Driving mosfet isolater (Si875x)

Thread Starter

fastbike

Joined Dec 29, 2020
120
I'm looking to use the dimmer circuit at Elliot Sound Products as the basis for designing a digitally controlled dimmer.
This makes use of an op amp running as a comparator to drive an SI875x optically isolated mosfet driver.

I'm curious as to why the author chose an op amp rather than a comparator. I've read the excellent TI paper "Op Amps and Comparators - Don't Confuse Them". This explains how the typical output stages differ: op amps run in the linear region, where as comparators run saturated and warns against using op amps in the way described in the circuit. "Texas Instruments applications have received a number of inquiries about using op amps as comparators in an open loop. The best advice [we] can give is to not do this. The very best a designer can expect is very poor performance; the worst is a circuit that does not work or even burns out" Ouch.

The snippet of the circuit is here
1621815420493.png

My guess is that it allows the simpler isolater part which runs as a simple diode emulator and does not require a separate decoupled 5v supply unlike the SI8751.

Here's my first cut at modifying the circuit to use a LM339 comparator and the Si8751.
The problem I can see is that the pull up resistor required by the open collector output from the comparator will turn the isolator on, at times where this may not be required. A transistor switch to invert the output logic could easily fix this.

1621817132463.png

Anyway looking forward to some feedback regarding the pros and cons of both options.
 

shortbus

Joined Sep 30, 2009
10,045
The problem I can see is that the pull up resistor required by the open collector output from the comparator will turn the isolator on, at times where this may not be required.
I may not be understanding what you asking/saying here. But isn't the output of the LM339 -393 at ground/common level until the voltages on the inputs make it change? If the out put is off(no output) then both the pull up and the input signal on Si8751 pin#3 should be at ground/ common. Unless the pull up resistor is so small that the pull up overpowers the grounding level of the comparator.
 

Thread Starter

fastbike

Joined Dec 29, 2020
120
If the out put is off(no output) then both the pull up and the input signal on Si8751 pin#3 should be at ground/ common.
The issue is that the comparator has an open collector (or drain) output so requires a pull up to provide a definitive on / off state.
The pull up will pull the output high in the absence of a signal.

I'm looking for a fail safe mode where the absence of the ramp signal will not allow the Si875x to drive the mosfets.
I have sketched on paper a circuit idea that uses a NPN transistor biased from the comparator output to effectively invert the comparator output so the default state is for a low signal to be feed to the Si875x.

I'll draw up a circuit schematic later tonight and post it here to explain my thinking.
 

ronsimpson

Joined Oct 7, 2019
2,986
The problem I can see is that the pull up resistor required by the open collector output from the comparator will turn the isolator on, at times where this may not be required.
"turn on at time where this may not be required" When? I don't understand when/how. The output goes high when (+) is higher then (-). (assuming there is a pull up resistor)
You can get voltage comparitors with an internal pull up transistor inside. (no resistor required) but that will not change things.
 

Thread Starter

fastbike

Joined Dec 29, 2020
120
Here's a schematic for the logic controlled version Si8751 (the IN pin expects to see a logic high signal to drive the mosfets). Decoupling caps are not shown for clarity.

1621923404841.png

When the comparator output transistor is off, the voltage at the base of Q4 will be 5v and will drive the transistor Q4 into saturation (ON). Close to 5v will be dropped across R41 so the voltage presented to the IN pin on U9 will be a logic low.

When the comparator output transistor is on, ~5v will be dropped across R40 so the voltage at the base of Q4 will be 0v and the transistor Q4 will be OFF. The voltage at the collector will be 5v so the IN pin on U9 will see a logic high.

I'm still trying to figure out what the default output state of LM339 is when both inputs are 0v.
 
Last edited:

Thread Starter

fastbike

Joined Dec 29, 2020
120
I've had a go at a schematic for the diode emulation version (Si8752). This part requires ~20mA and will drop about 2v between anode and cathode.
An additional PNP transistor is required to switch the diode on the driver so another stage is required to get the correct logic (open collector at comparator means mosfets are off).
1621927007355.png

The extra complexity of this circuit means I'll develop the Si8751 (logic input) as shown in #6 design further.
 
Last edited:

ronsimpson

Joined Oct 7, 2019
2,986
When the (+) input is higher than the (-) input the output will be open and no current will flow in the Si8752 A-C.
When the (+) input is lower than the (-) input, current will flow.
When the inputs are the same voltage then the output is unknown.
1621947685508.png
read this link
Or we could add some " hysteresis " to the circuit.
 

shortbus

Joined Sep 30, 2009
10,045
The issue is that the comparator has an open collector (or drain) output so requires a pull up to provide a definitive on / off state.
The pull up will pull the output high in the absence of a signal
No. You need to look closer at how the 339/393 comparators work. Ron gives a good description. As I said in my first post you will only get a voltage from the pull up if you don't use the correct value of resistor for the pull up, that will over power the ability to keep the output at zero volts.
 

Thread Starter

fastbike

Joined Dec 29, 2020
120
Here is a option where the LM339 shorts out the Si8752. If pin-1 of hte LM339 is low no current can flow in the Si8752. This inverts the logic.
View attachment 239551
Thanks for sharing an interesting variant. Although the 150R resistor (required to allow 20mA to flow through the LED which drops 2v) would allow 33mA to flow through the output transistor. In a quad part this would give dissipation of ~700mW which is getting close to the max power handling for the device (1W).

Also I've just noticed in the datasheet that the common mode voltage is Vcc - 1.5v so I'll need to power this from a higher supply than the 5v inputs. I have a 12v supply in the device so could use that.
 
Last edited:

Thread Starter

fastbike

Joined Dec 29, 2020
120
When the (+) input is higher than the (-) input the output will be open and no current will flow in the Si8752 A-C.
When the (+) input is lower than the (-) input, current will flow.
When the inputs are the same voltage then the output is unknown.
View attachment 239549
read this link
Or we could add some " hysteresis " to the circuit.
Thanks for the link. And for the discussion of hysteresis.
I'm tending to this circuit variant as the LED will be off unless the output of the comparator is conducting (logic low).
I am also looking at part MCP6549 which has a small amount of hysteresis built in, has common mode exceeding Vdd-Vss, and has an open drain output. Maximum current through the output is 30mA at up to 10v.
 

Thread Starter

fastbike

Joined Dec 29, 2020
120
Final circuit to be prototyped. Thanks for all of your feedback and suggestions.

1622013768918.png
NB: ramp and 0-5v dimmer input arranged so that mosfets are only driven when the dimmer control voltage exceeds the ramp voltage. I.e. output transistor is off when ramp voltage exceeds dimmer control voltage.
 
Last edited:

Thread Starter

fastbike

Joined Dec 29, 2020
120
As if you don't have enough to think about; If you want to invert things, switch the two inputs and the output will do the opposite thing.
Thx Ron. Design has moved on a little, and I will switch back to the Si8751 variant, driven via an AND gate. The other input on the AND gate is pulled low by an over current sensing circuit if excess current is detected on the output, to ensure a hard switch off in the case of a short circuit etc on the output mosfets.
 

rod-e

Joined Jun 17, 2016
9
Using LM358 opamps as comparators in the circuit is not an issue at all. The OP appears not to understand the difference between high and low-speed circuits, and the opamps are more than fast enough for the circuit. Nothing will 'blow up', and the entire circuit was bench tested to prove that it works as designed and shown on the P201 project page. In all respects, the opamp is much faster than the isolator.
 
Top