Totem Pole PFC is over-hyped?

[Faringdon]

Hi,
Ayk, Totem pole PFC  saves the losses in  the mains bridge rectifier, and the boost PFC diode.....it has its own losses, so its not a total winner.

So why not just use a normal Boost PFC, and make its output diode synchronous, and make the mains bridge diodes synchronous?....it would require far less effort than all the paraphinalia that goes with Totem Pole PFC circuits and controllers. Plus avoids the known common mode EMC problem that happens with Totem Pole PFC.

[Faringdon]

Hi,
344 people have read this as at 24 june 2023, and not one disagreed.
It is simply the case that we have all fallen for the hype....Totem Pole PFC is another "Y2K thing".
Fake news.

[temperance]

The reality is that  people are tired of answering your nonsensical "do you agree" topics with faulty spice simulations where you look with a magnifying glass into non existing glitches . But you don't seem to get the hints you've gotten from many people.

[beanflying]

Between your last clueless account and now this one you have made close to 1000 largely useless topics and nearly zero contribution to the forum outside of those.

Last time I darkened one of your threads I suggested you should get a clue. Clearly this hasn't happened in the last year since then and all the years before 

[Faringdon]

Thanks, the way to beat the so called Totem Pole PFC is just to do as in the top post.
The mains bridge is 50Hz and is easy to make synchronous.
For the Boost diode......just have a hi side supply which you use to turn on an NFET placed in parallel to the boost diode.
Refer the gate drive signal up with a high CMTI digital isolator.
Sense the load current, and when it goes so low that the inductor current goes discontinuous, , then disable the synch rect.

Also, when near the zero crossing, the boost inductor current may go discontinuous, so disable the boost diode's synch rect near the zero crossings.
Thats all thats needed to kick Totem Pole PFC into touch.

[temperance]

If you really think that what you state is true. Then why don't you come up with numbers and real working proto types made in your grandmothers basement to prove your point instead of fishing for answers by asking the "do you agree" questions.

[Faringdon]

Thanks, if someone wants to pay me to produce a standard Boost PFC , made synchronous, (synchronous mains bridge and boost diode) so its just as good as , and likely better than, than a Totem Pole PFC,  then i will gladly do it.....and give all rights to it to the payer.
(if you can have rights to something thats so simple and obvious)
It will be a lot simple than a Totem Pole PFC.....and wont suffer the known Totem Pole PFC Common mode EMC problem.

[temperance]

1. if if that's obvious, coming up with a proof will take you a few hours.
3. Why do people have to pay you to prove you right or wrong if you're the one asking questions?

Calculations please if things are that obvious. All other things can be classified as nonsense.

Totem pole PFC: 4 switches
Your PFC: 6 switches of which two are in series in the bridge. Good luck keeping the dissipation under that of a normal diode bridge in > 600W universal input 90...264Vac PFC without reverting to very expensive MOSFET's.

Thanks, if someone wants to pay me to produce a standard Boost PFC , made synchronous, (synchronous mains bridge and boost diode) so its just as good as , and likely better than, than a Totem Pole PFC,  then i will gladly do it.....and give all rights to it to the payer.

Just do it and find ways to turn ideas into cash. Those things are in high demand.

[Faringdon]

For Totem pole PFC you will need a very good software engineer, who also understands the finer detail of  high power SMPS.....and can implement software control loops for the current and voltage in the PFC......and every time you want to modify your power supply in some way, you'll have to have that software engineer still available to you.
Any junior Jo could do the Synchronous alternative PFC that i suggest.

Totem pole PFC
https://www.st.com/content/dam/is20/document/PE4-2_Erin_Wang_3-6kW_Totem-Pole_PFC_with_ICL_Solution_V02_EN.pdf


And High voltage SiC FETs with low enough RdsON are available now...so that would not be such an issue....and indeed, you are not going to have any switching loss in the 50Hz bridge sync FETs....and they are going to be generating far less common mode noise than the totem pole PFC FETs.....the front two FETs in totem pole PFC are known for very fast switching, even being GaN so they can switch real fast.

In a 3kW PFC at 90VAC input (worst case), each leg of the mains bridge is going to see 16.7Arms....so for a 20mR FET thats 5.5W....easily manageable....and even more so with a couple of parallel FETs.....noting that layout and heatsinking wont be crucial for these FETs from a noise point of view because they are essentially not switching other than at 100Hz......so little problem with those heatsinks becoming noise radiators.

And with Totem Pole........get ready for some very challenging visits to the EMC lab, both radiated and conducted...as their common mode noise problem is well noted....get some big common mode chokes on that board!

Heres a good FET for synchonising the 50Hz bridge
SIHG026N60EF-GE3
https://www.digikey.co.uk/en/products/detail/vishay-siliconix/SIHG026N60EF-GE3/14680918
$9 per 1000
or, as discussed, just use multiple cheaper fets in parallel.

Heres one at 40mR
https://www.digikey.co.uk/en/products/detail/vishay-siliconix/SIHW73N60E-GE3/4496328

The "Alternative synchronosied Boost PFC" that i suggest would turn out cheaper.
Ultimately, the Totem Pole PFC could possibly be slightly smaller, but it depends on how you manage EMC , and the need for more extensive common mode filtration will be there for the Totem Pole.

At the end of the day, most 3kW SMPS do not have to be super miniaturised anyway...so even if the "alternative to totem pole PFC"  did turn out slightly bigger...it would be an insignificant point.

As you know, with 3kW+ SMPS, that are offline and hard switched (both "totem" and "alternative" are hard switched).....there are significant issues with noise......and as discussed, the totem is the master of being noisy......Also, they may need to be paralleled...and the "alternative" type is more amenable to paralellisation than the "totem".
>>>>>>>
Below 400W and Totem pole PFC is deffo a waste of time

Up to 1kW and you might as well not bother with totem.....ive seen 800W offline 90-264 use just diode bridge on a 60x30x2mm alu plate heatsink, and even that was way away from any fan air...and those bridges never died  for >20 years.

So up to 1kW, you can, if you want, .... use an easy thing whereby you just make the bottom two diodes of the  mains bridge synchronous....realistically thats all thats needed.
Above about 1.5kW and its common to parallel two stages anyway, because a single 3kW pfc is going to be as noisy as heck.
Remember that the totem pole PFC is probably the king of all the hardest of hard switchers out there.

[temperance]

You're calculations are not incomplete:

-R-on increases with temperature and you've taken the typical rating. The real on resistance is 23mOhm 1.5 @ 125°C which is about 34mOhm. That's still less than the dynamic resistance of a typical bridge rectifier at currents above 20A. But the loss is significant @ 16.7A or 10W/MOSFET. Better MOSFET's quickly become extremely expensive. IXYS has a diode bridge with a dynamic resistance <3mOhm but the cost in €17 for two diodes in one package. This thing beats your MOSFET's and but the cost is probably the same as some better MOSFET's excluding their drive circuitry)

-You didn't calculate all switching losses. The synchronous boost rectifier Coss will be dissipated in the boost switch at turn on and the current spike will be a nasty one. This loss will be significant unless you find some way to temporarily story the energy else where and deliver it back to the input or out later on. Such things exist. But as usual, a free lunch doesn't exist and the gain/cost ratio is pretty bad.

-The bridge MOSFET will require measures to protect them from surges. The totem pole PFC has an inductor in series to dissipate some of the surge energy.


Software: if one is aiming at 100K units or more, software is not an issue.

Perhaps the two boost or bridge less boost in a more interesting option, eliminating two diodes. But the highest loss in any CCM PFC is still switching loss. Methods like an auxiliary switch to achieve ZVS turn on/off are not difficult to implement but the number of components required in a two boost/ bridge less PFC become unpractical and expensive. Interleaved DCM converters look promising until you look into the size of the required magnetics for a >500W PFC.

Gaining even the tiniest bit in efficiency is extremely difficult and the required effort is not for the faint hearth. (see the incredible amount of papers being published around ZVS/ZCS boost converters for solar inverters and EV chargers. That's why I said there is a marker for such things.)

[Faringdon]

The synchronous boost rectifier Coss will be dissipated in the boost switch at turn on and the current spike will be a nasty one.

Thanks, though a simple delay circuitry would ensure that  the boost diode's synch FET only turned on after its diode had  been brought into conduction....so thats ZVS pretty much.

-The bridge MOSFET will require measures to protect them from surges. The totem pole PFC has an inductor in series to dissipate some of the surge energy.

Thanks, the PFC output caps will provide a good deal of surge quenching........as they do with "standard" Boost PFCs.
Standard Boost PFC as you  know, most commonly has an inrush diode which shorts out its inducto to the inrush surge, and to any other surges.
Also, there are EMC filter inductors, as well as the MOVs at the front end.

....Yes the Totem does have the inductor upstream to add some more protection, but it will of course saturate when surge current becomes significantly high.
I dont think anyones touting the Totem on the basis of its surge resilience....though yes, i would give you that one to an extent. Though a standard PFC with equivalent inductance added in the way of EMC filter would be the same......and cheaper than Totem which, with all its extra circuitry, is more expensive in general.

I investigated a Totem design for an interview at a big disty near the M25 (you will have heard of them) where they were designing and building totems......the circuitry is way more extensive and expensive  than a standard boost PFC........They were of course doing this because their customers were not seeing the benefits of the Totem, so they were attempting to put together totem demo boards so that they could try and  show them to the cust....from the questions they were asking me at interview.....it was obvious they were having biggo problems with common mode noise.

Software: if one is aiming at 100K units or more, software is not an issue.

So it would be interesting to hear if you believe that the benefit_point for totem is when you can garantee orders of 100,000 per year or more?
There are not that many products that can be sure of sales of 1.5kW SMPS with those volumes....or anywhere near those volumes.

Thanks for the anlysis on FETs for the Mains bridge.....though as discussed, multiple cheap FETs in pllel would be they way round that........with cheap heatsink clips for easy assembly.....or even  multiple SMD D2PAK and thermal vias to heatsink......and poss metal bar across the top to give that bit more heat conduction away from the FETs.

[temperance]

Thanks, though a simple delay circuitry would ensure that  the boost diode's synch FET only turned on after its diode had  been brought into conduction....so thats ZVS pretty much.

I'm not talking about MOSFET turn on delays and things. That's pretty obvious. I'm talking about Coss of sync rectifier. This cap is being discharged every cycle and this power is being dissipated in the main switch. That's a substantial loss.

So it would be interesting to hear if you believe that the benefit_point for totem is when you can garantee orders of 100,000 per year or more?
There are not that many products that can be sure of sales of 1.5kW SMPS with those volumes....or anywhere near those volumes.

EV chargers...

Thanks for the anlysis on FETs for the Mains bridge.....though as discussed, multiple cheap FETs in pllel would be they way round that........with cheap heatsink clips for easy assembly.....or even  multiple SMD D2PAK and thermal vias to heatsink......and poss metal bar across the top to give that bit more heat conduction away from the FETs.

Not really. Two 100mOhm MOSFET's in parallel cost you  nearly the same as one 50mOhm MOSFET. Mounting cost increases and stray capacitance too.

[Faringdon]

I'm not talking about MOSFET turn on delays and things. That's pretty obvious. I'm talking about Coss of sync rectifier. This cap is being discharged every cycle and this power is being dissipated in the main switch. That's a substantial loss.

Thanks Yes, but the same such happens with the totem pole PFC.

Not really. Two 100mOhm MOSFET's in parallel cost you  nearly the same as one 50mOhm MOSFET. Mounting cost increases and stray capacitance too.

Thanks, as you know, Stray capacitance of the fets that make the mains bridge synchronous doesnt matter...they switch at 100Hz....and at the zero cross.

Not really. Two 100mOhm MOSFET's in parallel cost you  nearly the same as one 50mOhm MOSFET. Mounting cost increases and stray capacitance too.

Thanks, for 3kw PFC with syncd 50hz bridge, i would use four of  of these SMD fets on each diode...

IPL60R060CFD7AUMA1......
https://www.infineon.com/dgdl/Infineon-IPL60R060CFD7-DS-v02_00-EN.pdf?fileId=5546d46262b31d2e01633ed634654cdf

each SMD, thermal via'd to the single heatsink, which would be used for all the 16 of them...to make a single sync'd 50hz bridge heatsink assembly.
At 3kw and 90vac, they would worst case dissipate 1.74W each...but thats at 150degc junction.

With volumes in the 100,000's, i'd get them at well under a dollar each.

That would work out no more expensive that a totem pole at 90vac and 3kw.....as you know, totem uses the 4 fets.....two of  which have high switching and conduction loss...the other two have high conduction loss. OK , with my 50Hz sync'd thing i have two more fets and a diode (the diode wont conduct much).....but it'll work out cheaper overall than the totem.

I cannot find an example of either Totem or "syncd 50hz pfc" at 3kw, 90vac on the web though.
Its likely that 3kw at 90vac would be served by two 1.5kw pfc's in parallel...whether totem or "50hz sync'd".

I am going to search totem for the fets used in a 90vac, 1.5kw totem.
And of course, they (totem pole pfc's)  have far more extensive extra circuitry compared to the "50hz sync'd pfc" type

As seen below, the web examples of 1.5kw totem dont show it at 90vac and 1.5kw...only at 230vac nominal....
1.5kw totem pole pfc:-

https://gansystems.com/wp-content/uploads/2020/01/Optimal-Design-for-High-Frequency-GaN-Based-Totem-Pole-PFC.pdf

So, yes, been looking for some time for a ~3kW (or even ~1.5kW)  Totem pole PFC eval board on the web...or an app note on such...but cannot find it.......i believe that the FET situation for totem pole for 3kW PFC for 90VAC input would be very expensive. For 3kW @ 90VAC i believe those two front end HF switching FETs would be well toasted. They usually, ayk, make those GaN type........and they have high switching loss, as well as high conduction loss at 90vac in and 3kW.

[SlimEddie]

Had a chuckle reading this, while doing a bring up of 3.3kW TTPLPFC 

[Andy Chee]

make the mains bridge diodes synchronous?....

Isn’t that a totem PFC?

Otherwise I’m not picturing your proposed topology.

[temperance]

Had a chuckle reading this, while doing a bring up of 3.3kW TTPLPFC

What is your experience with the TTPL PFC in terms of efficiency and mains filters?

@ Faringdon

We are still waiting for proper calculations and a real world proto type demonstrating your sync boost PFC with an active bridge. There is a real market for those if what you wrote is true.

[coppercone2]

I think you need to invest in some power electronics lab and prototyping abilities to test some stuff because it seems like a big mess of theoretical spaghetti with this. hiding behind cost frowned upon in this forum, people here seem to figure out ways to do things that are hard

but, what stands out for me in this thread, as famous last words, is "software is not a issue." Hand wave a issue because a big number popped up and then follow that up with some figuring about the competition (and assuming their sane). A level V systems engineer wrote this?

some tips:
1) buy 10
2) make the circuit board repairable, modular, thick, whatever so when 1 explodes you don't go into cad crisis mode spamming the forum about simulators because some fragile ass prototype exploded unrepairably
3) attempt common sense based trouble shooting

[temperance]

but, what stands out for me in this thread, as famous last words, is "software is not a issue." Hand wave a issue because a big number popped up and then follow that up with some figuring about the competition (and assuming their sane). A level V systems engineer wrote this?

What exactly are you trying to say? Micro controller driven power supplies are a reality. Check manufacturer offerings from TI, Analog devices, Infineon,...

My point is was and will always be that improving efficiency is very difficult and every method has it's own set of problems, shortcomings and limitations. The statement made by the OP doesn't stand without a real working proto type.

3) attempt common sense based trouble shooting

Thanks for the excellent advice. I usually replace random stuff until things work.

[coppercone2]

and its not a problem for them because you managed all three companies and saw software implementation for a hardware power systems control go down smoothly?

its a crux if there ever was one. Treez is wanting to bleed less money. I can understand that, actually, and it makes sense that removing a software requirement would make things go smooth.

[temperance]

and its not a problem for them because you managed all three companies and saw software implementation for a hardware power systems control go down smoothly?

No I haven't and I certainly not stated it would be an easy task. On board chargers, server power supplies, gigantic LED walls with multi kW power supplies running parallel, 5...20 kW audio amplifiers for prof. use,... are the places where totem pole PFC's are employed. They all require a lot more functionality for which a micro controller is best placed. The shift to drive the complete power supply from a micro controller is nothing strange. TI has plug and play software available for a totem pole PFC which includes tools to perform a frequency response analysis without having to revert to soldering iron. I haven't used the TI tools but do have an infineon XMC dev kit to investigate those matters if time permits.


Here is what Faringdon stated:

For Totem pole PFC you will need a very good software engineer, who also understands the finer detail of  high power SMPS.....and can implement software control loops for the current and voltage in the PFC......and every time you want to modify your power supply in some way, you'll have to have that software engineer still available to you.

See above and:
https://www.ti.com/lit/ug/tidue54b/tidue54b.pdf?ts=1714263250483&ref_url=https%253A%252F%252Fwww.google.com%252F

its a crux if there ever was one. Treez is wanting to bleed less money. I can understand that, actually, and it makes sense that removing a software requirement would make things go smooth.

He doesn't need to invest in software and build a totem pole PFC. Dev boards for totem pole PFC's with efficiency plots can be found from TI, ST, Infineon,... He must build his proposed PFC and prove everyone wrong.

Meanwhile totem pole PFC controllers which don't require software became available.

[SlimEddie]

What is your experience with the TTPL PFC in terms of efficiency and mains filters?

So far this is my first time working with TTPL PFC. Now where do I begin...

First and foremost, TI's reference design code can be a mess, and the documentation is often skipping some steps.
From the get go I had issues with the code running from main processor, it would compile just fine and run on CLA (the co-processor on C2000 MCU family), but would fail on the main C28x core.
After some digging I found that they forgot to copy ISR functions into ram and the CPU just couldn't find the code where it expected it to be. Apparently TI discontinued the reference design that I was using and I think that the state of firmware was one of the reasons for it. Anyhow, after finishing my fight with the code, I successfully managed to run in in PFC mode whilst supplying it from DC supply instead of AC, I managed to verify that all measurements make sense and all the legs switch correctly.
Next issues came when I had to run TI's SFRA to get an open loop frequency response, since at that point I was still running from CLA instead of the main core that ended up being the bottleneck, which got resolved as soon as I was able to run somewhat stable fw on main core. Although the frequency response matched TI's reference in terms of shape, the gain was a tad lower than it should be. I shrugged it off and moved on to test with AC, where I am currently stuck, after fixing another bug in fw that had wrong defines used, I managed to get it switching with AC, but I am facing an issue with current loop instability, the shape of the curve has a dip which is not present in TI's reference design and overall gain is lower than it should be, so currently I am trying to figure out the root cause of this issue, which is difficult considering there is 3x switching frequency noise everywhere  you try to measure. Over all I expect to have a lot of issues with conducted and radiated emssions when I get to solving those issues in the future.

[temperance]

Thanks for your reply. So in short the TI solution doesn't work properly. I haven't had time to play around with the Infineon XMC solution. Maybe around summer or so.

Regarding the noise.

I have no idea about your level of experience with developing switching power supplies so I might suggest something you all ready know. If probing around with normal probes when almost no output power is delivered indicates to me that there is something wrong with how the power components are switching. If there is a problem, it is visible when probing the gate wave forms and inspecting the gate source miller plateau voltage at turn on. Any kind of wrinkles or sings of oscillations must be "cured" before you can continue because it will cause common noise all over the place.

[SlimEddie]

Thanks for your reply. So in short the TI solution doesn't work properly. I haven't had time to play around with the Infineon XMC solution. Maybe around summer or so.

Regarding the noise.

I have no idea about your level of experience with developing switching power supplies so I might suggest something you all ready know. If probing around with normal probes when almost no output power is delivered indicates to me that there is something wrong with how the power components are switching. If there is a problem, it is visible when probing the gate wave forms and inspecting the gate source miller plateau voltage at turn on. Any kind of wrinkles or sings of oscillations must be "cured" before you can continue because it will cause common noise all over the place.

Well this is one of a few solutions advertised as bidirectional, there are not many of those around and it fits our needs, so I am gonna have to fight with it till it works.

I have a bit experience with SMPS design, but most of those were <150W. The jump to 3.3kW is a tad scary and expensive in regards to equipment used.

On the bright side,I had some success fixing the shape of the current loop frequency response from this:

To this:

While it finally matches the shape of frequency response I was hoping for it is still 10dB lower in terms of magnitude compared to what should have been. Gonna need some work tweaking the values.

[Faringdon]

Sounds like you are fighting some good winning battles in getting it working.
You say you need bidi, and yes, for this reason the Totem pole PFC sounds good.

Otherwise, at 3.3kW , i wonder how much more efficient that would be when compared to a Dual interleaved PFC with UCC28070?
https://www.ti.com/product/UCC28070

...thats two PFC's each with  just 4.125A of output current.

Make the rectifier bridge synchronous and its even more efficient.
I doubt the totem pole is going to be much , if at all, more efficient than that.