Unidirectional Current Sense Transformer cannot be used in Phase Shift Full Bridge?

Hello,
Please can you confirm that the use of a Unidirectional Current Sense Transformer (CST) as shown on the first page of the UCC28950 (Phase shift full bridge converter) datasheet is totally wrong?
(It is a Phase Shift Full Bridge converter and the Current Sense Transformer should instead be a Bidirectional Current Sense Transformer. (-because the current in the primary of the CST is bi-directional due to the current back-flowing from the leakage inductance.)

UCC28950 datasheet
https://www.ti.com.cn/cn/lit/ds/symlink/ucc28950.pdf

The attached LTspice simulation and pdf schematic of a Phase Shift Full Bridge converter shows the situation. You can see that the magnetising current in the unidirectional CST is way too high.

I think the discussion is pointless without a current transformer specification. We have to assume that the transformer has been chosen to be free of saturation.

I don't understand the idea behind sensing the pulsed DC bus current instead of the output AC current as usual.

I think the discussion is pointless without a current transformer specification. We have to assume that the transformer has been chosen to be free of saturation.

Click to expand...

OK i take your point, but i am sure you agree that the magnetising current is hugely less for the Bidirectional CST placed in the bridge, than for the unidirectional one shown.
As such, there is no way anybody is going to pick the Unidirectional CST setup. Surely you agree?...i added the expressions for magnetising current in the LTspice simualtion, and its plain to see that the magnetising current is ridiculously large in the Unidirectional CST setup......even before you start picking ferrites , you are surely never going to pick a unidirectional setup over the shown bidirectional CST setup.

As you know, no spec is needed to make this choice... that whatever ferrite you will end up using, you would only use the Bidirectional CST setup...surely?

If I understand right, the peak flux is doubled in "unidirectional" CT, presuming complete reset during off time. Not more, not less. That's not necessarily "ridiculously large".

Yes but that is not the point here......the point is that ti.com are using a unidirectional CST in a circuit branch that contains bidirectional current (due to the back-flowing leakage inductance current). And..Its bidirectional but not symetrical current.....and this is producing the very high magnetising current that i speak of. You can either run the simulation of the top post, or i will run it and post you the magnetising current plots, and you will see what i am talking about.

- - - Updated - - -

The green plot shows the magnetising current in the bidirectional CST, which is placed (correctly) in the bridge branch. The Blue plot shows the magnetising current in the unidirectional CST, which is placed inbetween the input capacitor and the Primary switching bridge.

These positions are both shown in the top post…and the simulation of the top post is what I have used to produce this comparative plot of magnetising currents. Surely you now agree that ti.com have got there CST totally wrong, as discussed in the top post.

It's true that the primary current in the CT will be bidirectional, and the negative part of the waveform will not be sensed by the controller. For the purpose of CMC, this should be fine, as you only really need to detect the positive pulses.

As for saturation, I presume they've selected the burden resistor appropriately to ensure that it doesn't happen.

Thanks Mtwieg but I am afraid that is not relevant to this discussion.
What i am saying is the CST as shown in the ti.com document of the top post is obviously totally wrong. (reasons as per the discussion)

OK , its not exactly "totally wrong", but "totally and utterly unadviseable", and a bidirectional CST should be used instead, and placed in the PSFB bridge.

I am afraid that is not relevant to this discussion.

Click to expand...

The simulation in post #5 shows the relevance. You selected the circuit parameters so that the transformer is not reset during off-time. I presume that's not the intended operation.

As already said, I won't use the current transformer configuration in the datasheet. But I think it can work under circumstances, it's not so bad as you claim.

treez said:

Thanks Mtwieg but I am afraid that is not relevant to this discussion.
What i am saying is the CST as shown in the ti.com document of the top post is obviously totally wrong. (reasons as per the discussion)

OK , its not exactly "totally wrong", but "totally and utterly unadviseable", and a bidirectional CST should be used instead, and placed in the PSFB bridge.

Click to expand...

Your simulation does not accurate represent TI's design approach (they use a reset resistor with value 100x the sense resistor). Try simulating the design accurately before dismissing it. Adding a 200ohm resistor decreases the magnetizing flux greatly.

Also consider that if you were to use ACM control, then using a primary current sense would not be appropriate because the primary current during the freewheeling period does not describe the current being delivered to the output. Input current sensing does.

i have simulated it with a reset resistor being 100x that of the burden resistor, and once again, the magnetising current ends up being well higher with the unidirectional CST.....in fact , its over 5 times higher than with the bidirectional CST. So there's no way you would use a unidirectional CST with a PSFB, because you would just end up needing a much bigger CST torroid.

The solution with a unidirectional CST, (as in the top post) makes no sense.

Attached are the magnetising currents of the unidirectional (green), and bidirectional (blue), solutions with a 100x reset resistor in the unidirectional CST circuit (as the datasheet recomends).
There is no way anyone would rightly pick the unidirectional solution.

The other reason is that the unidirectional CST seriously makes the secondary current waveform inaccurate.......the secondary current is not conforming to the I(pri) * 0.01 expression....it is more than 20% out.....whereas the bidirectional CST is exact to the 100:1 turns ratio.

Yes it could be workable with a unidirectional CST, but there is no way anyone would pick that unidirectional CST that they show as in the top post. It should be a bidirectional CST with a PSFB.
I am saying this because i know you will find this as intriguing as i do.

- - - Updated - - -

Also consider that if you were to use ACM control, then using a primary current sense would not be appropriate because the primary current during the freewheeling period does not describe the current being delivered to the output. Input current sensing does.

Click to expand...

..The UCC28950 does not implement average currentmode control.....the top page says it does, but when you read in, it does not.
Also, if i wanted to do ACM on say a different chip, i would put the CST in the secondary side and stick with bidirectional CST...the unidirectional CST with 100x reset resistor doesnt accurately depict the current anyway.....the simulation shows this ...more than 20% out.

- - - Updated - - -

Another problem with the unidirectional CST solution is that the reset of the transformer is dependent on the amount of leakage inductance in the main transformer. –Because that determines the duration of the “back-flow” of the leakage current back through the CST primary.

All said and done though, the UCC28950 does look like a cracking good PSFB controller chip.

Without showing the actual circuit and current duty cycle, the comparison graph is meaningless, at least for me.

the circuit is as in the simulation shown here.(pdf schem and ltspice simulation attached here)
The circuit and sim shown here are aexactly the same as in the top post except that the reset resistor in the unidirectional CST has been changed to a 200R resistor, and all other reset components in that CST are removed. As you can see, its still bad. The bidirectional CST is clearly the best choice...by miles.

- - - Updated - - -

Without showing the actual circuit and current duty cycle, the comparison graph is meaningless, at least for me.

Click to expand...

The schematic and ltspice sim have both Unidirectional and bidirectional CST's connected in to the same converter, so that both can be contrasted and compared.

its not a very good app, but a CT will work in this manner without problem, just watch the diode reset voltage...

the lack of bus de-coupling in that schematic will cause more headaches....

Thanks, i wasn't sure which "manner" you were referring to?..the "unidirectional" CST? ....or the "bidirectional" CST?
You agree the bidirectional one is far, far better for the PhaseShiftFullBridge?
(yes i didnt bother with a bus cap as its not necessary in the simulation, but i agree with you its very necessary otherwise)

If you regard the sign of the accumulated magnetizing current in your simulation, you'll notice that's it's caused by the DC bus reverse rather than the forward current, apparently the CT termination could be improved. But in the first place I agree with Easy peasy that I won't make a bridge converter without bypass capacitors.

it will work as per the original app schematic, just need to limit the pwm to 95% say to allow some reset time on the CT, and limit the "flyback" or reset volts on the CT, also the CT winding needs to be low capacitance....

Thanks -the pdf in post#5 actually shows you the difference in the magnetising current for the unidirectional vs bidirectional CST. From this, i believe we agree that the bidirectioanl CST is far far better than the unidirectional CST for the Phase Shift Full Bridge?
Its more accurate aswell.

it makes a bit more sense to have the CT in series with the Tx, rather than in the power feed...

- - - Updated - - -

the Imag in the CT in the app note is a pulse train, there is no DC offset, the sim needs fixing, having 5% OFF time allows a full reset in the CT (the reset volts will be 20x the forward direction)

the Imag in the CT in the app note is a pulse train

Click to expand...

are you referring to the datasheet of ucc28950, which is linked in post #1?

the sim needs fixing

Click to expand...

I am sorry but the sim doesnt need fixing as it shows the point being made, that is that there is the build up of magnetising current in the unidirectional CST because the current in its primary is not unidirectional.

The fact that you agree with me that the CST should be in the bridge (ie bidirectional type), tells me that you actually agree with my point, just that we are talking at crossed purposes along the line somewhere

because CT's are usually toroidal, and have lots of turns on the o/p wdg, it takes very little DC current to saturate them, hence the sim shows a situation that simply won't work /doesn't occur, in a real cct, the Imag would reset every cycle for a sufficiently good diode and low Cwdg...