This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM5176: transient repsonse

Part Number: LM5176

Hi team,

my customer encounter large undershoot drop from 5.2V to 4.7V, I told him to increase the Rcomp ( from 2.49K to 25K),

and it did improve a bit from 5.2 to 5V, and they still want better.

but when they increase Rcomp to 47K, the signal is very noisy as shown below,

why, and is there any  other way we can improve ? 

BTW,  they added extra output cap 65uF, but transient didn't help much..

upload calculator, you can check the current compensation network 

5353.LM5176 Buck-Boost Quickstart Tool r1.0.xlsm

  • Hello Fred,

    Thanks for promoting our products to the customer.
    We can have a closer look after the weekend.

    In the meantime (just by looking at the data in the quickstart calculator) I have a few questions:
    - Why are they using a complex 4 switch Buck-Boost if they only need a Buck from 25V to 5V ?
    - Why did they select a much smaller inductor than recommended?
    - Does their board have footprints for Gate resistors (or is there at least some way to add resistors between the driver outputs and the FETs)?

    Can we please get more information:
    - What is the minimum and maximum input voltage?
    - What is the min/max output current?
    - Which upper and lower output current has been used to generate the transients?
    - Can we please get a full schematic?

    Thanks and regards,
    Harry

  • Hi Harry,

    it's for PD spec,

    Vin:24V

    Vout: 5V/3A, 9V/3A, 15V/3A, 20V/5A, 28V/5A,

    so transient is from 0 to 3A/5A.

    could you  give us a direction , like increase Rcomp or Cslope .....etc

    PD circuit.pdf

  • Hello Fred,

    Thanks for the information.
    There is one more thing that I would like to understand.

    They are using a DAC to modify the output voltage.
    So I assume that all the different output ratings shall be generated with one single design,
    WITHOUT changing the feedback network or any other component.
    Is this assumption correct?

    The overshoots have been seen with an output voltage of 5.2V.
    How does the output react at higher nominal output voltages? Which size of the spike do they see at 9V output or higher?

    As they are using a DAC, they can easily do this experiment:
    When coming down from higher output voltages, is there a point (voltage), where the spikes immediately become worse or are they continuously growing?

    Thanks again,
    Harry

  • Hi Harry,

    yes, they change output voltage by changing DAC output,

    the overshoot also seen in the higher voltage , and theoretically should be worse,

    but since the spec is not strict as 5V in higher voltage ,so the bottleneck is still 5V now,

    can we focus on the 5V/3A  transient improvement suggestion?

    I believe whether 5V, 9V or 20V, improving transient respsonse is the same direction, so we can just focus on 5V now

    Regards,

    Fred

  • Hello Fred,

    Thanks for the information.

    I will come back to you later today.

    Thanks,
    Harry

  • Hi  Harry,

    sure ,

     PCM transient undershoot.pdf

    BTW, I refer to this document, I think I got to increase the Rcomp and Ccomp based on the equation,

    hope you to double confirm this for me, or other alternative suggestion

    Regards,

    Fred

  • Hello Fred,

    I don't think that tweaking the compensation network (alone) can solve the problem.
    There are other unusual things going on and we will need to debug and understand those first.

    As a first step we would like to remove all uncertainties.

    - The DAC is directly connected to the feedback network and can maybe have a negative influence.
    Therefore, to make sure that this is not to blame, can we please ask the customer to remove R2322 and modify R2321 in a way that the converter is fixed to 5V out.

    - Looking at the first screenshot, we see the reaction of the converter to the load change between 0A and 3A. Is this correct?
    The output voltage breaks down from 5.3V to 5.3V and the converter can not manage to compensate for it.
    We should not see that voltage drop, no matter how bad the compensation is.
    The converter cannot manage to get the required power of 15..16W across. Why is this the case?
    - Is there any limitation on the input side that limits the current/power in some way?
    - Does the inductor go into saturation? (We would need screenshots of the inductor current)

    - When you increase the compensation resistor too much the feedback loop will become instable.
    The maximum value that I found using the quickstart calculator is 8K
    The "noise" that you see in the second image is probably an instable loop.
    - We would need to see an oscilloscope shot which is zoomed in, so we can see each cycle.

    - The slope compensation does not match with what the calculator recommends.
    - I would recommend using a slope capacitance PC2363 of 560pF


    In general:
    It does not help you much if you focus on one configuration and hope that the others will work even better.
    In the end all 5 voltages with high, mid, and low current requirements will need to work with the same setup/compensation.
    - You can use the quickstart calculator to go through all these combinations and find some good compromise.


    Here are the values that I have determined that way:
    - Slope capacitance PC2363: 560pF
    - Compensation resistor PR2329: 8k
    - Compensation capacitor PC2361: 33nF or 47nF
    - HF capacitor PC23B9: N.C.

    But these are all just theoretical values, not taking any parasitics or the influence of the PCB into account.
    So, probably they will need to get re-adjusted again.

    All information in this correspondence and in any related correspondence is provided “AS IS” and “with all faults”, and is subject to TI’s Important Notice (www.ti.com/.../important-notice.shtml).

    Best regards,
    Harry

  • Hi Harry 

    can you correct the comment "The output voltage breaks down from 5.3V to 5.3V" since I don't know where you are refer to in first picture.

    we already try PR2329 from 2.5K to 10K and to 25K now, and trend is that the undershoot is keep reducing,

    so I don't see why we should back to 8K.  increasing Rcomp equal to decrease the zero, the crossover frequency should be better

    and sorry, forgot to mention , Cslope was already 560pF then,

    please let me know your idea 

    I don't think with this normal compensation setting, the undershoot and overshoot should be this bad.

    do you think this undershoot level is expected?

     

    I would tell customer to try to remove the DAC and see if transient is still this bad

  • Hello Fred,

    I am sorry for the confusion caused by my typo.
    I meant to say: The output voltage breaks down from 5.3V to 5.2V

    I am talking about the fact that the output voltage is jumping up and down.
    The voltage @ 3A load (inside the red circle is permanently lower than without the load.
    This should not be the case. The system should be able to get back to the desired output voltage after some time, as indicated by the green curve.

    As this is not the case (although the output power of 15W is relatively low), there is something wrong.
    So, as the first thing to look at, I would doublecheck if there is any limitation of the input power.
    Please verify.

    As per the compensation: The phase margin has to be above 60° to allow for a stable operation. 21° is way too low.

    You can use the Quickstart Calculator yourself to find out where the border is:

    Values as provided in 5353.LM5176 Buck-Boost Quickstart Tool r1.0.xlsm:
    instable, phase margin = 21°.  Therefore the "noise" in your second image 

    Change output current to 3A:
    instable, phase margin = 22°

    Change Rcomp to 8K:
    stable, phase margin = 61°, just above 60° and also for a much lower output current.

    Best regards,
    Harry

  • Hi Harry,

    • as far as I know, there is supposed to be load line regulation, so the output DC level is supposed to drop a little with increasing loading.
    • but like said earlier, the crossover frequency is tradeoff to phase margin, although phase margin is low at 25K, but the crossover freq is high, and we did prove that the undershoot of 25K is better than 2.5K and 10K(which is close to 8K), so I think we should focus on crossover freq as long as the phase margin is not zero since zero means oscillate .

    please correct me if I'm wrong

  • Hello Fred,

    Please have a look at the training literature about loop compensation and you will find that 22° is considered too low.
    You can utilize the quichstart calculator to find values tha will match for all use cases.
    If the customer is happy with 25K, please go for it.
    I will stop this conversation here, as we are going in circles.

    Best regards,
    Harry

  • Hi Harry,

    I understand phase margin is important I understand the theory, but the practical issue right now is we have to pass the spec of undershoot and overshoot, 

    by changing all the Rcomp , seems like 8K Rcomp is just not enough. 

    the bottleneck now is not how to act like a normal stable controller but pass the spec. 

    or would you explain why increase phase margin and decrease crossover freq can help undershoot?

  • Hi Fred,

    i also would like to come back to the question from Harry before, where he asked about large drop on the output voltage due to the load:

    The controller will regulate at the point of the Feedback resistor connection to the output voltage, to keep this point stable. So we should not see any drop here ( at least not in that range).

    So i assume that the plot above has been done on another point - can you confirm that assumption?

    Can you also please ask the customer to measure the output voltage also on the top of the Feedback resistor divider (best as additional channel to the above measurement).
    - If there is a difference we need to understand and evaluate what is between the two measurement points.
    - If there is no difference we need to understand and evaluate why there is this large voltage drop not regulated.

    Best regards,

     Stefan

  • Hi Stefan,

    yes, although same node in the schematic, but the measurement point is actually far away from the FB divider, which could be the reason that cause IR drop and level gap. I'll ask them exactly how far and test the difference like you said tomorrow.

    aside from the DC level mismatch above, but do you think this measurement point location problem could cause transient worse as well?

  • Hi Fred,

    as the long trace is always resulting in some inductance this could lead to the voltage drop on the load step.

    Best regards,

     Stefan

  • Hi Fred,

    I have not see an update in this thread for the last 12 days.

    So I assume this can be closed. If there is still something open just reply to reopen it.

    Clicking on the resolved button helps to maintain this forum.

    Best regards,

     Stefan