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.

TPS92640: Reference design values and calculated values do not match

Expert 1350 points
Part Number: TPS92640
Other Parts Discussed in Thread: TPS92641, LM4040

Hello,

On an earlier post of mine Irwin had suggested that I use the TPS92640 or the TPS92641 part for a 10A driver that I'm trying to build. This was a good suggestion since this part seems to meet all my specs. I was looking at the 20A reference design using the TPS92641 and I plugged in the values from this to the design calculations in the data sheet and the values do not seem to match.

For e.g. with Con = 330p the calculated Ron value is 25.25k. However the value selected is 18.2k. Even with standard 1% tolerance values that is a lot of difference. With an Rvout2 of 100K the Rvout1 calculates to 104k but a 150k resistor is selected instead. Also the inductor selected in the design is a 1uH/30A one and this equates to a peak-to-peak inductor current swing of close to 11A. Is this correct? Is there a different calculator that I need to refer to?

  • Hello,

    The calculator should be fine. I would use the values that calculate the way you want your board to work. Note that the Vout resistors are also part of the switching frequency calculation.

    With 12V input and 3V output I calculate 7.5 Apkpk current at 300 KHz. Just using V = L*di/dt.
  • Thanks. So far here are the values I have:

    Rvout1    11.6K
    Rvout2    10K
    Con    1nF
    Ron    4.22K
    Rcs    25mR
    L    33uH
    Cout    100nF
    Cin    47uH
    Rudim1    100K
    Rudim2    14.7K
    Rudim3    17.8K
    MOSFET    FDMS86183

    The specs are here:

    Vin-min    11.1
    Vin(nom) 24
    Vin(max) 75
    Vled 5.2
    Iled 10
    Fsw 500KHz
    Vcs 0.2V
    ΔIled-pp 0.2A
    ΔVin-pp 1.5
    Vturn-on 10V
    Vhys 5V
    Efficiency (η)    0.85
    # of LEDs 1

    I calculated the MOSFET Pd as about 1.5W and conduction losses of about 4.5W. If I reduce the frequency, I'll need a bigger inductor and those at higher current values are really expensive. At the same time, I wanted to keep the ripple current as minimal as possible.

    Would this design work?

  • Looks like I messed up some of the units. Here are they are again:

    Rvout1 11.6K
    Rvout2 10K
    Con 1nF
    Ron 4.22K
    Rcs 25mR (I plan to have 4 x 100mR 1W in parallel since I have a bunch of those).
    L 33uH
    Cout 100nF
    Cin 47uF
    Rudim1 100K
    Rudim2 14.7K
    Rudim3 17.8K
    MOSFET FDMS86183

    Vin-min 11.1V
    Vin(nom) 24V
    Vin(max) 75V
    Vled 5.2V
    Iled 10A
    Fsw 500KHz
    Vcs 0.2V
    ΔIled-pp 0.2A
    ΔVin-pp 1.5V
    Vturn-on 10V
    Vhys 5V
    Efficiency (η) 0.85
    # of LEDs 1
  • Hello,

    Rvout1 11.6K
    Rvout2 10K
    Con 1nF
    Ron 4.22K
    Rcs 25mR (I plan to have 4 x 100mR 1W in parallel since I have a bunch of those).
    L 33uH
    Cout 100nF
    Cin 47uF
    Rudim1 100K
    Rudim2 14.7K
    Rudim3 17.8K
    MOSFET FDMS86183

    Vin-min 11.1V
    Vin(nom) 24V
    Vin(max) 75V
    Vled 5.2V
    Iled 10A
    Fsw 500KHz, I get 511 KHz, Ron at 4.32K gives 500 KHz
    Vcs 0.2V, If Rcs is 25 mohm current will be 8A, if you have Viadj at 2.0 volts, if you have it high then it will be close to 10A, 2.54V/10=0.254V
    ΔIled-pp 0.2A, I get 0.254A at nominal input
    ΔVin-pp 1.5V, I get about 0.2V peak to peak at low Vin, note this would be an ideal capacitor you don't want to use one electrolytic or one ceramic due to ripple current
    Vturn-on 10V, I get 9.956V
    Vhys 5V, I get turn-off at 4.94V, I would lower the hysteresis because you don't want it running this low something like 10V to 9V or 10V to 8V depending on what the input ripple looks like.
    Efficiency (η) 0.85
    # of LEDs 1

    I get about 3 watts conduction loss for 100% dutycycle which means it's going to be split between the two MOSFETs, 50% duty each would have 1.5W conduction loss (this is at high temperature and 7-8V gate drive).  Switching loss, I get about 2.7 watts for Vin at 75V, at Vin 24V I get 0.85 watts, this is per MOSFET without gate resistors.  I did not include Rgate but did use worst case gate resistance for the calculation.

    1)  I would reduce the UVLO hysteresis.

    2)  If Viadj is high output will be about 10A, at Viadj 2.0 it will be 8A.

    3)  MOSFETs will need good heatsinking to dissipate that much power.

    4) at 75 volts input 5.2 volts output and 500 KHz the minimum on-time will be less than the TPS92640 spec which means it will pulse skip.

    Regards,

  • Irwin Nederbragt said:
    Fsw 500KHz, I get 511 KHz, Ron at 4.32K gives 500 KHz

    Sure. I will update this.

    Irwin Nederbragt said:
    Vcs 0.2V, If Rcs is 25 mohm current will be 8A, if you have Viadj at 2.0 volts, if you have it high then it will be close to 10A, 2.54V/10=0.254V

    I plan to use a 2.5V voltage reference here for analog dimming. So yes it will be closed to 10A then.

    Irwin Nederbragt said:
    MOSFETs will need good heatsinking to dissipate that much power.

    Yes I was thinking about this. I really could have done without onboard heatsinks and I was hoping to use copper planes itself for the heatsinking.

    Irwin Nederbragt said:
    at 75 volts input 5.2 volts output and 500 KHz the minimum on-time will be less than the TPS92640 spec which means it will pulse skip.

    How did you calculate ton and toff?

    Thanks again for all the help on this. I really appreciate it.

  • Hello,

    duty cycle will be (Vout+Vcs)/Vin = 0.072

    ton will be dutycycle*1/Fsw = 144 nS

    toff will just be (1-dutycycle)*1/Fsw = 1.86 uS in this case.

    Regards,

  • Thank you.

  • Irwin Nederbragt said:
    Vhys 5V, I get turn-off at 4.94V, I would lower the hysteresis because you don't want it running this low something like 10V to 9V or 10V to 8V depending on what the input ripple looks like.
    Efficiency (η) 0.85

    Hi, do you mean lowering the Vhys value for calculation to 2V or increasing it to about 10V?

  • Hello,

    The easiest way to reduce the hysteresis is to make Rudim3 zero ohms.  The turn-on threshold will be 9.956V and the turn-off threshold will be 7.856V.  It will turn on at around 10V input and off at around 7.85V.  You can make the hysteresis lower by lowering the resistor divider values Rudim1 and Rudim2 if you want.

    Regards,

  • Hi,

    Thanks. The data sheet recommended the use of Rudim3 if PWM dimming was used. These are the values I had calculated: Rudim1 = 51K, Rudim2 = 7.5K and Rudim3 = 11.8K. This resulted in a Vhys of 3V. Would these values be appropriate?
  • Hello,

    Yes, that should be fine.

    Regards,

  • I appreciate all the assistance on this. Thank you.
  • Hello Irwin,

    I was able to put this together and do some testing but it does not seem to be working correctly. The driver is putting out 28A instead of the designed 10A. I've attached the schematic here and the specs and BOM is listed here. I'm not sure what seems to be wrong.

    A couple of notes about the schematic:

    1. C2, C3, C4, C5, C8 and C9 are simply footprint placeholders and are not actually populated. CIN1 and CIN2 are 100uF.
    2. Same goes for R6 and R7 where only R6 is populated and Cout1, 2, 3 and 4 where only Cout1 is populated.
    3. Not sure if the gate resistors are needed and hence I've chosen to place 0R resistors there.
    4. There is a LM4040 5V reference that is being stepped down to 0-2.5V at IADJ pin for analog output current control.

    Specs:

    • Vin-min 11.1V
    • Vin(nom) 24V
    • Vin(max) 60V
    • Vled 5.2V
    • Iled 10A
    • Fsw 500KHz
    • Vcs 0.2V
    • ΔIled-pp = ΔIL-pp 0.2V
    • ΔVin-pp 1.5V
    • Vturn-on 10V
    • Vhys 3V
    • Efficiency (η) 0.85
    • # of LEDs 1

     

    BOM:

    • Rvout1 11.5K
    • Rvout2 10K
    • Ron 4.3K
    • Rcs 25mR (100mR x 4 in parallel)
    • L 33uH
    • Rudim1 51K
    • Rudim2 7.5K
    • Rudim3 11.8K
    • Rf 1.5R
    • MOSFET TK22A10N1
    • Cout 100nF
    • Cin 47uF
    • Con 1nF
    • Ccomp 470nF
    • Cvcc 2.2uF
    • Cboot 470nF
    • Cvref 100nF
    • D1 BAS316
    • D2 FERD20S100SH

    PWM_CC_laserdriver_v9-alt.pdf

  • Hello,

    Can you look at some of the signals such as CS, Comp, Vout, Iout, Vin?
    Can you verify that the current sense resistors are actually 0.025 ohm X 4?
    Can you verify that the current measurement is accurate (can check with a lab supply shorted and see if it matches what it's putting out)?
    Viadj needs to be 2.5V to get 10A so 0.2 will only give 8A. Also check Viadj.
    If you have a potentiometer on Viadj does the output current change when you lower the resistance of the pot?
    Can you check to make sure the MOSFETs are connected correct, Left pin is gate, middle is the drain and right is the source?
    When the output is 28A what is the input voltage (measured)?

    Regards,
  • Hi,

    I can answer some of these right away. I'll come back with answers to signals on the pins you've suggested. Another point I missed mentioning earlier was. The heatsinks on the top two MOSFETs seem to be getting hot while the lower ones are stone cold. This probably suggests that the lower ones are not being turned on at all? 

    Can you verify that the current sense resistors are actually 0.025 ohm X 4? - These are actually 100mR x 4 in parallel. I don't have a milliohm meter so I'm not sure how to measure these.

    Can you verify that the current measurement is accurate (can check with a lab supply shorted and see if it matches what it's putting out)? - This was done using a clamp meter. The power supply is a 24V 15A brick. 

    Viadj needs to be 2.5V to get 10A so 0.2 will only give 8A. Also check Viadj. - Before connecting the load I set this to 2.5V using the potentiometer. 

    If you have a potentiometer on Viadj does the output current change when you lower the resistance of the pot? - No, it does not. 

    Can you check to make sure the MOSFETs are connected correct, Left pin is gate, middle is the drain and right is the source? - Yes this is connected correctly. In fact, this was the first thing I checked. 

  • Some more measured values:

    CS - 960mV
    COMP - 1.64V
    When Iout is 28A the VIN drops to 17.2V
    Vout is 5.5V
  • Hello,

    Sorry, I meant 0.100 ohms X4, just used the wrong number.

    Okay, if the potentiometer is not changing the current and CS is at 0.960 volts something is wrong.  Can you check if it is even switching?  It could be hitting the upper current limit and cycling but I think you'll find that it's just trying to be on and not regulating which would explain why the top MOSFETs are hot and the bottom  are not.  CS should be 0.25 volts if it is above it's not working.  I would check a few things, RT, the upper gate (this is a bit tough since it's floating) to see if it's almost always on.  Comp seems low for that high of a CS, makes me thing that the upper MOSFETs are damaged or connected in a way that they can be on when they are not supposed to be, I don't think I have a schematic but make sure the gate is connected to the correct pin.

    Regards,

  • Hi,

    Sorry about the delayed response. I'll try some of the troubleshooting methods you've suggested.