PSFB output ripple current issue

[pxidr]

Hi everyone,

A bit new here, seeking help for my issue after many attempts to solve the problem myself...

I'm currently developping a 100kHz, 24kW PSFB for an EV charger, using a full-bridge SiC MOSFET at the input (700VDC nominal), a custom-made planar transformer, and full-bridge SiC diode rectification at the output (4x2 STPSC30H12CWL arranged in parallel for current handling), the goal of the design is outputting a variable 50-500VDC at 64ADC max.

My output inductor is 33uH and my output capacitance is 200µF, using 4x50µF low ESR film caps (TDK B32776Z5506K000).

I'm currently testing my PSFB on a 300VDC lab power supply and a 2kW resistive load, voltage and current regulation at the output are fine.

At 110VDC output, 10ADC, my voltage ripple is acceptable (about 1,26V pk-pk), but my current ripple is way too high (776mA pk-pk).

I'm unable to resorb this current ripple. The weird thing is that current ripple doesn't increase if I increase the load, it is pretty constant at 2A, 5A, 10A etc.

Any suggestions for filtering out that current ripple ?

 

[cupoftea]

When you say "current ripple", presumably you mean the ripple in the load current, ie, that current that is going into your eLoad?
This is to do with the way that eloads work.....they arent like load resistors...they actually have the appearance of a capacitance....and that is why you see the "ripple current" going into it.

 

[pxidr]

When you say "current ripple", presumably you mean the ripple in the load current, ie, that current that is going into your eLoad?
This is to do with the way that eloads work.....they arent like load resistors...they actually have the appearance of a capacitance....and that is why you see the "ripple current" going into it.

Hi,

Yes I talk about the current ripple going into my load.
It is not an electronic load but a purely resistive load (heater elements connected in parallel).

 

[dick_freebird]

If the inductor is strictly inductive then you should see a
clean triangle wave. But looks like significant interwinding
C is letting switching edges push through to the filter, and
filter cap ESR/ESL too high to tamp it down.

I'd test this by swapping in bridge inductors of varying
core / winding design and see if that changes the
signature. Maybe look at better filter caps or adding
ones that seem appropriate to the edge-rate frequencies
(figure tr = 1/4 cycle) SRF-wise and low enough ESR/ESL
to squash the HF components of ripple?

 

[Easy peasy]

Suggestion: more caps is really the only way here ( and/or a bigger choke ) for charging, the batt does not really care about a wee bit of ripple ( and in fact acts like a pretty good capacitor itself if Li-ion ) - so this should really be a relaxed figure in the spec.

Also we would all be keen to see some waveforms at full power - doing 24kW @ 100kHz successfully at full power is rare - i.e. it going for more than 30 mins - I assume you have cold plate ( water ) cooling ... ?

p.s. the current ripple in the output choke will increase as you go up in Vout - which is when you will notice it getting very warm indeed ...

 

[pxidr]

If the inductor is strictly inductive then you should see a
clean triangle wave. But looks like significant interwinding
C is letting switching edges push through to the filter, and
filter cap ESR/ESL too high to tamp it down.

I'd test this by swapping in bridge inductors of varying
core / winding design and see if that changes the
signature. Maybe look at better filter caps or adding
ones that seem appropriate to the edge-rate frequencies
(figure tr = 1/4 cycle) SRF-wise and low enough ESR/ESL
to squash the HF components of ripple?

Yes, I can try to vary the inductance or the winding design.
BTW, I don't use any MLCC at the output, only 4 film capacitors (specs attached, note the low ESR).

Do these MLCCs can filter this 100kHz current ripple?

--- Updated Sep 28, 2022 ---

Suggestion: more caps is really the only way here ( and/or a bigger choke ) for charging, the batt does not really care about a wee bit of ripple ( and in fact acts like a pretty good capacitor itself if Li-ion ) - so this should really be a relaxed figure in the spec.

Also we would all be keen to see some waveforms at full power - doing 24kW @ 100kHz successfully at full power is rare - i.e. it going for more than 30 mins - I assume you have cold plate ( water ) cooling ... ?

p.s. the current ripple in the output choke will increase as you go up in Vout - which is when you will notice it getting very warm indeed ...

I can imagine that the EV Li-ion battery could smooth this out a bit (i didn't try on a battery, only on resistive elements).

It's more an EMI concern, because there will be some length of cable to the battery. Of course there will be an EMI filter at the output of the converter, not fitted at the moment.

I didn't pushed it to 24kW - I'm trying to resolve that current ripple issue first - but I successfully pushed it to 6kW with a 700VDC input (feeded by my PFC stage).

For the thermals, all power semiconductors and inductors are mounted on a big heatsink, with forced air cooling. And even at 24kW, total power losses should not exceed 800W.

 

[cupoftea]

When you put the battery load on...there will be loads of ripple going into it...because it is an enormous capacitor.
So i wouldnt be worrying about ripple here.

 

[cupoftea]

Probably most of your ripple will end up in the battery, rather than your 200uF

 

[Easy peasy]

All your EMI concerns will be in the 10MHz to 250MHz range - so don't worry too much about a bit of switching ripple in the cable - remember the cable has some L - which will assist.

Worry about the EMC caps to earth on the HVDC exit from the charger - these needed to try and limit the CM emissions.

MLCC's can be used - you need the stacked type, or a whole lot of 1210 630V types in series parallel on a separate solder in pcb - we have used this to good effect.

If the output ripple was clean - i.e. mostly single freq with nice rounded edges - your batt could easily tolerate 7A rms of ripple - it is the RFI noise that will cause an EMC fail ...

good luck above 6kW ...

--- Updated Sep 28, 2022 ---

@ cup of tea / flyback, the caps are close to the action while the batt is down a cable ( L ) - so actually these caps will see the majority of the ripple - it's just that the volt ripple will be low at the batt as it absorbs the remaining current ripple with ease ....

 

[pxidr]

All your EMI concerns will be in the 10MHz to 250MHz range - so don't worry too much about a bit of switching ripple in the cable - remember the cable has some L - which will assist.

Worry about the EMC caps to earth on the HVDC exit from the charger - these needed to try and limit the CM emissions.

MLCC's can be used - you need the stacked type, or a whole lot of 1210 630V types in series parallel on a separate solder in pcb - we have used this to good effect.

If the output ripple was clean - i.e. mostly single freq with nice rounded edges - your batt could easily tolerate 7A rms of ripple - it is the RFI noise that will cause an EMC fail ...

good luck above 6kW ...

--- Updated Sep 28, 2022 ---

@ cup of tea / flyback, the caps are close to the action while the batt is down a cable ( L ) - so actually these caps will see the majority of the ripple - it's just that the volt ripple will be low at the batt as it absorbs the remaining current ripple with ease ....

The full converter will be fully enclosed in a metallic box - also, there will be an output EMC filter (which is not the case right now).
Also, the cables going from the charger to the EV (via a CCS plug) are fully shielded.

I can try to put 1uF MLCC at the output (it's the only caps i got right now) before the output and post the waveforms here, to see if it changes anything.

 

[mtwieg]

At 110VDC output, 10ADC, my voltage ripple is acceptable (about 1,26V pk-pk), but my current ripple is way too high (776mA pk-pk).

It's strange that you think the load's ripple voltage is fine but the load's ripple current is not, when the load is a resistor and thus the two are directly related. In face, load ripple current is usually not something people put restrictions on.

But, as others are already telling you, your test setup with a resistive load is not going to have nearly the same ripple as when you use a battery as the load. Adding parallel capacitance can only reduce ripple if the impedance of that capacitance is lower than the load, and the battery impedance is probably far lower than any reasonable capacitor bank. Only realistic way to reduce ripple current on a battery is to increase inductance, or add another stage of LC filtering.

But I'm also not convinced that there's any point in reducing load ripple at all. At least not peak to peak ripple. Are there any standards limiting conducted emissions on charging cables?

 

[pxidr]

It's strange that you think the load's ripple voltage is fine but the load's ripple current is not, when the load is a resistor and thus the two are directly related. In face, load ripple current is usually not something people put restrictions on.

But, as others are already telling you, your test setup with a resistive load is not going to have nearly the same ripple as when you use a battery as the load. Adding parallel capacitance can only reduce ripple if the impedance of that capacitance is lower than the load, and the battery impedance is probably far lower than any reasonable capacitor bank. Only realistic way to reduce ripple current on a battery is to increase inductance, or add another stage of LC filtering.

But I'm also not convinced that there's any point in reducing load ripple at all. At least not peak to peak ripple. Are there any standards limiting conducted emissions on charging cables?

EMC standards for EV charging are defined in the IEC 61851-21-2 standard, the range of compliance is between 150kHz and 30MHz. So that 100kHz ripple should be fine I assume.

I also should try the output LC filter and see the results - a 4.7uH / 50µF low-pass has a cutoff frequency of around 10kHz, an order of magnitude less than 100kHz.

 

[Easy peasy]

Surely the standards refer to the noise on the mains only ?

conducted noise from the charger to the battery is mostly immaterial unless the radiated noise from same exceeds the standards required for mains side radiated noise ... usually 30MHz - 1GHz

excessive noise on the cables in the 500khz - 30MHz range can interfere with AM radio stations if the cables are such that they assist with antenna propagation, shielded cables do not assist, but CM noise on the outer will radiate.

 

[pxidr]

Surely the standards refer to the noise on the mains only ?

conducted noise from the charger to the battery is mostly immaterial unless the radiated noise from same exceeds the standards required for mains side radiated noise ... usually 30MHz - 1GHz

excessive noise on the cables in the 500khz - 30MHz range can interfere with AM radio stations if the cables are such that they assist with antenna propagation, shielded cables do not assist, but CM noise on the outer will radiate.

To the mains but also from the charger to the vehicle, and the charger itself.

I'm starting to suspect that conducted noise is causing my problem and perturbing my current probe, I will try to put an EMI filter at the output and see what happens.

 

[Easy peasy]

It's almost certain that CM noise is causing quite a bit of artifact on your scope traces.

 

[cupoftea]

There are Hall type currnt sensors with common mode noise cancellation features if you want.

 

[pxidr]

It's almost certain that CM noise is causing quite a bit of artifact on your scope traces.

So... I put an enclosed low-cost EMC filter that I found in my garbage bin. And the results are dramatic.

In the same conditions (110VDC out, 10ADC out) the current ripple is down to 23mA RMS and the voltage ripple to 90mV.

I still get some switching spikes tho.

 

[cupoftea]

How much is the series resistance of that?
Also, you kindly described that you are designign a 24W EV charger....i assure you that you have more to spend time on then this ripple as it doesnt matter.

As long as you filter enough ripple off to stop the battery cable from acting like an antenna then its job done...and for that, hardly any filtration is needed....if any at all.
What you must watch, is, when you connect the battery and cable.....does the mains PSU then fail on common mode emissions?

A ripple at the switchign frequency isnt going to radiate off the cable, no matter how big it is. Sorry to be nosey, but may i ask why you are worried abouth this battery ripple current?.....it isnt important.....what document tells you that it matters?

Also, as you know, for 24kW, i woudlnt be surprised if its cheaper to do it with multiple lower power current regulated SMPS chargers....as they are built with more commonly available, cheaper components.

 

[pxidr]

How much is the series resistance of that?
Also, you kindly described that you are designign a 24W EV charger....i assure you that you have more to spend time on then this ripple as it doesnt matter.

As long as you filter enough ripple off to stop the battery cable from acting like an antenna then its job done...and for that, hardly any filtration is needed....if any at all.
What you must watch, is, when you connect the battery and cable.....does the mains PSU then fail on common mode emissions?

A ripple at the switchign frequency isnt going to radiate off the cable, no matter how big it is. Sorry to be nosey, but may i ask why you are worried abouth this battery ripple current?.....it isnt important.....what document tells you that it matters?

Also, as you know, for 24kW, i woudlnt be surprised if its cheaper to do it with multiple lower power current regulated SMPS chargers....as they are built with more commonly available, cheaper components.

I was worried because I didn't understood where this relatively high ripple current came from, since my ripple voltage was relatively small.

In reality, the problem was the conducted CM noise made by my PSFB.

By adding an EMI filter at the output, all the current ripple was gone. I suspect that hall-effect current sensor probe I use is sensitive to this.

Now, with 20mA RMS of ripple current at 1kW output, it is clearly not an issue anymore.

And I think i'm not agreeing on your last affirmation. Pile-up multiple smaller power PSUs is more expensive than a single 24kW one. You don't have as much redundant components. Also SiC permits smaller magnetics and cooling requirements.

 

[cupoftea]

Thanks, so your output is going to be eg 370V at 64A (24kW)....and your output rectifier is just diodes made of pairs of TO247 SiC diodes? And you will get 96.6% efficiency? And theres no synchronous rectification?
But dont use water cooling. I think you have made a big breakthrough if this converter isnt very very large. It sounds like your sepc needs water cooling.

As far as diff mode filtering is concerned going to your battery....i woudlnt think much significant diff mode filtration was necessary...so presumably your EMC filter was pretty much totally a common mode filter.

I think when you have 24kW running like that for many hours in ambient temps above zero degrees C, then Tesla and co will be inviting you for "coffee-with-biscuits"!