2
\$\begingroup\$

I have a problem with short current-pulse widths through the diode of an opto-isolator -- with widths less than \$1\:\mu\$s. The charge in each pulse is approximately 3 nC. The current height of the pulse is adjustable, as I have a series resistor available for setting the peak current. The pulse shape starts at the peak value and then follows the usual RC decay for the trailing edge. So the current pulse takes on a sharp onset followed by an RC decay. Roughly speaking, that can be taken as a triangle shape, where the following relationship exists:

$$3\:\textrm{nC} = \frac{1}{2}I_{peak}\cdot t_{width}$$

If I set \$I_{peak}=10\:\textrm{mA}\$ then \$t_{width}=600\:\textrm{ns}\$. I can increase \$I_{peak}\$, but only at the expense of \$t_{width}\$. And visa versa. So that's the problem. I need to convert this narrow, optically isolated current pulse into a low-impedance square-wave output with a pulse width of \$250\:\mu\textrm{s}\$, or so (in order to accommodate the \$1600\:\textrm{Hz}\$ mentioned earlier.) Eventually, a \$75\:\Omega\$ output would be perfect, but ten times that would be okay.

The 6N136/HCPL4502 isn't spec'd with fast enough rise and fall times. The IL610 (see: datasheet) is fast enough. But those things are expensive and I've no experience with them, or their availability. I am going to buy some of the 6N136 and IL610 parts to play with. I'm not excluding them. But in the meantime I'm interested in ideas about using more pedestrian methods using parts like the 4N25.

Here is a schematic I'm considering:

schematic

simulate this circuit – Schematic created using CircuitLab

That's in broad-strokes. \$R_2\$ and \$C_1\$ form a time constant that is very much larger than the rise and fall time of the 4N25A -- on the order of tens of microseconds. But this also creates a \$\frac{\textrm{d}V}{\textrm{d}t}\$ through \$C_1\$ for a short time with a peak of about \$40\:\mu\$A, which pulls out any stored charge in \$Q_2\$, turning it quickly off. The compensated current source provides a meager \$1\:\mu\$A drive for \$Q_2\$. (I could just replace the whole thing with a single resistor that I tweak in, I suppose.) The pulse width will be on the order of \$3.3\:\textrm{M}\Omega\cdot 390\:\textrm{pF}\$ or \$1-2\:\textrm{ms}\$. A bit long for my needs, but I'm trapped between the CTR of the 4N25A and how little of my narrow pulse will drive across the optocoupler.

I just started thinking about this problem, today. And I may already be way off-base (still need to lower the output impedance.) I may also be guilty of not yet spending enough of my own time thinking about this, but I'm interested in any obvious difficulties I missed or preliminary thoughts about different directions to point in regarding the use of a 4N25A for something like this. I have those laying about and nothing else, yet. (Yes, I will be getting those fancier devices and trying them out.)

For now, are there any ideas about how to flog a 4N25A into a predictable circuit for something like this? (The CTR is spec'd given enough time, time I don't have. So I'm assuming very little of the pulse gets through and that I have to set up a sensitive circuit.)


If interested, this is driven by a microwatt VFC I've designed, which develops periodic \$3\:\textrm{nC}\$ discharges through the optocoupler. The VFC is attached to an arbitrary, different supply under observation and must present a \$25\:\textrm{M}\Omega\$ load to a \$15\:\textrm{V}\$ source, delivering a pulse frequency of \$200\:\textrm{Hz}\$. It's easy to compute the allowable discharge per pulse here:

$$\frac{15\:\textrm{V}}{25\:\textrm{M}\Omega\cdot 200\:\textrm{Hz}}=3\:\textrm{nC}$$

The allowed target supply variation is \$4-120\:\textrm{V}\$. (Think of this like powering a circuit from an old POTS phone line service in the US, where each on-hook phone is only permitted to load the line by \$5\:\textrm{M}\Omega\$.) This means the frequency can reach about \$1600\:\textrm{Hz}\$. (It goes a little bit non-linear as you approach the lower voltage limit, of course, and the frequency drops to perhaps \$10\:\textrm{Hz}\$.)


In case it helps, here's the VFC sketch:

schematic

simulate this circuit

(Those \$4.5\:\textrm{M}\Omega\$ resistors are as big as they get and still be cheap and widely available.)

I'm also thinking about smearing things out with a \$100\:\mu\$H inductor in series with the LED. Makes a nice round bump out of the pulse. Might make things a lot easier with the 4N25A. However, there may be a cost with the SCR circuit, dragging things out like that. I need to worry over that bit, as well. (Particularly the positive feedback aspect through \$C_2\$.)

\$\endgroup\$
  • \$\begingroup\$ are you stuck with optos or can you use magnetics to broaden your choice of isolators? \$\endgroup\$ – Neil_UK Oct 10 '16 at 8:44
  • \$\begingroup\$ @Neil_UK Yes, I can consider a custom inductor design. I think the IL610 is a specialized inductor device, in fact (designed to be almost a drop in replacement for optos.) It's just that it costs a lot and it's very boutique. I haven't yet had time to think about using a transformer approach, although I did consider the idea of inserting an inductor in series with the diode and resistor to smear out the pulse a bit. I just didn't write about those thoughts here, yet. I'd love some thoughts on magnetics, though. \$\endgroup\$ – jonk Oct 10 '16 at 8:48
  • \$\begingroup\$ I didn't mean 'couplers', which is a transformer + other stuff to get DC levels through, so expensive. I meant a transformer, as you are only dealing with pulses, and are putting your own stuff round it. Even a few turns of wire put over a leaded open inductor for instance. \$\endgroup\$ – Neil_UK Oct 10 '16 at 8:58
  • \$\begingroup\$ @Neil_UK I've added the VFC, in case it helps. I think I understood what you meant. I just haven't spent much time considering it, yet. If you have anything specific, especially now that the VFC is added, feel free. I'm going to go to go AFK for a while, though. \$\endgroup\$ – jonk Oct 10 '16 at 9:06
  • \$\begingroup\$ Do you need those absolute frequencies, or just the range? You could trivially increase the pulse energy and perhaps duration by increasing the size of C1. Thanks for the circuit, I've often mused how to make a micro-power self triggering thyristor for just such an application. \$\endgroup\$ – Neil_UK Oct 10 '16 at 9:48
2
\$\begingroup\$

Vishay's application note Faster Switching from Standard Couplers says:

PUSHING THE “SPEED ENVELOPE”
If less than 1 μs toff is required, there are some additional measures that can be taken to further increase the switching time of a standard coupler. […]

PHOTODIODE OPERATION
One possibility is to give up current gain for the sake of speed in the extreme, and operate a standard transistor coupler in photodiode mode by only using the base collector pins of a 6 pin coupler. This will vastly increase the switching response of the device but will require an output buffering stage to produce a practical signal level. Also, it is possible to further increase the switching speed of this type of coupler by applying an increasingly large reverse bias voltage on the photodiode detector. The effect is to reduce the junction area and capacitance and increase the frequency response. […] Figure 30 illustrates a photodiode configuration which makes use of a trans-impedance amplifier to provide better gain and noise performance. The key principle of this design is that one can now take advantage of the lower junction capacitance of a simple reverse biased diode rather than the large capacitance of a phototransistor to drastically improve transient performance.

Trans-Impedance Amplifier Photodiode Circuit

But don't bother with all this complexity; it also says:

At this point, however, economics come into the equation, and a designer needs to consider whether the increased complexity of the design is justified. The other option is to consider one of various parts that are specifically designed for high-speed operation.

For example, the common 6N137 requires 5 mA to switch on, and has a propagation delay of 100 ns, or the TLP2361, 1.3 mA, and 80 ns.

This results in a nice digital signal; you then still need a circuit to make the pulse longer.

\$\endgroup\$
  • \$\begingroup\$ Okay. I'm already considering the 6N136, so I'll add in those and buy a few. Also, I'll look into the BC aspect. I hadn't considered that option! Lots of good in that white paper. Thanks! \$\endgroup\$ – jonk Oct 10 '16 at 9:07
  • \$\begingroup\$ Hmm. TLP2361 is cheap enough. I could go there. I need to buy some. \$\endgroup\$ – jonk Oct 10 '16 at 10:01
  • \$\begingroup\$ Once you have a reliable pulse detected, use a fast comparator to produce a (more or less) 5 volt pulse, then use that logic-level pulse to drive a monostable such as a 74HC4538 with a 250 usec time constant. Put a high-current buffer such as a FET gate driver on the monostable, and you should be able to drive 75 ohms easily. \$\endgroup\$ – WhatRoughBeast May 22 '17 at 18:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.