UCC28070: UCC28070

Hello Stephen,

Thank you for your application using the UCC28070 PFC controller. 

I agree that your efficiency results are disappointing, and I would expect something in the neighborhood of 95% +/- a little (mostly +). 
I suggest to set up a fixed DC load and conduct a loss audit to try to pin-point which components are dissipating more than expected. 
 
To start, I suggest to investigate the temperature rise of the boost inductors.  I think maybe their core-loss may be higher than expected. 
Micrometal's powdered iron material is good material for many applications, but the high flux swings of the PFC currents may result in high core loss. 

I've noticed that audio companies prefer toroidal cores to maintain low external magnetic fields from your magnetic devices, and I assume that you also prefer that.
For PFC inductors, I've had good results using "sendust" type material (Magnetics, Inc Kool-mu, for example), and newer materials have even lower losses.
Powdered metal can have "soft-saturation" characteristics which reduces inductance at higher currents.  This allows for a small inductor size, but may allow significant DCM (discontinuous conduction mode) currents in the inductor.  A small amount of DCM around the AC-line zero-crossings has a small effect on THDi, but significant DCM can really degrade THDi and drop the PF. 

Most the low PF at light loads comes from reactive current in your line filter X-caps, though.  

Another suggestion to investigate is the parallel IGBTs.  I don't have any real experience with these devices, and I'm surprised that they may be driven at 100kHz.  
Assuming that you have used them before, I defer to your experience.  My concern would be current-sharing during turn-on and turn-off edges.  MOSFETs have a positive temp-co of Rds(on) so they split the current fairly evenly all the time.  I don't know how IGBT's behave in this regard, and if they have a negative temp-co Vce, then perhaps the switching losses can be higher than expected. 

If you wish to post your schematic (it can be only the PFC part, doesn't have to be the entire PSU), you can embed an image or a file into your post by clicking on the "Insert" menu tab at the bottom of the post window, and then clicking the "image/video/file" selection.  In the dialog box that pops up, click on the "Upload" link to browse for your file.
Although your schematic can be in the form of a .jpg or screen-capture image, I prefer a PDF file, if you can generate it, because it allows me to print if necessary and mark it up with notes.  

Since your efficiency results stay relatively constant despite a ~3X increase in load, it suggests that rms conductive losses are not particularly significant (at these load levels) and that switching losses (including core loss) may be the main culprit(s).

Regards,
Ulrich 

Dear Ulrich,

Thank you for your prompt response.

I shall convert my schematic to PDF this evening and send it to you. 

I shall contact Micrometals this morning and see what they can offer. I am using core size T250-18 at 250uH with a single layer winding structure.

I have studied many docs concerning the value of the PFC inductor versus frequency and current and it seems that everybody has a different calculation. My logic based on an overlay of what I have been able to find was to choose this 250uH to reduce ripple. Maybe I am in the wrong ballpark or even zip code in my analysis!

Can the value of the inductors influence efficiency?

In your Power Designer the value of the inductors are 16uH and order of magnitude smaller than what I am using. Your input here would be most valuable. I use 375v as output and 18A as output current with a 100KHz

I have some PDFs of designs from Ti of 1.2Kw, 3Kw and 5Kw. 

They are attached for your reference.

I have changed Cin from 0.1mfd to 2mfd with zero change in efficiency.

I have changed the CAOA and CAOB parts for 50KHz operation with no change in efficiency. (My UCC28070 drive board is on a plug in module and I have modules stuffed for 50 and 100KHz so easy to change and measure!

I did mention that I see what I believe is an error in your Power Designer where the capacitors on the CSA and CSB pins are shown as 220nf = 0.22mfd which is much larger than the 220pF which I see in the UCC28070 PDF files.

I have used IGBTs for many years with good results. The HGTG30N60A4D are rated to 200KHz and I have tried other high speed IGBTs with the same efficiency results.

When I have done the efficiency tests (especially now in the cold weather) the ambient is about 14C. I run the tests and the core temp does not even rise close to 10deg C so I am thinking that it is not core temperature.

This does not work as there is no "upload" link:

If you wish to post your schematic (it can be only the PFC part, doesn't have to be the entire PSU), you can embed an image or a file into your post by clicking on the "Insert" menu tab at the bottom of the post window, and then clicking the "image/video/file" selection.  In the dialog box that pops up, click on the "Upload" link to browse for your file.

Hello Stephen, 

I agree with you, there seems to be some problem with the Webench Power Designer tool when using UCC28070 controller. 
Not only is 220nF much to high for a current sense filter, but the boost inductance also comes out too low. 
I'll notify the Webench team to investigate, but i can't predict when it can be fixed. 

Please use 220pF for the CSx input filtering. 
Your boost inductor value of 250uH is a little conservative but reasonable. 

For attaching files, I'm not sure what your interface looks like.  It may be different than what I see in my E2E screens here. 
This is what I see: 
 


 

Click on the "Upload" button and type in the filename or browse and select from your file folder.

As for low efficiency, if everything is running cool, then maybe the measurement of calculation is erroneous.
From your example above, 1450W output at 88.9% eff = 181W dissipation.  This is close to expected loss at full power. 

This 181W loss is spread amongst several power components: mainly diode bridge, inductors, IGBTs, output diodes.  
Basically 7 major components, so 181/7 = ~25W each (very roughly speaking).  This is non-negligible heat to dissipate, so something must be getting warm.  If nothing is inordinately warm, then the measurement must be suspect.

Are you using a dedicated power meter, or separately measuring input Irms and Vrms and multiplying together for input power?    
Lower power factor at light loads makes the VA product higher than the actual Pin, and can make the efficiency look worse than it is. 
Please check your power component temperatures and also review your power measurement method. 

Regards,
Ulrich 

I sent you my schematics last night in reply to the email you sent to me. I have attached them here. I have ordered both some Kool Mu and Micrometals Sendust cores to try.

I am using a Valhalla 2110A PFC/Current/Voltage/Wattage meter so I can choose at any time during measurement what I want to see on the display. I have cross checked the accuracy of the Volt/Amp/Watt readings against other meters and I have loaded the mains supply with a pure resistive load (I have 3 Kw load resistors) and they all check out.2068.UCC28070.pdf

Hello Stephen, 

I do not recall emailing you on this matter. I have searched my Sent box, Inbox, Deleted, and internal email files and can find no record of sending an email to you.  And I haven't received any email reply from you.  All my correspondence so far has been through this E2E forum.  
If the email address supposedly from me ends in anything other than "@ti.com" then it is a fraudulent email.  And I don't think that anyone else at TI would email you claiming to be me.  I suspect that someone reading this forum has taken your email address (from your first post) and is posing as me to obtain your design information.  You should block that address and never answer it again. 

Getting back to the original problem: I'd like to verify that the matter of concern is one of poor efficiency in your PFC converter.  
Please correct me if I'm wrong. 

I'm glad that you are using an AC power meter to measure input power.  For output power, I assume that you are measuring DC output voltage and DC output current of the PFC converter only, not the output of a down-stream DC/DC converter attached and not including cable-drop to a load.  From your first post, 534W, 850W, and 1450W represent 10.7%, 17%, and 29% of full load respectively. 

I have data from a 1200W evaluation board showing 95.2%,  96.2%, and 96.7% efficiencies for 10%, 20% and 25% loading with 200Vac input (I didn't take  208Vac data).  The design used normal parts and did not use extraordinary means to raise efficiency, although it did not have an EMI filter.  But I would not expect a filter to lose an additional 7% especially at light loads.

If your calculation of (Vdc*Idc)/Pac = ~89% for 10-29% loading, then I must conclude that either some power processing components have excessive losses or there are additional circuits that are dissipating power that are not directly part of the PFC conversion.  Do you have high-power cooling fans running, for example?  Or, a bias supply that is powering circuits other than the PFC control and gate-drive? 
If 1450W output measures ~180W loss, and we expect maybe 94% efficiency which is ~93W loss, then almost 90W is not accounted for.

The schematic you provided in your latest post shows only the controller daughter-board of the PFC, none of the power path.  
I don't want you to reveal your entire design in this public forum, but I do want to ensure that only the pertinent PFC circuits are being measured for efficiency and unrelated circuits are not being included.  If that is confirmed, then the excess losses have to be in one or more of the PFC power-path components.

Regards,
Ulrich