Floating probe! For $2.50

[JS]

  Hey, I'm trying to build a floating probe, this design is based on the tek A6902B, for the optocoupler LF insulator. Tweaking the design a bit I was able to get 230kHz BW from it using a TL082, a NE5532 and two 4N35. The cost I mention is for those four ICs, of course, the complete build cost will be higher but still quite cheap compared to anything on the market. 200kHz BW is nothing to write home about but quite useful for any audio applications among others.

  I leave you a schematic and the rising edges with two different compensations, for 100kHz and 230kHz -3dB bandwidth. As you can see for the 230kHz BW there's a bit of a jiggle so I tweaked the filter to get the other compensation. Also, for the 230kHz the delay in the signal makes way over 90º phase shift at -3dB point. I didn't tried to trim for gain or offset, that shouldn't be too hard to make and having a spare amplifier in each side leaves the option for tweaking. I didn't tested linearity either, but it doesn't look too bad from the waveforms I've observed.

  There are faster linear optocouplers, like the 6N136 I'd like to test but I don't have the sim model and can't source the part locally. I'd like to run a test with it!

  I think I could find this quite useful, as for higher frequencies using a transformer is not too hard to get around. As you could observe in the 6902 schematics, TEK is using a transformer to get from the optocoupler (which goes up to a few kHz) to 20 MHz, so something similar could be added, but I can't source HF opamps locally either so I gave up about that for now. It doesn't seem to have trouble running from 6V or even lower (other than the clipping level) so it could very well run from two 9V batteries to makes things easier, or build a few channels with a 9V battery for each input and a dedicated PSU for the output. The battery on the floating side should last about 100 hours in this condition, I guess I could try to get that longer using lower DC currents for the optocouplers or use two 18650 for the input and a dedicated supply for the output, which could also work without much hassle.

JS

[mikerj]

It's not an isolated design if the input and output op-amps are powered from the same voltage rails.

[StillTrying]

Interesting, reminds me of my flashing LEDs at a photo diode experiments!

Why is it inverting, intuition says non-inverting would have better stability, a bit better bandwidth, and as high an input impedance as you want.

[JS]

It's not an isolated design if the input and output op-amps are powered from the same voltage rails.

I did mentioned the battery for the input side at least, not much of a different running the sim to check how it went but essential for the actual device. That's the next step and time to check CMRR.

Interesting, reminds me of my flashing LEDs at a photo diode experiments!

Why is it inverting, intuition says non-inverting would have better stability, a bit better bandwidth, and as high an input impedance as you want.

I have an extra amplifier to buffer the signal at the input, in the prototype I used a tl082, but I wanted the virtual earth circuit to keep the output of the optocoupler to that node, making equivalent circuits in both sides to get the best tracking between them and get the best transfer response as the output of the optocouplerd only have signal current but not signal voltage.

JS

[edavid]

I think you would need to select matching optocouplers.  They have a very wide spread of CTR.

[StillTrying]

I wanted the virtual earth circuit to keep the output of the optocoupler to that node, making equivalent circuits in both sides to get the best tracking between them and get the best transfer response

I see, makes a bit of sense.

[schmitt trigger]

If I remember properly, Avago, or whatever is nowadays called, had a dual matched optocoupler for precisely this type of circuits.

As other posters have mentioned the CTR tracking of independent optocouplers will  show significant differences depending on their bias point.

[Gyro]

You did pretty well for 4N35s.

One of those linear optocouplers with the built-in feedback phototransistor is the TIL300. It's quite an old part now but on paper, it's slightly faster than the 4N35, and with none of the CTR variation issues of course. I'd be surprised if there isn't a faster equivalent available these days.

[JS]

You did pretty well for 4N35s.

One of those linear optocouplers with the built-in feedback phototransistor is the TIL300. It's quite an old part now but on paper, it's slightly faster than the 4N35, and with none of the CTR variation issues of course. I'd be surprised if there isn't a faster equivalent available these days.

Thanks! I was doubious about this design but I had the things to set a breadboard and test, it did surprised me. I wish I could source new and good parts but here is pretty painful to do. Once in a while I make an order overseas for specialized parts but for now I have to do whith what I can find aroubd here.

If I remember properly, Avago, or whatever is nowadays called, had a dual matched optocoupler for precisely this type of circuits.

As other posters have mentioned the CTR tracking of independent optocouplers will  show significant differences depending on their bias point.

  I couldn't find any locally, I should see what parts I can get and test again. The concept working with this is a nice start, and even to an useful degree I think, so I might build one or two and upgrade when a better solution appears. I should check for stability but trimming the offset in each use only takes a few seconds and I expect gain to be even more stable.

  Also, one think I like to take in my projects are easy to get parts, if I can source them locally most likely anyone will so they can get a good use of my design. TL082 and 4N35 are jelly beans that can be found anywhere... If something useful comes from them seems great for me, this could help a lot of beginners not blowing their scope and getting much more comfortable with them.
  Then I of course would like a better behaved version, with faster response and better stability.
  The limiting factor for the speed seems to be the input side compensation, I should run some more tests to see how far I can improve this, the virtual ground makes a much better reaponse than just the ootocoupler, as there's no signal voltage at the outout device and that takes the BW from about 40kHz up to over 200kHz (simulated and real world confirmation) but at that frequency the input side compensation starts to be a problem, having a peak which gets compensated by the LPF of the output side to get a flatish response up to 230kHz. At 100kHz the output LPF dominates, and the response is slower but much smoother.
  Having the optocoupler inside the feedback loop makes it tricky, as it's expected, let's see how far I can take it...

JS

[Yansi]

I don't see too many uses (if any) for this thing.

At least switch the coupler to something appropriate, such as IL300.

Or just buy an analog isolation amplifier!  For example Si8920 costs under $4 and has pretty darn good linearity, is even differential and the bandwidth is at least as good as your circuit. (probably better).

//EDIT: BW is a bit under 1MHz. Also the CMTI of 75kV/us is far better than what you would ever get from a nonshielded (even shielded) optocoupler.

[JS]

Other use could be injection amplifier, for the LF wiuld be better than any transformer coming down to DC.

I don't see too many uses (if any) for this thing.

At least switch the coupler to something appropriate, such as IL300.

Or just buy an analog isolation amplifier!  For example Si8920 costs under $4 and has pretty darn good linearity, is even differential and the bandwidth is at least as good as your circuit. (probably better).

//EDIT: BW is a bit under 1MHz. Also the CMTI of 75kV/us is far better than what you would ever get from a nonshielded (even shielded) optocoupler.

Yes, better and specialized parts makes for better circuit and easier to design, no doubts! If I could get some I will.

JS

[Gyro]

If you want to continue with the opto approach, the Broadcom HNCR200 linear opto is good for about 1.5MHz.

[JS]

If you want to continue with the opto approach, the Broadcom HNCR200 linear opto is good for about 1.5MHz.

I couldn't find any reference to the propagation delay in the data-sheet, that will limit the application of this circuit inside the feedback loop. Taking it out of the feedback loop will make tracking and offset compensation a real problem, there are some designs with that approach and might work reasonable fine for some applications, but drift will be bad. In any case it looks like I'm close to the limit of the 4N35. 6N136 is much faster too and I think I can get them, but not locally either. I've seen a project here with one of those, I should check how well it did, the switching takes 0.3µs typ and 1µs max, but this is not switching so specs means very little. All specs are with output voltage with a finite resistnace, not current in a virtual earth config as I'm using them, so I expect to get a bit over of those specs.

JS

[Gyro]

No, you're right the datasheet is light on information. The 1.5MHz bandwidth is quoted for the Fig 16 application circuit (the transistor one). I'm not sure how linear that one would be.

It sounds as if sourcing the part isn't going to be an option for you at the moment anyway.

[JS]

No, you're right the datasheet is light on information. The 1.5MHz bandwidth is quoted for the Fig 16 application circuit (the transistor one). I'm not sure how linear that one would be.

It sounds as if sourcing the part isn't going to be an option for you at the moment anyway.

Fig 12 makes a more similar thing to what I've made, using opamps would make for a more libear design I guess. Not too crazy about linearity for an 8 bit converter anyway... If they are using it like this, the propagation delay must be really low, otherwhise it will become unstable, note the lack of compensation other than the optocoupler itself, I think that's pushing a bit too much.

I could get a HCNR200 for the small amount of $20 plus shipping... The 4N35 is $0.50 here for reference. I will anotate that part for my next order overseas, but it's likely to be for the next year or close to that...

JS

[Yansi]

HCNR200 is well under $4  as far as I can tell. (Mouser, TME)

If you go for speed, then don't  4N35. It is a "standard slow" coupler if I remember correctly.  Any photo-transistor coupler will be likely very very slow with low collector currents.

//What kind of distributors are there in Argentina?

[JS]

HCNR200 is well under $4  as far as I can tell. (Mouser, TME)

If you go for speed, then don't  4N35. It is a "standard slow" coupler if I remember correctly.  Any photo-transistor coupler will be likely very very slow with low collector currents.

//What kind of distributors are there in Argentina?

4N36 was what I had in the bin. The kind of distributors we have here are mainly the old man who used to repair TVs so they know eventeverithing about vertical deflection but very little about any modern-ish components. Then you have the bigger stores which probably have the same old man in the desk behind the scenes and a bunch of kids dealing with the public, then you have some individual selling dedicated components for 20x what it should cost and some major importer that will get you anything you want for 10x what it costs but within the month (or two). In my city only the first two options, and 300km from here, in Buenos Aires, all of the mentioned flavors.

When I was building guitar pedals and tube amps everithing was great, find an old man in a slow day or a kid like me interested I what I was doing and they would allow me to go into the pile of junk searching for all the good NOS they had... JRC4558DD, TA7136P, NE3101, silver mia caps, a wide variety of germanium devices, etc.

  With all that, any serious development here gets really frustrating pretty fast and working on my own becomes a thing about what can I do with the components I can find knowing it will be pretty easy with modern components. Building small gadgets to sell as a kickstarter seems like a very distant future seen from here, so to calm my thist of development I run into this projects useful for the beginers to help them in tooling their lab. For a day job I do industrial automation where all this is not usually a problem and I can work more freely.

JS

[StillTrying]

Using 2 optos in push-pull instead of any feedback , and using just the photo diode part, I can get the BW well over 1MHz.
At least in the simulation it's linear with IN between +/- 2V @ +/- 10mA, OUT is unfortunately +/- 18uA, +/- 80mV.

[JS]

Thanks for that, I'll try it in the breadboard!

Did you observed the propagation dekay?

JS

[StillTrying]

"Did you observed the propagation delay?"

I think it depends on how you define it. If I put a 1us wide 1.5V pulse through the sim it comes out looking like that, the half signal height delay is 120ns.

There some real 1us pulses through red, green and white 5mm LEDs viewed by a photo diode here.- the last image.
https://www.eevblog.com/forum/chat/20w-halogen-bulb-viewed-by-a-photodiode/msg1751210/#msg1751210

I still don't know why the light shape(and current) through a red LED is so curved for the first ~0.5us, I suspect IR LEDs might be similar, but not tried one yet.

"I'll try it in the breadboard!"
In the sim the current through the LEDs side is about 6mA, +/-5mA on the signal peaks, and on the photodiodes side about 7uA +/- 6uA, good luck!

[JS]

I'll run a sim now to define a full circuit to take to the breadboard. I guess some discrete transistor should be used to take those 6µA up.

The fast response might make it suitable to work inside a feedback loop as well, but the differential approach also makes sense to retrieve higher signal at the output, many things to check now, thanks for the idea!

JS

[StillTrying]

"I guess some discrete transistor should be used to take those 6µA up."

I'd just view the +/- 80mV across a 5k1 or 5k6 resistor for now myself. You're going to need at least some resistance from the photo diodes mid point to 0V because a 0.5uA difference in the PD quiescent currents would make that point go to near one of the rails.
It's a good place for a CFA amp, but I've not had a massive amount of luck on a breadboard where every pF at the summing point counts.

"the differential approach also makes sense to retrieve higher signal at the output"

I used 2 optos working in anti-phase to hopefully cancel their non linearity between 2mA-12mA, of course I've no idea how well it might work in practice.

If I swap the sim to use the photo transistors rather than the base-collector photo diodes the BW is only 12kHz.

[JS]

"I guess some discrete transistor should be used to take those 6µA up."

I'd just view the +/- 80mV across a 5k1 or 5k6 resistor for now myself. You're going to need at least some resistance from the photo diodes mid point to 0V because a 0.5uA difference in the PD quiescent currents would make that point go to near one of the rails.
It's a good place for a CFA amp, but I've not had a massive amount of luck on a breadboard where every pF at the summing point counts.

"the differential approach also makes sense to retrieve higher signal at the output"

I used 2 optos working in anti-phase to hopefully cancel their non linearity between 2mA-12mA, of course I've no idea how well it might work in practice.

If I swap the sim to use the photo transistors rather than the base-collector photo diodes the BW is only 12kHz.

  Well, I made it, I came up with this circuit which simulates nicely, ~600kHz BW, good linearity, decent step response and all that. I leave the schematic here.

  I started with another approach which didn't polarized the output of the optocouplers properly, on the sim seemed all great but it worked only up to 40kHz or so in real life.
  Once on the breadboard I started messing around and came up with the circuit in the schematic, which simulates up to 600kHz but I can only observe a bit over 200kHz in the breadboard. The good thing is that the step response is much better than the one in the previous 200kHz circuit, no ringing at all. Changing R6 and R12 from 10k to 1k takes the BW up to 250kHz in real life, to 2MHz in the simulation. Rise time still at about 1.2µs. I'm observing this response at the output of the optocouplers, so the bandwidth limit is there, didn't improved much using it as diodes, next test could be a dual virtual earth on the output, try with diodes or transistor output connection. I don't know if changing Q4 and Q5 for jfets will make any difference. I also leave a capture of the step response for your joy!

  Thanks for all the help and I listen to any ideas of how to take the BW even further, I'm liking the open loop approach, much cleaner step response and seems it can be pushed a bit more. I'm happy with it as it is so this will end in a PCB inside a box, but the more BW the better! 

JS

[Gyro]

Nice iteration. 

If the bandwidth that you are seeing is still at the outputs of the optocouplers, then you could go ahead and box it with them socketed and then replace them with faster ones when you can obtain them as they are pretty much all pin compatible.

There is a circuit floating around somewhere on the web that splits the LF and HF frequency components, the DC/LF being handled by a linear opto (an IL300 I assume). The HF portion is coupled by a very low value, high isolation voltage capacitor coupled circuit. Both are then re-combined to produce a high fidelity waveform. I though I had seen the circuit on either the EDN or Electronic Design publications' websites but haven't been able to locate it again yet.

Of course this combined approach would reduce the CMRR that your purely opto approach gives you.

[JS]

Nice iteration. 

If the bandwidth that you are seeing is still at the outputs of the optocouplers, then you could go ahead and box it with them socketed and then replace them with faster ones when you can obtain them as they are pretty much all pin compatible.

There is a circuit floating around somewhere on the web that splits the LF and HF frequency components, the DC/LF being handled by a linear opto (an IL300 I assume). The HF portion is coupled by a very low value, high isolation voltage capacitor coupled circuit. Both are then re-combined to produce a high fidelity waveform. I though I had seen the circuit on either the EDN or Electronic Design publications' websites but haven't been able to locate it again yet.

Of course this combined approach would reduce the CMRR that your purely opto approach gives you.

The tek uses a transformer for the HF, to get from a few kHz upto 20MHz, I haven't seen a cap approach.

CMRR is a major factor for this amplifiers, the problem for HF is the capacitance between all the input amp and the rest of the world, hence the optic fiber coupling seen in Dave's vid lately. You don't need that separation for 1kV insulation but the higher the capacitance the lower the CMRR and the higher the load for fast common mode signals as they were measuring at the gate driver.

JS