DRV8701: Intermediate spikes on SO & SNSOUT lines

Hi Brad,

Thanks, please capture (and share what you deem necessary and possible) scope captures of GLx, GHx, SHx, and IN1 and IN2 signals at the time range you have shared. We need to determine if it comes from PWM, from the load, or from something in between.

Has a test been done with motor and without the load? Further variable isolation.

Hi Hector,

Additional captures...

GL1 Blue/GL2 Red full capture

GL1 Blue/GL2 Red 1st zoom:

GL1 Blue/GL2 Red 2nd zoom:

GL1 Blue/GL2 Red No Load:

GH1 Blue/ GH2 Red Start:

GH1 Blue/ GH2 Red Zoom 1:

GH1 Blue/ GH2 Red Zoom 2:

GH1 Blue/GH2 Red Zoom 3:

GH1 Blue/GH2 Red End Zoom:

GH1/GH2 No Load: Looks similar on/off characteristic as GL1/GL2 No Load.

SH1 Blue/SH2 Red Zoom 1

SH1 Blue/SH2 Red Zoom 2

SH1 Blue/SH2 Red End Zoom

SH1 Blue/SH2 Red No Load:

We are using the 8701E, so there is no IN1/IN2 signal.

Hi Hector,

Yes, the GH1/GH2 zoom 3 shows the additional "failed start" nicely.

1. It is my understanding that the 8701E handles the PWM internally, whereas the 8701P requires external PWM.

2. I'll grab that one next... should have it tomorrow.

3. Originally, we didn't have any current limiting using the VREF or SP inputs. We had large current surges during pump-start, which is what spurred on this change; no 100% PWM D.C.

HI Brad,

Thank you for the clarifications. If you are driving the DRV and depending on current regulation as you appear to be doing, the current spike stems from the load you utilize. There are no spikes when the load is removed, and I interpret load as the compressor and the motor is still connected. Current chopping is activating, and it is dependent on the value of VREF, which I think you know about.

All of this means that, the amount of current seen while the load is being driven is large and, when the fixed PWM on the Gate signals is goes high, SO passes the VREF value and current chopping activates. You can see that after every one of those "spikes", the off time is the Toff established in the datasheet (25 us).

It appears the application circuit can be designed to meet a VREF value small enough to handle the amount of current surge on pump start yet large enough to avoid the current chopping you see when driving the load.

Thanks Hector.

That makes sense, that it's the load causing the additional spikes. Our concern was that the intermediate spikes of a much smaller width may be indicative that we are driving something incorrectly. Do you believe this not to be the case, based on all of the signal traces? That it's just a characteristic of the load current draw?

Hi Brad,

The concern I see is in the capture of End Zoom, such as GH1 Blue/GH2 Red End Zoom. When I look at the signals of the explictly stated capture, and compare them to, say, GH1 Blue/GH2 Red Zoom 3, I wonder if there is a difference between both of these capture states that we should be aware of? The end zoom capture has negative spikes on GH1 Blue and the toff time is not met, so I am not certain what is the setup or what is happening right now.

Hi Brad,

See replies to points/questions below:

But I think I see your point... it appears that near the end of the run, the device stops it's internal PWM behaviour.

-> When you say end of the run, how is the run ended in the application?

That brings the question of whether the current regulation PWM mode specified in the data sheet Section 7.3.3 is the same sort of PWM behaviour you would expect when operating with the DRV8701P device.

-> It is. We have not seen behaviour to indicate otherwise.

It is interesting that it takes a bit of time at the beginning of each cycle before the PWM behaviour starts, as evident in "SH1 Blue/SH2 Red Zoom 2" and "GH1 Blue/ GH2 Red Zoom 2". Why is that?

-> It can the time it takes to turn on the regulation. I need to look further into the reason why it happens.

Looking at the frequency of operation in "GH1 Blue/GH2 Red Zoom 3", you can see two modes... if you look at the cycle time between a "large pulse" at 658us to the "small pulse" at 718us, you have a 16 kHz PWM, whereas from the "small pulse" at 718us to the next "large pulse" at 746us, it's closer to the design value of 35.7 kHz (f_pwm in the datasheet shows ~ 38 kHz). Is that because if the current limiter doesn't need to PWM, it will override and apply constant drive to the compressor?

-> On the 658 us to 718 us drive period, there was not enough SO voltage for it to pass the VREF voltage, so the PWM worked as designed. You can see the PWM off time is 25us.

-> On the 718 us to 746 us drive period, there was enough SO voltage for it to pass the VREF voltage, so the current regulation worked as designed: The high time of the gate signal is cut off due to current regulation. You can see the PWM off time is 25us as well.

Hi Hector,

See replies to points/questions below:

But I think I see your point... it appears that near the end of the run, the device stops it's internal PWM behaviour.

-> When you say end of the run, how is the run ended in the application?

--> A tank is filled to a specific pressure and then the EN pin is toggled

That brings the question of whether the current regulation PWM mode specified in the data sheet Section 7.3.3 is the same sort of PWM behaviour you would expect when operating with the DRV8701P device.

-> It is. We have not seen behaviour to indicate otherwise.

It is interesting that it takes a bit of time at the beginning of each cycle before the PWM behaviour starts, as evident in "SH1 Blue/SH2 Red Zoom 2" and "GH1 Blue/ GH2 Red Zoom 2". Why is that?

-> It can the time it takes to turn on the regulation. I need to look further into the reason why it happens.

--> Any update on this Hector?

Looking at the frequency of operation in "GH1 Blue/GH2 Red Zoom 3", you can see two modes... if you look at the cycle time between a "large pulse" at 658us to the "small pulse" at 718us, you have a 16 kHz PWM, whereas from the "small pulse" at 718us to the next "large pulse" at 746us, it's closer to the design value of 35.7 kHz (f_pwm in the datasheet shows ~ 38 kHz). Is that because if the current limiter doesn't need to PWM, it will override and apply constant drive to the compressor?

-> On the 658 us to 718 us drive period, there was not enough SO voltage for it to pass the VREF voltage, so the PWM worked as designed. You can see the PWM off time is 25us.

-> On the 718 us to 746 us drive period, there was enough SO voltage for it to pass the VREF voltage, so the current regulation worked as designed: The high time of the gate signal is cut off due to current regulation. You can see the PWM off time is 25us as well.

Hi Brad,

To clarify on the second point: the moment the gate signals turn on is the moment right after EN is toggled on in your circuit, is that correct? (using "SH1 Blue/SH2 Red Zoom 2" as the example)

On the third point, the concern is related to wether or not your system is designed to be pulling as much current as it does that SO trips VREF and causes those shortened drive periods due to current regulation kicking in. If you system is designed for the SO values you are seeing but want to limit the inrush current at the beginning, then the VREF value needs to be increased so SO does not trip VREF during drive time but it does trip it for inrush current to avoid the inrush spike.

Hi Hector:

To clarify on the second point: the moment the gate signals turn on is the moment right after EN is toggled on in your circuit, is that correct? (using "SH1 Blue/SH2 Red Zoom 2" as the example)

I do not have a trace overlaying SH on EN, but I presume so. It couldn't be off by more than ms though. Would a gap here be indicative of something?

On the third point, the concern is related to wether or not your system is designed to be pulling as much current as it does that SO trips VREF and causes those shortened drive periods due to current regulation kicking in. If you system is designed for the SO values you are seeing but want to limit the inrush current at the beginning, then the VREF value needs to be increased so SO does not trip VREF during drive time but it does trip it for inrush current to avoid the inrush spike.

This statement suggests that we should assess increasing our VREF to remove these short bursts. Is the reason for this because the decay in the t_off time is insufficient to bleed through the low-side FETs? I'm curious why it recovers after the next cycle though.

Hi Hector, 

I'm still hoping we can get some clarity on what's happening.

Here are captures for SH and EN overlaid at the start, middle and end of a compressor cycle. We had filtering disabled to ensure we didn't hide something telling, so our probes are likely picking up switching and 60 Hz noise as well.

In an effort to see if we could increase the size of the intermediate spike, we experimented with increasing the voltage reference (assuming that it was getting cut off because it went beyond the current surge threshold). Our threshold was originally set at 2.76 V for a surge of ~18 A. However it seemed that there was very little shift in the size of the intermediate spike even if we increased to 3.02 V. Also, the top of the pulse is much lower than the peak of the larger pulse on either side. We confirmed that there is no filtering on this scope capture of SO. (Vref  increased to 2.85V). We assumed that if we were hitting VREF with the intermediate spike, it would be obvious on the SO capture, but that doesn't appear to be the case. 

Any further ideas?

Brad

Hi Brad,

Can you explicitly state the questions and doubts you have? I see the doubt of whether or not SO trips VREF. Any other doubts?

I am looking into this point.

Hi Hector,

Sure, just looking to understand the behaviour.

I suppose it boils down to "Why is the h-bridge PWM pulse getting cut so short even though SO did not reach the threshold voltage?"

Brad.

Hi Hector,

Here are the captures you request...

SNSOUT with SO

SNSOUT with VREF

SNSOUT with SP

SNSOUT with SN

Hi Brad,

Please, keep me informed on your findings. I have a theory as to what might be happening, and the EVM is the next in the verification process.

Hi Brad,

Any updates on the EVM trial? 

If possible, please share design files via private message.

Hi Hector,

I'm checking on status of the EVM now. Will share files now as well.

Brad

Hi Brad,

Keep me posted. Checking design files now.

Hi Brad,

I verified the design files. Please, see the private response.

Do you have an update on the EVM testing?

Hi Hector,

EVM tests are underway, should have results shortly.

Brad

Hi Brad,

Please, inform me of what you decide will be the next steps and how you will move forward. Thank you.

Hi Brad,

Pinging you for an update here. Thanks.

Hi Hector,

Please find attached snapshots of the EVM board (SNSOUT vs SO signal lines). EVM board does not seem to have those "spikes" we observed on our design. Are there any other signals we should capture on the EVM board? 

Hi Akzhana,

Since the issue appears to only be ocurring on the customer board, no need to further probe the EVM.

It appears the scope is not capturing potential noise on VREF or SO or other lines. I suggest a trial of probing with a different scope to see if there is noise on customer PCB that has not been captured. If the noise is on SO or VREF or both, the trace(s) is/are cut on the PCB and blue wired to see if the noise reduces and if the problem goes away.

Layout improvements have been recommended in the private conversation with Brad.

Hi Hector,

Thanks for your response. Just wanted to clarify - 

If the noise is on SO or VREF or both, the trace(s) is/are cut on the PCB and blue wired to see if the noise reduces and if the problem goes away.

Are you suggesting cutting SO or VREF traces on our PCB and wire them to EVM and see if we are still seeing the noise? 

Thanks,

Akzhana

Hi Akzhana,

Apologies: I did not click the Reply button on the draft response I wrote here last week.

The suggestion is to cut the the traces with the noise the scope is possibly not capturing (after verifying the noise with another scope or voltage probe) and blue wiring the trace. All on the same board. No EVM in the equation.

Let me know if any further clarification is required.

Hi Akzhama,

Have you decided or proceeded with next steps? We can close this thread and open a new one once further questions come up. Keep me posted. Thank you very much.