LMR36520: LMR36520ADDA, Confirmation for 3.3V and 9V, Audible noise @ -9V

Hello David,

In the bottom-most design, you are experiencing some level of  instability.

Given in IBB design, a RHP zero is introduced into the transfer function, these devices, which internal compensation is geared towards 1/10-1/5 crossover frequency of 400k-2.1M, now requires often a large feedforward capacitor to increase the phase prior to crossover frequency, which I can assume is sub 20k.

It looks like you have injection resistor placeholder in FB network. Do you have frequency response analyzer available to use?

I would like to determine actual crossover frequency, then we can see the required CFF to stabilize.

Hi Marshall,

i'm assuming that the comment are about the -9V design ( IBB) and the 3.3V and 9V design are behaving as expected?

Yes we have in the layout a footprint where we can place a cff. Also there is R287 = 0R ( this is what you mean by injection resistor placeholder?).  We currently don't have a frequency response analyzer available to use.

Is it possible to determine the crossover frequency with a oscilloscope? If yes, how can this be done?

What about the negative peak in switching waveform when the Vin = 24V?

Best Regards,

David.

Hello David,

yes, my comments are in regard to the IBB. Yes, from the waveforms you shared and the schematics I analyzed for the other designs appear OK.

It is possible to determine crossover frequency with oscilloscope. "Easiest" way would be apply a load transient which is bandlimited by the control loop. IE: 500mA/uS slew, 1A step (ensure you dont hit current limit).

Crossover frequency's period is the time it takes for the converter to begin to recover back to nominal set voltage.

See fig49 

https://www.ti.com/seclit/ml/slup340/slup340.pdf

Before worrying about the negative peak on SW, I would first like to ensure we have a stable design. It seems to me we do not at this time.

Hi Marshall,

here are the measurements, please note that we had to set -9V offset on the scope in order catch the transient response.

If we are measuring correct the f_cross = 22kHz, according to SLVA289B

Page 5 figure 7.

Best Regards,

David.

Hi Marshal,

1)

we did the calculation per equation 3 on page 4 -> and got 36pF so we used 33pF on the board

and here are the measurements ( the blue is the load step 1A = 10V / 10Ohm)

zoomed in

There is still some oscillation,  also for the noise, last time the noise started ad 600mA load and now it starts at 800mA 

So there is an improvement. One thing, i was going through my calculation when i designed the schematic and i found out that i designed 

I_max=0.5A and not 1A ( my mistake), and we have a stable design up to 0.5A and no noise sound. Maybe the noise is because the inductor is not suited for 1A operation?

2) Newer the less  we have negative voltage spike on sw point when vin = 24V and Iout is below the max 0.5A

Best Regards,

David.

Hi Marshal,

i noted one more thing.

The measurements form my last two post where taken with 1A step and Vin=24V

Now i changed the Vin to 12V ( because we saw instability under 12V vin)

Here is a Vin= 12V and 0.5A step and we have fc= 20kHz like before ( no cff)

And here is a Vin=12V and 1A step and now is the output oscillating and we have fc=10kHz ( no cff)

Which one should we look at?

Best Regards,

David.

Hi Marshall,

we measured the current through inductor ( we soldered 100mOhm resistor between IC switching pin and inductor).

1) Vin = 12V, I_load = 0A (no load) , no cff

2) Vin = 12V, I_load = 100mA  , no cff

3) Vin = 12V, I_load = 500mA  , no cff

4) Vin = 12V, I_load = 600mA  , no cff, transition

5) Vin = 12V, I_load = 650mA  , no cff ( current hits the 0)

8) Vin = 12V, I_load = 850mA  , no cff ( current hits the 0)

Best Regards,

David.

Hi Marshall,

I'm David's colleague, nice to meet you.

Unfortunately currently we don't have injection transformer in company but we're planning to get it, anyway we done measurement with step response (470mA Step, V_ in = 12V).

*Note this converter works in inverting buck mode, V_out is -9V

 (Ch1 - V_out transient, Ch2 - Load Voltage transient)

 
Regarding definition of crossover frequency following Application Report SLVA289B for our case that is 15.62kHz

otherwise you said before;

Marshall_Beck said:

Crossover frequency's period is the time it takes for the converter to begin to recover back to nominal set voltage.

For Calculation I found a few equations and I'm not sure witch one I need to use

1) From  LMR36520ADDA Datasheet Cff section say use Application Report SLVA289B



Solve this equations I got value off Cff ≈ 306pF (R1 = 100k, R2= 12k4)

2) Application Report SLVA834 equations 

Solve this equations I got value off Cff ≈ 50pF (R1 = 100k)

We also tried to blindly put some values of Cff between 15pF and 300pf, but without success. We got even worse cases then without Cff capacitor, the response was with many oscillations or totally unstably system. 

Regards Mateo

Hello Mateo,

I understand the frustration here. You maybe a little bit light on output cap if CFF is not allowing you to get anywhere stable.

I would recommend changing ceramic to 4 x 22uF 16VDC.

Alternatively, you can try adding low ESR electrolytic in addition to the current ceramics ie: 100uF.

You could do away with CFF by utilizing ESR of electrolytic ie: 50mOhm, but that way can be a bit messy when you consider the tempCo of this ESR.  

Hi Marshall,

Yes you are right about output capacitors, we increased output capacitor from 4x 10uF to 8x 10uF and now we have stable system on output current of  500mA and higher (I checked up to 900mA, and there is system stable)

1) 24 Vin, 500mA step response

2) 12 Vin, 500mA step response

I have another question about negative spike at the switching point, looks image below 

3)Vin 24V, Without load

It's that normal behavior? Lower value should be -Vout, he have additional 5V in this spike 

Maybe we should  use parallel diode like on image below? This image is from Application Report SNVA856A page 6., we didn't find any explanation regarding this parallel diode which impact it has on waveform at the switching point. 


Regards Mateo