# [SOLVED]PSFB current sensing

Status
Not open for further replies.

#### Porsche

##### Newbie level 6
Hello everyone! I am working on PSFB based on UCC3895. My design is practically similar to the DS’s one. Before turning it on at full power i decided to look at current waveform, so i put a load resistor at the output of the converter and have disabled feedback loop. So PWM controller provides 100% phase shift at any condition.
First, I turned on PWM controller and then I started to raise HV voltage slow. As I have some load resistor at the output, I expected to saw current waveform at Rcs also rises as I raising HV voltage. But instead, I saw that current waveform is not a right shape and it has low amplitude. What could be the reason?
My current sensor has a 1:50 transfer ratio, the secondary inductance is 3,4 mH.
Here what i got:
1. Input current 0.522A Yellow - measured across Rcs (20R); green and red at QA and QD gates.
2. Input current 0.517A Yellow - measured across Rr (20K)
1. Input current 1.016A Yellow - measured across Rcs (20R); green and red at QA and QD gates.
2. Input current 1.012A Yellow - measured across Rr (20K)
Simulation shows current waveform should look more like sawtooth.
P.S. I am sorry for my grammar. I hope it's not so terrible.

#### Attachments

• Schematic_PSFB_2021-07-16.pdf
191.3 KB · Views: 6

Solution
1. Input current 0.522A Yellow - measured across Rcs (20R); green and red at QA and QD gates.
2. Input current 0.517A Yellow - measured across Rr (20K)
View attachment 170798View attachment 170799
1. Input current 1.016A Yellow - measured across Rcs (20R); green and red at QA and QD gates.
2. Input current 1.012A Yellow - measured across Rr (20K)
View attachment 170800View attachment 170801
The thing that sticks out to me is that the CT swings negative than positive, whereas I expect the opposite. Could the CT be wired backwards?

#### Easy peasy

have you sized the CT properly?

you need to allow reset time on the CT, so you can't run more than 160 deg phase shift, 0 ph shift = zero power.

have you sized the main Tx properly? what is B max ?

#### Porsche

##### Newbie level 6
Shure, I have sized main transformer properly, it has ETD54 core with equivalent crossection area of 280 mm^2 and 25 turns in primary side. It sized to work fine at much higher voltages that I am testing right now.

The CT is a discrete one. I harvested it from some junk PCB. I opened it and counted turns. It has 1 turn in primary and 50 turns in secondary, crossection area is approximately 8 mm^2. I didn’t get why is maximum phaseshift is like 160 deg? Why not 170 or 180? I have dead time about 500 ns in both A/B C/D pairs. So it should be enough time to reset CT even at 180 phase shift.
Maximum duty cycle is:
$$D_(clamp)=(1/2f_s-t_d)*2f_s=0.9$$
delta B is:
$∆B=(U*dt)/(N*A_e)$
Where U (voltage on secondary) can be rewriten as:
$U=(Ip/N*R_cs+V_f)$
where vf -forward diode voltage
rewrite dt as 1/2fs and use D:
$∆B=(I_p/N*R+V_f)*D/(N*A_e*2f_s)=0,037 T$
which gives us:
$B_max= ∆B/2 =0.019 T$

#### Easy peasy

Something is limiting the reset on the CT, for starters remove the 20k resistor, make sure the CT diode is 100V - ideally a 1A schottky, it needs to be very fast ( Trr < 30nS )

further, if we assume 4A pk in the pri ( 1600W for 400V bus ) the sec current in the CT is 80mA x 20E = 1.6 plus the diode drop = 2.4 volts, say 2.6 to allow for IR drop ( p.s. CT res of 20R = 128mW worst case dissipation ).

If we assume a 4uS ON time for the PWM applied to the Tx and if we assume the CT core is 4mm x 4mm ( from photo )

then from dB / dt = V / ( N . AE ) we get Bpk = 13mT - which is OK

If we need to reset in 200nS say, the reset volts need to be at least 4uS/0.2uS x 2.6V = 52V for the whole 200nS ( the capacitance of the CT wdg may inhibit this - requiring a longer reset time ).

We can possibly assume the CT is damaged, i.e. nearly shorted turn due to flash-over from being run open ckt.

and/or the CT diode may be damaged / inappropriate.

#### Easy peasy

Just looking at the ETD54, 25T, 400V (?) 292mm^2 Ae best case, gives 110mT each way for flux

at 125kHz, this will give rise to some core heating - unless you have some very good ferrite, e.g. 3C96 -> 4.3 watts min core heating.

The Lpri will be in the order of 2.5 - 2.75 mH, so the Imag will be in the order of +/- 320mA on 4A pk, i.e. 4.32A pk plus the contribution of the output choke ( approx 0.5 amp ) = 4.82 Apk in the pri. ( Assuming 25:4 turns ratio giving 64Vpk out of the Tx )

p.s. the diodes D9,10 will need snubbers to avoid going bang due to over volts, as will the o/p diodes ....

#### cupoftea

##### Full Member level 2
Yes, as pointed out above by Easy Peasy you are using unipolar CST and you have no reset...you need diode/zener reset on the secondary, to give you the reset volts that Easy Peasy describes.
The attached gives some theory on CSTs

Note that the CST for a Phase Shift Full Bridge (PSFB) SMPS should be placed into the bridge of the PSFB, and should be a “bi-directional” type of CST setup, as shown in the schematic of this folder.

You are best not using the “Unidirectional” type of CST, which is also shown in the schematic, because the current in the branch between the input capacitor and the primary FETs is NOT unidirectional, but it is bi-directional, and even worse, it is non-symetrical. (-the current is not unidirectional because of the “back-flowing” leakage inductance current.)
--- Updated ---

Please also find a LTspice sim of a PSFB....it shows both types of CSTs in use.
LTspice is free
--- Updated ---

I also hope that you have a gap in your transformer core (ETD54)......its required for PSFB's...because they are lovers of magnetising current....which they slosh back and forth and get zero voltage switch on.
--- Updated ---

Actually, rather than jam up the servers with more attache's....pse find here a link to a free SMPS course, which has a folder on PSFB (and CSTs)
--- Updated ---

You dont say your vin....but you have caps in series at the input...so are you running off a rectified three phase output?....if so, there is some good knowledge on that here, and on the PSFB

...you could benefit from caps across the fets as discussed there, (in fact, by Easy Peasy)

#### Attachments

• Current sense transformers.zip
9.7 MB · Views: 1
• Phase Shift Full Bridge.zip
4.3 KB · Views: 2
Last edited:

#### Easy peasy

CT's can be successfully used in series with the main Tx, and in the opinion of some PE engineers - work better than in the DC feed - mainly due to AC flux swing and that the power stage can go to 180 out of phase....
--- Updated ---

p.s. even with the CT in the DC - you still need a cap (film / foil ) right at the H bridge to reduce turn off stress, and for RFI.

#### Porsche

##### Newbie level 6
Something is limiting the reset on the CT, for starters remove the 20k resistor, make sure the CT diode is 100V - ideally a 1A schottky, it needs to be very fast ( Trr < 30nS )
If I understood you correctly, as I remove 20K resistor I am not limiting the reset voltage anymore? So if the problem is saturation this idea should work for me. Reset voltage will develop at stray capacitance of CT and if my diode is the tough one, it should survive it. Am I got you right?
CT diode is a fast one Trr = 35nS, Vf=1.7, Vr=600V, If=1A. Also i tried to put 100V 1A schottky in there - nothing changed.
If we need to reset in 200nS say, the reset volts need to be at least 4uS/0.2uS x 2.6V = 52V for the whole 200nS ( the capacitance of the CT wdg may inhibit this - requiring a longer reset time ).
As I mentioned above, I set dead time at 500 ns for now, so it should be enough?
We can possibly assume the CT is damaged, i.e. nearly shorted turn due to flash-over from being run open ckt.

and/or the CT diode may be damaged / inappropriate.
Yes I checked transformer using pulse generator and scope, transfer ratio is 50:1 @ 200 kHz. I also double checked the diode.
Just looking at the ETD54, 25T, 400V (?) 292mm^2 Ae best case, gives 110mT each way for flux

at 125kHz, this will give rise to some core heating - unless you have some very good ferrite, e.g. 3C96 -> 4.3 watts min core heating.
Main transformer will work @ 350VDC at primary as nominal value. Now it is less than 200 volts, so I don't think this is problem for current sensor.
Yes, as pointed out above by Easy Peasy you are using unipolar CST and you have no reset...you need diode/zener reset on the secondary, to give you the reset volts that Easy Peasy describes.
The attached gives some theory on CSTs
Thank you so much for all these materials, it will be useful for me in the future!
I don't think I got your point about zener diode at reset. The zener clamp will only protect the diode against high voltage but not providing the reset. I have a 600V diode at CT so it should be fine almost at the all of reset conditions.
You are best not using the “Unidirectional” type of CST, which is also shown in the schematic, because the current in the branch between the input capacitor and the primary FETs is NOT unidirectional, but it is bi-directional, and even worse, it is non-symetrical. (-the current is not unidirectional because of the “back-flowing” leakage inductance current.)
I picked the unidirectional solution because I saw it in the datasheet and it should work fine for my PCMC purposes. I know, there's a lot of methods to sense the current but I want to understand the unidirectional solution first.
I also hope that you have a gap in your transformer core (ETD54)......its required for PSFB's...because they are lovers of magnetising current....which they slosh back and forth and get zero voltage switch on.
I don't have a gap in ETD54 core, I have shim inductor 30uH in series with main transformer.
You dont say your vin....but you have caps in series at the input...so are you running off a rectified three phase output?....if so, there is some good knowledge on that here, and on the PSFB
My Vin is 350VDC nominal and 380VDC max. I am running from a TDK Lambda PSU for now, and then I will be using PFC circuit. Thank you for a useful information once again!
CT's can be successfully used in series with the main Tx, and in the opinion of some PE engineers - work better than in the DC feed - mainly due to AC flux swing and that the power stage can go to 180 out of phase....
In my opinion crucial disadvantage of this "AC" solution is that CT in series with main transformer is not detecting a shot through currents in one of the legs.

#### cupoftea

##### Full Member level 2
To ensure a transformer is reset, you need , for any coil, volt*seconds ON = volt.seconds OFF.

I am not sure what you meant by saying you have a 600V diode involved in reset of your unipolar CST? (or have i got you wrong?..if so, apologies)....the docs i sent you show the diode/zener reset method being used.

#### Porsche

##### Newbie level 6
To ensure a transformer is reset, you need , for any coil, volt*seconds ON = volt.seconds OFF.

I am not sure what you meant by saying you have a 600V diode involved in reset of your unipolar CST? (or have i got you wrong?..if so, apologies)....the docs i sent you show the diode/zener reset method being used.
Yes, please see my picture in previous post. For any coil, volt*seconds ON = volt.seconds OFF and this mean that areas under the curve must be equal. As I have a dead time 500 ns, I must provide a sufficient amplitude of voltage on the secondary of my CT during this 500ns reset time (to make V*s equal). A 600V diode can handle up to 600V during this reset time so V*s can be huge. It is the way i see things. Zener could clamp this voltage to ensure diode will not be damaged during this reset time. This is what I meant.

#### mtwieg

1. Input current 0.522A Yellow - measured across Rcs (20R); green and red at QA and QD gates.
2. Input current 0.517A Yellow - measured across Rr (20K)
View attachment 170798View attachment 170799
1. Input current 1.016A Yellow - measured across Rcs (20R); green and red at QA and QD gates.
2. Input current 1.012A Yellow - measured across Rr (20K)
View attachment 170800View attachment 170801
The thing that sticks out to me is that the CT swings negative than positive, whereas I expect the opposite. Could the CT be wired backwards?

#### Porsche

##### Newbie level 6
The thing that sticks out to me is that the CT swings negative than positive, whereas I expect the opposite. Could the CT be wired backwards?
Hello, mtwieg. Strange, It is seems fine to me. CT swings positive first and negative then. Am I wrong?

#### Easy peasy

There appears to be no phase shift .... ?

#### Porsche

##### Newbie level 6
There appears to be no phase shift .... ?
Yes, im sorry, if i didn't make it clear. QA and QD have no phase shift (diagonal pair). Tx exposed by maximum duty cycle - D clamp.

#### Easy peasy

try showing the volts across the Tx, CH1 - CH2 .... you won't see much without some actual phase shift.

#### cupoftea

##### Full Member level 2
Hi,

Attached is a basic LTspice simulation of your PSFB with the standard unipolar CST.

This one is open loop.

You can adjust the phase shift by adjusting the value “phase_shift_degrees” in the simulation schematic window.

At a phase shift of 148 degrees, the output into the 2.304 Ohm load is 48V at 1kW .

You can see the operation of your unipolar CST. As we know, the CST primary has bidirectional current in it (even though its in the DC supply rail of the PSFB). You can clearly see that the secondary current in the CST is a mirror image of the primary, but shifted up significantly. As we know, this is all to do with the fact that your primary current is not unipolar. I think you should change to a bidirectional CST and put it in the bridge.

Also attached is an LTspice sim which uses both unipolar and bipolar CSTs so you can compare them.

You could run the simulation and change the couplings in the transformers to see how your waveforms change. Change the coupling by adjusting the k value beteeen 0 and 1.

I believe you need to get a representative simulation working before going to the bench. A working simulation is meaningless in many ways…But the rule of simulation is that if you can’t get a basic representative simulation working, then generally you will never get anything working on the bench.

#### Attachments

• PSFB_with unipolar CST1.zip
2.8 KB · Views: 1
• PSFB_avec unipolar CST and bidirec CST1.zip
2.9 KB · Views: 1
• PSFB schematic1.pdf
177.8 KB · Views: 1
Last edited:

#### Velkarn

##### Full Member level 3
R72 helps to demagnetize the CT. Recommendation is R72=100*R74 as i remember. why you put 20k?
--- Updated ---

is the R73,C62 time constant ok for used freq?

Last edited:

#### Porsche

##### Newbie level 6
Hello again, cupoftea! I am just really touched with you care! Thank you so much for these awesome simulations you prepared for me! You can be sure this stuff will help me a lot!

A am agreed with you at 100% that «get a representative simulation working before going to the bench». My bad, I am really not aware person about simulators. Just started to learn LTspice. It is always looked kind a complicated program to me, I never really knew how to start to use it. Those simulation you made for me is perfect thing to start with.

Concerning unipolar CST vs bipolar CST situation. As you said there is difference in amplitude of voltage between two waveforms. Well, I am agreeing, unipolar solution is not the most precise current sensing technic, nevertheless I don’t need accurate current sensor at primary side of my converter. This CST is only generating for me a sawtooth to support PCMC operation and it also feeds the sensed current to the CS pin. In both these cases I can easily adjust value of my RCS (using pot for example) to make current limit trip at power level that I need.

Shape of these two waveforms looks quite similar for me.
The disadvantage of the bipolar solution is much more critical one. This bipolar CST simply do not measure short circuit current of AB/CD transistors. So, if for some reason both QA and QB are in the ON state, bipolar CST will sense zero current and will not trigger the current limit. However current through QA and QB will cause a big smoke.

So, if I am got you right and difference between these two solutions in accuracy of current measuring only, I will still prefer a unipolar CST that I can trim as I need.

try showing the volts across the Tx, CH1 - CH2 .... you won't see much without some actual phase shift.
I will repeat my measurements on Tx this time, sorry for confusing you. I thought gate signals on QA and QD will be enough.
R72 helps to demagnetize the CT. Recommendation is R72=100*R74 as i remember. why you put 20k?
Yes, you are totally right, my mistake is to put 20k instead a 2k resistor. Will change it.
is the R73,C62 time constant ok for used freq?
I took this values from DS. DS says this filter has a pole @ 482 kHz. Seems fine to me.

#### Easy peasy

making R72 smaller lengthens the reset time for the CT, you want more volts to reset quickly, not less.

#### cupoftea

##### Full Member level 2
Sorry to harp on, but Regarding this problem of needing the unipolar CST because it detects shoot-thru current, wouldnt you still think it better to use a bipolar CST in the bridge, and just use some sort of latching overcurrent protection for shoot thru?

Status
Not open for further replies.