LM51521-Q1: How to measure loop response

Hello Brigitte,

Thank you for your swift answer.
I understand.

My customer said they couldn't  measure loop response correct by adding resistor at discussed point.

I'm requesting them to share their measured result and test condition.

Best regards,
Kazuki Kuramochi

Hi Kazuki,

Thank you for your reply, yeah please do share that and also let us know what value of resistor did they use there for the loop response measurement?

BR,

Haroon

Hi Haroon,

Below is their measurement result.

As you can confirm there is phase inversion around 35k.

Their circuit is as below and measuring loop response by inserting injection resistor between VOUT/SENSE and output.

I suspect that it may be from the resonance of inductor(L1708) and input cap due to resonance frequency is almost same (1/2pi*sqrt(21uF*1uH))=35kHz.

Would you provide your opinion for root cause of this phase inversion?

Best regards,
Kazuki Kuramochi

Hi Kazuki,

Could you please fill out this quickstart calculator tool with the design specs so I can compare the loop response they got with the one from calculator?

https://www.ti.com/tool/download/SLVRBL6 

Thank you

BR,

Haroon

Haroon,

Please find following customer's calculation sheet.

LM51521-Q1_Excel_Quickstart_Calculator_for_Boost_Controller_Design-マーキュリー対応.xlsx

Best regards,
Kazuki Kuramochi

Hi Kazuki,

Thank you for the reply, It is hard to comment on the phase inversion. Is it possible to change the compensation circuit and then retake this bode plot of the loop response? with Rcomp = 400k, Ccomp = 0.2nF and CHF = 2pF?

As for the resonance I count Cinput as 31uF, am I missing something?

Thank you

BR,

Haroon

Hi Haroon,

I'm requesting them to provide measured waveform under requested condition.

For inject resistor they put in between VOUT/SENSE and output by cutting pattern as we talked so there isn't it on schematics.
So It is exactly same location against EVM.

By the way, they ,modified board as follows and tested it.

1. Add 2pcs 10F MLCC at Vin side
2. Add 2pcs 10F MLCC and 180uF E-CAP at Vin side
3. 180uF E-CAP and change Rcomp to 30kOhm

I'm requesting them to show actual measured waveform but they said that abnormal phase shifting is disappeared but the gain margin is too small.

As you can confirm, the measurement result and calculation sheet are completely different.
So they worry about reliability of calculation sheet.
Therefore, please identify why there is this difference about how to improve this performance.

Best regards,
Kazuki Kurmaochi 

Hi Kazuki,

I see, yeah just make sure that there is no capacitor near that injection resistor and it is the same as the EVM.

The actual measurement and the theoretical results from the calculator could be a bit different, but there are some external factors and layout issues that the excel sheet does not take into account, so in general the excel sheet loop response bode plot is the closest one can get to the actual measurement, they could also try using the power stage designer and check the loop response there as well. it would be good if they can share their loop response result after the modification so that I can compare them with the excel sheet and see the difference.

In general to improve the gain margin, they could try to increase the CHF or C1999 and see if their gain margin increases.

Hope this helps.

BR,

Haroon

Hi Haroon,

I understand that calculation sheet's loop response should be similar as actual measurement result generally but some external factor and uncovered parameter in calculation sheet may cause difference.
As we observed, their first measurement result has huge difference.
Could you provide anticipated factor which cause this huge difference based on provide information?
If provided information is not enough to estimate the factor of causing this difference, could you please tell me what additional information is needed?

I understand increasing CHF will increase gain margin.
However, if this changing will be able to satisfy their target, it is not approaching root cause directly.
So, this root cause may cause other trouble if it is remained.
Therefore, we'd like to focus clarifying root cause at first.

Best regards,
Kazuki Kuramochi

Hi Kazuki,

I am understanding what you are saying, but could you please share the latest loop response from the customer? I cannot say what was the root cause if I did not see the before and after loop response of the system. 

BR,

Haroon

Hi Haroon,

I received their waveform when they changed Rcomp to 30kohm and Romp to 30kohm and adding 180uF E-cap.

Could you confirm expected root cause by following loop response?

Changing Rcomp to 30kohm

Changing Rcomp to 30kohm and adding 180uF E-cap at Vin

Best regards,
Kazuki Kuramochi

Hi Kazuki,

Thank you for the bode plots, as you can see this last waveform that is stable, is very similar to the bode plot you see on the excel sheet calculator here below:

Now this tool or calculator will show you a loop response based on theory, so it cannot take into account the external factors such as layout or resonance between the inductor and input capacitor, for that only the actual bode plot would be the most accurate.

One question: this loop response is not of the EVM, but from the customer board, correct?

BR,

Haroon

Hi Haroon,

They are measuring this waveform on their board.
Also, they provide loop response it is similar as your required condition(Rcomp = 390k, Ccomp = 0.22nF, CHF = 2.2pF).

Could you please provide your estimated factor which is causing this abnormal loop response?

We are taking 2 weeks for this discussion so they want to know your opinion as soon as possible even if it is not solid estimation.

Best regards,
Kazuki Kuramochi 

Hi Haroon,

Thank you for your explanation.
I explain them as it is our current estimation.

As I talked, they already tested increasing input capacitor as below and they cannot observed difference at phase response.

1. Add 2pcs 10F MLCC at Vin side
2. Add 2pcs 10F MLCC and 180uF E-CAP at Vin side

Does it mean, the expected root cause is only any parasitic factor and they need to change layout to improve this response?

Fortunately, they are planning to improve layout.

Best regards,
Kazuki Kuramochi

Hi Haroon,

Thank you for your confirmation.
I explained our opinion and be waiting feedback from them.

I'll let you know when I get update or comment from customer.

Best regards,
Kazuki Kuramochi

Hi Haroon,

Our customer confirmed  COMP value which can satisfy their target PM=60deg/GM=10dB at full load.
Then Rcomp=56k and CHF=22pg could satisfy their target as follows.

However, they measured half load condition, PM and GM cannot satisfy their target as below.

As we talked, they will modify proto board.
Then, is there risk of happing similar significant PM/GM difference depending on load current condition even if they changed their layout?

Frankly speaking, I haven't seen such a bad measurement result when we successfully measured PM/GM.
So, I worry  that it may not from schematics or layout.

Best regards,
Kazuki Kuramochi

Hi Kazuki,

Thanks for the update.
Due to high workload, answers may currently delay by a bit.
Haroon will get back to you within 1-2 days.

Thank you very much for your patience.

Best regards,
Niklas

Hi Haroon,

Thank you for your reply.
I understand that load current condition should have smaller impact for loop response than their measured result if the layout is appropriate.
Also, we guess it should be improved at next proto if their design is appropriate.

By the way, this is just confirmation, do you have loop response measurement data at different load current condition?
This request just want to see ideal loop response against load condition.

Best regards,
Kazuki Kuramochi