LM5176: After replacing short circuit input capacitor (1 of 15), circuit no longer starts (issue caused by non-resistive load)

Sometimes the buck-boost does start and can the load and input voltages can be changed within range. Other times it stops when adjusting voltages. Adjusting load seems without issues.

Most of the times it just doesn't start to output a potential at Vout.

And now the short circuit situation at Vin is here again.

For Vin, I have tried: 34.01, 35.01, 36.01, 37.01, 38.01, 39.01, 40.01, 41.01, 42.01, 43.01, 44.01, 45.01, 46.01, 47.01, 48.01, 49.01, 50.01, 51.01, 52.01, 53.01, 54.01 and 55.01 V from a Picotest P9611A. P9611A specs say: < 0.35 mVRMS and < 2 mV p-p.

For SW1, please explain a rookie more detailed what you need. When SW1 is the leg of the inductor (L1) near M2, there I sometimes see 0.00000V @100kHz. I do see a voltage at the gate of M2: Udc=2.2V and Upp=3.67V. At SW1 on other times I do see: Udc=0.06V, Uac=0.005V, Upp=0.43V;

While appling power to Vin, there is a short potential visible at the gate of M1. Afterwards the M1 gate voltage stays 0.00000V

Bitscope 5 sees this at SW1 (GND=Vin minus pole) @ 50us/Div versus 100mV/Div [Trigger 39.984kHz]:

A little while later Bitscope shows:

I hope that I understood your request for more information correctly.

I wish I haven't told anything about pedalling. I don't see how voltage spikes will arrive at Vin: at Vin will be the battery. A more likely candidate for voltage spikes from pedalling is Vout, because that is where the motor is connected to.

The issue now is that the LM5176 based circuit won't start to output voltage at Vout. I can't find any unexpected short circuits at Vin, neither are the positive and negative poles of Vin being shorted.

When may I expect a follow up here?

Without knowing your reactive load property, it is hard to predict what is really needed. Actually the low ESR cap does not help much here.  What you need is to damp out the possible resonance by introducing some power dissipation element. Can you measure the loop property when pedaling?  Need to see where the resonant frequency is in order to properly damp the reactive load effect.

You should ask the vendor who produces the motor/pedal module about its input impedance curve, or measure it yourself with an impedance analyzer. Consult the impedance analyzer user guide for how to measure the impedance. The input impedance of the module appears to a load impedance on the LM5176 output rail.

Impedance analyzers have DC bias and signal voltage settings. Red Pitaya signal is only 0.4V having 0 bias, while the more expensive ones have ranges from 10 mV - 2V DC drive/signal and (external) BIAS potentials of up to 200 volts. Should the motor+controller be tested at a specific voltage in this case? Or is even a DC bias neceassary to lift the voltage to 44VDC?

And should the signal even be a DC signal, or should the signal be an AC signal?

Note: the "pedal assist motor" is an integrated (dc to sine wave) controller and motor in one housing.

It seems that a Bode Gain plot is desired here...

www.youtube.com/watch

An impedance plot of the motor with integrated controller is not going to fix the LM5176 that no longer starts.

I have replaced the shorted cer.cap., M1, and U1 (the LM5176). There is still no starting LM5176 circuit with a (resistive) E-load.

What can be wrong?

@Youhao Xi: I don't see much change in the 1...100kHz (step 1kHz) sinus signal response: ± 1 dBm using a Velleman HPG1 as signal generator and read back device. What next?

You may be right about the capacitors, but it is not the ESR mismatching but the lack of resonance damping. If this is the case, you may have larger line inductance in your input line on the bike than on your bench, and the line inductance can resonate with the input capacitors. Adding the elco helps damp the resonance and prevent over voltage. From theoretical point of view, the issue is usually referred to as the Middlebrook Criteria for the switch mode power supply's input filter, which you can search on the internet and find a lot of literature addressing this issue. For experiment, can you add a couple of your original elco capacitors and see if the problem is gone?

Hi Youhao,

Thank you for the quick reply.

For my understanding: add the 4 original elco's or replace the the 12 MLCC with the 4 original elco's?

Note: the field line input is a Li-Ion battery pack, tested with 10S5P.

I don't have an impedance measurement device.

Is it (most likely) better to:

1. Add the 4 original elco's parallel to the existing 12 pieces of ceramic caps?

or

2. Replace the the 12 MLCC with the 4 elco's (to best match WeBench design)?

Switching to the LTC3779 is a lot simpler solution. Even the evaluation board DC2456A with only Rfb1 replaced with 412kOhm (for 42.0V output) results in zero (0) issues at all. I could cycle fifteen (15) minutes without any issue, even up hill with maximum pedalling assistance and the buck boost PCB is still working.

I see 4 possibe causes:

1. the incorrect ELCO with MLCC substitution (our engineering; will try to test this after the board has been repaired)
2. an error in the PCB design (our engineering)
3. a WeBench design/configuration/schematic/component selection error (TI)
4. an LM5176 design error (TI)