• State
  • Replies 3 replies
  • Subscribers 30 subscribers
  • Views 44 views
  • Users 0 members are here
Support feedback
Options
Options
Related

DRV8323: Jitter on Driver Output

Mathias Leptin
Prodigy 10 points
Part Number: DRV8323
Other Parts Discussed in Thread: DRV8334,

Tool/software:

Hello,

I am using the DRV8323S MOSFET driver for a motor controller. I see a noticeable jitter of around 70ns on the outputs of the driver IC, even when I use a fixed PWM (fixed frequency and duty-cycle) as input. 

This constant change in propagation delay causes some noise in the motor winding currents, which then induces mechanical noise in my application. I am using 48V bus voltage and 3-PWM mode. Most of the setting are at default. I have played with some of the driver's settings (change of dead time, driver strength, 6-PWM mode), but the change in the jitter was very little. 

I think the problem might come from the IC's internal clock, which is used to synchronize the input signals and generate internal timings. Unfortunately, the datasheet does not contain much information on this.

C1: Output H-Bridge of Phase A

C2: Output H-Bridge of Phase B

C3: Output H-Bridge of Phase C

C4: Input signal of all phases (all connected together)

When I searched for this phenomenon, I came across several forum posts describing similar things. For example this one:

https://e2e.ti.com/support/motor-drivers-group/motor-drivers/f/motor-drivers-forum/1467705/drv8353m-reducing-pwm-jitter-on-mosfet-output-causes-solutions

My questions now are:

1. is there a way to reduce this noise? (for example by disabling the internal synchronisation)

2. if not, is there a way to synchronize the driver's internal clock with my system clock to get a predictable behavior?

3. are there alternative driver ICs without this issue? 

Best regards,
Mathias

  • 0
    TI__Expert 4515 points

    Hi Mathias,

    1. Yes, I believe your assumption is correct that this is due to the internal input synchronization of the DRV. Sadly, this is not a feature that can be disabled here.

    2. There is no way to synchronize inputs with the internal clock as we don't have any FSYNC or readout pins for this clock available externally.

    3. You can look into the DRV8334, which has input referred synchronization which will mitigate the jitter seen here.

    You can see this from the DRV8334's register settings, it will be input-referred by default.

    Thanks,

    Joseph

  • 0 in reply to Joseph Ferri
    Prodigy 10 points

    Hi Joseph,

    thank you for the replay. The DRV8334 sounds interesting, but does this feature only effect how the dead time is "triggered" or does it also effect how the input-to-output propagation delay is controlled? So no internal synchronization with an internal clock? Because I also need a high relsoution PWM and the internal clock reduces my time resoluton and causes other timing issues. 

    Are there any information about the input noise of the internal current sense amplifier? Because I found out that the CSA of the DRV8323 has relatively high noise. 

    Best regards,
    Mathias

  • 0 in reply to Mathias Leptin
    TI__Expert 4515 points

    Hi Mathias, 

    Basically, on the DRV8334, the dead-time will be inserted once it sees an input go low. What this means is, the variation (jitter) on the output SHx nodes will be lower, as the inputs are much more consistent and easy to monitor than the GHx or GLx signals to insert dead time, as you can expect. 

    While it does not change the actual propagation delay, it will help to mitigate the jitter. For our automotive customers with rigorous system requirements, they much prefer this input-referred dead time insertion. I think you would find it much more accurate and it may even eliminate the noise in your motor. But I understand you are already using the 8323 so I know this change may be difficult.

    As for the CSA accuracy, the DRV8323 should perform to the specs outlined here in the table. If you see results from outside this table range, then I would guess that there is probably some noise on the SPx/SNx lines that would need to be reduced.

    Keep in mind that the accuracy will also drift across temperature, so if your device is getting hot, the CSA will be less accurate.

    Thank you,

    Joseph