Flyback LED driver has poor quality feedback loop?

[crutschow]

i agree it appears to work on a general scale...but how about the transfer function of the opamp...is it accurate when operated like this...and if not, then that means the feedback loop calculation for the smps has gone awry....bad news.......

As long as the open loop gain is sufficiently high, the "accuracy" of the op amps transfer function is not a factor. The transfer function is determined by the feedback loop, not the op amps open loop transfer function.
The open loop gain of that op amp does drop when the common-mode voltage is zero (as per the data sheet) but it still is more than high enough to have no significant effect on the closed loop response.... good news...

 

[Flyback]

i guess i am convinced, but doesnt it seem a little odd to you that the opamp has to "know" when the inv input has gone below its noninv input in order to know how to drive its output...and i still do not believe that the opamp actually "recongises" any voltage level that is below its negative supply pin voltage....in other words, the opamp can never know that its inv input has gone below its noninv input.......i am of a mind that its just noise on the noninv input thats actually making the noninv input seem above ground, then when the inv input is at ground, then the opamp "knows" that the inv input is below the noninv input....i wouldnt mind betting that its just ambient noise thats keeping this opamp working properly.

 

[alec_t]

in other words, the opamp can never know that its inv input has gone below its noninv input.

It can, and frequently does know. How else would the output ever be driven towards the +ve rail?

 

[ChrisP58]

i guess i am convinced, but doesnt it seem a little odd to you that the opamp has to "know" when the inv input has gone below its noninv input in order to know how to drive its output..

But isn't that exactly what an op-amp is supposed to do?

 

[Flyback]

yes, my mis-phrasing, i worded it wrong, and yes, that is what the opamp is supposed to do.....but in this case, the inv input cannot go below the noninv input...at least not recogniseably so to the opamp

 

[crutschow]

yes, my mis-phrasing, i worded it wrong, and yes, that is what the opamp is supposed to do.....but in this case, the inv input cannot go below the noninv input...at least not recogniseably so to the opamp

Untrue. The inv input has to go below the noninv input by a (very small) voltage equal to the positive output voltage divided by the open loop gain (ignoring any input offsets). And that has nothing to do with any noise in the system.
That's how an op amp works no matter what the common-mode voltage is (as long as it's within the operating common-mode range).
I don't understand why that's difficult to comprehend.

 

[ronsimpson]

from experiance most op-amps will work fine if one input is inside the operating range and the other is outside. (by fine; it will know if one a above/below the other)
In your case; one input is at 0V which is inside the range. The other input might be -100mV which is outside the range but the amp can deal with that. By the time the second input is at 1mV the part is probably working correctly. (the part will work at -1mV from experience)

I think I had problems with J-FET op-amps that when one input is outside the range (by some large amount) the part went nuts. (inverted)

 

[Flyback]

as long as it's within the operating common-mode range

...thanks, though when the inv input goes below 0v, its no longer inside the common mode range.
As seen by the datashete of the LT6220 opamp..

LT6220
http://cds.linear.com/docs/en/datasheet/622012fc.pdf

If we are saying that we can have opamp inputs outside of the supply rails, then exactly how far outside the rails are we saying we can go?, and what documentary evidence do we have for this? Whats the tolerance on the figure of how far outside the rails we can go?...and again what documentary evidence do we have?.......Its never necessary to go outside the supply rails of an opamp....so why do it when there's easy other ways to do it?

Going outside the rails with opamp inputs is like going off-piste when skiiing, -you can do it, you can get away with it for a long long time, but when it goes wrong, and you get caught in that avalanche.....oh dear.

Why do we want our opamp inputs to be able to go outside the rails when there's no need?...there are other ways.

 

[crutschow]

Obviously you can't be convinced that a few mV below the SPEC LIMIT is okay, so you'll just have to live with your rigid view of what works.
If you don't want to use a rail-rail op amp at it's rated 0V input then don't. But don't expect the rest of us to adhere to that narrow view.
A few mV at the op amp input is hardly comparable to going off-piste in skiing.

 

[Flyback]

Obviously you can't be convinced that a few mV below the SPEC LIMIT is okay

If the feedback components are "slow", then it could be more than a few millivolts, surely?......i suppose i see your point though, that their is a potential divider affect there, (R8 and R7) and that any voltage below ground across the sense resistor is divided down by the divider, so the amount below ground that it goes isnt "generally" going to be that much.

 

[Les Jones]

You do not seem to understand to the circuit works. The voltage across R7 (3.5 x 2.4/35.4 = 0.259 volts) is the fixed reference voltage that is being compared with the voltage drop across the current sense resistor.

 

[ronsimpson]

...thanks, though when the inv input goes below 0v, its no longer inside the common mode range.

OK..OK but it will work. With one input within range and the other out of range it will work. With the inv input at -0.2V it will understand it needs to pull the output high. Good enough. By the time things are back in balance both inputs will be at 0V. (inside range)

 

[Flyback]

You do not seem to understand to the circuit works. The voltage across R7 (3.5 x 2.4/35.4 = 0.259 volts) is the fixed reference voltage that is being compared with the voltage drop across the current sense resistor.

Sorry but the reference voltage is 0V here.
The voltage at the divider point of R7/R8 is an "image" of the voltage thats being controlled....the voltage that is being controlled is , as you know, the voltage across the current sense resistor.

In some ways your statement is what anyone would think when looking at the circuit at a first glance.......but when you take your second look, you will see, as i did, that in fact there's a bit of didgereedoo going on here.....a bit of "skiing off piste" as i would put it.......a circuit that has an opamp input going outside the rails, when there are other ways to do it...other clearer and simpler ways to do it....i'd just throw in a sot23-5 hi side current monitor and do it via that way.

There's definetely "something of the night" about having an opamp control circuit where the voltage being controlled can and does go outside the rails of the opamp....when theres other ways.

 

[tcmtech]

i'd just throw in a sot23-5 hi side current monitor and do it via that way.

And what will happen if one of the high side current sensing inputs go above the positive rail voltage?

Personally I sort of understand why your company chose to hire this out to an external source rather than have you do it in house.

 

[ronsimpson]

sot23-5 hi side current monitor=$
Why not use a amp that is good to -0.3V?

 

[ChrisP58]

Flyback.

Although I feel the existing design is OK, I understand that if you just can't accept that, you won't feel comfortable without fixing it. And, whereas you could use a highside current sense amp, there is a much simpler and cheaper solution.

Just move the move the non-inverting pin slightly above ground. And that can be done with a single resistor.

In my simulation, I changed R10 to 2.4K, and added a 1.5Meg resistor from the non-inverting pin to your 3.5V ref at the top of R8. This moved the DC point of that node up to +5.6mV. And the waveform at the inverting pin now runs 5.4mV to 5.7mV peak to peak. So now everything is above ground.


Cost = 1 resistor.

 

[Flyback]

Hello ChrisP58.

I really like your suggestion.
That clinch's it for me.

What you are saying is the way its done on page 8, fig 5 of the following , with TSM101..

Figure 5 page 8.....
https://www.st.com/web/en/resource/technical/document/datasheet/CD00001116.pdf


...I beleive the whole point of the current source in the TSM101 is to enable the excellent way that ChrisP58 has described.
So i think not only is ChrisP58 got it bang on, but Chris has also given the bona fide way to do this sort of thing.

Isnt it interesting that absolutley none of the TSM101 app notes have the opamps ever having inputs going below the TSM101 supply rails...they always use the internal current source to bias the noninv input above ground as ChrisP58 has suggested.

And what will happen if one of the high side current sensing inputs go above the positive rail voltage?

that wouldnt happen, as you can see from the ltc6102
https://www.linear.com/product/LTC6102
..the way its connected disallows that

 

[ChrisP58]

It looks like that LTC6102 would have cost you about USD 2.50. I'll take 10% of the money I just saved you.

 

[Flyback]

thanks, there are much cheaper ones from diodes.com

 

[MrAl]

Hi,

When a design is done the result is the result of the whitepaper studies as well as the experience of the engineer. It could be that the guy who did this design had enough experience with this kind of design to know it would be successful. You cant measure that by looking at the data sheets.
So just because the design in questionable on paper doesnt mean it wont work in real life.

Here we have the question of the operation of the op amp with various inputs that seem to go outside it's specified range of operation. There are however general rules you can follow that help when this kind of thing happens, and it is not too severe. There are two basic rules that come into play here.
1. The PNP stages allow input operation below ground. This happens regardless of the 0v spec.
2. The input can go as low as 0.4v, and for low currents would be protected from going lower than about 0.5v or so by the ESD diodes, which can usually take some milliamps.

The question of input offset also comes into play however. If the input offset is positive on the non inverting terminal (and it probably is) then the design works every time. If it is negative however and the other input is designed so that it can not go lower than this value (which seems to be what you said earlier) then the design doesnt work at all. If it is close to the right value, then yes it might work sometimes and not other times.

But since the input offset is the last question to be answered, a simple test reveals the go or no-go of the design. After all, we just want to know for sure if the design works all the time or not, for any particular op amp we pull off the shelf of that part number.
The test is to bias the non inverting terminal slightly negative, by 5mv. If the unit still operates as it did before, then the design passes. If it fails or acts erratic, then you really do have to bias it positive and there should be no doubt about the need for this. The reason for the 5mv is because that would be the worst case for this op amp according to the data sheet plus a tiny bit more.

Now you might still question the design because it goes below the spec of 0v, but that would be somewhat normal, as long as the input current is always low (like 1ma).

It would be interesting to see the results of the -5mv bias test regardless how you intend to proceed as this would tell us if the designer was really on his toes or not.