• State Resolved
  • Locked Locked
  • Replies 9 replies
  • Answers 1 answer
  • Subscribers 64 subscribers
  • Views 284 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

Original question:

TPS23756: Audible Noise issue PMP40514 Based PCB

TPS23756: Power Sag when Loaded

eljeffo
Intellectual 560 points
Part Number: TPS23756
Other Parts Discussed in Thread: PMP40514, , , LMV431, TPS23758EVM-080

Hi, I wanted to leave a post in case this situation looked familar.

I have 3 prototype (custom pcb) boards modeled after PMP40514.  I changed the voltage divider to the voltage ref to drop Vout from 12v to 5v.   

https://www.ti.com/lit/df/tidryo5/tidryo5.pdf

I have 1 working board, and 2 non-working boards.

Do I need to change the feedback resistors/caps just because Vout changed?  I don't think so since that only changes the frequency response, but I'm not really trying to change that (but maybe I did since Vout changed?)  In any case I'm starting with the PMP40514 values and one board works.

My working TPS23756 outputs 5.0v just fine at no load with 0.4V peak to peak.  The output looks fairly smooth .. sinewave like similar to the TPS23756EVM output as far as I can recall.

My non-working boards output a sawtooth waveform 5.0v average, but about 1v peak to peak. As soon as I load the board with nearly anything at all, the sawtooth starts to "dissolve" like someone poured acid on the waveform.    Needless to say, follow in boost/buck regulators depending on the 5v input blink on and off.  The POE board starts to click and pop (as a boost/buck would do during a short-circuit)

I admit, I'm still trying to learn all the feedback equations for what looks like a type 2 opamp compensator circuit.  But the fact that one board is working makes me think that at least part of the circuit is working, but can't respond fast enough to change under load?

In case you've seen something like this before, is there any conceptual thing I should know that would lead me to find the source of error quicker?

So far here's what I've tried:

Checked for any non-soldered pads (seems okay, even added solder to be sure)

Changed all large capacitors, inductors, transformer, diodes, TP23756 chip, voltage ref LMV431AIMFX-NOPB

Conceptually I'm thinking, hey maybe I'm drawing too much power, but 500ma @ 5v is well within reason.  

On the way home I remembered that I tripled up on C24 (active flyback clamp capacitor between the two mosfets) on the working board. Could that be the reason my working board survived, but the other two are browning out?  I noticed on PMP40514 Schematic tidryo5.pdf that C24 is noted at 6 * 0.1uF -- does it really mean there are 6 caps if they aren't drawn?  I don't think so since the TPS23756EVM and the PCB Plot for the PMP40514 don't show more than one cap.

Anyways, if this situation sounds familiar, please let me know.  I'm headed back to the office when I wake up to see if I can figure out what I missed.

  • 0
    Intellectual 560 points

    (as a note, power sag on non-working boards makes voltage dip to 4.1v or so... then spikes back up to 5.1/5.2v on the rising edge of the sawtooth output...)

  • 0 in reply to eljeffo
    Intellectual 560 points

    Ah yes, I've also tried fiddling with values/removing the RC on the integrator for the LMV431, but that only seems to change the slope of the sawtooth. 

    Tried to remove just the lone C28, but that just makes the sawtooth fuzzy.  Tried 12pF, no luck.

    Also tried to adjust R23 (resistor to Opto) from 330 Ohm all the way to 5k ohm, no luck.

    Also tried to put in R21/C29 on the opto receiver side like TPS23756EVM (10kR + 10nF closest I had) no luck.

    Tested all removed caps/diodes, they seem to have been operational.

    Oh changed primary side Q2 MOSFET, no change. (just in case ... it had a stuck valve or something hahaha)

    Clearly I'm just shooting in the dark, but wanted to let you know I tried :)

  • 0 in reply to eljeffo
    Intellectual 560 points

    Ah, also tried removing R25... really doesn't help :)

  • 0 in reply to eljeffo
    Intellectual 560 points

    Ah this is my opto: https://www.everlight.com/file/ProductFile/EL3H7-G.pdf  EL3H7D (had them lying around) CTR 300-600... seemed "close enough"?

  • 0
    TI__Mastermind 21320 points

    Hello,

    If you are changing the output design from 12V to 5V, the first thing that needs to change is the transformer. Since the transformer sets the primary and secondary current peaks and ripple, the output capacitance and compensation components must change also. To get an acceptable design, I think you will need to first start with the transformer. 

    Designing DC-DC power converters is no simple task. If you are new and have a short deadline, I would highly encourage you to copy the TPS23756EVM. 

    If you have more time on your hands, the following paper outlines how to design an ACF with another TI DC_DC controller. Simply use the TPS23756 IC values where the UCC part values are used. 

    If this post answers your question, please indicate so by marking this thread as resolved. Thank you.

     

    Regards, 

     

    Michael P.

    Applications Engineer

    Texas Instruments 

  • 0 in reply to Michael Pahl
    Intellectual 560 points

    Hi Michael,

    I don't think the ACF paper you tried to link came though... would you be kind enough to share again? :)

    I should have mentioned I changed the transformer too!  So sorry about that.  I'm using a Coilcraft POE13P-50L designed especially for 5v.

    And yes you are right, I am new and... have a moderate deadline.  

    Thanks Michael!

    FYI after a lot of reading and watching videos, I'm getting closer... I've found that the LMV431 needed a bit more input current at both Cathode and Vref and that made the response time much better on the secondary side.  Before that I guess the high resistor values at Vref "current starved" the voltage reference.

    However, I'm finding that it takes "too long" for the primary side to respond (Vout takes about 17-44ms to increase after the signal from the opto).  

    I'm trying to reduce the time it takes for CTL to rise again.  Not sure why only 1 of 3 boards has CTL/Vout which responds in a reasonable nanosecond timeframe.  Maybe my opto is still pulling down CTL for too long (since all values on the primary CTL side for R/C are already super small).  I'll order some new optos that match the EVM just in case there's a rating that I missed.

    Thanks so much for trickling what you can to me, it really means a lot!  

    (And yes I wanted to copy the EVM directly but my application doesn't have enough space for all the parts to drive the mosfets on the secondary side, so followed the PMP40514 instead... totally agree with you which is why I bought the EVM as a reference.  It's really nice!)

  • 0 in reply to eljeffo
    TI__Mastermind 21320 points

    Hello,

    Here is the paper: https://www.ti.com/lit/an/slua535a/slua535a.pdf?ts=1593721311227&ref_url=https%253A%252F%252Fwww.google.com%252F

    Another thing is we design the opto-CTR around 80%-160%, so if you are coping the feedback components you will not get the same response. 

     If you do not have all the space for the 5V ACF design, there are plenty of 5V synchronous flyback designs for 13W. The one with the least space is the TPS23758EVM-080. It uses primary side regulation, so there is no opto-feedback. It also integrates the switching FET on the primary side. 

    We have other parts that use the traditional opto-feedback if you want to use it. 

    If this post answers your question, please indicate so by marking this thread as resolved. Thank you.

     

    Regards, 

     

    Michael P.

    Applications Engineer

    Texas Instruments 

  • 0 in reply to Michael Pahl
    Intellectual 560 points

    Hi Michael,

    You're absolutely correct, my response curves are totally nuts.  I was able to get my Vout to be a little better via experimentation, but it's not stable under load.  I'll read the paper you noted above and see if a little more insight will help me set things straight.

    Thanks for the pointers on the other products and options.  I'll look into it!

    Jeff

  • 0 in reply to eljeffo
    Intellectual 560 points

    Michael,

    Closing for the same reason as other thread.  Raised output inductor to correct value and operation is more or less... "sane" as a starting point now.

    Thanks!

    Jeff