Opto isolated feedback: Opto pulls down COMP

[cupoftea]

Hi,
We wish to do an SMPS with SG3524 as primary side controller. The opto transistor will pull down the COMP pin of SG3524, to give regulation.
So we will connect the + input of the SG3524 error amp to 2.5V (divided down 5vref)....and the "-" input of the internal error amp to GND.
Then the COMP pin will be nice for us to pull down with the opto.

But the error amp in SG3524 has common mode input range of 1.8 to 3.5V....so should we in fact best
connect the "-" input of the internal error amp to , say 2v?

What would be the worst that could happen if we connect "-" input to GND? Surely not phase reversal?

SG3524

https://www.ti.com/lit/ds/symlink/sg3524.pdf?ts=1714417635113&ref_url=https%253A%252F%252Fwww.google.com%252F

 

[cupoftea]

Do you know why the error amp in the SG3524 is so poor?
Why does it have a common mode input range of just 1.8 to 3.5V?

..It means we cant ground the inv pin as part of getting it to drive COMP high so that we can then pull it down with our regulation opto.

I mean, the UC2843 is yonks old as well, but its error amp isnt that bad......why did they put such a lousy one in SG3524?
___-----____
Also, i cant believe how bad the SG3524 is....there are two drive transistors...but there is nothing to pull down the FET gate to ground to keep it off.

 

[Easy peasy]

it is what it is - a 45 year old chip in its design - used for simple low power converters by the million, < 12c US in volume

also;

which can be useful.

 

[crutschow]

Do you know why the error amp in the SG3524 is so poor?

It's not "poor".

Why does it have a common mode input range of just 1.8 to 3.5V?

Because that's all it needs to have for normal operation.
The 5V reference voltage is normally connected to a divide-by-two resistor pair to give a 2.5V reference at one input.
The other input then has a divider from the output to generate 2.5V at the desired regulated voltage.

If you use opto feedback, then you just configure it to generate 2.5V at the desired output voltage.
(Example circuit below using a TL431 reference at the output to provide the feedback signal to the opto coupler.
The opto output would pull the voltage through Rs down to 2.5V at the regulated output voltage.)

Why do you think you need the input to be referenced to 0V (which would not work)?

 

[cupoftea]

Why do you think you need the input to be referenced to 0V (which would not work)?

...Thankyou, that is what we thought.

You kindly tell us it would not work..but they tell us it has worked fine for year like that (with "-" input grounded)

The customer has sent us a schem of a 200W pushpull 24vin to 24vout, isolated. They have asked us if we would have a look at it. According to them it works and has been sold for years, (but they just changed the transformer and so want someone to have a look)

The "-" pin of the SG3524 has been grounded.........the "+" pin has got 2.5v on it from a divider from vref. They then pull down the COMP pin with the opto collector (opto in common emitter)
Presumably, they grounded the "-" pin so as to make the error amp output go high, and then they can pull it down to do regulation with the opto as discussed.
(ie, thats like the way you do it with the UCC28C43)

..But yes, ground is outside the common mode input range...so why have they done this we wonder?.....they say it has worked for years and , they say, has had high sales over the years.

 

[danadakk]

As a prior production EE and later a test EE and FAE we characterized parts to set datasheet
limits and then cranked in, typically, not always, additional margin in test programs.
But ultimate authority of "what if" belongs in designer lap. That being said due to
poor simulators and incomplete models (even today) back in the day even they did
not always know what would happen if limits were violated. Thats when field engineers
would get involved with unexpected behavior in parts, and back they would go to rel
for eval.

As always "safest" tack is meet data sheet specs. You can always explain to the boss when
brought onto the carpet "why did you do this" you met all specs. Otherwise you get to
light up the room with bright red face explaining your shortcomings.


Regards, stay safe, Dana.

 

[andre_luis]

Why does it have a common mode input range of just 1.8 to 3.5V?

Just out of curiosity, 2.5 = sqrt( 1.8 x 3.5 ), perhaps not by accident.

 

[D.A.(Tony)Stewart]

Do you know why the error amp in the SG3524 is so poor?
Why does it have a common mode input range of just 1.8 to 3.5V?

..It means we cant ground the inv pin as part of getting it to drive COMP high so that we can then pull it down with our regulation opto.

I mean, the UC2843 is yonks old as well, but its error amp isnt that bad......why did they put such a lousy one in SG3524?
___-----____
Also, i cant believe how bad the SG3524 is....there are two drive transistors...but there is nothing to pull down the FET gate to ground to keep it off.

What do you define as "poor" for an error amp? Av = 10 k nom. as a transconductance amp which varies gain with load R.

It uses BJT's with a CC sink to the differential amp which needs 1.8V min. above Vee or gnd. that was typical for BJT Diff Amps.

If Vin+ is fixed at Vcc/2 or 2.5V then the error feedback to Vin- will also be at 2.5V.

So what's the beef?

 

[cupoftea]

If Vin+ is fixed at Vcc/2 or 2.5V then the error feedback to Vin- will also be at 2.5V.

Thanks, in the schem that we have been sent, the inverting pin of the error amp is hard conncted to ground.....the non inverting input is connected to a 2.5V...divider off vref.

 

[Easy peasy]

common mode range - refers exclusively to both inputs at that level - so - looking at the internals, connecting one terminal to zero and another to 2v5 does not violate the CM rule.

A pedant might set up a divider string to have the inputs at 2V & 3V to force the EA o/p high.

 

[cupoftea]

common mode range - refers exclusively to both inputs at that level

Thanks, this is interesting.....suppose the opto had been set up as Crutschow kindly says....ie in CC....with emitter pouring current into a grounded resistor...and this then put into the SG3524 with its error amp set up as an inverting amp....would this be preferable to simply pulling the COMP pin down with opto in CE connection?...(and the - input grounded so as to make the error amp behave as a current source with output high when in max load

 

[crutschow]

It make work with (-) input grounded, but that violates the input voltage limits.

--- Updated May 14, 2024 ---

common mode range - refers exclusively to both inputs at that level - so - looking at the internals, connecting one terminal to zero and another to 2v5 does not violate the CM rule.

When the device is regulating the output, both inputs are at (nearly) the same voltage.

 

[Easy peasy]

When the device is regulating the output, both inputs are at (nearly) the same voltage.

it's not though, here, if you read the text, the o/p of the EA is forced high, and an opto or similar is used to pull down the current limited o/p of the EA.

This is used a lot when the EA has to be on the other side of the isolation boundary

a quick read would have rendered your comment unnecessary

 

[crutschow]

he o/p of the EA is forced high, and an opto or similar is used to pull down the current limited o/p of the EA.

Sorry, but that makes little sense to me.
The opto doen't normally affect how the converter is regulated, it just provides isolation.

 

[FvM]

Sorry, but that makes little sense to me.

It does, though. In an opto isolated feedback topology, error amplifier gain (and usually also compensation) needs to be located at the secondary side, for obvious reasons. SG3524 and other controllers allow to bypass the internal EA in this case by feeding the amplifier output node directly. EA output must be forced high, the only question of this thread is if the chosen input voltage combination does.

My assumption is that the EA doesn't involve phase reversal and it's sufficient to have one input within in CM range, but strictly read the datasheet doesn't specify, you should test yourself.