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.

TPA3251: Post filter feed back issue

Part Number: TPA3251
Other Parts Discussed in Thread: TPA3255

Hi,

I'm developing a new board with the TPA3251/55 amplifier and doing some components tests

I have really good results without the PFFB --> -90db / 0.003% THD+N with 7uH output inductor

When I mount the PFFB with the recommended value in the SLAA788A, the THD+N is really better but at the same time I get a big limitation in output power.

I couldn't go furhter than 10W, after that the THD+N is getting bad as if we reach the max output of the amp.

Did anyone have this problem?

  • To add some debug notes, I've done the following test:

    all PFFB assembled : power limitation

    no R_fb / c_fb_out 1nF : same problem

    no R_fb / c_fb_out 220pF : same problem

    no R_fb /no c_fb_out : work fine

    R_fb / no c_fb_out : TPA3251 fault error

    First question, is the system working without c_fb_out?

    What could i test more?

    I've put the PFFB circuit close to the last ampOp that drive the input of the TPA325x, do you think it can be an issue?

    I've also test the PFFB on the TPA3255 with 10uH and it work better, I've a little power limitation befor THD goes bad but not so much

  • Hi,

    We are looking into your question. As some team mates are on National holiday in China, reply may be delayed. Thanks for the understanding!

    Regards,

    Hao

  • No worry, thank you!

    I have some update too:

    I've done different test with 4 inductors that are recommended in the slaa701A LC filter design

    I used the recommended PFFB value from SLAA788A 

    - no PFFB coilcraft MA-5173 10uH --> power limitation 180W for THD < 0.1%

    - coilcraft MA-5172 7uH --> power limitation 70W for THD < 0.1%

    - coilcraft MA-5173 10uH --> power limitation 140W for THD < 0.1%

    - Wurth 7443630700 7uH --> power limitation 10W for THD < 0.1%

    - Wurth 7443631000 10uH --> power limitation 10W for THD < 0.1%

    It seems that the output inductor affect the result of the feedback and I don't understand with what is written in the different application notes.

    Is the feedback circuit only dependant from TPA325x version or also LC output filter

    Also I don't understand because the Wurth inductor without feedback give a better result than the coilcraft.

    But when the feedback is applied it get worth

    And also it seems that the feedback problem is more visible on the TPA3251 than TPA3255 with the same config and PVDD supply @ 36V

    I have some graph to show what I'm telling

  • Hi Florent,

    First of all, the reason for why PFFB implementation has better THD+N performance than without it is, it helps to eliminate the non-linear factors of the inductor. That's why you get almost the same results using different inductors with PFFB, while you get different results without PFFB. 

    You also mentioned there is some power limitation when PFFB gets involved. I think this is because the PFFB introduce some changes in the control loop stability. Generally speaking, the closed control loop tends to be unstable as the power ramps up, which also matches your observation. 

    From this perspective, we may need to do some optimization on the PFFB feedback path.

    Layout can be another potential factor here. The higher the power is, the more noise will be on the board.

    Thanks!

    Regards,

    Hao

  • Hi Hao,

    Thanks for your reply! I have done 8 different layout of the same schematics to see the impact of it on the performance / quality.

    What I can see is that the type of inductor has more impact than layout difference.

    I have done a lot of measurement THD+N vs Power and THD+N vs Freq to observe the comportment with the TPA3255 and also TPA3251.

    What I see is that without the PFFB .

    The Wurth inductors 7uH and 10uH gives the best THD+N with TPA3255 and TPA3251 but with already a power limitation (curve goes up at 70W RMS) compare to the coilcraft

    As you mentioned in the application notes the 7uH is better for quality and 10uH for power.

    With the same value on the same IC, the Wurth is always better than the coilcraft in terms of quality but not in power.

    Then the apllication will decide wich combination to use.

    Now the problem I'm facing is that your PFFB setttings almost work with the coilcraft inductor but not at all with the Wurth that should gives the best quality and it's what I'm looking for.

    Is there a way for you to give me other settings or a way to simulate the PFFB to get some data?

    I have a lot of graphs that I can send you to check what I have

    Warm regards 

  • Hi Florent,

    I went to check the spec of both WE and Coilcraft inductors. The difference I found was saturation current of Coilcraft parts are twice higher than WE parts.

    As you know, the inductance of a inductor varies towards the current it is carrying. Looks like Coilcraft parts has little inductance change at high power conditions as it has higher Isat. That's the reason why it has better THD+N performance in high power conditions.

    So, if you want to use WE parts, please find a part with higher Isat value. If you would like to use PFFB, coilcraft part should also work for you. If you got any instability issue, please try to decrease the open-loop gain, I mean decreasing the component value on PFFB path a little bit in same scale.

    Thanks!

    Regards,

    Hao

  • Ok thanks!

    Two question :

    Do you think using 7uh coilcraft or wurth with pffb will be the  same result?

    decrease the open loop gain means reducing the pffb gain? or increase it?

  • Hi Florent,

    1. I think so. Bench verification is recommended for this.

    2. I mean decrease the gain, cause open loop gain is formed by the product of forward gain and feedback path gain(also know as G*H).

    Thanks!

    Regards,

    Hao

  • Hi Hao,

    I've tried what you purpose and here are the result :

    If I reduce the FB with higher value of R_fb, the result in general are worst and also the power limitation.

    I try to increase the feedback gain to 10db with a 12K resistor instead of 18K, the THD increase well and the power limitation is still the same as with 18K, 33K, or other value.

    I also see that changing the value of R_fb_gnd have an impact. Increasing it gives better power performance, could you explain me why? I'm not sure to understand the T Filter impact.

    Also the PFFB have a bigger limitation on power with the 7uh coilcraft inductor than the default 10uH that is on the eval board.

    All the value given in the application notes are for a 10uH?

    Or the value of component for the feedback should not be impacted by the value of the output inductor?

    Thanks a lot for your help!

    Warm regards Florent

  • Hi Florent,

    The open-loop gain we talked contains two part outside the IC, first is the output L-C filter, second is feedback network. What we are trying to do in PFFB here, is to decrease the open-loop gain, or we make it more clear by saying, limit the bandwidth. We suppose the power limitation has something to do with the loop stability, and limiting the loop bandwidth could help on the stability.

    For LC filter, if we take other considerations away, 7-uH inductor has higher LC double pole frequency, which means it has lower bandwidth. So you can get more power with 10uH indutor, am I right?

    For feedback loop, decreasing the R_fb and increase the R_fb_gnd helps to decrease the gain and limit the bandwidth. Let's regard the C_fb_x as impedance for AC analysis, then the effect of T-type network can be understood as, the impedance of the T-type branch equals to  (Zc_fbin+Zc_fbout)*(1+(Zc_fbin//Zc_fbout)/Rfb_gnd). The higher Rfb_gnd, the lower banch impedance, the lower gain.

    All the curves are based on the EVM configurations. Component value on feedback also depends on the LC filter, cause the the whole open loop gain is the product of the two stages.

    Regards,

    Hao