UC2625-EP: BLDC motor RPM Variation issue

Hi Booshan,

We will look into your question and aim to provide a response by the end of next week. 

Regards,

Anthony Lodi

Hello Anthony,

Thanks for your response. Looking forward for your answers ASAP.

Regards,

Booshan S

Hi Anthony,

Any updates?

Regards,

Booshan S

Hi Booshan,

Sorry for the delay in responding. Could you email me your schematic? It is difficult to understand what could be going on without seeing the complete feedback loop for the circuit. 

Have you tried an ABA swap to make sure that the issue follows the part and not the PCB? I will check with the team on the lot trace code to see if there is any information that can be discovered from that.

Regards,

Anthony Lodi

Hi Anthony,

The UC2625-EP circuit section as shown in below image.

/resized-image/__size/4288x3240/__key/communityserver-discussions-components-files/38/UC2625-_2D00_-Schematic.png

also we performed the ABA swap of UC2625-EP chip with working (98Z05GV batch) board & Non-working (34Z01GV batch) board. With 98Z05GV batch chip, the Non-working board passed the acceptance criteria.

Regards,

Booshan S

Hi Booshan,

Thanks for the schematic and the confirmation on the ABA swap! 

I will aim to provide an update by middle of next week as I am travelling this week at a conference.

Regards,

Anthony Lodi

Hi Anthony,

Waiting for your response.

Regards,

Booshan S

Hi Booshan,

The team member that is experienced to look into the lot trace code just got back in office today, I will check with him and provide a timeline by the end of the day tomorrow on when he is able to review it.

Regards,

Anthony Lodi

Hi Booshan,

The team is looking into the 2 lots to see if there is any noticeable differences between the 2. I hope to have some additional information by the middle of next week.

Regards,

Anthony Lodi

Hi Booshan,

From the data that we have available, lot 34Z01GV didn't appear to have any obvious yield abnormalities. The team is working to pull the data for lot 98Z05GV and will look further to see if they can find anything. I will aim to provide an update next week.

Regards,

Anthony Lodi

Hi Anthony,

Thanks for your response.

How the 34Z01GV batches were verified for abnormalities?

As mentioned earlier in this post why the signal difference's observed between the working & non-working boards even though the command inputs are same. kindly comment.

Regards,

Booshan S

Hi Booshan,

What I mean is that there was no unusually high yield loss for batch 34Z01GV.

I am not yet certain that the issue is in the error amplifier, since in the waveforms you provided, you measured EA IN+, EA IN-, and EA OUT. EA OUT is not as much a function of EAIN-, but rather more of a function of V_{SPEED_ACTUAL} (shown in the schematic below) and V_{SPEED_SET}. More specifically, EA OUT = -3.2*(V_{SPEED_ACTUAL} - V_{SPEED_SET}) + V_{SPEED_SET}. 

Would you be able to provide a waveform showing V_{SPEED_SET}, V_{SPEED_ACTUAL}, EA OUT, and TACH_OUT? As shown in the above schematic, V_{SPEED_ACTUAL} is the output of U4.

Another important thing I saw in the schematic is that the R value for RC_BRAKE (8.06k ohm) is lower then the minimum recommended R value (10k). The C value is also at the highest end of the recommended range, and as the datasheet mentions, the best accuracy and stability are achieved with values in
the centers of the recommended ranges. Can you adjust the R and C values to be closer to the center of the recommended ranges and see if this improves the performance? It appears that the values you are using could be resulting in lower accuracy and stability which would ultimately effect the perceived speed of the motor since the TACH_OUT on time is determined by this. 

The green dot in the below graph shows approximately your current RT and CT values (resulting in an on time of around 540us) , which is outside the recommended RT value and on the edge of the recommended CT value. I would recommend moving the values closer to the center (possibly around the location of the red dot) to improve stability and accuracy. You can still achieve the same desired 540us on time with more stable R and C values. You might try a C value of 10nF and an R value of 80.6k ohms.

I would be interested in seeing a waveform measuring the fall time of the RC Brake output on one board vs the other board. I suspect that you will find that the fall time on one is longer than the other (one closer to 540us and one farther away), which could be the root cause of the issue since this would result in the TACH OUT average voltage changing between boards leading to the slower RPM.

Regards,

Anthony Lodi

Hi Anthony,

We will provide the waveforms of V_{SPEED_SET}, V_{SPEED_ACTUAL}, EA OUT, and TACH_OUT as soon as possible when the testing completes.

As part of root cause analysis the RT values were modified & tested, that doesn't help in terms of RPM difference's. also the same values of RT & CT (as per schematic) is working in almost 50% of the boards. But still will give a try to your recommendation RT & CT values & will share the results.

Even though the RT value is marginally outside the datasheet recommendations still the RT value is higher than the internal series resistor value (2k) which will not create current loading onto the chip.

The RC Brake signal also will be captured with rise time, fall time & shared shortly.

Regards,

Booshan S

Hi Booshan,

Thanks for your comments, I look forward to the test results.

Regards,

Anthony Lodi

Hi Anthony,

To capture the above waveform will take some more time & will share you the results as soon as possible.

did you able to look into the batch differences? and also can you share the details about the manufactured location or other batch difference details?

What is your comments/thoughts onto the other signal difference's which i have mentioned earlier & also any other thoughts onto the RPM variation cause?

Regards,

Booshan S

Booshan,

Anthony is out of office Thursday/Friday of this week so we will respond early next week, hopefully by Tuesday.

Regards,

-Adam

Thanks for the response Adam.

Hi Anthony,

Hope you are back. Any further inputs or updates onto the batch differences?

Regards,

Booshan S

Hi Booshan,

Looking forward to the waveforms! 

Regarding the batch differences, I just followed up yesterday with the team to check for an update, I will hopefully have an update from them by the end of the week. At this point I don't have any other things on the RPM variation cause, I hope to get some more insight once I get your waveforms.

Regards,

Anthony 

Hi Anthony,

The waveforms V_{SPEED_SET}, V_{SPEED_ACTUAL}, EA OUT, and TACH_OUT are captured for both working and non working boards. The snippets are attached for the same.

Working Board:

/resized-image/__size/800x400/__key/communityserver-discussions-components-files/38/Working-Board_2D00_2.png

/resized-image/__size/800x400/__key/communityserver-discussions-components-files/38/Working-Board_2D00_3.png

Non working Board:

/resized-image/__size/800x400/__key/communityserver-discussions-components-files/38/Non-working-Board_2D00_2.png

/resized-image/__size/800x400/__key/communityserver-discussions-components-files/38/Non-working-Board_2D00_3.png

Note:- 

Channel 1: V_{SPEED_ACTUAL}

Channel 2: V_{SPEED_SET}

Channel 3: EA OUT

Channel 4: TACH_OUT

Also the RT and CT values of the RC_Brake was modified to 80.6k ohms & 10nF respectively. The RPM of the Non working board further decreased by 500+ counts from the previously failed value. 

Required RPM: 4800

Non working Board RPM: 4500 (approx.)

Non working board after RT - 80.6k ohms & CT  - 10nF (RC_Brake) mod: 3960 RPM (approx.)

Kindly check & provide feedback. 

Regards,

Booshan S

Hi Booshan,

Thanks for the data! I will look over this in more detail and hope to provide more comments next week. For the RT and CT value modification based on my recommendation, Did you do the modification on both a non-working board and a working board? I would like to confirm that there is not a similar drop in RPM on the good board if the RT and CT values are changed to 80.6k ohms and 10nF. If there is not a similar drop in speed on a good board when using the new RT and CT values, then that could indicate that the behavior is related to the TACH_OUT circuitry.

Regards,

Anthony Lodi

Hi Booshan,

I took a look over the waveforms and I noticed that the VSPEED_SET was different for the non-working board (2.37V) vs the working board (2.33V). Would you be able to repeat the test showing VSPEED_ACTUAL, VSPEED_SET, TACH_OUT, and EA OUT with identical VSPEED_SET voltages? That will help me understand how TACH_OUT frequency and VSPEED_ACTUAL vary between a good board and a bad board under the exact same speed reference. It is also helpful to take the measurements once the motor speed has stabilized. For measuring the TACH_OUT frequency, for best resolution it would be good to zoom in to show 2-3 pulses in 1 scope shot so that I can see the frequency with more precision. Also measuring the TACH_OUT high time with the cursors for the good board and bad board would also be helpful to see if it is aligning with the on time set by the RC values (see below waveforms, in between red marks). 

Is it possible to decrease the volts/division from 500mV to 250mV to provide even more granularity for measurement? That helps better read the change in voltages that are occurring especially when calculating the expected EA_OUT from the VSPEED_SET and VSPEED_ACTUAL. If 250mV goes off the screen then maybe 300mV/div could work?

is 2.37V for VSPEED_SET the voltage reference that is needed to get the desired speed on the bad board?

Regards,

Anthony Lodi

Hi Anthony,

Thanks for the response.

The RT and CT values were modified in the working board to 80.6k ohms and 10nF. The RPM decreased in a similar trend to that of non working board & will not work for our application and details as follows:-

Required RPM: 4800

Working Board RPM: 4840 (approx.)

Working board after RT - 80.6k ohms & CT  - 10nF (RC_Brake) mod: 4300 RPM (approx.)

we also have verified the VSPEED_SET tolerances between the multiple working and Non-working boards. The VSPEED_SET voltage levels are 2.26 VDC ± 10 mV where 10mV = 20 - 25 RPM (approx.). This calculated tolerance RPM is well within our acceptable limits. The waveforms which was attached earlier has the voltage difference's of 2.37 VDC or 2.33 VDC due to the measurements offsets & tolerances of the oscilloscope. Also based on Timescale the Vmax value has minor variation.

To rule out the measurement difference's on VSPEED_SET. The Pin: 1 of the UC2625-EP was disconnected from the internal circuit & a external precision DC source was connected to provide VSPEED_SET of 2.260 VDC. The RPM value of working board & non-working board was verified for repeatability for old RPM values with internal VSPEEDSET supplied before capturing waveforms. with this configuration the below waveforms are captured.

Based on our application and test results the TACH_OUT is varying with a tolerance's of 940 ± 30 Hz (approx.) for working board after stabilization of motor. In the captured waveforms the zoomed TACH_OUT frequency will be at any instance of 940 ± 30 Hz(approx.). This TACH_OUT frequency band is almost 40 - 50Hz lesser for Non working board.

Working Board:

/resized-image/__size/800x480/__key/communityserver-discussions-components-files/38/Working-Board_2D00_4.png

/resized-image/__size/800x480/__key/communityserver-discussions-components-files/38/Working-Board_2D00_5.png

/resized-image/__size/800x480/__key/communityserver-discussions-components-files/38/Working-Board_2D00_6.png

Non working Board:

/resized-image/__size/800x480/__key/communityserver-discussions-components-files/38/Non-working-Board_2D00_4.png

/resized-image/__size/800x480/__key/communityserver-discussions-components-files/38/Non-working-Board_2D00_5.png

/resized-image/__size/800x480/__key/communityserver-discussions-components-files/38/Non-working-Board_2D00_6.png

Regards,

Booshan S

Hi Booshan,

Thanks for the additional information! Let me look over this in more detail and hope to provide a response next week.

Regards,

Anthony Lodi