OPA860: Voltage to current LED driver

Hi Chris,

What voltages are your pulses? What kind of drive strength will your require for your LED? The OPA860 buffer output can sink and source 60mA of current, will that be enough for your LED? Can you provide me with a schematic of what you are intending to do?

Thanks!
-Karan

Hi Karan,

I haven't finalized the LED selection yet, cause it will depend on the output capabilities of the driver, but it'll be UV or blue, requiring at least 3V forward voltage, but only a few mA. To give a concrete, but overkill example, say our LED will require 3.5V and 20mA peak. So it looks like the OPA860 should be able to produce plenty of current if it can do so at the voltage required to turn on an LED at all.

What we're trying to do in the end is characterize a charged particle detectors read out with PMTs, so we don't need much light, but we do want to tightly control the time domain shape of the light/current pulse, i.e. make it proportional to the input voltage. The pulse generator I've got in mind can provide an input voltage pulse anywhere in the range +-2V into 50 ohms.

Do you mean a block diagram of the whole setup, or my first crack at a circuit on the OPA860 demo board?

Thanks,

Chris

Hi Chris,

in the 90ies when I studied physics we tried the same. We worked with GaAs microstrip detectors for the ATLAS detector at CERN. But our GaAs microstrip detectors were very fast and produced pulses of only a couple of nanosenconds when being hit by a mip. It turned out that a LED was not fast enough to generate a nominal detector pulse. If I remember correctly they had too much junction capacitance.

Would the LED's junction capacitance also be an issue in your application?

Kai

Hi Chris,

here is some material for you. Do you know the datasheet of (obsolete) OPA660? And this appnote?

opa660.pdf

01b7cd566115c542b887f956a5e2af638d2f.pdf

And you might find this thread useful. It contains some TINA-TI simulations with the OPA860:

Are you familiar with TINA-TI? It's a very powerful TI's simulation tool. It's for free.

Kai

Hi Kai,

I had not seen those documents for the old OPA660. Looks like they contain quite a bit of useful detail!

I do have TINA-TI installed and ready to go. I put constructing a model on hold when I began to wonder if I was completely off track with the OPA860, but I'll try and get something working over the weekend to make sure I don't post something obviously silly in response to Karan.

Thanks,

Chris

Hi Chris,

seems to work:

To prevent negative going undershot of the LED current I would add R8, though. Find out the proper value by trial and error:

R7 was only inserted to be able to measure the LED current in the simulation.

And this is my simulation file. Play with it:

chris1.TSC

Kai

Hi Kai,

Thanks for the model! I'm still trying to get measurements out of TINA, e.g. figure out the difference-in-practice between a sense resistor and an ammeter, so it probably would have been a while before I had anything useful.

I can't say I fully understand all the details of the circuit yet, but I'm working on it. Looks like this is essentially the circuit from Figure 20/Figure 15 of the old 660 data sheet, but without the additional current mirror. It looks like your model shows the output current following VG1 right to zero, but is that TINA's way of drawing a negative value on an axis that stops at zero? What's the difference between the 660 and the 860 that obviates the need for the mirror? Also, does the parallel configuration just double the output current, or is it serving some other function as well?

In any case, I'm gonna take a break from trying to understand the bias currents and go build this beast and see if I can make it work without letting the smoke out of anything. If two 860s are required, any chance I can get away with mounting the second one dead bug style on top of the first on a single demo board?

Thanks,

Chris

Hi Chris,

the 1mR shunt resistor instead of a current meter was inserted by me, because the oscilloscope tool of TINA-TI cannot display currents but only voltages. So, when using the oscilloscope tool of TINA-TI I can take a look at the current by displaying the voltage drop across the shunt resistor. That's why I prefer the shunt resistor for current measurements.

The OPA660 was the original diamand transistor chip developed by Burr-Brown (now TI). I have worked a lot with it in the 90ies. Many more appnotes and example circuits were published for the OPA660 than for the OPA860 and others.

I think the current mirror (which is intended to work as a stable current source) was originally added to the LED driver to allow for sophisticated modulation applications, where the LED current shall be modulated arround a stable DC bias LED current. But in your application I see no need for a current mirror. The poor man's current source R8 is enough to prevent the LED from being totally turned-off. That said, I don't know whether R8 is important in your application at all...

Yes, if your LED is below 30mA then even only one OPA860 might be enough to drive the LED:

Please note the modified component values, though.

No, TINA-TI's negative values do not stop at zero on an axis:

Take care, the current through the LED is one thing. But I think you are more interested in the light actually being emitted from the LED? So, you should not only simulate and measure the LED current, but also measure the LED light. This can be done with the help of a simple TIA. A FET OPAmp in combination with a BPW34 should be able to accomplish this.

One last word about the OPA860. This is a very very fast chip but precision is not its thing. So, don't be surprised, if you notice some offset voltage, input bias current or drift issues. Happily, in the most cases this has little impact on performance, especially if you wire the OPA860 consequently low ohmic as in your application.

Kai

Hi Kai,

Thanks for the explanation on the shunt resistor. I was afraid it might be a workaround for some SPICE subtlety or TINA bug.

I thought the current mirror might have been intended to trim out any offset in the output current, but modulation around a non-zero DC level makes sense. In my application, the LED would ideally be completely off in between excitation pulses. I have been assuming that at low forward currents, my LED will have quite linear optical power out per current input, and hopefully won't have much of a non-linear emission threshold. I've got a little optical splitter that will send most of the light to a PD/TIA, so I will try and verify that assumption soon and shop for a better LED if need be. I certainly don't expect to heat it enough to go non-linear on the high end. In any case, you are correct, the thing I care about at the end of the day is a linear correspondence between the voltage pulse input and optical output.

As for stability and drift, I'm optimistic that your prediction is correct. Eventually, I'll mount this driver in a temperature controlled box, which should help a little, too. But pulse-to-pulse repeatability will be more important than long term stability.

In re component values, what does the reduction of R3 to 50Ohm accomplish? 

Regarding the selection of R8 and the negative undershoot: in the presence of D1, if what I'm concerned about is light output from the LED, would it actually be desirable to err on the side of little bit of undershoot? It should mostly go through D1, I think.