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.

WEBENCH® Tools/TPS92640: Design Software for TPS92640

Part Number: TPS92640
Other Parts Discussed in Thread: LM3409, LM3409HV

Tool/software: WEBENCH® Design Tools

I want to use the TPS92640 for a new LED-Project (28V LED, 0.6A), because of the very good analog dimming range. Unfortunatelly, i didn't find a webbench application for the device. Is there a software-development tool for calculating the TPS92640?

Thanks in advance,

best regards

Norbert

  • Hello Norbert,

    I am not aware of any design calculators for this device. The design guide in the datasheet is pretty easy to follow however. If you get stuck on anything feel free to post here and I would be happy to help. Thanks.

    Regards,

    Clint

  • Hello Clint,
    thank you for the offer to help.
    I'll first try with a LM3409 design, which i did before. I have some boards form this design left, so i can easily do some tests.
    If this it not good enough, i will come back for using the TPS92640.
    Best regards
    Norbert
  • Hello Norbert,

    That sounds good. The LM3409 is also a very capable device.

    Regards,

    Clint

  • Hello Clint,

    now i did my sample circuits with the LM3409HV with analog dimming. The dimming range is very fine.

    It works quite good - with a few issues:

    First i killed the LM3409, when i put the 48V input voltage hard in the lab power supply, while the Iadj Pin was connected to another lab ps with about 0,5V on the pin. At this time R8 was 0 Ohm and R7 was not populated. When i first put 48V on, and then voltage on Iadj, everything is ok. But nevertheless it's not clear for me, why LM3409 can be killed when voltage is on Iadj pin, before main power is connected?

    Second issue is, that i have radio noise form the circuit (DAB-Radio is disturbed). When i put my oszilloskop on the gate-pin (6) an on gnd, the noise is off. Do you have any suggestions, how i can solve this practical? (can't sell oszi with every circuit ;-))

    To get a small temperatur rising on the circuit, i took a big inductance (2x 68uH in series). Is there another possibility to lower the temperature of the circuit? (since circuit is mounted in LED-case with 50°C ambient). Can this big inductor be a issue for bad regulation (LED-ripple)?

    I attached circuit:

    and schematic:

    Thanks in advance,

    best regards

    Norbert

  • Hello Norbert,

    I checked with the IC designer and it looks like having a voltage on IADJ should not hurt anything. We do recommend the 1k resistance externally just for extra protection but that is unlikely the cause of the damage. I would be more concerned with the input voltage transients caused by applying the input voltage hard. You have to be extra careful when you have an inductor (L2) on the input. If you are applying the voltage fast enough you can easily cause overshoots well above the abs max VIN rating of the device. Often times the internal SCR latch used for ESD protection can protect it through some cycles but will weaken over time and fail eventually. If this is a condition you expect it may be a good idea to use a TVS to protect VIN.

    Radio noise is a tricky EMI issue. It is hard to diagnose or fix without an actual circuit, but it seems like slowing the gate down a little seems to help when you are applying the probe. So it might require some sort of EMI mitigation within the circuit or filtering on the input/output of the circuit. But it could simply be that the layout isn't ideal and you are creating larger current/return current loops than you should. A good layout helps a lot with EMI.

    A large inductor can be an issue, but only if it is large enough so that you get less than 24mV of ripple on the current sense resistor. I don't think you are at that limit with your switching frequencies. Higher inductor values (given the same case size) have higher ESR though and the 0.44ohm specified for those is pretty high and will generate heat. Is there a reason you need to go this high? You can get the same LED current ripple with less L and more C on the output. You would get significantly less heating with a lower L value and you could also just use one inductor. It's is also possible that there is mutual heating between the FET, diode, and inductors. So you could reduce heating by using a FET with a lower Rds(on) assuming it doesn't have significantly more gate charge.

    I hope some of that helps.

    Regards,

    Clint

  • Hello Clint,
    thank you for your hints.
    As you can see in the circuit diagram there is a TVS-Diode fitted (P4SMA56CA, D2).... This is a bidirectional one, only symbol is wrong.

    I did some layouts with LM3409 before, most equal component placement - there was no EMI problem. But switching frequency was very different (much lower). Could a small capacitor from gate to gnd be a solution? Or will this cause other problems?

    I selected the big inductor value for keeping the temperature of the LM3409 lower. (extreme values: 50Khz switching frequency will get a Tj of 37°C and 1MHz switching frequency will get a Tj of 79°C). This is because my circuit is fitted in the LED-housing with an Tambient of 50°C.
    The FET i used is the best compromise i found for Rds(on) and GateCharge... Do you know a better one (from TI?)

    Best regards
    Norbert
  • Hi Norbert,

    I see the TVS now. The only problem is that it will not protect the LM3409 where it is placed. The overshoot would be caused by the input capacitors and input inductor which are after the TVS. But it is possible there may be some other transient going on that I'm not aware of with IADJ (getting pulled below ground due to the fast input transient or something), but populating those resistors would help with things like that.

    EMI does tend to get worse as the frequency goes up, particularly if the grounding isn't solid and there is some distance (or interruptions) between the input cap grounds and the output diode ground. I can look over your layout if you would like to provide layer plots. It's likely your issue if you are radiating that much noise, but that is up to you. You can use a cap from the gate to ground or the gate to VIN and/or add a gate resistor to slow the turn on/off of the FET if that helps. The only thing it would do would be to cause the FET to run a little warmer.

    So I am not sure how using higher inductor values is supposed to help with the LM3409 IC temperature, could you explain this to me? As far as I can tell the relatively high resistance introduced by those two FETs in series (almost 1ohm) will actually result in the duty cycle increasing to compensate for the losses. The result is the FET is on for a longer time and will get hotter as a result but it shouldn't affect the LM3409 thermally at all. As for TI FETs I don't believe we have any PFETs with high enough voltage. We have great NFETs though not that that helps here.

    Regards,

    Clint

  • Hello Clint,

    thank you for your kind support.

    You are right with the TVS - i'll place a second one behind the inductor.

    Here's the layout and schematic:

    As i calculated the circuit in webbench, for a switching frequency of 50KHz a Tj of 37°C and for a switching frequency of 1MHz a Tj of 79°C is given for the LM3409. Therefore i think, a bigger inductor => lower switching frequency => lower LM3409 temperature? Am i right?

    As mentioned before, this is because my circuit is fitted in the LED-housing with an Tambient of 50°C.

    Best regards

    Norbert

  • Hello Norbert,

    Your layout isn't too bad, but it isn't ideal so I could see how it could radiate noise. Refer to my sloppy paint attachment to see what I mean. I think it would be ok if you just filled all free space on the bottom layer with a ground plane and put vias next to the ground of the input caps, rectifier diode, and beneath the IC in the DAP (which will also keep the device much cooler as it uses the ground plane for heatsinking).

    Otherwise you are correct, the lower the switching frequency the less VCC current draw there is so the device will run cooler. The FET will run cooler as well since the AC losses in it are directly proportional to switching frequency. You will get a huge benefit for the IC over what you have now though if you do attach the DAP to a bottom side ground plane as mentioned.

    Regards,

    Clint

  • Hi Clint,
    thanks for information - sounds good.
    I'll put your suggestions in the second version of my layout.
    One last question: Would it make sense shortening the yellow path e.g. turning the freewheeling diode 90° an moving its anode nearer to LM3409 (below the FET)? Then i also could turn the right choke 90° and move the led-connection-pads more to the left?
    Thanks for your help.
    best regards
    Norbert
  • Hi Norbert,

    I don't think that would be necessary if you use a bottom layer ground plane and place vias where I suggested. I would re-route the EN trace maybe to the left side, even on top some if needed, so that it doesn't cut the bottom layer ground plane beneath all of the current carrying components. 

    Ideally the grounds of Cin, D, Rcs, and the IC are connected directly to the same plane on top and to a bottom layer ground plane. But your layout is pretty good so if you use a good ground plane and vias it should work really well and radiate less noise.

    Regards,

    Clint