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.

DRV8837: out-of-specs observed Rdson

Part Number: DRV8837

Hello,

We currently use the DRV8837 in the following manner : it is a secondary low-voltage driver for a high-power torque motor. Its purpose is to allow a finer control of the motor current, when the torque needed is low. The DRV8837 is meant to supply loop control currents up to 1 A, after which the main driver takes over and drives up to 10 A in the motor.

Since the main driver works with 28VDC, the two drivers need to be separated. This is done with 2 MOSFET transistors : they are passing when the DRV8837 is in use, and open when the main driver is operating, to avoid busting through the low voltage rail.

We have however noticed a very high power loss in the DRV8837 when powered with its supply rail, P3V3A, when P3V3A = 3,3V. So we changed the board a bit to have P3V3A = 5.3 V, so we can compare with the Rdson data in the datasheet.

The motor is 2.35 ohms, 13.8 mH, no back-emf.

The command supplied to the IN1/IN2 pins is a 70% duty cycle forward/brake command : IN1 = 3,3V, and IN2=3,3V 30% of the time, and 0V 70% of the time. The PWM frequency is 20 kHz.

Below is the OUT1 (yellow) and OUT2(blue) waveforms. When in forward, V(OUT1)=4,4 V & V(OUT2) = 0.55V. When in brake, V(OUT1) = -0.55V & V(OUT2) = 0.5V.

The average voltage at the motor terminals is then approx. 2.35V, which is confirmed by our current probe which is showing approx. 1 A.

Why are we seeing such a high Rdson on both the high side and low side ? The calcultated total Rdson (TOP+BOT) is 1,5 ohms !

We then tried the same command type (forward/brake) with 16% PWM and the results are :

  • V(out1) = 5.1V
  • V(out2) = 100 mV

This corresponds to a motor current of 340 mA, and a total Rdson (TOP+BOT) of almost 1 ohm.

Even if our PCB was badly designed (which it isn't), there is no way to make a +600 mohms power or ground trace without having the full intention to do so.

There is approx. 120 µF (V-bias and worst-case tolerance applied) on the P3V3A rail, and the DRV8837 is the ONLY IC on that rail.

V(out1) and V(out2) are measured directly at the output of the DRV8837, and we also measured V(HB1) and V(HB2) (after the separation transistors), to make sure they weren't at fault. We have V(OUT1)=V(HB1) and V(OUT2)=V(HB2) within a few 10 mV.

  • Hello!

    Quick question on the scope capture.  Output seems to be swinging below GND.  Is there some offset in this channel and output is actually higher if this offset is corrected?  

    One thing that can explain a higher than expected Rdson is thermal.  Have you ensured the powerpad is soldered to the PCB?  If you have a thermal camera or thermocouple, can you take a measurement of the case temp?  

    Regards,

    Ryan

  • About the scope capture : there is no offset, checked multiple times and calibrated each probe right before measurement. The probe GND is the net R3V3A. It swings below GND because it enters brake mode : low side of half bridge 2 is always ON, and half bridge 1 alternates between TOP and BOT being ON. So during brake, because of the motor's inductance, current stays somewhat the same and the current flows from R3V3A to the motor through the low side of half bridge 1. I don't think we are seeing a body diode drop (even though the value is similar), because the low-side of the half bridge 1 is supposed to be conducting. 

    This is further illustrated in the fact that in low current (16% PWM, 340 mA in motor), we have V(OUT1) in brake mode at approx. -200 mV (sorry I didn't mention that in the OP).

    This is the layout :

    It's far from perfect (too few vias in GND plane). I can't really ensure that the exposed pad is well soldered, but I trust the reflow process of my subcontractor. 

    In the low current (340 mA) case, we measured 45°C (Ta + 15°C) on the top of the case. We then inferred that at most the junction temperature would be 70°C given our estimation of power dissipation inside the chip. This would correspond to an increase of approx. 25% of the total Rdson according to the datasheet. Instead we have (seemingly) 200%. 

  • I actually don't see any thermal vias in the pattern under the IC.  So, that is the only copper in that square for dissipation?  Or maybe the vias aren't properly displaying?  

    Regarding brake, I am still confused by what I am seeing.  Only during switching should you see the output go below GND as the the output is in the HIZ state and current flows thru the body diode.  Can you plot inputs on the same scope capture as outputs so I can see if device is functioning properly.  

    Regards,

    Ryan

  • You're correct, there aren't any thermal vias under the pad, because we thought it would make reflow difficult, and changing the component nigh impossible. In hindsight, there should at least be 2 to 3 times more vias in close proximity to the ground pad to decrease the thermal resistance from pad to ground plane.

    Here are the scope readings of (1) a 12% command and (2) a 70% command. Signals from 1 to 4 (yellow, blue, pink, green) are : IN1, IN2, OUT1, OUT2. Note the different vertical scales.

    As you can see in (1), during brake (IN1=IN2=1), OUT1 is negative but far too close to 0V to be a body diode drop. This indicates, to me, that it is braking properly since in this mode both OUT1 and OUT2 are supposed to be L (not high-Z as you suggested, that would be Coast mode).

    This is more visible on the OUT2 change : we go from approx. 100 mV in (1) to 600 mV in (2). The motor is still the same, and doesn't heat up with 1 A (a few degrees, not enough to be significant). The calculated current using only the readings and the 2,35 ohms in the motor gives, for (2), 1,16 A : this would mean Rdson on the low side is some 500 mohms. This is far beyond what we would expect even for a junction temp of 100 or 150 °C.

    The voltage reference for these measurements is the GND pin of the DRV8837 (we soldered a wire on the exposed side).

  • Ok, this is what I would expect in terms of logic.  If you read my message above again, I wasn't implying that brake is HIZ...HIZ state occurs during deadtime and switching.

    Can you please check the part number on the IC?  Might be a long shot, but there is a pin-to-pin compatible IC called the DRV8837C.  This device has Rdson values similar to what you are seeing.

    https://www.ti.com/lit/ds/symlink/drv8837c.pdf

    Regards,

    Ryan

  • Replying to this since I can't reply to your last message.

    That was it, a simple mistake when ordering that passed through the checks because I was unaware of the existence of the "C" version. Had I known about it, I would have checked. Thank you very much, I am relieved but still vexated that this possibility didn't cross my mind.

    Have a nice day !

  • Great!  Took me some time to think about it as well...glad it is solved!  Will close this case.

    Have a nice day as well!