Fast but small optocouplers?

[Joerg]

Loads of different types, eg
http://www.nve.com/webstore/catalog/default.php?cPath=30_25

But not opto...

A bit much in quiescent current and very pricey but they sure are fast.
However, one has to be careful. Several clients have been burned by
fancy new-tech isolator chips (not from NVE though). In all cases the
comeuppance happened at the EMC test facility.

 

[Joerg]

I you are willing to power the receive side, just buy a logic
optocoupler with CMOS output. There are tons of them, some very fast.

Thing is, I've only got a (somewhat) regulated 10V on that side and not
much power. A milliamp or two. So that'll add more parts.

 

[Joerg]

Jan Panteltje wrote:

[...]

??

bottom page 4

That's CM rejection, not an important parameter here.

 

[Klaus Kragelund]

No need for a one-shot, a comparator will do (NRZ scheme). But

unfortunately transformers with isolation ratings are large.

What about spiral turn PCB transformer, the prepreg constitutes approved isolation between the two turns. Needs HF (above 10MHz) though due to low magnetizing inductance if you need it that small

Cheers

Klaus

 

[Joerg]

What about spiral turn PCB transformer, the prepreg constitutes
approved isolation between the two turns. Needs HF (above 10MHz)
though due to low magnetizing inductance if you need it that small

That works. But it would make EMC a challenge. I'll try the opto thing,
looks like the Sanyo couplers might do the job. If not then I guess
it'll have to be a transformer. Or maybe an ultrasound piezo transfer,
that would be really high tech

 

[Joerg]

On a sunny day (Sat, 01 Sep 2012 12:35:06 -0700) it happened Joerg

Jan Panteltje wrote: [...]


Do not confuse wuth common mode rejection in 30kV /uS

??
bottom page 4

That's CM rejection, not an important parameter here.

I knowm, but I thought you took it maybe for output rise time
This thing really is not Ferry Vast, some years ago we discussed that issue here
and somebody came up with really fast ones, maybe google still has it.

The Sanyo is fast enough, so most likely I'll use that one.

 

[Tim Williams]

The usual 2kV RMS. It's not that I can't make a small one, done it many
times. The issue is that on most such projects it has to come with "the
papers".

Meantime I've found this from Sharp, requires a voltage and no
uncommitted emitter but might do the trick here:

http://www.sharpsma.com/webfm_send/1118

Cute! It's a miniature SFH6345 (which is, in turn, a 6N136 without the
base pin...and its pesky capacitance). That'll come in handy for IGBT
drivers!

Hmm, offhand, Digikey and Mouser don't stock it.. will have to look
around.

Tim

 

[spamtrap1888]

Hello Folks,

So far I've never had size constraints and my staples were logic devices
such as the HCNW137 series. Now I need one that ideally doesn't require
a logic supply on the output, rise/fall times well under 1usec even when
not fed more than 1mA into the LED, as small as possible. No more than
about 0.150" wide and 0.250" long. Isolation around 2kV RMS is fine.

This would be at the size pain threshold but as usual is way too slow:

http://www.avagotech.com/docs/AV02-0774EN

I went through scores of them and all in the several usec. Any
sub-microsecond ones?

Unfortunately the bases aren't pipe out on these so not speed tricks :-(

What I want to do is ferry a fast PWM across a barrier, 100kHz range,
needs to be somewhat accurate. I could do the transformer plus NRZ thing
but that gets old and not enough space.

Did you look at this one? They tout it as a pulse transformer
replacement:

http://www.avagotech.com/docs/AV02-1302EN

 

[Tim Williams]

Well, I've seen them simply get into the wrong state. Most have
incremental/edge coupling so can get confused.

The ones from Analog brag about their active refresh, which makes sure
that, if they flip, they flip back to the correct state within some period
of time.

We've used them (the ADI parts) with MOSFET drivers, where the dV/dt
pushes the datasheet limits. We haven't seen them chatter at all.

Tim

 

[Tim Williams]

Don't insult the base pin. A b-e resistor really speeds up an opto.

I have full right to insult anything that breaks!

Resistors aren't necessary with these types, and taking it out to a pin is
far more detrimental in typical applications: big edges really screw them
up.

Note carefully, for example, how Fairchild rates the dV/dt of their 6N136.
http://www.fairchildsemi.com/ds/6N/6N135.pdf
The dV/dt could be 1 or 100kV/us and it won't matter, the charge injection
isn't even enough to turn on the transistor at a mere 10Vpp. Terrible
liars. Other manufacturers rate theirs at only 1kV/us, though still a 10V
edge. Compare that to the SFH6345, which is rated over 10kV/us (by any
manufacturer), and tested with a 1kVpp edge (Avago tests theirs at
1.5kVpp!).

As for empirical results, a 6N136 will do this in a gate driver:
http://t3sl4.dnsdynamic.net/Images/Induction902.jpg
The full waveform is,
http://t3sl4.dnsdynamic.net/Images/Induction901.jpg
which as you can see isn't very much delta V or delta T as gate drivers
go. A little shielding wrapped around the 6N136 got it working up to
320V, but I still don't trust it. And the shield reduces the peak voltage
rating in a non-approved manner.

Tim

P.S. Mmmm, Sprecher Abbey Triple. Geez, some fruit flies got in through
the window, they're curious about it. Heck, I would be too. Now where's
the fly swatter...

 

[Tim Williams]

Don't insult the base pin. A b-e resistor really speeds up an opto.

I have full right to insult anything that breaks!

Resistors aren't necessary with these types, and taking it out to a pin is
far more detrimental in typical applications: big edges really screw them
up.

Note carefully, for example, how Fairchild rates the dV/dt of their 6N136.
http://www.fairchildsemi.com/ds/6N/6N135.pdf
The dV/dt could be 1 or 100kV/us and it won't matter, the charge injection
isn't even enough to turn on the transistor at a mere 10Vpp. Terrible
liars. Other manufacturers rate theirs at only 1kV/us, though still a 10V
edge. Compare that to the SFH6345, which is rated over 10kV/us (by any
manufacturer), and tested with a 1kVpp edge (Avago tests theirs at
1.5kVpp!).

As for empirical results, a 6N136 will do this in a gate driver:
http://t3sl4.dnsdynamic.net/Images/Induction902.jpg
The full waveform is,
http://t3sl4.dnsdynamic.net/Images/Induction901.jpg
which as you can see isn't very much delta V or delta T as gate drivers
go. A little shielding wrapped around the 6N136 got it working up to
320V, but I still don't trust it. And the shield reduces the peak voltage
rating in a non-approved manner.

Tim

P.S. Mmmm, Sprecher Abbey Triple. Geez, some fruit flies got in through
the window, they're curious about it. Heck, I would be too. Now where's
the fly swatter...

 

[[email protected]]

The usual 2kV RMS. It's not that I can't make a small one, done it many
times. The issue is that on most such projects it has to come with "the
papers".

You claim that everything suggested is too big, but I just wonder, how
you are going to handle the PCB creapage distances with very small
(narrow) packages, if you intend to keep the _system_ level insulation
ratings in the kV range.

 

[Nico Coesel]

Hello Folks,

So far I've never had size constraints and my staples were logic devices
such as the HCNW137 series. Now I need one that ideally doesn't require
a logic supply on the output, rise/fall times well under 1usec even when
not fed more than 1mA into the LED, as small as possible. No more than
about 0.150" wide and 0.250" long. Isolation around 2kV RMS is fine.

This would be at the size pain threshold but as usual is way too slow:

http://www.avagotech.com/docs/AV02-0774EN

I went through scores of them and all in the several usec. Any
sub-microsecond ones?

Unfortunately the bases aren't pipe out on these so not speed tricks :-(

What I want to do is ferry a fast PWM across a barrier, 100kHz range,
needs to be somewhat accurate. I could do the transformer plus NRZ thing
but that gets old and not enough space.

I've used an Avago HCPL-0631 for a similar situation. Its blazingly
fast. It does require 5V-ish power on the secondary side. In my case I
solved that with a simple resistor + zener diode.

 

[Joerg]

Cute! It's a miniature SFH6345 (which is, in turn, a 6N136 without the
base pin...and its pesky capacitance). That'll come in handy for IGBT
drivers!

Hmm, offhand, Digikey and Mouser don't stock it.. will have to look
around.

They hold 21,000 of them right now. Is that not enough stock?

 

[Joerg]

You claim that everything suggested is too big, but I just wonder, how
you are going to handle the PCB creapage distances with very small
(narrow) packages, if you intend to keep the _system_ level insulation
ratings in the kV range.

Partially by potting. I have space, but not as much as needed for a big
fat HCNW137 chip or the like.

 

[Joerg]

Did you look at this one? They tout it as a pulse transformer
replacement:

http://www.avagotech.com/docs/AV02-1302EN

Yes, I looked at those. They have the usual problem, a whopping 0.500"
length. It would be like trying to squeeze a Ford F-350 Crew Cab with a
full length bed into a parking space for a compact car.

 

[Joerg]

I've used an Avago HCPL-0631 for a similar situation. Its blazingly
fast. It does require 5V-ish power on the secondary side. In my case I
solved that with a simple resistor + zener diode.

But you also had the space for these rather fat IC packages

 

[Nico Coesel]

But you also had the space for these rather fat IC packages

It also comes in an SO-8 compatible housing. Its a bit high but the
footprint is small and it is a dual-opto so the footpring for each
optocoupler is tiny

 

[Joerg]

It also comes in an SO-8 compatible housing. Its a bit high but the
footprint is small and it is a dual-opto so the footpring for each
optocoupler is tiny

But it's still too big and we only need one.

 

[Tim Williams]

They hold 21,000 of them right now. Is that not enough stock?

Weird, I had found the PC457L0NIT0F, which has the exact same datasheet,
which is the -P0F, not -T0F.

Digikey has 23k and change of the -NIP0F, and 58k and change of
the -YIP0F, which is VDE marked. Again, same datasheet and link.

Now I want to know what the hell is the difference! Nothing about part
numbering in the datasheet. Sharp's website isn't helpful; datasheet
specifies "corresponding Optoelectronic Application Notes", but there are
none listed under this product type. And their listing shows the
45*6*L0NIP0F, but only the 457 -YIP0F.

Tim