EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: T3sl4co1l on April 07, 2014, 08:01:37 pm
-
Anyone ever tried speeding up a 4N35?
Example: using the detector as a saturated switch, the switching delay is a whopping 20us. Even with a rather low resistance load (to keep it from saturating), it's microseconds.
I've tried various approaches over the years. Have a circuit on the breadboard right now getting 1.85us delay, 3.83us rise time, and good linearity. It's a four transistor TIA circuit. It's very nice, but still rather slow.
What's your favorite way?
Tim
-
On the 4n35 you can bias the output transistor up a bit into the "just" active range. It will switch a pinch faster. Pin 6 is the base I believe.
-
What the point to do this rather than just select faster optocoupler? I see no point but you can improve switching off time by pulling down base to the emitter with a resistor. You won't be able to improve switching on time thou.
-
Not much point no, but simple solutions provide some value in those applications where you need just a little something, and the signaling isn't correct from another device (e.g., wide voltage swing, odd levels). Good example would be, a B-E resistor, or a one transistor buffer or something.
And where speed is a concern, I'd rather grab a 6N136 or one of those logic couplers, but then, maybe you have something not so fast (say, the gate drive to a motor speed control) and you're already using a 4N35 (or something like it) in an isolated power supply. Why clutter your BOM with more parts?
And if nothing else, anyone who categorically denies the value of a mental exercise is... shall we say, a prime example of the result of such negligence. ;D
Tim
-
By point I mean that to speed up that optocoupler you already need to add more parts to the schematic. Playing around is nice, however it is not good for using in a real application IMHO. As beyond datasheet specs you cannot guarantee that each device will work as intended. If it is a one-off device then there is even less point to shrink your BOM in a such way.
-
I agree with wraper. There's nothing wrong with doing an exercise, definitely encouraged for newbies, but you indicated that this is
a real project, and I'm sure you already know all the options and limitations. The bigger problems with OC transistor optos are
jitter and temp variations. As you say, there are drop in replacements with logic OP, how easy is that?
I've mostly moved over to the Silabs RF isolators, Si8442s etc. Good to 150MbS at fixed low power.
Edit: oops, replaced by Si8640 - 100MB
-
Real project? What's that?
I've mostly moved over to the Silabs RF isolators, Si8442s etc. Good to 150MbS at fixed low power.
Edit: oops, replaced by Si8640 - 100MB
I don't like those so much: I'd have to check with manufacturer and series I'm thinking of (might not be Si), but a lot of them aren't DC stable. I really like the ADI parts, which are DC stable. The kind of thing that, if you've got error correction in the link (data or whatever), you're probably fine, but you wouldn't want to use it for a gate driver coupler.
The ADI parts of course carry an ADI pricetag, but it's about the same as an optocoupler of the same speed, so, *shrug*.
Tim
-
... a lot of them aren't DC stable. I really like the ADI parts, which are DC stable .... you wouldn't want to use it for a
gate driver coupler.
Not sure what you refer to ... being "DC stable" ? I put 1MBaud through them and have never seen a retry.
No picking on what you say, just interested on something that I may have missed.
-
One of the ADI appnotes goes into detail -- they can be upset by ambient noise, intense yes, but nonetheless possible. The ADI parts have a missing pulse detector (also a failsafe condition!) and pulse repeatedly, so the longest the input and output will be out of sync (ideally) is a few hundred ns or something like that. I haven't seen the same properties touted in competitive parts, but I haven't looked in a while either.
I love the isolation dV/dt on these parts. Great for gate drives on fast, high power applications.
Tim
-
-- they can be upset by ambient noise ... The ADI parts have a missing pulse detector (also a failsafe condition!)
OK Thanks, should be an interesting read.
-
Anyone ever tried speeding up a 4N35?
Example: using the detector as a saturated switch, the switching delay is a whopping 20us. Even with a rather low resistance load (to keep it from saturating), it's microseconds.
I've tried various approaches over the years. Have a circuit on the breadboard right now getting 1.85us delay, 3.83us rise time, and good linearity. It's a four transistor TIA circuit. It's very nice, but still rather slow.
What's your favorite way?
Tim
Either use the transistor as a photodiode or shunt the base-emitter junction with a resistor to reduce the storage time. If used as a transistor, it may also be worth driving a cascode to reduce the Miller effect which would also prevent saturation.
In a saturated switching application, it seems like baker clamping the collector to the base with a schottky or germanium diode would help a lot. I will have to try that.
-
One method I have seen is to use a pair of optocouplers together, in what I suppose you would call a "quasi-complementary" arrangement. LEDs in inverse parallel (for bipolar drive), one transistor from supply to output, the other from output to ground. This was in a cheap RS-232 isolator from DX, so probably used whatever optocouplers they swept off the floor that day... Seems to work OK, though I only tested it at 9600bps.