two optocouplers instead of one and
will smoke if the input is unconnected.

No, it won't smoke for the same reason as above. Remember, these are
being run at especially low drive currents, which changes the game.
In this case, two LED drops plus two limiting resistors set the LEDs'
current to much less than half if the input is open. The two output
transistors can easily take it.

BTW I wouldn't be concerned
about the speed of the LEDs compared to that of the photo transistor.

Agreed, at least at this level of performance.

Once we had a problem with decoupling of RS-232 via H11L1. There was a
big difference in the speed (several times) depending on the particular
make and the particular part. The board was already made; the trivial
solution was just pick the fastest parts from the box

That's one way, for sure!

Cheers,
James Arthur

 

[Vladimir Vassilevsky]

James Arthur wrote:

Not so. The opto's current transfer ratio (CTR) limits the
phototransistor's current.

I feel suspiciously about the circuits with the potentially unlimited
escalation of the current. The datasheet guarantees the minimal CTR but
it says nothing about the maximal.

Since the LED current is about 2mA, the
output current is about the same, and the opto's phototransistor will
dissipate 10mW, typical. The detector in the specified opto can
safely dissipate 150mW, which would be quite impossible given the
combination of limited LED current, CTR of roughly 100%, and 5v
supply.

Yes, it won't smoke at once, but it will smoke eventually.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[Rich Grise]

opto channels and bipolar drive for their LEDs, but it's brutally
elegant for speed.

"Brutally elegant"?

Sounds a little creepy. ;-)

Cheers!
Rich

 

[Rich Grise]

It's amazing how hard it is to get every last technical and
spelling/punctuation error out of datasheets and web pages. I sure
hope this one hasn't cost us a lot in sales.

I'm available to do proofreading - I'm the self-appointed chief of
the Apostrophe Police, after all. ;-)

And, of course, at times I branch out to the Spelling Police (I
used to win contests), Grammar Police, and "Does-it-make-any-sense"
police. ;-)

Cheers!
Rich

 

[Jim Thompson]

James Arthur wrote:



I feel suspiciously about the circuits with the potentially unlimited
escalation of the current. The datasheet guarantees the minimal CTR but
it says nothing about the maximal.


Yes, it won't smoke at once, but it will smoke eventually.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

My maxim is a twist on Murphy... if there's ANY chance of smoke, no
matter how minute, there WILL be smoke.

...Jim Thompson

 

[James Arthur]

I feel suspiciously about the circuits with the potentially unlimited
escalation of the current. The datasheet guarantees the minimal CTR but
it says nothing about the maximal.

Minimum CTR at 10mA is 100%, typical is 130%. From the datasheet(*),
at 2mA LED current CTR is typically 75% for an output current would be
about 1.5mA. 30mA would be the maximum allowed, dissipation-wise.

(*) http://www.fairchildsemi.com/ds/MO/MOCD217-M.pdf

Those spreads look pretty tight to me, so I'm perfectly comfortable
with my 2,000% safety margin, but if you aren't, you could add a small
collector resistor to the opto. I wouldn't, but you could.

Yes, it won't smoke at once, but it will smoke eventually.

Let's do the math. With input floating, if John used 1.5k resistors,
the LED current would be from 4v - (2 led drops) passing through 2 x
1.5k ohm resistors. That amounts to (4v - 2.2v) / 3k ohms = 600uA.
At these currents the opto's CTR drops, so output current will be just
220uA (typical), for a safety margin of 13,600%. John's version is
_very_ safe.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultanthttp://www.abvolt.com

Cheers,
James Arthur