• State TI Thinks Resolved
  • Locked Locked
  • Replies 8 replies
  • Answers 4 answers
  • Subscribers 63 subscribers
  • Views 1254 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

TPS92692: TPS 92692 spreadsheet values v TPS92691 Webench

Philip Hodgson23
Prodigy 250 points
Part Number: TPS92692
Other Parts Discussed in Thread: TPS92691,

I'm having some trouble with the TPS92692 tripping the fault detection, and I've discovered that there's a big difference between the source current monitoring resistor for the TPS92692 in the the spreadsheet and the same resistor for the TPS92691 in Webench. I know the threshold for overcurrent detection is higher in the TPS92691 (525mV, compared with 250mV for the TPS92692), but I wouldnt have expected the two values to be so far apart - the spreadsheet gives 90mohm for the TPS92692, and Webench gives 8mohm for the TPS92691. When I used 90mohm in my circuit, it wouldn't run. I was getting a brief flash every 50ms or so. I dropped the resistor to 25mohm, and it runs much better, but is still hitting the over-current trip at supply voltages below 9.25V. Anything less than 25mohm fries the MOSFET. The supply voltage range entered into the spreadsheet and Webench was 8 - 30V. The load is a string of 18 white LEDs, running at 500mA. Am I asking too much of the circuit?

  • 0
    TI__Mastermind 25500 points

    Hello Philip,

    I am not surprised the 91 and 92 would have different values since the thresholds are different and the 92 has adjustable slope. I am not sure about the webench discrepancies however. Webench is a good starting point, but it is still a model so I would double check values against the datasheet equations.

    To answer your question, no, you are not asking too much. These devices are capable of the power level you are trying to reach if designed correctly. If you are blowing up FETs with lower values then either the FET is running away thermally (due to high Rds(on) and/or high Fsw) or you are reaching the inductor saturation current rating at lower VIN. I would check the inductor rating first and if it is sufficient you may want to lower Fsw or select another FET.

    Regards,

    Clint

  • 0 in reply to Clinton Jensen
    Prodigy 250 points
    I'm not surprised that they're different, but I am surprised at the magnitude of the difference, unless I'm misunderstanding the datasheet. The spreadsheet gives 90milliohms for the 92, which has a threshold of 250mV, and Webench gives 8milliohms for the 91, which has a threshold of 525mV. The difference is more than 10 times, with the more sensitive 92 requiring the higher resistance.

    My inductor is a Bourns SRR1260A 22µH, which has a saturation current (at the point where the inductance drops by 25%) of 3.7A. The spreadsheet calculates a maximum inductor current of 2.78A. I checked the values in the spreadsheet against the calculations in the datasheet, and they are very similar.

    The MOSFET is a Vishay SQJA68EP-T1-GE3, which has an Rds of 117milliohms with a 4.5V gate drive. Probably not the very best, but it has to be automotive spec, which restricts the choice somewhat, and its total switching and conductive losses are 2.11W, which are well within the device's permissible dissipation. Although I don't know the thermal resistance of the board exactly, we have a reasonable amount of copper on the back of the board (140 sq mm) thermally connected to the mounting pad using an array of thermal vias, as well as an area of copper on the top side. I can reduce the switching losses by dropping the frequency, which I will try, but would this not allow the inductor current to rise to a higher value, as the switch will be on for longer?
  • 0 in reply to Philip Hodgson23
    TI__Mastermind 25500 points

    Hello Philip,

    I'm still not sure about webench. Perhaps it is choosing the value based on the current rating of the inductor it chooses? It's hard to say, but bugs are not unheard of in the models.

    Are you basing all of your equations on the minimum input voltage? I think your current and power dissipation will be higher than mentioned above. I don't have all of the exact specs such as the Vf of the LEDs used. But as an example, a string of 18 LEDs will be approaching 60V so your output is about 30W? The efficiency will also likely be <90% with such a high duty cycle and high Rds(on) switch.

    If I assume 30W and 90% efficiency then my input power will be 30W/0.9 which is 33.3W. So at 10V input your average inductor current will be 3.33A and the peak will be some higher value depending on your switching frequency. If you reduce Vin to 9V then your average inductor current is already at 3.7A and the peak is well above the saturation current rating.

    I would double check how you are doing the calculations, if my assumptions are even close that is. If not then let me know. You may also want to consider trying the FET used on the EVM as its Rds(on) is much lower so you could really cut down on power dissipation. It is Q grade.

    Regards,

    Clint

  • 0 in reply to Clinton Jensen
    Prodigy 250 points

    Hi Clint

    I think the bug is in the spreadsheet, rather than Webench (although Webench only supports the TPS9291). Webench chooses an 8.7A Isat inductor - the spreadsheet claims a peak current of 2.78A, so I chose a 3.7A one to give me some margin. When I did the calculation using the formula in the datasheet, I got a peak current of 3.87A, which explains the melting MOSFETs when I turn the volts down. I think I may have to  write my own spreadsheet - we'll be using more of these parts, I suspect. I've ordered some MSS1210 parts, so I should know if it's fixed by tomorrow.

    Regards

    Phil

  • 0 in reply to Philip Hodgson23
    Prodigy 250 points

    Hi Clint

    With a 6.7A inductor, it works fine at full output, regardless of the input voltage, so it looks as though the inductor wasn't man enough for the job. However, when I turn on the PWM ( by connecting 1.2V to the DIM/PWM pin), I get a mixture of reset and PWM on the output, and the lamp flickers horribly, instead of dropping to 10% LED output. I can see spikes on the IS pin at the point when the PWM FET is turned off. These spikes exceed the 250mV limit. It seems as though the system is taking a long time to reach full output, and is getting interrupted by the PWM signal. Do you have any idea why I should be getting spikes when the PWM FET switches off?

  • 0 in reply to Philip Hodgson23
    TI__Mastermind 25500 points

    Hello Philip,

    There are a lot of transients involved during PWM dimming and they can lead to the need to adjust some things. What PWM frequency are you using? How much output capacitance are you using? The first thing to make sure of is that you have a good amount of output capacitance so that it holds its charge well each off cycle.

    The next thing would be to try and increase the error amp bandwidth by adjusting the COMP values and/or trying to lower COMP just a bit each cycle. The second thing can be accomplished by connecting a high value resistor from COMP to ground (perhaps 300k to 1M).

    Sometimes in the end you need to reduce the IS resistor value further.

    Regards,

    Clint

  • 0 in reply to Clinton Jensen
    Prodigy 250 points

    Hi Clinton

    We have four 4.7µF 100V ceramic capacitors in parallel on the output of the switcher, before the LED current sense resistor. After the p-MOSFET there's another 100nF.

    We are running the PWM at 240Hz.

    I will have a go at increasing the amplifier bandwidth as per your suggestion. I have noticed that the evaluation board also flickers badly when the supply is reduced below 8.5V when running at 350mA output, so is it possible that the same mechanism is causing the flickering here too?

    Regards

    Phil Hodgson

  • 0 in reply to Philip Hodgson23
    TI__Mastermind 25500 points

    Hello Phil,

    It is likely it is the same issue with the EVM. It depends on the output voltage at 350mA. The EVM is designed for a maximum of 25W. Even then if you are PWM dimming the peak switch currents will be higher. So at lower input voltages while PWM dimming you may need to increase the bandwidth and/or decrease the IS resistor on the EVM. You can see some of this in the graphs where the minimum input voltage is limited even during normal operation if the LED stack and current are high enough.

    Regards,

    Clint