12v 80A switch noise and snubber cap

[tango17]

I recently switched to Lithium batteries in the RV.  They can accept charge current at up to 400A, but my primary charger is limited to 125A.  Because I charge with a a noisy generator, I need to minimize generator run time by increasing the charge rate.

To achieve that, I added an 80A switch mode charger.  It worked great for a week, but suddenly the battery monitor system (BMS) for the lithium began to spit errors. It was the 80A charger.  When I removed the charger, the errors stopped.  The errors were from the temperature monitors. I was certain the temperature was not excessive as the error code claimed.

However, the 80A charger was still producing it's rated current and voltage.  I suspected it was producing noise that was interfering with the BMS temperature monitoring network and sensors.

When I opened the charger I found that the snubber capacitor was split in two.  When I got home from vacation, I replaced the snubber cap.  The cap was definitely bad (open).

I've attached a schematic from the patent for the device (PD9280).  The snubber is C24.

My problem is that I looked at the output before and after the repair and it looks unchanged.  The switching frequency is 4.5 kHz.  Every 200 to 300 ms there is a ringing spike of about 2 volts peak to peak.  It rings down and decays in about 2 ms with about 10 cycles of ring (I could put a scope pic up if it helps.  This is with no load. 

I haven't tried reconnecting it, and I don't have any easy way of testing it with a load until I reinstall it.

I'm just not familiar enough with the noise characteristics of switching supplies to be sure, but 2 volts on a nominal 12v system seems excessive.  I've also attached a schematic of the output with its filtering.  The filter caps and choke seem OK.

I'd appreciate comments on whether the bad snubber might have caused other damage.  Is the noise level reasonable in light of the filter circuit shown in the attachment?  How can I reduce the switching noise from the 2 volts P-P to a few hundred millivolts.  I don't need perfection, but I'd really like better than I'm seeing now.  Thanks.

The patent link is:
https://patents.google.com/patent/US7782002B2/en?oq=7782002

The patent describes circuit operation and has most of the component values listed at the very end.

[T3sl4co1l]

Any pictures of the unit, internals?

Also, the snubber resistor is still nominal value?  I'd think that's more likely to open up.  Good reason to use a low-loss capacitor (polypropylene film) for replacement, in any case.

Curious what core type L2 has.

More C11-C18 will help with filtering, but if there's a lot of common mode noise, really all you can do is add a common mode choke, and another round of capacitors.  The choke will be massive at these currents; a somewhat expensive nanocrystalline core material will help save cable, and maybe cost in the end.  Make sure the capacitors are returned to ground as well as earth.  The problem in this case is voltage generated between input and output grounds, so filtering must be done with respect to that path.

Don't leave out input-side filtering, either.  Same thing applies, more parallel capacitors, series choke if necessary.  (Should only need one common mode choke between both sides, preferably on the higher voltage, lower current side, since that saves cable ampacity.)  Note that series (differential) filter chokes need to be made of low permeability stuff, not common-mode-choke stuff.  Powder cores and gapped ferrites are typical (including simple rod cores).

Tim

[tango17]

The snubber resistor R37 is 100R  10W and I measured it as 99.6 ohms.  I was surprised that the resistor was so large and the cap so small.  It looked like they'd left space on the board for a larger cap, but when I looked for a replacement 500V cap they were all similarly sized.  I may take 4 of them and arrange them as two in series and those papers in parallel if the single replacement fails.

I've attached some more photos.

[tango17]

The white block below the current sensor 109813 is the snubber resistor R37.  You can just see the split broken blue snubber cap to its left (this photo was taken before the repair.

Is it worth checking the input filtering components?

[T3sl4co1l]

Yeah, not a bad idea.  Or a 1kV+ C0G type ceramic should also be fine.

Also, that patent is preposterously insipid; it essentially claims what was already standard (if uncommon) in PC supplies for a decade at that time, and what remains standard to this day.  One boggles that they bought an expiration extension!  It's also oddly written: basically the schematic and BOM (but not layout) of their actual product -- who even does that?

I'm guessing they've never asserted it in court...

Anyway, C11 is suspect.  It's embarrassingly small.  I'd put... at least four more in parallel?  Same voltage rating, at least 1000uF each, rated for high ripple current / low ESR.  Should be easy enough to do, there's a lot of free space on the board.  Can even drill holes and mount them into the board all proper-like!

Input filter doesn't look too terrible; it's not as well filtered as it could be (they left an extra footprint there), but presumably, they tested it to FCC Part 15 this way, so shouldn't be too awful.  I would then still suspect differential output noise (ripple), which goes with the one puny capacitor handling that.

Tim

[David Hess]

A small output capacitor C11 will also get trashed more quickly by excessive ripple current.  I suspect they made it only large enough to handle the ripple current requirements assuming a low impedance load from the battery which is reasonable.  So maybe C11 has failed without visibly showing it.

I suspect snubber capacitor C24 failing was a fluke.  Or maybe it was not properly derated.  I would be careful about replacing it with a capacitor type that that tends to fail short.

[tango17]

I'm glad you mentioned C11.  I thought it was small and replaced it with one 5 times larger.  It helped a bit.  Reduced the ringing 10% or so. 


Remember that all my scope measurements were taken at no load.  But in use, it will almost always be near Max rated 80A.  Am I correct in expecting the problem to get worse at high current levels?


Would you just reinstall it and see what happens (put a scope on it) or would you suggest trying to add more filtering first? Perhaps I should scope the primary charger on the battery and see what kind of noise it has while charging at 80 to 100A.  That would give me a baseline comparison so I could see what a similar charger is doing.

[David Hess]

Remember that all my scope measurements were taken at no load.  But in use, it will almost always be near Max rated 80A.  Am I correct in expecting the problem to get worse at high current levels?

I do not think it matters because the design relies on the low impedance of the battery bank for proper operation.  Noise without a load is irrelevant.

Would you just reinstall it and see what happens (put a scope on it) or would you suggest trying to add more filtering first? Perhaps I should scope the primary charger on the battery and see what kind of noise it has while charging at 80 to 100A.  That would give me a baseline comparison so I could see what a similar charger is doing.

Unless it involves too much inconvenience. I would make the noise measurement under actual operation conditions before messing with the output filtering just to have something for comparison.

[tango17]

 The 125A charger (Magnum MS2812) produces 4V peak to peak ringdown when charging at 90A.  It's running at 100 kHz as compared to the PD9280 that runs at closer to 5kHz I didn't realize how noisy the chargers are.

I think I'll reconnect it and see what happens.  Perhaps the problem was fixed when I replaced the snubber.  The PD9280 worked fine for a week or so, and I didn't have a good chance to look at the output when it was broken.

[T3sl4co1l]

I do not think it matters because the design relies on the low impedance of the battery bank for proper operation.  Noise without a load is irrelevant.

Assuming short lead length, of course.  There's just about no RFI filtering on there, making that a concern already, but the EMI from ripple will be significant I think into a modest length of cable (more than a few feet), and that could very well be upsetting the BMS.

Ideally, the BMS is immune to such sources as well, but the charger still shouldn't be making noise.

Tim

[David Hess]

I do not think it matters because the design relies on the low impedance of the battery bank for proper operation.  Noise without a load is irrelevant.

Assuming short lead length, of course.  There's just about no RFI filtering on there, making that a concern already, but the EMI from ripple will be significant I think into a modest length of cable (more than a few feet), and that could very well be upsetting the BMS.

Ideally, the BMS is immune to such sources as well, but the charger still shouldn't be making noise.

I agree.  As you pointed out however, filtering at that current level is difficult which I assume is why they hardly bothered.

I might try using one or more of those big "DC Link" film capacitors with the spade terminals for low inductance to filter the outputs at the chassis.

[tango17]

 I added some more capacitance at C11 and repaired the snubber cap, then reinstalled the charger.  It's still causing the BMS to throw temp sensor errors within seconds of turning on.

I don't have much info about the temp sensor system.  It's supposed to be a series connected optical fiber connection between the multiple sensors.

The charger ran correctly without producing BMS temp sensor errors when I first installed it a few months ago, so I'm still assuming the problem is in the charger, but I can't figure out what the problem is now.  The voltage is correct and it is producing 80A.  There is still switching noise, but no worse than the primary charger 125A, which charges just fine.  The main difference is the switching frequency 100kHz for the main vs. 5-6kHz for the PD9280.

Ultimately, I'd like even more charge rate, so one option is to buy a second PD9280.  That would let me compared outputs and test voltages to see how they differ.  Of course, I don't want a second charger that won't work with my system, but I have circuit diagrams for the PD9280, so it's much easier to work on, and it needs to be modified slightly to charge the lithium batteries.

[tango17]

I got a chance to put my scope on another PD9280.  It's newer, but appears to be identical to mine.  It is running at 500 kHz and has about 400mV P-P switching noise.


That compares to mine that seems to be running at 5-6 kHz with 2 V P-P noise.  I haven't returned home yet to check mine, but I assume I should look at the RC timing for the current controller switching chip?  Any thoughts or comments?

[tango17]

The patent describes a 0.12 uF cap and a 1.82K resistor. That works out to 4.6 kHz, close to what I measured in mine (~200 uSec between pulses.)  His newer one showed pulses every 2 uSec.  Maybe they have redesigned it?

[tango17]

The datasheet for the uc3846 controller chip shows a typical application with 0.5K ohms and 0.005 uF capacitor. That's 400kHz. However the chart they provide ranges from 100Hz to 900kHz, with 5kHz near the middle.  So both 5kHz (mine) and 500kHz (his) are within reasonable operating frequency limits of the chip.

[T3sl4co1l]

Hah, some big changes in the design indeed, it would seem.

I wouldn't suggest trying to convert one into the other.  I would expect the PCB layout may have changed, and if nothing else the transformer, capacitors and snubber are likely different.

Note that output ripple is doubled, because of the full wave rectifier; it's actually running at 200kHz and 2.5kHz, I would suppose.

Tim

[tango17]

It's very frustrating.  I'd hoped to be able to at least learn whether my unit is working right ( in which case I'd need to determine how to improve performance) or is busted (in which case I need to fix it).

Mine is producing the right voltages and currents, but makes the BMS unhappy.  However, it worked fine for a few weeks, before beginning to make the BMS unhappy.  That seems to indicate it's broken, but I can't be sure.  I'd expected by testing a friend's PD9280 on my installation to learn which it was: broken or not.  Instead, I'm still uncertain.  I'd like to open up my friend's PD9280, but it's a bit much to ask to tear his open.

[NiHaoMike]

Have you tried adding some inductance in series with the output?

[tango17]

 No.  How do you suggest I add inductance?  I don't have any 80A inductors.  Do I buy a core and wind my own?

[Circlotron]

No.  How do you suggest I add inductance?  I don't have any 80A inductors.  Do I buy a core and wind my own?

For a common mode choke, try this.
https://www.richtek.com/Design%20Support/Technical%20Document/AN008?sc_lang=en

[NiHaoMike]

Literally just take some wire and coil it up, ideally with something to act as a core. Keep in mind that really low loss is not desirable for a noise filter - you want some loss to damp out ringing. Something like a bolt might make a good core for your use.

[tango17]

Thanks for the suggestions and link about CMC.  I'm on vacation now, but brought along the power supply and an assortment of components.  I'll let you know how it goes.

FYI, my friend's PD9280 turned out to be a totally different design.  The heat sink and case were about 2" shorter than mine.  He was willing to let me open it up, but I decided it wouldn't help me much.

[tango17]

I've been on the road for the last two weeks.  I brought the PD9280 with components needed to increase capacitance and inductance for filtering.  Before I attempted that, however, I decided to modify the current limit circuit connected to the UC3846 current mode controller chip.  I reduced the limit from 80A to 75A.  I also adjusted the current foldback so it kicked in just a bit earlier. I had previously modified it to remove the "charge wizard" microcontroller circuit that was designed for adjusting output voltage when charging lead acid batteries.  I wanted a straight 14.6V output, current limited with foldback when shorted.

I connected it up and my problems with interference were gone!  I don't have my scope with me, so I can't comment on the changes in output. I'll see what it looks like when I get back home in a few months.

Thanks for the help.