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.

  • Resolved

TPS7H3301-SP: ESR Requirement and Stability

Prodigy 190 points

Replies: 8

Views: 378

Part Number: TPS7H3301-SP

Hello,
These 3 questions are primarily related to the 2mOhm ESR requirement for stability of this regulator. Compared to the recommended low ESR polymer capacitors used in the datasheet example, typical CWR29 grade caps can have significantly higher ESR. In some cases it becomes impractical to meet the ESR guidelines set forth because of the amount of high reliability caps it would take to get under this limit.

1.) It appears to me that the primary benefit of this filter (based on the loop responses shown in the datasheet) is to improve gain margin of the system. Without it the overall loop gain can experience a magnitude with a flat slope (after the zero introduced by the ESR) sometime after the crossover frequency and this might only be 10dB below unity gain (Figure 23). In addition the filter probably helps to address the potential problems caused by inductive peaking of the output filter's impedance that could cause the magnitude to re-cross the unity gain line without this filter. If the purpose of this filter differs from this, please explain.

2.) In simulations, I'm finding that a potential downside of this filter is that it degrades the load disturbance rejection of the VTT output. Basically the filter reduces the bandwidth of the regulator's loop and in some cases worsens the phase margin as well. So how necessary is this filter if the 2mOhm ESR requirement isn't met? Can the PSPICE model for this regulator be used reliably to predict the phase margin and stability of the VTT regulator if the ESR is greater than 2mOhm (without the R-C filter)? One concern I have with such an approach is that the datasheet doesn't provide a tolerance of the gm for this device even at a few bias points across temperature. And the SPICE model probably only give a typical gm behavior (maybe with temperature dependence). Can someone comment on what the variation of gm from part to part might be like?

3.) In my particular application, my worst case bias current is below 300mA. I'm assuming that the 2mOhm requirement is based on using the full current range of the device. Is it a fair statement that operating at lower currents could mitigate the ESR requirement some? For instance a lower bias current would likewise lower the gm of the regulator and the dc value of the open loop gain which would directly affect the gain margin of the loop.

Thanks,
John

  • Hey John,

    The 2mOhm requirement is there to ensure that the zero caused by the ESR/Cout interaction is placed at a high frequency.
    As you suggest with the ESR/Cout zero being placed at a lower frequency there may be issues with gain margin.
    The filter can be placed to add another pole to the topology and account for this.

    While I wouldn't say overall that the requirements change for small loads, what you are asking about is a very small load.
    You can somewhat see the effect of the gm decreasing in the EVM that was made for the part.
    In figures 7, 8, 9, and 10 of the user's guide for the EVM you can see that as the current decreases as does the crossover frequency and gain margin (although slightly).

    With how low the current you are requiring is I would expect that the crossover frequency would significantly decrease and the gain margin increase quite a bit. What you will need to do is figure out what gain margin requirement you need to make and then plan the ESR zero appropriately.

    Your best option will be to add a pole to the frequency response to have a good gain margin, which is essentially what the datasheet is trying to suggest.

    Thanks,
    Daniel
  • In reply to Daniel Hartung:

    Daniel,

    Thank you. I think that confirms my understanding of the filter's purpose. The real issue I'm having is that we have a design were the RC filter wasn't included and the worst case ESR for the bulk caps is fairly high (~30mOhm). The system works as is and we haven't seen any stability issues and are debating whether we need to re-spin the board to add the filter for robustness. I think in part the issue has been mitigated by the low load currents in our application. In simulation we appear to have plenty of gain margin across loads for stability. However, the amount that the dc gain can shift and still maintain about 60deg or P.M. can be as low as ~2x at our max load. That's why I inquired about the tolerance on the gm of the device at given biases. If the tolerance is fairly tight, .e.g 10%, then we may be able to live without the filter. If on the other hand the gm can vary by 2x from one device to another, then I would be concerned with wildly varying transient performance between boards. Do you have any feel for what the variation in the gm parameter can be like for say 270mA? Along those lines, the variation in the current loop's high frequency pole (~200+ KHz per the App Note Calculator/Excel Worksheet) would also be of interest in deciding what we want to do.

    Thanks,
    John
  • In reply to John Rodriguez:

    Hey John,

    Unfortunately the information you are asking for simply isn't specified on our end as gm isn't part of the electrical characteristics of the part.

    Thus I cannot comment on the variation from part to part.

    Thanks,

    Daniel

  • In reply to Daniel Hartung:

    Hi Daniel,

    By your response, I gather that you mean it isn't specified explicitly in "section 6.5 Electrical Characteristics".

    However the Gm is clearly an electrical characteristic for this regulator, it's mentioned in a number of places throughout the datasheet. For instance, Page 18 states that "The typical Gm is 250 S at 3 A and changes with respect to the load in order to conserve the quiescent current (that is, the Gm is very low at no load condition)." The App note / excel calculator also has a sample curve as a function of output voltage and current. The use of the phase "Typical Gm" suggests that the author of the datasheet, had some understanding of how this parameter might vary and chose to give something like an average value. Are you sure that there is no other statistical information that may be available or someone who might have this information that you can direct me towards?I would greatly appreciate it.

    Thank you,
    John
  • In reply to John Rodriguez:

    Hey John,

    Yes I mean "Isn't specified explicitly in section 6.5 Electrical Characteristics"

    As you can see from the calculator that we have for the device, the gm varies widely.
    This is why it is not specified in section 6.5 and why I cannot comment on the variation from part to part.

    All of the data in that calculator is typical values.

    Thanks,
    Daniel
  • In reply to Daniel Hartung:

    HI Daniel,

    Thank you, I understand why you can't comment on my question the way I posed it. I appreciate the answers you've given me so far, they've been helpful. If you would humor me, is it reasonable to infer an upper bound on how much the gm could change (in some instances) based on Figure 23?

    Here's the rational. This Bode plot appears to be taken from an Eval board under a 2A load, the loop gain has a gain margin of about ~10dB (~3x's). An inference might be that the Gm at 2A most likely can't vary by more than ~3x's (at 2A) across parts because then you would run the risk of some of your eval boards being unstable while using the datasheet's recommended output filter to guarantee stability. It seems that the Gm variation can not exceed a factor of three (roughly) for the reasons stated. Would you agree with this? Even this simple bound would be extremely helpful for me to know.

    Thanks,
    John
  • In reply to John Rodriguez:

    Hey John,

    I can't give anything but typical values.
    All of the maximums and minimums we provide on the part show up in the electrical characteristics.

    There is a lot of testing is put in to allow us to specify a maximum and minimum of the part.
    Without all of that testing it is not possible for me to comment on what a possible maximum would look like.

    Is there a limit to how high that gm can get in the part?
    Yes there is likely a limit.

    Is the gm from part to part likely going to vary by 3x?
    No, the variation from part to part is likely not going to vary by 3x. The part would be very hard to use.

    Is it something I can guarantee or give a maximum value on?
    No.

    Thanks,
    Daniel
  • In reply to Daniel Hartung:

    That answers my question. I know you're not in a position to make such guarantees without statistical testing to back it up. Thanks again, it's much appreciated!

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.