• Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Isolated SMPS with cascoded opto feedback is superior?

Status
Not open for further replies.

Flyback

Well-Known Member
Please note these two equivalent offline flyback simulations.
-one has cascoded opto feedback , and the other does not.
Note how the non-cascode version has 2V of overshoot on vout at start-up. Due to the lack of a cascode, this overshoot is virtually impossible to get rid of whilst keeping sufficient gain and phase margin. The cascaded version has no overshoot, and achieving this was simple.
Cascodeing is extremely cheap to do...why does everyone not do it?
Schematics and LTspice simulations of offline flybacks given. (one cascode and one no-cascode, but they are the same power level, vin, vout, np/ns, Fsw.)
 

Attachments

Last edited:

ronsimpson

Well-Known Member
Most Helpful Member
these two equivalent
Comparing apples and oranges:

1)The two output caps are 4.7:1 different.
2)One has a second error amp in the LT1243 with a gain of 2.
3)The "slow one" has a 10nF cap across the 'slow' opto. The fast one has no slow down cap.
4)R15 being different by 3:1 gives me the idea there is a big difference in opto current (translate speed) and a gain difference of about 3 .
5)There are many differences around the first error amp. R9&C5 (lead) and R14&C7&C9.

All the above is blames on 'to cascode or not to cascode'.

It seems we have been here before.
 

Flyback

Well-Known Member
1)The two output caps are 4.7:1 different.
2)One has a second error amp in the LT1243 with a gain of 2.
3)The "slow one" has a 10nF cap across the 'slow' opto. The fast one has no slow down cap.
4)R15 being different by 3:1 gives me the idea there is a big difference in opto current (translate speed) and a gain difference of about 3 .
5)There are many differences around the first error amp. R9&C5 (lead) and R14&C7&C9.
...all of these differences had to be put in place because otherwise the non-cascoded feedback network is impossible to stabilise.....this in fact emphasises the great value of casc0ding.

Why I ask, is cascoding so uncommon.?

Also, this forums auto text editor keeps turning "casc0de" into "cascade"

2)One has a second error amp in the LT1243 with a gain of 2.
the error amp in the LT1243 is bypassed in both cases
 

ronsimpson

Well-Known Member
Most Helpful Member
I removed a non functioning op-amp.
Changed a couple of things mostly R11. Needs some work but this is a first try. Looks more stable.
Note V(vout)/10 so the vout and the error amp(s) output can be seen on the same scale.
upload_2014-1-4_18-50-18.png
 

Flyback

Well-Known Member
R10/R12/C6 set the gain and frequency response of the error amp.
....I see what you mean, I totally forgot that in the casc0ded version, I had used the "error amp" internal to the LT1243 as a simple gain stage, as you know, it is not being used as an "error amplifier", the error amplifier is outside the chip, and is done by U3 opamp...but I do see what you mean, you are lumping that gain in with the gain of the error amplifier which is fine.....to be honest, I just wanted to use the internal amplifier in the LT1243 as a unity gain follower, but there was no access to the noninverting pin, so I had to use it as an amplifier....that's why I forgot I had used it as an amplifier.

I removed a non functioning op-amp.
..its not non functioning, it is as shown on page 16 of this...
http://www.onsemi.com/pub_link/Collateral/AND8273-D.PDF
...pages 10 and 11 describe the use of that opamp...figures 15 and 17....that extra opamp makes it a current source that feeds the opto-diode, which makes for better transient response.

Thanks for your work here, You do appear to have got a good response there, and I am going to make your schem and investigate it in LTspice.

..OK just simulated your version, without the casc0de, and it was excellent. I am going to see if adding the casc0de to what you did makes it even better.

Also, I adjusted your version as attached, to make the opto transistor in the common emitter configuration, (instead of the common collector arrangement) and it oscillates wildly. I am wondering what it is about common collector configuration of the opto transistor that is so good? (and common emitter connection so bad) Is it something to do with the base-collector capacitance of the opto transistor seeming less when in the common collector configuration?, (since collector voltage is fixed when in common collector configuration.)

Also, do you believe that the onsemi app note (AND8273) that I linked to here is talking nonsense when they suggest the casc0de?..is just simply using the opto transistor in the common collector mode sufficient to mitigate the problematic low opto pole frequency?..the common base stage of the casc0de not being needed?
 

Attachments

Last edited:

ronsimpson

Well-Known Member
Most Helpful Member
Second op-amp: Yours does nothing. I don't think it is the same as Mot's.
Cascode: The point is to have the Vc-e stay constant. (almost) This fights the Miller effect. By adding the transistor the Vce of the opto is probably constant to 50mV. I usually use the error amp in the IC and set the gain to 2 to 10. (there might be a problem at start up, but when running....) If the gain is 10 then the delta voltage across the opto is 1/10 and thus the Miller effect is 1/10 and the speed of the opto is faster. I think by adding gain you can get to the speed of the cascode.
 

Flyback

Well-Known Member
I think by adding gain you can get to the speed of the cascode.
...I think I agree, but without the casc0de, your speed is at the expense of reduced gain and phase margins(?)
Basso discusses opto feedback between pages 288 and 308 of his book "switch mode power supplys", and I am wondering why he does not make mention of the points which you raise in post #8.
I don't think it is the same as Mot's.
sorry I didn't understand what is meant by "Mot's"?

At the moment I am confused as to whether or not the casc0de really does help, and also, whether common collector opto connection is as good as using the casc0de stage, or whether the common emitter and common collector connections are just the same as each other.
Basso certainly speaks widely about common collector and common emitter opto connections, but does not suggest that one is any better than the other.....certainly , on the simulator, the common collector configuration, is far easier to stabilise than when using the common emitter connection.
 

ronsimpson

Well-Known Member
Most Helpful Member
Motorola/ Mot./ On Semi. same thing.

You can test the speed of op-amp/opto/amp (or cascode). Just build the error amp section and do a frequency sweep. Plot gain and phase verses frequency. I test the error amps by plotting the gain and phase of just the amps . Break down the power supply into small sections.
 

Flyback

Well-Known Member
looking at the onsemi app note referenced above, surely my connection of that second opamp is just the same as theirs?

Also, regarding comparison of common collector opto connection, and casc0ded opto connection, I ran the following simulation comparing each, and sucessively reduced the output capacitor until one went unstable....the results are dramatic, the common collector configuration goes unstable whereas the casc0ded version rings but the ringing dies away and it goes stable.
These circuits are otherwise exactly the same,and I would conclude from this that the casc0ded optocoupler feedback is superior?

LTspice simulation and schematic attached.
 

Attachments

ronsimpson

Well-Known Member
Most Helpful Member
You changed the output cap from 100u to 30u. (the cap was 470u at one point)
This effects the loop greatly. I changed the caps back to 100u and it look much more stable.

Half the loop is inductor (transformer), filter cap and switching frequency.
The other half is the R&C around the two amplifiers. (three is you count the opto)
If you change something in 1/2 the loop you probably should make a correction in the other 1/2.
 

Flyback

Well-Known Member
You changed the output cap from 100u to 30u. (the cap was 470u at one point)
This effects the loop greatly.
...Absolutely , that was the whole point of doing it, to creep each of the smps's (both casc0ded and non casc0ded versions) toward instability and see which one goes unstable first.
The change to 30u was made in both versions, and the non casc0ded version couldn't cope with it.
The casc0ded version was clearly superior.

By the way, your thoughts on the "second op amp" appear to be collaborated by the simulator......the one *with* the second opamp doesn't appear to be any better in transient response than the one without this "second op amp".
I am not sure what the onsemi app note was on about in relation to that opamp?

By the way RonSimpson, you did an excellent job of compensating the feedback loop of the flyback with optocoupler in common collector connection. In common emitter connection, I believe you would agree that getting such a great response is really hard. I have made an attempt here, but it overshoots by more than 10%...wheras your common collecter version had no overshoot.
I bet it is simply not possible to do it with common emitter opto connection and get no overshoot?
LT spice sim and schem of my attempt at compensating the feedback loop of the flyback with common emitter feedback connection are attached.

Page 17 of this states how "The common-collector configuration eliminates the miller
effect of the output transistor’s collector-to-base capacitance"
and generally increases achievable loop bandwidth...
http://cds.linear.com/docs/en/datasheet/4430fc.pdf
 

Attachments

Last edited:

Flyback

Well-Known Member
Regarding Cascoded optocoupler feedback for SMPS's, on page 9, figure 13 of the following document concerning casc0ded optocoupler feedback, which transistor in figure 13 should be prevented from going into saturation?...is it Q1 or the opto-transistor?......the document says its the opto-transistor that should be stopped from saturating, but surely that can't be right?...its surely Q1 which must be prevented from going into saturation?
http://www.onsemi.com/pub_link/Collateral/AND8273-D.PDF
 
Status
Not open for further replies.

EE World Online Articles

Loading
Top