LED as detector for high speed link

[Joerg]

On 2019-05-20 14:48, klaus.kragelund@gmail.com wrote:

Hi

For a test setup I am toying with an isolated interface where I need
1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an
indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

Hey, German in a datasheet! Haven't seen that in a while.

If I use the same dirt cheap LED as transmitter I can get about
10kBit with small amplitude. However, I need a large amplitude, to be
able to detect bit level with logic 3V. (so over 2.4V)

3V is a tall order, almost needs a nuclear fireball as a transmit pulse.

This is off the cuff so use with a grain of salt:

What drives the LED? If a uC, maybe you could spare another pin and hook
the cathode of the LED to that instead of ground. For receive you could
switch the normal drive pin to low, then set this new pin to open drain
input with pull-up resistor, hoping the uC has internal pull-ups and
allows that. This would save the cost of an extra resistor which you'd
have to provide if there is no interal pull-up feature. Now the LED
would operate in photoconductive mode and when enough light falls on it
the voltage on this nwew pin will be pulled down. Since internal
pull-ups are usually in the 50-100k range (or a similar current source)
this method would only require sufficient light to generate 50uA of
photo current.

The price of the emitter is not relevant. So I was thinking to use a
white LED (to cover the entire spectrum of the RED LED detector), a
big one, and driving it with a lot of current, to get sufficient
photon level at the red LED to have both high speed and logic IO
level

An LED often have a different emission wavelength than the
responsivity wavelength, so that is why I use a white LED to be sure
wavelenghts overlap

They are often sensitive slightly below the wavelength of emitting
operation. So in your case possibly something around 600nm.

I will drive the emitter LED with high current, and then have a
sufficient low load resistor for the red detector LED to have high
bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is
there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf

If you do that make sure there are no humans or animals close by who
could retain eye damage from that.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[Cydrome Leader]

John Larkin <jjlarkin@highland_snip_technology.com> wrote:

On Mon, 20 May 2019 23:33:33 +0000 (UTC), Cydrome Leader
presence@MUNGEpanix.com> wrote:

klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

If I use the same dirt cheap LED as transmitter I can get about 10kBit with small amplitude. However, I need a large amplitude, to be able to detect bit level with logic 3V. (so over 2.4V)

Why not amplify the LED current output, into a comparator maybe?



The price of the emitter is not relevant. So I was thinking to use a white LED (to cover the entire spectrum of the RED LED detector), a big one, and driving it with a lot of current, to get sufficient photon level at the red LED to have both high speed and logic IO level

An LED often have a different emission wavelength than the responsivity wavelength, so that is why I use a white LED to be sure wavelenghts overlap

I will drive the emitter LED with high current, and then have a sufficient low load resistor for the red detector LED to have high bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf

I suspect you'll run into problems. White LEDs use a phosphor to convert
blue light to white. There will likely be a persistence issue, not unlike
CRTs or florescent light bulbs. So, to be able to transmit data in the
1Mhz range, the rise and decay time of that phosphor they use better be
really, really short. IR seems to be the norm for data transmission with
LEDs.


I tested some Cree white LEDs for speed, and they had nanosecond
response. The phosphors seem to be very fast.

that's good to know. That migth be one reason some LEF lighting is so
flickery and terrible looking, not counting bad CRI.


> But I'd guess that the best driver of a red LED is a red LED.

Ha

 

On Tuesday, May 21, 2019 at 10:46:30 AM UTC-4, John Larkin wrote:

On Mon, 20 May 2019 22:53:42 -0700 (PDT), dagmargoodboat@yahoo.com
wrote:

On Monday, May 20, 2019 at 8:09:25 PM UTC-4, Klaus Kragelund wrote:
I cannot add circuitry to the detector side, since needs to be low cost

Cheers

Klaus

Then it's going to be slow. LEDs make rotten photodiodes. Functional,
but just.

I tried it years ago. But rather than dig for my notes, I just
measured two combinations.

LEDs nose-to-nose, clear cases, If = 11.2mA, Vr = 0V.

emitter 'detector' photocurrent
------- -------------- ------------
GaN green AlGaInP red 1.7uA
AlGaInP red same as emitter 62uA

That's actually not too bad. If we assume I
was lucky enough to couple 5mW optical, that's
12mA/W photocurrent, or about 2% as good as a
PIN diode.

Cheers,
James Arthur

The target is awfully small, so most of the photons miss. I bet a lens
would help a lot.


--

John Larkin Highland Technology, Inc

lunatic fringe electronics

The LEDs I used last night had lensed cases, excepting the green
emitter (who lost her head in an unfortunate--for her--encounter
with a Dremel tool).

So, nose-to-nose, I think there's a fair chance I coupled
something like 25-50% of the photons.

If Klaus wants to transmit over any distance, he's going to
have to compromise on cost or range and speed.

Cheers,
James

 

On Tuesday, May 21, 2019 at 5:01:29 AM UTC-4, piglet wrote:

On 20/05/2019 10:48 pm, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

If I use the same dirt cheap LED as transmitter I can get about 10kBit with small amplitude. However, I need a large amplitude, to be able to detect bit level with logic 3V. (so over 2.4V)

The price of the emitter is not relevant. So I was thinking to use a white LED (to cover the entire spectrum of the RED LED detector), a big one, and driving it with a lot of current, to get sufficient photon level at the red LED to have both high speed and logic IO level

An LED often have a different emission wavelength than the responsivity wavelength, so that is why I use a white LED to be sure wavelenghts overlap

I will drive the emitter LED with high current, and then have a sufficient low load resistor for the red detector LED to have high bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf

Cheers

Klaus


My experience was that Red LED responds very badly to red but much
better to higher energy/shorter wavelength green or blue light.

That's my recollection too, which would seem to make sense
band-gap-wise. So, I was surprised when last night's red LEDs
did so poorly detecting a GaN green source.

White
phosphors will only lose you efficiency and worsen timings.

Not sure what your detector circuit will be but I cannot see how you can
get 2.4V (and isn't that above the Vf?) output without some
amplification even if only a penny BC847.

piglet

Yep. A one-cent one-transistor common base amplifier would do
wonders for speeding things up.

Cheers,
James Arthur

 

[George Herold]

On Tuesday, May 21, 2019 at 12:29:40 PM UTC-4, Joerg wrote:

On 2019-05-20 14:48, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need
1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an
indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf


Hey, German in a datasheet! Haven't seen that in a while.


If I use the same dirt cheap LED as transmitter I can get about
10kBit with small amplitude. However, I need a large amplitude, to be
able to detect bit level with logic 3V. (so over 2.4V)


3V is a tall order, almost needs a nuclear fireball as a transmit pulse.

This is off the cuff so use with a grain of salt:

What drives the LED? If a uC, maybe you could spare another pin and hook
the cathode of the LED to that instead of ground. For receive you could
switch the normal drive pin to low, then set this new pin to open drain
input with pull-up resistor, hoping the uC has internal pull-ups and
allows that. This would save the cost of an extra resistor which you'd
have to provide if there is no interal pull-up feature. Now the LED
would operate in photoconductive mode and when enough light falls on it
the voltage on this nwew pin will be pulled down. Since internal
pull-ups are usually in the 50-100k range (or a similar current source)
this method would only require sufficient light to generate 50uA of
photo current.

Hmm I just did the 'silly' measurement on the bench, some red led
(through hole) +15V with 10 meg ohm resistor to ground. Sitting
in the room lights I got about 0.4 V, pulling my desk light right on top
of it I got 3-4 Volts (0.3 uA) (desk lamp is 60W incandescent)

50 uA seems like a lot.

George H.

The price of the emitter is not relevant. So I was thinking to use a
white LED (to cover the entire spectrum of the RED LED detector), a
big one, and driving it with a lot of current, to get sufficient
photon level at the red LED to have both high speed and logic IO
level

An LED often have a different emission wavelength than the
responsivity wavelength, so that is why I use a white LED to be sure
wavelenghts overlap


They are often sensitive slightly below the wavelength of emitting
operation. So in your case possibly something around 600nm.


I will drive the emitter LED with high current, and then have a
sufficient low load resistor for the red detector LED to have high
bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is
there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf




If you do that make sure there are no humans or animals close by who
could retain eye damage from that.

--
Regards, Joerg

http://www.analogconsultants.com/

 

[George Herold]

On Tuesday, May 21, 2019 at 3:21:13 PM UTC-4, dagmarg...@yahoo.com wrote:

On Tuesday, May 21, 2019 at 5:01:29 AM UTC-4, piglet wrote:
On 20/05/2019 10:48 pm, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

If I use the same dirt cheap LED as transmitter I can get about 10kBit with small amplitude. However, I need a large amplitude, to be able to detect bit level with logic 3V. (so over 2.4V)

The price of the emitter is not relevant. So I was thinking to use a white LED (to cover the entire spectrum of the RED LED detector), a big one, and driving it with a lot of current, to get sufficient photon level at the red LED to have both high speed and logic IO level

An LED often have a different emission wavelength than the responsivity wavelength, so that is why I use a white LED to be sure wavelenghts overlap

I will drive the emitter LED with high current, and then have a sufficient low load resistor for the red detector LED to have high bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf

Cheers

Klaus


My experience was that Red LED responds very badly to red but much
better to higher energy/shorter wavelength green or blue light.

That's my recollection too, which would seem to make sense
band-gap-wise. So, I was surprised when last night's red LEDs
did so poorly detecting a GaN green source.

Did the LED have a clear lens? I used a red led with a red lens
and it worked best with an orange (amber) led... less well with
yellow and green. Even when I sanded lens down to a few mil of the
surface.

George H.

White
phosphors will only lose you efficiency and worsen timings.

Not sure what your detector circuit will be but I cannot see how you can
get 2.4V (and isn't that above the Vf?) output without some
amplification even if only a penny BC847.

piglet

Yep. A one-cent one-transistor common base amplifier would do
wonders for speeding things up.

Cheers,
James Arthur

 

On Tuesday, May 21, 2019 at 3:29:12 PM UTC-4, George Herold wrote:

On Tuesday, May 21, 2019 at 3:21:13 PM UTC-4, dagmarg...@yahoo.com wrote:
On Tuesday, May 21, 2019 at 5:01:29 AM UTC-4, piglet wrote:
On 20/05/2019 10:48 pm, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

If I use the same dirt cheap LED as transmitter I can get about 10kBit with small amplitude. However, I need a large amplitude, to be able to detect bit level with logic 3V. (so over 2.4V)

The price of the emitter is not relevant. So I was thinking to use a white LED (to cover the entire spectrum of the RED LED detector), a big one, and driving it with a lot of current, to get sufficient photon level at the red LED to have both high speed and logic IO level

An LED often have a different emission wavelength than the responsivity wavelength, so that is why I use a white LED to be sure wavelenghts overlap

I will drive the emitter LED with high current, and then have a sufficient low load resistor for the red detector LED to have high bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf

Cheers

Klaus


My experience was that Red LED responds very badly to red but much
better to higher energy/shorter wavelength green or blue light.

That's my recollection too, which would seem to make sense
band-gap-wise. So, I was surprised when last night's red LEDs
did so poorly detecting a GaN green source.

Did the LED have a clear lens? I used a red led with a red lens
and it worked best with an orange (amber) led... less well with
yellow and green. Even when I sanded lens down to a few mil of the
surface.

George H.

All cases were water-clear. I'd hacked the green emitter's lens off
to make it a better point-source in another experiment, but here I
don't think it was a big loss, since that allowed me to press the
(red) detector LED almost smack up against the emitter's die.

I sent John some of those magnificent Agilent GaN greens a decade
or two ago, and he posted about them here.

Cheers,
James

 

On Tuesday, 21 May 2019 11:01:29 UTC+2, piglet wrote:

On 20/05/2019 10:48 pm, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

If I use the same dirt cheap LED as transmitter I can get about 10kBit with small amplitude. However, I need a large amplitude, to be able to detect bit level with logic 3V. (so over 2.4V)

The price of the emitter is not relevant. So I was thinking to use a white LED (to cover the entire spectrum of the RED LED detector), a big one, and driving it with a lot of current, to get sufficient photon level at the red LED to have both high speed and logic IO level

An LED often have a different emission wavelength than the responsivity wavelength, so that is why I use a white LED to be sure wavelenghts overlap

I will drive the emitter LED with high current, and then have a sufficient low load resistor for the red detector LED to have high bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf

Cheers

Klaus


My experience was that Red LED responds very badly to red but much
better to higher energy/shorter wavelength green or blue light. White
phosphors will only lose you efficiency and worsen timings.

Yes, I saw that in a datasheet, that the emission wavelenght is not the same as the receptivity wavelength, and therefore other colors than red might be better

Not sure what your detector circuit will be but I cannot see how you can
get 2.4V (and isn't that above the Vf?) output without some
amplification even if only a penny BC847.

A BC847 may be acceptable

Cheers

Klaus

 

On Tuesday, 21 May 2019 07:26:56 UTC+2, whit3rd wrote:

On Monday, May 20, 2019 at 2:48:13 PM UTC-7, klaus.k...@gmail.com wrote:

For a test setup I am toying with an isolated interface where I need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an indicator when the link is not active)
...
If I use the same dirt cheap LED as transmitter I can get about 10kBit with small amplitude

Why not just use an IeDA transmit/receive pair? Speed is good, parts are available. And,
because it's a standard interface there are laptops and PDAs and laser printers that have it
already built in. A 'LED' might be tecnically capable of receiving, but it isn't
optimal (not a lot of receiver area to pick up light).

I cannot add significant cost, since this is only a link for debugging. Any add on would increase the product price

It's even easier to transformer-couple an interface; Ethernet magnetics
are mass produced and can outperform your target throughput handily, at quite reasonable
cost.

I am working on that also, in parallel with the LED interface

Cheers

Klaus

 

On Tuesday, 21 May 2019 02:09:42 UTC+2, Lasse Langwadt Christensen wrote:

tirsdag den 21. maj 2019 kl. 01.33.38 UTC+2 skrev Cydrome Leader:
klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

If I use the same dirt cheap LED as transmitter I can get about 10kBit with small amplitude. However, I need a large amplitude, to be able to detect bit level with logic 3V. (so over 2.4V)

The price of the emitter is not relevant. So I was thinking to use a white LED (to cover the entire spectrum of the RED LED detector), a big one, and driving it with a lot of current, to get sufficient photon level at the red LED to have both high speed and logic IO level

An LED often have a different emission wavelength than the responsivity wavelength, so that is why I use a white LED to be sure wavelenghts overlap

I will drive the emitter LED with high current, and then have a sufficient low load resistor for the red detector LED to have high bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf

I suspect you'll run into problems. White LEDs use a phosphor to convert
blue light to white. There will likely be a persistence issue, not unlike
CRTs or florescent light bulbs. So, to be able to transmit data in the
1Mhz range, the rise and decay time of that phosphor they use better be
really, really short. IR seems to be the norm for data transmission with
LEDs.

https://books.google.dk/books?id=xR8uDwAAQBAJ&pg=PA449&lpg=PA449&dq=white+led+slow+phosfor&source=bl&ots=k7rVDUHoDo&sig=ACfU3U0oTbJQZwhXZN90ghs5k0GcpPKmvA&hl=da&sa=X&ved=2ahUKEwi2_r_ynKviAhVPJlAKHUHjCF0Q6AEwCHoECAcQAQ#v=onepage&q=white%20led%20slow%20phosfor&f=false

That also explains the funny "white" emission curve. Thanks

 

[bitrex]

On 5/21/19 12:29 PM, Joerg wrote:

On 2019-05-20 14:48, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need
1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an
indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf



Hey, German in a datasheet! Haven't seen that in a while.

I hear Wernher von Braun said the same when he started work on the
Redstone program...

 

On Tuesday, 21 May 2019 17:38:42 UTC+2, bitrex wrote:

On 5/21/19 10:48 AM, John Larkin wrote:
On Tue, 21 May 2019 02:46:54 -0400, bitrex <user@example.net> wrote:

On 5/20/19 5:48 PM, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

If I use the same dirt cheap LED as transmitter I can get about 10kBit with small amplitude. However, I need a large amplitude, to be able to detect bit level with logic 3V. (so over 2.4V)

The price of the emitter is not relevant. So I was thinking to use a white LED (to cover the entire spectrum of the RED LED detector), a big one, and driving it with a lot of current, to get sufficient photon level at the red LED to have both high speed and logic IO level

An LED often have a different emission wavelength than the responsivity wavelength, so that is why I use a white LED to be sure wavelenghts overlap

I will drive the emitter LED with high current, and then have a sufficient low load resistor for the red detector LED to have high bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf

Cheers

Klaus


If the price of the emitter isn't relevant why not use a big tri-color
RGB LED. I'm assuming there is some kind of software on the receiver end
too, to transmit data back.

Run an "auto tune" test data sequence on the emitter side that cycles
around thru the area of the color palette where you figure the best
responsivity of the red LED is and have the receiver side communicate
back where the best data rate happens. Then save that for use with that
particular LED.

All LEDs of a given part number will have about the same spectral
response, so he only has to figure that out once.



In the case of real bargain-basement LEDs I'm not sure there are any
guarantees of that. When you buy a bag of cheap LEDs from China their
characteristics are all over the place and stuff like forward voltage
vs. current varies wildly.

I tried one time to test a bag full and there's no nice bell curve to be
found on IV parameters of parts like that; all the good shit that can be
parameter-matched has been pulled by the mfgr already to make e.g. those
LED christmas lights strings where the LEDs can be run in parallel with
a single current-limiting resistor for the string because all the LEDs
in the set have matched forward voltages.

Perhaps spectral response would be the one thing that would be invariant
regardless but doesn't seem to apply to anything else at that price point..

This is the LED as written in the OP:

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

It's cheap, but also a good brand (Osram)

 

[Joerg]

On 2019-05-21 12:03, George Herold wrote:

On Tuesday, May 21, 2019 at 12:29:40 PM UTC-4, Joerg wrote:
On 2019-05-20 14:48, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need
1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an
indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf


Hey, German in a datasheet! Haven't seen that in a while.


If I use the same dirt cheap LED as transmitter I can get about
10kBit with small amplitude. However, I need a large amplitude, to be
able to detect bit level with logic 3V. (so over 2.4V)


3V is a tall order, almost needs a nuclear fireball as a transmit pulse.

This is off the cuff so use with a grain of salt:

What drives the LED? If a uC, maybe you could spare another pin and hook
the cathode of the LED to that instead of ground. For receive you could
switch the normal drive pin to low, then set this new pin to open drain
input with pull-up resistor, hoping the uC has internal pull-ups and
allows that. This would save the cost of an extra resistor which you'd
have to provide if there is no interal pull-up feature. Now the LED
would operate in photoconductive mode and when enough light falls on it
the voltage on this nwew pin will be pulled down. Since internal
pull-ups are usually in the 50-100k range (or a similar current source)
this method would only require sufficient light to generate 50uA of
photo current.

Hmm I just did the 'silly' measurement on the bench, some red led
(through hole) +15V with 10 meg ohm resistor to ground. Sitting
in the room lights I got about 0.4 V, pulling my desk light right on top
of it I got 3-4 Volts (0.3 uA) (desk lamp is 60W incandescent)

50 uA seems like a lot.

Yes, if you want to do that with internal pull-ups you'd have to blast
it with light. It must be focussed and the measly watt that comes out of
the 60W incandescent bulb is wasted in all directions. The other 59W
just heat the room.

Klaus is aware that this method is going to be inefficient, it just has
to be cheap. 1Mbit/sec pretty much precludes resistor values above the
typical 50-100k internal pull-up.

Time to consider switching to LED or at least CFL

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

 

On Tuesday, 21 May 2019 18:29:40 UTC+2, Joerg wrote:

On 2019-05-20 14:48, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need
1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an
indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf


Hey, German in a datasheet! Haven't seen that in a while.

Jawohl, sehr gut

However, I cannot get used to the German patent lingo

If I use the same dirt cheap LED as transmitter I can get about
10kBit with small amplitude. However, I need a large amplitude, to be
able to detect bit level with logic 3V. (so over 2.4V)


3V is a tall order, almost needs a nuclear fireball as a transmit pulse.

This is off the cuff so use with a grain of salt:

What drives the LED? If a uC, maybe you could spare another pin and hook
the cathode of the LED to that instead of ground. For receive you could
switch the normal drive pin to low, then set this new pin to open drain
input with pull-up resistor, hoping the uC has internal pull-ups and
allows that. This would save the cost of an extra resistor which you'd
have to provide if there is no interal pull-up feature. Now the LED
would operate in photoconductive mode and when enough light falls on it
the voltage on this nwew pin will be pulled down. Since internal
pull-ups are usually in the 50-100k range (or a similar current source)
this method would only require sufficient light to generate 50uA of
photo current.

That is actually the way I was thinking it could be hooked up. I just need a low loading resistance to increase the bandwidth (combating capacitive effects)

The price of the emitter is not relevant. So I was thinking to use a
white LED (to cover the entire spectrum of the RED LED detector), a
big one, and driving it with a lot of current, to get sufficient
photon level at the red LED to have both high speed and logic IO
level

An LED often have a different emission wavelength than the
responsivity wavelength, so that is why I use a white LED to be sure
wavelenghts overlap


They are often sensitive slightly below the wavelength of emitting
operation. So in your case possibly something around 600nm.

Yes

I will drive the emitter LED with high current, and then have a
sufficient low load resistor for the red detector LED to have high
bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is
there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf




If you do that make sure there are no humans or animals close by who
could retain eye damage from that.

I know I need to be carefull. It's actually also written quite explicitly in the datasheet, keep away from eyes

Cheers

Klaus

 

[Johann Klammer]

On 05/21/2019 11:49 PM, klaus.kragelund@gmail.com wrote:

On Tuesday, 21 May 2019 18:29:40 UTC+2, Joerg wrote:
On 2019-05-20 14:48, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need
1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an
indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf


Hey, German in a datasheet! Haven't seen that in a while.

Jawohl, sehr gut

However, I cannot get used to the German patent lingo



If I use the same dirt cheap LED as transmitter I can get about
10kBit with small amplitude. However, I need a large amplitude, to be
able to detect bit level with logic 3V. (so over 2.4V)


3V is a tall order, almost needs a nuclear fireball as a transmit pulse.

This is off the cuff so use with a grain of salt:

What drives the LED? If a uC, maybe you could spare another pin and hook
the cathode of the LED to that instead of ground. For receive you could
switch the normal drive pin to low, then set this new pin to open drain
input with pull-up resistor, hoping the uC has internal pull-ups and
allows that. This would save the cost of an extra resistor which you'd
have to provide if there is no interal pull-up feature. Now the LED
would operate in photoconductive mode and when enough light falls on it
the voltage on this nwew pin will be pulled down. Since internal
pull-ups are usually in the 50-100k range (or a similar current source)
this method would only require sufficient light to generate 50uA of
photo current.


That is actually the way I was thinking it could be hooked up. I just need a low loading resistance to increase the bandwidth (combating capacitive effects)


The price of the emitter is not relevant. So I was thinking to use a
white LED (to cover the entire spectrum of the RED LED detector), a
big one, and driving it with a lot of current, to get sufficient
photon level at the red LED to have both high speed and logic IO
level

An LED often have a different emission wavelength than the
responsivity wavelength, so that is why I use a white LED to be sure
wavelenghts overlap


They are often sensitive slightly below the wavelength of emitting
operation. So in your case possibly something around 600nm.


Yes

I will drive the emitter LED with high current, and then have a
sufficient low load resistor for the red detector LED to have high
bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is
there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf




If you do that make sure there are no humans or animals close by who
could retain eye damage from that.


I know I need to be carefull. It's actually also written quite explicitly in the datasheet, keep away from eyes

Cheers

Klaus

Here's my guesses for SFH350 photodiode
and also an SFH756 IR LED... unsure where I got the capacitance guess for that...
maybe here. their datasheet specs a minimum (on) photocurrent of
1.6uA.
That and the typical offset voltage of uC internal comparators
motivated the choice of resistor. I wanted around 300mV out.

RC corner freq:
detector SFH250 11+6pF
emitter SFH756 30+6pf
180k series R

SFH250 fc=52kHz ( 24kbaud) 19200
SFH756 fc=24kHz ( 12kbaud) 9600

with 5V rev bias
SFH250 3.5p+6p fc=93kHz (50kbaud) 38400
SFH756 9.6p+6p(guess) fc=56KHz (25kbaud) 19200

High speed it is not.
Also, stronger signals reduce reverse bias so will worsen bandwidth.

 

[George Herold]

On Tuesday, May 21, 2019 at 5:47:58 PM UTC-4, Joerg wrote:

On 2019-05-21 12:03, George Herold wrote:
On Tuesday, May 21, 2019 at 12:29:40 PM UTC-4, Joerg wrote:
On 2019-05-20 14:48, klaus.kragelund@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I need
1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an
indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf


Hey, German in a datasheet! Haven't seen that in a while.


If I use the same dirt cheap LED as transmitter I can get about
10kBit with small amplitude. However, I need a large amplitude, to be
able to detect bit level with logic 3V. (so over 2.4V)


3V is a tall order, almost needs a nuclear fireball as a transmit pulse.

This is off the cuff so use with a grain of salt:

What drives the LED? If a uC, maybe you could spare another pin and hook
the cathode of the LED to that instead of ground. For receive you could
switch the normal drive pin to low, then set this new pin to open drain
input with pull-up resistor, hoping the uC has internal pull-ups and
allows that. This would save the cost of an extra resistor which you'd
have to provide if there is no interal pull-up feature. Now the LED
would operate in photoconductive mode and when enough light falls on it
the voltage on this nwew pin will be pulled down. Since internal
pull-ups are usually in the 50-100k range (or a similar current source)
this method would only require sufficient light to generate 50uA of
photo current.

Hmm I just did the 'silly' measurement on the bench, some red led
(through hole) +15V with 10 meg ohm resistor to ground. Sitting
in the room lights I got about 0.4 V, pulling my desk light right on top
of it I got 3-4 Volts (0.3 uA) (desk lamp is 60W incandescent)

50 uA seems like a lot.


Yes, if you want to do that with internal pull-ups you'd have to blast
it with light. It must be focussed and the measly watt that comes out of
the 60W incandescent bulb is wasted in all directions. The other 59W
just heat the room.

Right, Just a silly test and Klaus will have to measure his LED photo-
response.
I was thinking that a laser, with maybe a diffusor right in front of the
LED/detector, might give more light.

Klaus is aware that this method is going to be inefficient, it just has
to be cheap. 1Mbit/sec pretty much precludes resistor values above the
typical 50-100k internal pull-up.

Time to consider switching to LED or at least CFL

I've got an LED light, but I do like an incandescent lamp too.
Nice broad-band black-body source. And it extends into the IR...
which is good for checking some things.

George H.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

 

On Monday, 20 May 2019 23:48:13 UTC+2, klaus.k...@gmail.com wrote:

Hi

For a test setup I am toying with an isolated interface where I need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

If I use the same dirt cheap LED as transmitter I can get about 10kBit with small amplitude. However, I need a large amplitude, to be able to detect bit level with logic 3V. (so over 2.4V)

The price of the emitter is not relevant. So I was thinking to use a white LED (to cover the entire spectrum of the RED LED detector), a big one, and driving it with a lot of current, to get sufficient photon level at the red LED to have both high speed and logic IO level

An LED often have a different emission wavelength than the responsivity wavelength, so that is why I use a white LED to be sure wavelenghts overlap

I will drive the emitter LED with high current, and then have a sufficient low load resistor for the red detector LED to have high bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED, is there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf

Just an update:

I got it working now at 10MBit, allthough only with a smaller amplitude (I will use a "free" comparator in the micro to detect the 100mV signal

Solution: Blasting a cheap Red LED with a big Red Led, with a resistor on the detector side to decrease fall times

Cheers

Klaus

 

[Joerg]

On 2019-05-29 06:42, klaus.kragelund@gmail.com wrote:

On Monday, 20 May 2019 23:48:13 UTC+2, klaus.k...@gmail.com wrote:
Hi

For a test setup I am toying with an isolated interface where I
need 1 Mbit speed if possible

The receiver is a dirt cheap red LED: (which is also used as an
indicator when the link is not active)

https://dammedia.osram.info/media/resource/hires/osram-dam-2493888/LH%20R974.pdf

If I use the same dirt cheap LED as transmitter I can get about 10kBit
with small amplitude. However, I need a large amplitude, to be able to
detect bit level with logic 3V. (so over 2.4V)

The price of the emitter is not relevant. So I was thinking to use
a white LED (to cover the entire spectrum of the RED LED detector),
a big one, and driving it with a lot of current, to get sufficient
photon level at the red LED to have both high speed and logic IO
level

An LED often have a different emission wavelength than the
responsivity wavelength, so that is why I use a white LED to be
sure wavelenghts overlap

I will drive the emitter LED with high current, and then have a
sufficient low load resistor for the red detector LED to have high
bandwidth (1MHz)

But, is this a good idea? Will I be overdriving the detector LED,
is there some kind of saturation effect?

For the setup, I could use Win Hills 200A pulser:

https://www.dropbox.com/sh/svr7q23agwuvtat/AABdQSJq4d-7Qye96NLCCb9pa?dl=0&preview=200A-LED-pulse.pdf



Just an update:

I got it working now at 10MBit, allthough only with a smaller
amplitude (I will use a "free" comparator in the micro to detect the
100mV signal

Solution: Blasting a cheap Red LED with a big Red Led, with a
resistor on the detector side to decrease fall times

Great! Thanks for the feedback. 100mV with a resistor small enough to
allow 10Mbit/sec is a lot. Probably the big red LED has enough power to
set a table cloth on fire.

--
Regards, Joerg

http://www.analogconsultants.com/

 

And by the way, at 1kbit I have full 3V amplitude, no need for a comparator (Rl xxx times larger than at 10MBit)

Cheers

Klaus

 

It's a 700mA LED, it can perform even better if I use a tube to concentrate the light