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.

tps23754

Other Parts Discussed in Thread: TPS23754EVM-383, TLV431A, TL431, TPS23754

we have designed custom proto board based on tps23754evm-383.
we have followed the layouts as per the EVM

we dont see any output at the Tp13 with Tp3.. output voltage shows zero
we have measured voltages on EVM and our board and details are attached in the
excel spreadsheet.

looks like feedback is not working as gate is not firing and CTL is always 5v,
and T2P is not going low
Also we are not seeing any pulses on GAT or GAT2 pins (TP7)

Another interesting observation is that when we remove 1mH inductor(L3 in EVM) in Aux path,
we see GAT pulses but still the output is not seen

what could be the problem, how do we proceed with this debugging, need help

do we need to replace TL431A (U3 in EVM) with TLV431A as mentioned in this issue

http://e2e.ti.com/support/power_management/power_interface/f/204/t/132254.aspx

-venu

Yds_board_poe_voltages.xls
  • Hi venu,

    The EVM uses the TL431. So as long as you use the same feedback resistor divider and TL431, it should be okay.

    Do you have a schematic I can look at of your design? If you copied the EVM schematic and layout, then the only issue I can think of is a possible wrong component value or connection on the board. This would be good double check.

  • we performed continuity checks for shorts and also connections, it looks ok. anyway we will double check.

    component values we have checked, and also checked the impedence compared with EVM, looks correct.

    find enclosed the schematics. Q5 connection was wrong, we have fixed in the board.

    thanks

    -venu

    J3208-final-POE-v1.0.pdf
  • Hi Venu,

    I couldn't find anything wrong with the schematic.

    Are you testing with a PSE? or power supply?

  • we are testing with 48v Adaptor, 1A . we have also checked with PSE(Tplink).

    can you provide some slot today, may be 9-10am your time, we are at india, Bangalore

    we can chat online with skype so that we can quick interaction to get to the root of the problem...need help

  • we observed constant GAT pulses (looks like reset cycles) on test points and output was 1.8v

    when we connect from PSE, the leds on the TPlink box keeps blinking (resets)

    we captured screen shots on the scope and enclosed the same.

    Refer to our schematic, file names represent the reference designator and pin nos. pls review and let us know your observations

    what is not clear to us is why is the Q6 is not switching, the pins 1,2,3 are always showing zero in scope. what would be the root cause, and how to go about in debugging the same

    thanks

    -venu

    scopeimages-27Nov2014.rar
  • Hi Venu,

    Yes, these are very unusual switching waveforms. Can you also attach a layout or .brd file?

    Are you loading the output (perhaps theres an overload)? Can you test with light load?

    Is the VC reaching it's UVLO voltage at startup? It must reach ~15V before the controller starts switching and the output ramps up.

  • we havenot connected any load, its just the LEDs D16, D25 as per the schematic. enclosed the layer plots.

    what can load a GAT ?, we have checked for shorts with GAT pin, couldnt find anything..

    how do we check VC its UVLO at startup ?,

    we tried removing R66, CS, the GAT pulse is NOT expanding and added a POT instead of R76 and still GAT pulse doesnt seem to expand or contract... what could be the reason 

    let us know

    J3208-LAYERPLOTS.pdf
  • we observed that R66 in series with the CS pin is placed on the solder side and rest of the connections for CS is on Component side.. does this cause any overload ?

    we also tried connecting a resistor directly from the C46 to Mosfet(Q7) pin 1 with 1k ohm resistor and remove the R66 from the solder side.(shortest path for CS signal)

     we tried removing R65, R56 (GAT, GAT2) series resistors to see whether the controller is generating the pulse, we observed that GAT2 is getting generated but not GAT signal.

    looks like CS is the cause of this issue, can we reduce the Rcs from 0.12ohm to lesser value ? we expect 2A peak current, how do we compute the Rcs resistor value ? 0.55/2.0 = 0.275ohm ?

    does current slope can cause this ? how is current slope compensation computed in evm as 1k ? we did refer Pg 29 of TPS23754 datasheet, what is the desired slope ? Vslope_d in the equation and how do we assume this ?

  • Hi Venu,

    I don't see an issue with R66 being placed on a different layer as long as the distance isn't too long. This shouldn't cause any overload.

    The only issue I've seen on placement for this resistor is if it is tapped on the noisy FET pin rather than the sense resistor pin (even if it's the same node, the FET pin is noisy and can couple into the CS pin. Rather than air wiring it to the mosfet, air wire it to the sense resistor. Although I don't think this is the issue here.

    The equation you have for the sense resistor is correct.

    I don't think it's an issue with the slope compensation. I've seen this resistance go as high as 5k in some designs.

    Be sure your controller is not damaged, maybe even replacing it. Same with the primary FET.

    What is the inductance you are getting between pins 1 and 2 of T1 on your PCB? Can you measure with an LCR meter? 

  • thanks for your inputs Darwin. 

    T1 Pin 1 & 2 inductance was 103uH, 4&5 - 34uH, 7&10 - 33uH.

    couple of more observations

    1. we air wired to sense resistor, didnt help, we see output voltage as 2.0xxx volts

    2. we replaced R62 to 100k instead of 80.6k (BLNK), thinking that there might be poweron surge, it didnt help

    3. we removed the T1 and powered up with adaptor, the GAT pulses were same as what we observed earlier, 

    4. we removed both Q7 (primary, Si4848DY), Q2 (Si2325D5) and found that controller was generating the GAT, GAT2 pulses cleanly. we tried replacing a new Mosfet, Q2, Q7 but didnt help

    is Q7 mosfet loading CS ?,  what do we conclude ? any fix, you can suggest

  • Hi Venu,

    What is your company email? Let's work this offline.

  • Hi Darwin, 

    venu.k@yashaswinies.com.

    Appreciate a quick response.

  • we found the issue, we removed the R66 (1k) resistor in series with the CS and Gate source. we see output to be around 11.87v.
    can you pls clarify why this was loading the controller ?, as per EVM-383, this resistor is there..is this a something modified in the EVM and not reflected in schematics ?.. pls clarify
  • Hi Ven,
    I thought you tested air wiring from the CS pin to the sense resistor? Did you replace 1k ohms with a 0ohm?

    I checked the latest REV of the EVM, it is populated and EVM is working solution.

    Is the 47pF populated on your board? Or could be a higher value? I've seen in the past where 1uF was placed on CS and the output couldn't startup.
  • we resolved the issue, we removed the R66 (1k) resistor in series with the CS and Gate source. we see output to be around 11.87v.
    can you pls clarify why this was loading the controller ?, as per EVM-383, this resistor is there.. pls clarify whether this a a schematic issue in EVM ?
  • we will check with 0ohm. we have checked 47pF, it looks correct, anyway i will verify it again...
    how is the overall current for our design computed, do we need add R66 + R67 for the overall current ?, as this seems to be our issue...
  • Hi Venu,

    Please clarify what you mean by "we resolved the issue, we removed the R66 (1k) resistor in series with the CS ". This could not mean that it's open circuited. Does this mean you shorted R66? Or did air wiring it work?

    In general, the slope compensation is a resistor is sized on bench setup. Typically, you operate at your lowest operating input voltage and highest output power. With a pot you vary the R66 and check gate pulse waveforms for a consistent duty cycle.

  • Hi Darwin,
    i meant resolved, as we could see the output voltage... but still the question remains as to how this is working as R66 is open, we havenot shorted it.. we have 1 board working without R66 resistor, this board there is no path from Mosfet source to CS pin, CS Pin has only C46 capacitor ...
    we have tried replacing 0ohm on another board but still we DONOT see GAT pulses, GAT2 is getting generated. this is really strange, we have removed both the mosfets and R65 gate resistor to check whether this is causing any load, but very strangely we see the GAT pulses are very narrow as what we captured earlier and sent to you. it is not wide enough. how do we check what is loading the CS pin ?, we donot see any shorts in the board between the controller pins .. pls suggest
  • Hi Darwin,
    we added a POT in place of R66 (1K) and tried changing the values, we didnt notice any change in GAT pulses. Yes we have checked the C46 (47pf), is correctly assembled. we need to know, how is Vslope_D assumed ? as per EVM it looks like it is 255mv, as per equation in datasheet pg 29, Current Slope Compensation. we dont know whether we should increase or decrease R66 value. pls provide details on Current Slope compensation, looks like this may be the cause of our issue.
  • Hi Venu;

    I have a same problem. I use TP-LINK PoE Injector as a psu (It is not a class 4 device).

    I changed Rcls to 1270R from 63.4R (class 4 to class 0). But it did not work either.

    I wonder how it fixed.

  • There is nothing wrong about the TI Schematics. My problem was resolved.

    (I had very tiny short between DT and Blink resistors)

    Please double check your PCB design.
  • thats good. we have checked but there are no shorts in controller. we suspect Bias Supply circuit. we have used 1mh Coilcraft and Transfer from Wurth(part number). its very strange we have a working board with 1k resistor removed. Darwin should suggest what could be the issue
  • By the way, I used 750311320 (Midcom / Würth Electronic) as a main transformer.