LTC4020 battery charger charging current unstable

[hoi]

Hi everyone,

I am making a battery charger with LTC4020 to charge 12S li-lon battery but having problems for months and still could not fix it.

Input 24V
output 43V 2.5A charge current
battery voltage range 24V-43V (cell voltage 2V - 3.6V)

schematic setting
Maximum input current 10A
charging current 2.5A
battery float voltage 42.86V
Maximum output voltage 43.7V
Precharge voltage 28V

Problem
The charge current is lower than expected value, it should be maximum 2.5A but I could only get 0.5-1A charging current. And the waveform is strange.

I tried to tuned Vc and ITH value for weeks, it change a little bit with different value but could not solve the problem.

I tried to contact local FAE but did not get any useful information, tried to post the question in AnalogDevices forum but no one reply. Google searching LTC4020 and it seems some people here are using this chip so I think maybe I could get help from here. I would really appreciate if anyone could help.

The attachment is schematic, waveform of SW2 and charge current.

Thanks

 

[KlausST]

Hi,

my hints:
* PCB layout problems
* inductor selection (saturation)
* capacitor selection
* maybe wiring

So if you show us details to the items above we can validate them.

Klaus

 

[D.A.(Tony)Stewart]

One pix does not tell the whole story .

Maybe 6 to 10 more with photo of measurement method of all the feedback and control signals and different load conditions. This could be a very high Q series resonant problem with good parts getting crosstalk or shootthru from improper deadtime or .... lack of damping from DNI

open circuit resonance is easiest to compute.

 

[D.A.(Tony)Stewart]

You posted a haystack and we're looking for a needle that is unable to regulate your charge current even if demand and supply exist. That means it getting false sensing signals or generating false demand current from noise. Validate your probe and layout methods by demonstrating a flatline on all your ground connections with adequate SNR..

 

[Easy peasy]

What is your power source ?

 

[D.A.(Tony)Stewart]

we're assuming 24V 10A source is not disturbed by the voltage boosting CC source by something limiting the output to 430 W and nothing is obviously getting too hot.

 

[hoi]

Thanks so much the the replies!

Here is the details of the inductor.

74435582200 Würth Elektronik | Inductors, Coils, Chokes | DigiKey

Order today, ships today. 74435582200 – 22 µH Shielded Drum Core, Wirewound Inductor 15 A 7mOhm Nonstandard from Würth Elektronik. Pricing and Availability on millions of electronic components from Digi-Key Electronics.

www.digikey.jp

Here is the layout and waveform
I tried to follow the evaluation board as much as possible.

All the waveform was captured in no load. I would take more if it could be helpful. Thanks
Sorry I only have 2 channel scope.

I use a DC power supply maximum 32V 10A, and set the input voltage to 24V during the test all time.
And I use a electric load to simulate the battery, I tried to connect to the battery but the perform was the same.

The inductor is very hot, 4 MOSFET are not that hot but still hot especially the 2 low side MOSFET.

Here is the efficiency I measured.

24	0.163	20.29	0.092	0.4771676892	Precharge mode
24	0.164	21.63	0.093	51.10746951	Precharge mode
24	0.164	22.93	0.094	0.5476168699	Precharge mode
24	0.165	22.58	0.094	0.535989899	Precharge mode
24	0.191	27.19	0.117	0.6939856021	Precharge mode
24	0.193	27.66	0.119	0.7106088083	Precharge mode
24	0.193	28.09	0.116	0.703462867	Precharge mode
24	2.024	30.99	1.311	0.8363778409	Quick charge mode
24	2.004	31.99	1.259	0.8373962492	Quick charge mode
24	1.98	32.99	1.207	0.8379404461	Quick charge mode
24	1.96	33.99	1.16	0.8381887755	Quick charge mode
24	1.944	34.99	1.118	0.8384520748	Quick charge mode
24	1.926	35.98	1.078	0.8390974386	Quick charge mode
24	1.905	36.98	1	0.8088363955	Quick charge mode
24	1.885	37.98	0.999	0.8386830239	Quick charge mode
24	1.867	38.99	0.965	0.8397016158	Quick charge mode
24	1.851	39.98	0.933	0.8396663965	Quick charge mode
24	1.833	40.98	0.902	0.8402427714	Quick charge mode
24	1.815	41.98	0.871	0.8394072544	Quick charge mode
24	0.437	42.1	0.193	0.7747234935	Stop

 

[KlausST]

Hi,

Please show your PCB layout.

Please tell us what the columns in your table are.

You use a lot of 0-ohm resistors. On some places they introduce impedance that will degrade performance.
What's the use of D11? (Here I expect problems)
Why only one diode (D13) across the bridge?
What's CHARGER_VCC voltage?

Klaus

--- Updated Oct 2, 2021 ---

Added:
Although no bad idea, but doesn't the cut in GND plane around the LTC increases loop length? Maybe could be optimized.

 

[FvM]

You are showing two different "SW2 no load" waveforms without explaining the test conditions.

According to SW2 high voltage, no load condition involves > 40 V battery voltage. SW1 waveform should be mostly on with 100 ns low pulse under this condition (d.c. > 99%), but it's actually switching with< 20% d.c. I notice that feedback compensation at pin VC is far off from suggested dimensioning. This may be a reason for unstable voltage feedback loop.

If I understand right, your major complains are about insufficient output current, but you don't show waveforms for loaded operation.

In post #1 waveform, how are you measuring output current? Differential voltage probe across R83? Additional ground referenced shunt? Current probe? Unlikely that the actual battery current has +/- 5A spikes.

 

[hoi]

Thanks for the replies!

According to datasheet,
A low current 6V Zener (0.1A) in parallel with the BST2-SW2 capacitor will also effectively shunt any errant charge and prevent excessive voltage build-up.
I tried to remove this and nothing change.

Since the circuit is step up most of the time, so I only place 1 diode across high side MOSFET (Same as the datasheet's typical application in last page). I tried to remove it but nothing change.
Did not tried to place diode across low side MOSFET. Will try it later. Thanks for the hint.

CHARGER_VCC is 5V

I tried to isolate SGND from other GND, don't know if it is a good idea.

Sorry for the confusing table. Here is the update

Vin 24	Iin 0.163	VBAT 20.29	IBAT 0.092	Efficiency 0.4771676892	Precharge mode
24	0.164	21.63	0.093	0.5110746951	Precharge mode
24	0.164	22.93	0.094	0.5476168699	Precharge mode
24	0.165	22.58	0.094	0.535989899	Precharge mode
24	0.191	27.19	0.117	0.6939856021	Precharge mode
24	0.193	27.66	0.119	0.7106088083	Precharge mode
24	0.193	28.09	0.116	0.703462867	Precharge mode
24	2.024	30.99	1.311	0.8363778409	Quick charge mode
24	2.004	31.99	1.259	0.8373962492	Quick charge mode
24	1.98	32.99	1.207	0.8379404461	Quick charge mode
24	1.96	33.99	1.16	0.8381887755	Quick charge mode
24	1.944	34.99	1.118	0.8384520748	Quick charge mode
24	1.926	35.98	1.078	0.8390974386	Quick charge mode
24	1.905	36.98	1	0.8088363955	Quick charge mode
24	1.885	37.98	0.999	0.8386830239	Quick charge mode
24	1.867	38.99	0.965	0.8397016158	Quick charge mode
24	1.851	39.98	0.933	0.8396663965	Quick charge mode
24	1.833	40.98	0.902	0.8402427714	Quick charge mode
24	1.815	41.98	0.871	0.8394072544	Quick charge mode
24	0.437	42.1	0.193	0.7747234935	Stop

Sorry for the waveform. I made one mistake. one is actually SW2 and TG2 with load (upload here). And I will try to take waveform with load later. Thanks.

I tried many different value in Vc and ITH including the one shown in datasheet and LTC4020 evaluation board. It change a little bit but still could not solve the problem.

The output current was measured with current probe. It was a chinese made current probe with the scope set to 100x. Actually I do not trust that too much, I think the probe itself already pickup many noise.

 

[D.A.(Tony)Stewart]

I wonder what the peak current is in L . What does V (R90) say ? 50 kHz 15% d.f. ? 3 us ON Time? ( =dt ), V= ? L2 = 22uH dI= V dt / L = ~ V / 7 [A] should not cause excess heat (?) unless in continuous mode

You only producing about 40W of charging yet have 8W of heat loss. Where and why is the bigger question , not why is it limiting prematurely. And yet you have low RdsOn and DCR and ESR components. Something must be out of sync from false sensing or some other error.

For a LTC4020 battery charger, I would expect a CC control then CV control with a cutoff mechanism.

1. Why are the DNI parts not used for damping ?
2. What are all your design specs ?
3. Which ones are not met for each component ?
4. Which parts differ from simulation and why?

--- Updated Oct 3, 2021 ---

You should be getting 98% effic. not 84%. Please validate each stage from the input

--- Updated Oct 3, 2021 ---

Design Note 531: 55V High Efficiency Buck-Boost Power Manager and Multi-Chemistry Battery Charger | Analog Devices

IntroductionToday, battery chargers are expected to easily support a variety of battery chemistries and accept a range of voltage inputs, including wide-ranging solar panels. It is increasingly common for input voltage ranges to span above and below the output battery voltage, requiring both...

www.analog.com

--- Updated Oct 3, 2021 ---

If you are not getting a nice clean triangular waveform for L2 current, then calibrate your probes up to 20 MHz for a differential Ch1-2 flatline using spring probes. Twist the probe coax too, if you have too. Eliminate these parasitic measurement errors. Then work backwards to root cause of timing, transient and resonant errors.

 

[D.A.(Tony)Stewart]

 

[hoi]

Thanks so much for the hints!

I did not measure R90 since I think I could not see the voltage difference clear of R90. I will try to measure it using the method you provide.

1. I tried to place RC snubber for MOSFET and one more capacitor in Vc ang ITH pin, did not help. (tried many combination)
2. design specs 24V input 2.5A charge current, battery 24V to 43V.
3. The charge current did not meet the spec. Refer to the schematic, it should be 2.5A charge current, and the charge could be adjust by RNG/SS pin but it cannot in my case. And as you said, all MOSFET and inductor and low ESR so there should be not much heat.

4. I could not get the simulation to work at all. Even for the example file provided by LTSPICE or the evaluation board DC2134A file I download from AnalogDevices forum, the switching is strange and would stop eventually, and no charge current.

 

[D.A.(Tony)Stewart]

There is no load. I added 10 ohms and here is the simulation.

 

[hoi]

Thanks for the simulation

Are you using the dc2044a_Model_0 simulation file?
If just place a resistor in VOUT, LTC4020 actually stop switching if you check SW2 and TG2, and all the current is actually from the voltage source V2, which is different from what the datasheet said, it will only draw current from battery only if the current demand is more than LTC4020 can supply.

Also if remove the voltage source V2, LTC4020 is switching quite well. The regulated voltage is 36V, then drop slowly. Actually it is similar to my PCB (although I did not check the waveform in this situation)

Here is my simulation file

 

[FvM]

What is I(R21)?

--- Updated Oct 3, 2021 ---

O.k., load current, just averaged output voltage.

In fact, I have doubts that the switcher can work in simulation. The observed effects are probably more "real world" problems.

 

[D.A.(Tony)Stewart]

Your schematic looks quite different to this one. Notice the logical flow and details with small current loops.

--- Updated Oct 3, 2021 ---

What is I(R21)?

--- Updated Oct 3, 2021 ---

O.k., load current, just averaged output voltage.

In fact, I have doubts that the switcher can work in simulation. The observed effects are probably more "real world" problems.

Switcher works fine in simulation for me. after 1 ms for 2 ms or longer if I make drastic changes to FB

I added R21 to put a load onto the demo board sim. Here are traces of currents for {input , L, output 10 ohms}

I don't have any answer or solutions, but my method of debug would be to verify every assumption I made to deviate from the demo board BOM and layout.

 

[hoi]

It should be similar to DC2134A except I adjust some resistor value for my application

https://www.analog.com/media/en/technical-documentation/user-guides/DC2134AF.PDF

What did you mean logical flow and details with small current loops? Could you tell me more about this?
Thanks

 

[D.A.(Tony)Stewart]

The problem with very low DCR is poor damping and resonance with parasitic capacitance across L , and all switching FETs connected to L. If you knew from Saturn PCB.exe or other tools how to model ESL and parasitic C which can easily be 3 to 5 pF / cm for large L pads and Drain pads to ground plane underneath , it greatly adds to losses with reactive high current resonance but is supposed to reduce EMI. But does it? with such resonance. No. If you could get your simulator to work and add parasitic L and C and compare DCR and ESR of your parts vs the ones the designer chose, you might understand the difference. Although their 8 layer layout is better than yours that also has too much parasitic capacitance near L. Try to reduce about 3 to 5 pF of parasitic C around your next layout on the switching elements . So too high a Q or low RdsOn can also be a problem with the opposite in the x to xx MHz range.

here I added 2 parasitic L's and 5 parasitic C's and you can see the effect on current. Also the lower RdsOn has much larger Coss, so this too is a tradeoff.

 

[D.A.(Tony)Stewart]

What is I(R21)?

--- Updated Oct 3, 2021 ---

O.k., load current, just averaged output voltage.

In fact, I have doubts that the switcher can work in simulation. The observed effects are probably more "real world" problems.

Switcher works fine in simulation for me. after 1 ms for 2 ms or longer if I make drastic changes to FB. The demo sim was with no load, so I added it to measure sense current R I(R90) equivalent to View attachment 172205

--- Updated Oct 4, 2021 ---

The most critical signals are the Full bridge drain Volltages across L2 due to parasitic capacitance. Even 5 pF is too much. The ground plane under the pads will do that as I demonstrate in this simulation with 3 pF 3 mohm parasitic caps across each FET but with C across L removed then added below it. Notice the losses increase on input current.

You design is far different from the demo board in choice of L+DCR,C+ESR, f and layout parasitics. That makes all the difference with the extra pulses in the bridge and higher current.

Parasitic C's also make my CPU run 100% and takes 10x longer than 30 seconds normal for this design.
I have 8x i7 + 16GB