Why is my opto-coupler so slow?

[kenw232]

Maybe someone can suggest something.  I'm using a PVI5050N octocoupler. Very straight forward. I've attached my circuit, I did exactly this which works great, but only at very low frequency. This allows me to do all the things I want such as control the frequency and duty-cycle with my waveform generator, adjust the 120 Volts down and up (as high as 1400V) as I see fit with my high voltage power supply dial, and provides voltage isolation for the load when the IGBT is in the off state. My load is a 120V lamp.

I scoped the lamp, you'll see at 1Hz it works great, it flashes on and off. But as I turn up the frequency (using my waveform generator) the quality of the wave the IGBT is switching degrades, and its also dropping in voltage and the lamp therefore gets dimmer with anything above 80Hz or so. At 1KHz the circuit for some reason is off, and at 5KHz its just wide open giving me 120V straight DC again.

Does anyone know what I might be able to do to improve my speed here, and waveform integrity at higher frequencies? Is it my 2N222 transistor that is giving me these frequency limitations? The Opto-Coupler? The results are even worse with other IGBT's and MOSFET's I've tried. I suspected it would be slow, but this is terrible. I was hoping to hit 10KHz.









Waveform generator pulsing 2N222 at 5KHz:

[dannyf]

It is discussed in your original thread - this part is fundamentally unsuited for your application.

it only takes a cursory reading of the datasheet to know that.

your wrong schematic didn't exactly help either.

[wraper]

Because look into datasheet. Minimum Short Circuit Current @ ILED = 10mA, 25°C (see figures 1 to 2) 5.0 uA. How fast you are going to charge IGBT gate with 5 uA current?

[dannyf]

The great thing about the internet is that you get everyone's view. Lots of diverse experience, design mistakes, etc. so you can learn really fast.

The bad thing about the internet is that you get everyone's view: a monkey with a keyboard can come across just as eloquently as Shakespeare. It is up to you to see which bit of information offered on the screen is credible or not.

If yo don't have that capability, you are far better of learning slowly, through schools or books.

Or you end up where you are.

I said this back then: your design is difficult because you made it so.

[linux-works]

I did not check your opto part, but I did a test a year or two ago with 2 opto parts.  one was dog-slow and the other was a lot faster.  it was all about the part:

slow part with delay:



much faster part that did the trick:




I first tried some junk part I had onhand, then went with a faster part and that made all the diff.

not sure if your part is bad or if there are other issues, but the opto parts do definitely vary quite a lot and parts choice is important.

[kenw232]

I guess the PVI5050N just doesn't cut it.  I'll have to try to find something else. 

I have others but its difficult to plugin Surface Mount IC's with a normal breadboard.   How do you all handle that?  SMT on a breadboard...

[GK]

What is the application? Does the load have to be referenced to GND? Why not reference it to the +120V DC supply configure the IGBT as a "low side" switch, with the source connected directly to GND? Doing that would simplify things significantly.

[kenw232]

Load has to referenced to ground so that when the IGBT is off I have voltage isloation for the load. Off must be off.

[linux-works]

if you want smd to dip adapters, search on 'brown dog'.

or ebay.

stuff like this, I assume, is what you want?



those little green boards are 'browndogs'.

http://cimarrontechnology.com/so8to8-pindipadapterpn970601-3.aspx

[GK]

Load has to referenced to ground so that when the IGBT is off I have voltage isloation for the load. Off must be off.

Well the load is ostensibly "off" when the current through it is interrupted. Why is "off" isolation from the +120V necessary?

[pmbrunelle]

Load has to referenced to ground so that when the IGBT is off I have voltage isloation for the load. Off must be off.

Well the load is ostensibly "off" when the current through it is interrupted. Why is "off" isolation from the +120V necessary?

Assuming "GND" in his schematic is at the same potential as Earth, I would assume the same reasoning that drives people to switch (with a single-pole switch) the hot wire of a circuit, rather than the neutral.

[Harvs]

The term "isolation", and an IGBT driven with an opto-coupler really shouldn't be used in the same sentence.  While "technically" correct, it does isolate the current flow if off, usually when we're talking isolation its about separating part of the circuit in order to not end up with a potential conduction path. I.e. not kill someone when it all turns to poo.

If this lamp is going to potentially have unwary people poking at it with their fingers (e.g. changing the bulb), you need a hard power switch as well, that provides real isolation.

[GK]

Load has to referenced to ground so that when the IGBT is off I have voltage isloation for the load. Off must be off.

Well the load is ostensibly "off" when the current through it is interrupted. Why is "off" isolation from the +120V necessary?

Assuming "GND" in his schematic is at the same potential as Earth, I would assume the same reasoning that drives people to switch (with a single-pole switch) the hot wire of a circuit, rather than the neutral.

He's not switching ac mains here - it's a (apparently variable) DC supply the details of which have not been given. BTW, seen a commercial lamp dimmer circuit lately? The switching device (TRIAC) is almost always put in the neutral leg, which is the "cold" leg of the mains as it is terminated back at the switchboard to earth. 

[GK]

If this lamp is going to potentially have unwary people poking at it with their fingers (e.g. changing the bulb), you need a hard power switch as well, that provides real isolation.

Which is precisely why it is likely that it would not matter which leg the IGBT switching was done in. In any case, is the real load a lamp, or is the lamp just for testing? Not sure why one would want to duty cycle control the intensity of a lamp at a repetition rate as high as 10 kHz......

[pmbrunelle]

Load has to referenced to ground so that when the IGBT is off I have voltage isloation for the load. Off must be off.

Well the load is ostensibly "off" when the current through it is interrupted. Why is "off" isolation from the +120V necessary?

Assuming "GND" in his schematic is at the same potential as Earth, I would assume the same reasoning that drives people to switch (with a single-pole switch) the hot wire of a circuit, rather than the neutral.

He's not switching ac mains here - it's a (apparently variable) DC supply the details of which have not been given. BTW, seen a commercial lamp dimmer circuit lately? The switching device (TRIAC) is almost always put in the neutral leg, which is the "cold" leg of the mains as it is terminated back at the switchboard to earth.

Actually, the last two lamp dimmers I bought in the last 5 years or so were two-terminal devices.

[David Hess]

It is discussed in your original thread - this part is fundamentally unsuited for your application.

it only takes a cursory reading of the datasheet to know that.

The bad thing about the internet is that you get everyone's view: a monkey with a keyboard can come across just as eloquently as Shakespeare. It is up to you to see which bit of information offered on the screen is credible or not.

From my post where I suggested this solution:

If switching speed is not an issue, they make optically isolated MOSFET drivers which simply consist of an LED on one side and a photovoltaic pile on the other side which will generate the voltage necessary to fully turn on a MOSFET.

The original post does not identify any speed requirements and involved an SCR and they are not known for fast switching anyway.

As far as speeding up the circuit as is, I can think of a couple of ways:

1. Reducing the input and miller capacitance of the high voltage transistor would help a lot with both transition time and maximum frequency.  This could be done by switching a low voltage MOSFET with one photovoltaic optocoupler while driving the gate of the high voltage MOSFET continuously with a second one.  The high voltage transistor is then switched from its source instead of its gate.  Similar switching circuits are used with a MOSFET driving a cascoded bipolar transistor to achieve better performance than either alone but I have never seen one used as a high side floating switch.

2. Use the photovoltaic optocoupler as a floating power source with a capacitor and another optocoupler and a signal transistor to drive the high voltage MOSFET gate hard.  This would help with the transition times but the frequency would still be limited by the power available from the photovoltaic optocoupler.

[GK]

There are high-side MOSFET driver IC’s commonly available that can handle 600V DC for peanuts. My last batch of FAN73832 set me back <20c ea in tube qty. Unless the OP really needs to go to 1400V there is no point in stuffing around with photovoltaic opto couplers gate drivers that were originally designed for switching speeds comparable to mechanical contactors.

And a compelling reason (although there may be a valid one) has yet to be given for not simply switching the ground leg instead.

[kenw232]

And a compelling reason (although there may be a valid one) has yet to be given for not simply switching the ground leg instead.

I'm not sure what you mean by ground leg.  You mean in a normal low-side configuration?

[David Hess]

And a compelling reason (although there may be a valid one) has yet to be given for not simply switching the ground leg instead.

I'm not sure what you mean by ground leg.  You mean in a normal low-side configuration?

Yes.  Normally an N-channel MOSFET would be placed on the ground side of the load.  Being ground referenced instead of high supply referenced makes switching it trivial.  Reasons for not doing this include loads where the ground may not be disconnected for safety reasons which may be the case working with high voltages.