Episode 1849 can you make the transfer curve of an optocoupler linear? docs.broadcom.com/doc/HCNR200... docs.broadcom.com/doc/AV00-02... www.vishay.com/docs/83708/app... Be a Patron: / imsaiguy
Still, 1 kilo-ohm resistors have kept it simple for me with LEDs on 12 Volts DC power (in that particular case, my modified Halogen spotlight which I chucked in metal halide HID lamp, and I opted to use LED-illuminated rocker switches - it worked fine with voltage up to almost 15 Volts DC from the Lithium-ion battery pack). Some LEDs are quite efficient nowadays, so it doesn't matter if it consume 1 or 20 mA, the outcome is still the same.
@@aviandragon1390 In a professional environment, appearing to 'guess' might not be the best career move. Being able to guess correctly is great, but it perhaps needs to be backed-up with page of 'Design File Memo' supporting data. I've seen six pages of math to select a rivet, so EEs should be seen to take as much care. Up to you !!
In a professional product, of course you want to calculate for whatever parameter is most important to the design. This guy is just spitballing a circuit for demonstration. I'm not sure the same level of care is required here.
Been there, done that. Used a bulkhead (bolt-shaped) optocoupler to couple audio into a system where we had to ensure that absolutely nothing went the other way. The LED and Phototransistor were given the old Y=mX+b treatment to linearize and center them within the available dynamic range. We didn't even spend much time analyzing it, but very simply employed trimmer pots, with very generous range, to set the bias (offset) and gain on both sides of the bulkhead. Twiddled four trimmer pots, while viewing with a 'scope. It worked like an absolute champ.
For the first simple schematic you should use a current source not a voltage buffer with a resistor. It's easy: just connect the feedback to in- to the point between the diode and the resistor. This should already improve linearity.
Sure, but from an educational point of view the first, naive approach is good to show you how it does NOT work and why. And leaves space for improvement and success stories 😉
Yep, to a point. Nothing is literally linear, unfortunately. Not to mention the subtle differences between a few different transistors in an OP-AMP chip causing rare instance of weird events to happen in a specific application (a few people, including me, have been bitten by it). Analog wizardry is rarely simple.
It was my idea at the very first glance. Thanks you wrote it. Voltage to current converter, just one resistor in negative feedback of first op-amp, thus voltage produces directly current of LED, LED light is ~propotional to the current and light intensity on the Base of transistor works like Base current, at the end Emitter current is proportional to Base current (light intensity). Almost perfect isolated follower. Then later we could add some extra circuits, current mirrors, thermal compensation and even optocoupled negative feedback.
This is a good compensation circuit which reminds also of how to work with thermocouples to measure rf power by comparing it with dc power. It depends on finding matching couplers or the special matched dual coupler. An alternative to transmit an analog signal with optocouples is to convert the analog signal with analog means to analog pwm, which can be in fact analog but looks of course very much digital and is not quite as fast as with the optocouples.
I designed some 4-20mA transmitters using IL300's. Been in production for years now. The IL300's are binned according to their CTR which helps reduce the variation from opto to opto. They are a little pricey but work well.
Those are called vactrols in the audio electronics world. You can also diy it, quite a simple and powerful part. An alternative to it is the FET output opto H11F1. Have you seen that part youtube.com/@IMSAIGuy ?
"Pro tips" from the stone age! OMG! The two separate optocouplers are obsolete for decades! IL300 is friend if you REALLY wanne do it plain analog, which is still OK in many situations. Modern isolation amplifiers all work digital using a delta sigma converter on the input, transmitting digital (optical, inductive or capacitive) and converting back to analog using a filter. Works like a charm! If you REALLY learn something usefull, get an IL300 and play with it!
To achieve linearity you should drive the LED using current rather than voltage. That could be achieved by moving the OpAmp negative feedback to the top of the resistor.
thanks for the video. when we talk about isolation we should have 2 separate isolated power supplies for the circuit another thing, input voltage of this circuit is limited by the power supply back in the old days we used cpc5710 linear opamp with a gain of 6 we used it for telephone DC voltage monitoring.
Interestingly my older series 3478a uses little transformers to isolate the control from the analogue side. The later models used opto couplers. Boringly, I just looked at the circuit and it is just two micro controllers sharing the connection via one pin and a differential driver for the transformer coupling. No analogue magic.
this reminds me of a regen receiver kit i built that used an hp opto isolator - “Build the OCR II receiver” in the september 2000 issue of QST. the optocoupler isolated the regeneration from the antenna. How did *that* work? because it did (sort of) work. aha. iirc it used an IL 300…
By using TWO optos, you will run into trouble in the long runs as they 1) degrade over time & 2) optos are very temperature sensitive. Devices like the IL300 likely mitigate the issue. Looking forward to part 2 & testing with things like temperature.
What about the load on the right hand side, that might change but the other opto on the left won't react to it as the feedback is sorta fudged ? or not lol !.....cheers.
This video is well explained and but still.....I wonder what would be the kind of analog (practical examples) of such applications where you strictly need an identical signal produced by the opto-coupler ? Thank you.
I'm curious to see how this is done for HFC (Hybrid Fiber Coax) cable TV systems? In that application, from the cable TV head-end, there's a distribution of the signal over fiber optic transmission out into the CATV distribution system. There's a laser that's analog modulated by the RF carriers coming out of the head-end, probably going from about 30 MHz up to 1 GHz or more - the whole forward passband of the CATV system. And this has to be pretty linear given the fancy modulation used by ATSC HDTV video transmission and a pretty dense (256-QAM) constellation for your DOCSIS 3 Internet cable modems. Even more demanding with DOCSIS 3.1 and beyond using OFDM and squeezing the bit even harder..
@@JacquesMartini Then you try it out in circuit and see what value works best. Capacitor substitution boxes are very handy for lower frequency circuits.
Why would you put the optos in series and not just parallel? Because now the non-linearity at the feedback opto is basically doubled, since they are in series?
if in parallel, you have two LEDs in parallel. since the Vf curve will be not the same there will be current hogging of one LED before the other and they will not match well
I can imagine there are still applications for these kind of circuits if you want to keep the parts count and complexity down (by not going to A/D and D/A etc...). If it doesn't have to be super accurate on the isolated side then these analog solutions should work fine.
6:40 The output of the second photo transistor is feeded by the +V of the high voltage, so how is that isolated if you bring this back to the input of your OpAmp ? Or is the +V of the second Opto-coupler another powersource then the +V of the high voltage side ? Also, you have 3 ground points, if all 3 are connected, there isn't any isolation between the low and high voltage side, or am i missing something ?
That's an interesting circuit, using an op-amp's negative feedback to liberalize the transfer function. But it is just unipolar. There are linear, bipolar optoisolators designs that use a pwm signal.
@@IMSAIGuy now my turn! ('???') What I meant was placing the second LED on the right and dioide on the left to complete the feedback loop. I edited my post to clarify it. This design is liken a transmiter/receiver set which only measures the power needed at the transmitter side. Today, the radio (specifically TV) stations measure the transmitted power some distance away in multiple points and send the power readings back to the transmitter to adjust the power accordingly. For example, for rainy/snowy or sunny days or reducing power at night . Then power would not be wasted. In PSUs, I think there are opto couplers to adjust the PWM. But the transformer's output voltage is used to compare and adjustments.
I thought purpose of opto-coupler is input/output isolation , but it seems like you are defeating that purpose once you bring the connection on the input side . May be I am missing the something in this experiment , can you help me to understand , if possible
yeah this method has been around, but still relies on the opto component matching which is crap. Trimpots also no good for a professional design. Thanks anyhow on the refresher.
@@IMSAIGuy Thnks, some people use 6N137 for MIDI, it's almost 2€ piece, got some PC900 (obsolete, used by Roland), they all have some logic inside. a PC817 + 74LVC1G175 (Single Schmitt-Trigger Buffer) would be cheaper, although it might cost a litle more more pcb real estate
Excellent work. I'm just wondering if you could do the same with "True Feedback" like use 2 optos in opposite directions and cancel each other's nonlinearity. That way you wouldn't have to have matching optos and achieve true output feedback. Hmmm.... what would the circuit look like... I'm going to work it out I think.... :)
Wow, ... that let me think of an 'isolation-device, for amateur-radio use. How about to 'build' an isolation 'USB-thingy', to 'de-couple' hum and noise (spikes and 'deathly voltages'), from the pc or cell-phone to the 'HAM-radio devices? Like an isolation-transformer for Audio-Frequencies, but for USB, instead? That would solve some problems, that a lot of HAM's have, when using electronical devices, with the computer and the radio. Is it possible to make such an 'USB -transformer' with the opto-couplers ? 73 de Markus - db9pz - (Loc: JN39fq - 5km/3miles east of LX !)
A wise man once advised, "Never use a 1k resistor; people will assume that you just guessed."
Still, 1 kilo-ohm resistors have kept it simple for me with LEDs on 12 Volts DC power (in that particular case, my modified Halogen spotlight which I chucked in metal halide HID lamp, and I opted to use LED-illuminated rocker switches - it worked fine with voltage up to almost 15 Volts DC from the Lithium-ion battery pack). Some LEDs are quite efficient nowadays, so it doesn't matter if it consume 1 or 20 mA, the outcome is still the same.
I have a box of 1.1k just for that reason :))))
If it works, who cares if you guessed? 🤔
@@aviandragon1390 In a professional environment, appearing to 'guess' might not be the best career move. Being able to guess correctly is great, but it perhaps needs to be backed-up with page of 'Design File Memo' supporting data.
I've seen six pages of math to select a rivet, so EEs should be seen to take as much care.
Up to you !!
In a professional product, of course you want to calculate for whatever parameter is most important to the design. This guy is just spitballing a circuit for demonstration. I'm not sure the same level of care is required here.
Been there, done that.
Used a bulkhead (bolt-shaped) optocoupler to couple audio into a system where we had to ensure that absolutely nothing went the other way.
The LED and Phototransistor were given the old Y=mX+b treatment to linearize and center them within the available dynamic range.
We didn't even spend much time analyzing it, but very simply employed trimmer pots, with very generous range, to set the bias (offset) and gain on both sides of the bulkhead.
Twiddled four trimmer pots, while viewing with a 'scope. It worked like an absolute champ.
For the first simple schematic you should use a current source not a voltage buffer with a resistor. It's easy: just connect the feedback to in- to the point between the diode and the resistor.
This should already improve linearity.
Sure, but from an educational point of view the first, naive approach is good to show you how it does NOT work and why. And leaves space for improvement and success stories 😉
Yep, to a point. Nothing is literally linear, unfortunately. Not to mention the subtle differences between a few different transistors in an OP-AMP chip causing rare instance of weird events to happen in a specific application (a few people, including me, have been bitten by it). Analog wizardry is rarely simple.
It was my idea at the very first glance. Thanks you wrote it. Voltage to current converter, just one resistor in negative feedback of first op-amp, thus voltage produces directly current of LED, LED light is ~propotional to the current and light intensity on the Base of transistor works like Base current, at the end Emitter current is proportional to Base current (light intensity). Almost perfect isolated follower. Then later we could add some extra circuits, current mirrors, thermal compensation and even optocoupled negative feedback.
This is a good compensation circuit which reminds also of how to work with thermocouples to measure rf power by comparing it with dc power. It depends on finding matching couplers or the special matched dual coupler.
An alternative to transmit an analog signal with optocouples is to convert the analog signal with analog means to analog pwm, which can be in fact analog but looks of course very much digital and is not quite as fast as with the optocouples.
Thank-you for this video - you unlocked something that was a mystery to me for some time.
Perfect timing! I was just pondering about isolated analog inputs the other day :D
Burr Brown ISO100 was a very fancy version of that scheme, in one huge ceramic package.
With one huge price tag too, I'm sure.
@@tonyfremont Yup. Ceramic and metal-capped chips are usually insanely expensive, even before the pointless economic inflation.
I designed some 4-20mA transmitters using IL300's. Been in production for years now. The IL300's are binned according to their CTR which helps reduce the variation from opto to opto. They are a little pricey but work well.
Nice work ! Many thanks.
Your movies are thrilling like a crime. 😬
Quite a nice part Iv used in the audio world before is the NSL-32, resistive output opto, basically a IR led and a LDR
Those are called vactrols in the audio electronics world. You can also diy it, quite a simple and powerful part.
An alternative to it is the FET output opto H11F1. Have you seen that part youtube.com/@IMSAIGuy ?
"Pro tips" from the stone age! OMG! The two separate optocouplers are obsolete for decades! IL300 is friend if you REALLY wanne do it plain analog, which is still OK in many situations. Modern isolation amplifiers all work digital using a delta sigma converter on the input, transmitting digital (optical, inductive or capacitive) and converting back to analog using a filter. Works like a charm! If you REALLY learn something usefull, get an IL300 and play with it!
This was what I was thinking they should do. You could even have a 555 on one side for simple pwm. Half retro junk.
A professional designer should also consider the cost of the product. Do you know the price difference between these two optocouplers ?
To achieve linearity you should drive the LED using current rather than voltage. That could be achieved by moving the OpAmp negative feedback to the top of the resistor.
I thought you'd be talking about resistive optocouplers (vactrols), but no... Pretty interesting.
thanks for the video. when we talk about isolation we should have 2 separate isolated power supplies for the circuit another thing, input voltage of this circuit is limited by the power supply back in the old days we used cpc5710 linear opamp with a gain of 6 we used it for telephone DC voltage monitoring.
Interestingly my older series 3478a uses little transformers to isolate the control from the analogue side. The later models used opto couplers. Boringly, I just looked at the circuit and it is just two micro controllers sharing the connection via one pin and a differential driver for the transformer coupling. No analogue magic.
I thought about this and concluded it could not be done,
so I am amazed by the solution.
this reminds me of a regen receiver kit i built that used an hp opto isolator - “Build the OCR II receiver” in the september 2000 issue of QST. the optocoupler isolated the regeneration from the antenna. How did *that* work? because it did (sort of) work. aha. iirc it used an IL 300…
Half way between full-analog and full-digital (ADC-DAC) would be PWM-unPWM with a suitable pulse frequency.
A IC like that IL300 was used in many telephone modem.
By using TWO optos, you will run into trouble in the long runs as they 1) degrade over time & 2) optos are very temperature sensitive.
Devices like the IL300 likely mitigate the issue. Looking forward to part 2 & testing with things like temperature.
That is indeed a cunning circuit.
Wow! Thank you for this video! I think you might just have solved my issue hehehe
What about the load on the right hand side, that might change but the other opto on the left won't react to it as the feedback is sorta fudged ? or not lol !.....cheers.
This video is well explained and but still.....I wonder what would be the kind of analog (practical examples) of such applications where you strictly need an identical signal produced by the opto-coupler ? Thank you.
regulation circuits of high voltage power supplies or transmitters
I'm curious to see how this is done for HFC (Hybrid Fiber Coax) cable TV systems? In that application, from the cable TV head-end, there's a distribution of the signal over fiber optic transmission out into the CATV distribution system. There's a laser that's analog modulated by the RF carriers coming out of the head-end, probably going from about 30 MHz up to 1 GHz or more - the whole forward passband of the CATV system. And this has to be pretty linear given the fancy modulation used by ATSC HDTV video transmission and a pretty dense (256-QAM) constellation for your DOCSIS 3 Internet cable modems. Even more demanding with DOCSIS 3.1 and beyond using OFDM and squeezing the bit even harder..
Oh man. I ordered some LM358's, and they came in the same package.
The circuit will get complicated but there are another way to send PWM through the second optocoupler from output to input.
Hey, you have a four channel scope, so it would be a good idea to show input an output at the same time!
or even use XY mode and look for a straight line. Great things for you to try
Well there goes your isolation!!! :)
This scheme has been known for more than 10 years.
What happens if you introduce a little DC offset to the signal to help overcome the diode forward biasing ? Does that reduce the signal corruption ?
yes, you can try and find a more linear portion of the LED and only use that. not as good as this, but might be fine in some applications
That's interesting, how did you calculate the value of the capacitor to make it stable?
Experience and rule of thumb!
@@JacquesMartini Then you try it out in circuit and see what value works best. Capacitor substitution boxes are very handy for lower frequency circuits.
Why would you put the optos in series and not just parallel? Because now the non-linearity at the feedback opto is basically doubled, since they are in series?
if in parallel, you have two LEDs in parallel. since the Vf curve will be not the same there will be current hogging of one LED before the other and they will not match well
@@IMSAIGuysomething that could be taken care of with a current mirror? 😊
I can imagine there are still applications for these kind of circuits if you want to keep the parts count and complexity down (by not going to A/D and D/A etc...). If it doesn't have to be super accurate on the isolated side then these analog solutions should work fine.
All this stuff is available fully integrated in a single, cost effective IC nowadays.
6:40 The output of the second photo transistor is feeded by the +V of the high voltage, so how is that isolated if you bring this back to the input of your OpAmp ?
Or is the +V of the second Opto-coupler another powersource then the +V of the high voltage side ?
Also, you have 3 ground points, if all 3 are connected, there isn't any isolation between the low and high voltage side, or am i missing something ?
the two sides have different supplies, total isolation. see part 2
@@IMSAIGuy thank you for the quick reply, i didn't see part 2 yet untill now. Grtz
Can I use PC817 optos for it ?
Neat
That's an interesting circuit, using an op-amp's negative feedback to liberalize the transfer function. But it is just unipolar. There are linear, bipolar optoisolators designs that use a pwm signal.
this would also if you just DC offset the signal to 1/2 Vcc
@@IMSAIGuy Then it wouldn't be linear.
@@williamogilvie6909 I just showed it is
@@IMSAIGuy You don't know the definition of linear. The function f(x) = x + A is not linear.
Y = Ax + B seems to be a straight line, no? that IS the definition of linear.
Use a cheap opamp instead of fancy one?
For some applications like switching power supply, wouldn't be better to place the 2nd opto LED on the other side?
??? there is an input side and output side, just turn the paper 180 degrees
@@IMSAIGuy now my turn! ('???') What I meant was placing the second LED on the right and dioide on the left to complete the feedback loop. I edited my post to clarify it. This design is liken a transmiter/receiver set which only measures the power needed at the transmitter side. Today, the radio (specifically TV) stations measure the transmitted power some distance away in multiple points and send the power readings back to the transmitter to adjust the power accordingly. For example, for rainy/snowy or sunny days or reducing power at night . Then power would not be wasted.
In PSUs, I think there are opto couplers to adjust the PWM. But the transformer's output voltage is used to compare and adjustments.
@@bayareapianist compete the feedback? remember this is high voltage isolated.
HCPL-7800 work well!
I would have thought that an optocoupler driven by a PWM signal would have been a better approach.
Yes, non linearity then becomes an advantage. The trade off is bandwidth though, as optocouplers are typically slow.
I thought purpose of opto-coupler is input/output isolation , but it seems like you are defeating that purpose once you bring the connection on the input side . May be I am missing the something in this experiment , can you help me to understand , if possible
your are missing. see part 2
yeah this method has been around, but still relies on the opto component matching which is crap. Trimpots also no good for a professional design. Thanks anyhow on the refresher.
What's the optocoupler model?
Could they substitute 6n137 / 6n139 (in midi circuits)?
I used a PC817 but any optocoupler should work fine
@@IMSAIGuy Thnks, some people use 6N137 for MIDI, it's almost 2€ piece, got some PC900 (obsolete, used by Roland), they all have some logic inside. a PC817 + 74LVC1G175 (Single Schmitt-Trigger Buffer) would be cheaper, although it might cost a litle more more pcb real estate
Excellent work.
I'm just wondering if you could do the same with "True Feedback" like use 2 optos in opposite directions and cancel each other's nonlinearity. That way you wouldn't have to have matching optos and achieve true output feedback. Hmmm.... what would the circuit look like... I'm going to work it out I think.... :)
Wow, ... that let me think of an 'isolation-device, for amateur-radio use.
How about to 'build' an isolation 'USB-thingy', to 'de-couple' hum and noise (spikes and 'deathly voltages'), from the pc or cell-phone to the 'HAM-radio devices?
Like an isolation-transformer for Audio-Frequencies, but for USB, instead? That would solve some problems, that a lot of HAM's have, when using electronical devices, with the computer and the radio.
Is it possible to make such an 'USB -transformer' with the opto-couplers ?
73 de Markus - db9pz -
(Loc: JN39fq - 5km/3miles east of LX !)
.01uF is called 10nF overseas . . .
nF is for the youngsters 😎
@JacquesMartini I'm overseas (NZ) and we call it by both.
If you go back farther, pF is for youngsters. uuF for the old timers!
Can't you just feed it into a logic gate? I am assuming you want logic in and out without linearity?
Doing analog linear isolation. Logic is not linear. Using A/D, then D/A would introduce delays and other artifacts to the signal
The IL300 is somewhat common and cheap.
One popular application was isolated 4-20mA current loops. Good use there.