## The Martins' Method

Ask for and share advice on using the PicoScope kit to fix Electric Vehicles
ben.martins
Pico Staff Member
Posts: 317
Joined: Tue May 16, 2017 1:02 pm

### The Martins' Method

Hi all,

Hope we are all safe and well?

You might all be intrigued by the subject of this post. The reason for it is because I want to share a technique that I discovered whilst spending a lot of time researching better ways to diagnose 3 phase motors. I found that actually there is a very simply way we can do this and after sharing with the team here at Pico, Steve Smith kindly named it the Martins' Method after yours truly. So let's get started.

The Martins' Method drew upon existing math for plotting the relationship between current phases of 3 phase motors. Known as the Park Vector Approach it comprises two formulas -
• iD = (sqrt(2)/sqrt(3))A-(1/sqrt(6))B-(1/sqrt(6))C
• iQ = (1/sqrt(2))B-(1/sqrt(2))C
Where i = current and D and Q are just components for the formula.

These two formula's have to be set up correctly in the math channel wizard and have to take in account the range used in the channels and has too match. Other complications also arise from the fact that it's math which means if watching it live, there is a delay whilst the software draws the math from the raw data. The most important part of the Park Vector Approach is that you have to use XY to view the result. At time of writing XY hasn't yet been implemented into PS7 Automotive so the following images will all still be in PS6 Auto. Needless to say it is coming and will be awesome when it arrives but for the time being in order to use XY we will have to stay with PS6 Auto. The tests carried out the following Park Vector captures were done in controlled conditions using a 3 Phase motor powered by a VFD unit and not from a vehicle.
The first capture shows a 3 phase motor during startup and what we see is a spiral as the current increases. What we are seeing is the current relationship between the phases. This is further highlighted when we load the motor further.
This was great to see as it's so clear to see issues that might arise but the hurdle is the fact it's math. When PicoScope is capturing data in block mode, typically anything under 200ms/div, we do get an update at the time but shifting to streaming the math is created at the end of the buffer and waits till the next before is full before drawing again. Not only this but you have to reset the math for each motor as the current levels might be different. Not so much of a problem post capture but we want to view issues in real time. This got me thinking, what would the current phases look like if there compared directly with no maths.
It may not be a perfect circle but you can see the relationship between each of the phases relevant to another. We have V & W against U, U & W against V and finally U & V against X. The great thing about using XY is that it is drawn in real time. This means if you were to start a motor from rest and were streaming data at say 500ms/div, you will actually see the pattern being drawn on the screen. These type of waveform is more typically known as a lissajous pattern. They are typically used to calculate the phase difference between two sinusoidal signals having same frequency but I've yet to see them being used for motor analysis.

The use of XY can help to spot issues with motor balancing as the waveforms should sit on top of each other. Below is an example of when the current is different in a phase.
And also what happens when you lose a phase.
I know 3 phase motor testing is difficult on some vehicles and impossible on others, but where it is possible then applying the Martins' Method could offer some quick detail on the motors you are troubleshooting.

It doesn't stop there though as we can apply the Martins' Method to another component found on the majority of electrified vehicles, the resolver. You may have seen in our newsletter on Resolvers - https://www.picoauto.com/library/traini ... -unwrapped and right near the bottom I'd included an animation demonstrating how the two signals are used to determine position.
In the corner you can see the circle being drawn in, yes you guessed it, XY view. Whilst this isn't possible in PS7 we can do this in PS6 but we have the issue with math's again.
The nice thing about the math though is we have removed the excitation frequency to leave us with the actually signal the ECU is seeing. Details on this math can be found in the PS7 Guided Test electric vehicle section. This results in a perfect circle and is very clear when something goes wrong. We can though still apply the Martins' Method to the raw signal. As the signals are 90 degrees apart a circle will still be drawn but it just appears filled in!
An important note to make here is that you have to watch that the grid for the XY view is perfectly square. You can adjust the size by clicking the line between the two views and adjusting as you can get a more of an ellipse than a circle! This is important as when you have a bad one as in the one featured in Miki Diosy case study video - https://youtu.be/-0aPawW5vGg. Below is the capture of the resolver in XY view.
As you can see, it's quick to determine something is wrong and best of all you can draw a mask in XY which means a known good can be loaded, XY drawn and then carry out the capture to quickly determine an issue.

From here you may need to adjust the grid layout. This is made a lot simpler in PS7 but for PS6 you will need to click Views > Grid Layout > Custom Layout
Here you can amend the layout and I find 1 Row x 2 Columns works best for this type of view.

Once you have the XY view where you want you now need to select the channel you would like to put on the X axis. You can do this by right clicking on the XY grid select X-Axis and then the channel you would like to use. You can then hide any other channels not of interest which in the resolver file would be the Channel A and the math channel.
I hope this helps and I've attached a resolver capture which you can have a play with in PS6 and see if the Martins' Method could help with the diagnosis.

Kind regards

Ben Martins
Attachments
gt885-motor-resolver with amt.psdata
Last edited by ben.martins on Fri Apr 30, 2021 11:20 am, edited 1 time in total.

Jfixmycar
Newbie
Posts: 3
Joined: Tue Nov 01, 2016 2:07 pm

### Re: Martins' Method

This is amazing and I love it, I can only hope that a small part of this makes it to the automotive presets for fuel pumps especially, since we are seeing a lot of new petrol cars with 3 phase intake pumps. Outstanding work and I am just blown away.

ben.martins
Pico Staff Member
Posts: 317
Joined: Tue May 16, 2017 1:02 pm

### Re: The Martins' Method

Thank you, I do think this could help simplify the way we view 3 phase signals. We will of course be looking at the presets for the guided to tests and how this technique can be added in especially for the EV guided tests.

The brushless DC motors is a guided test on our list and something we need to take a look at with this method. In theory it should work the same but I've not applied it to these motors. I know there are plenty of captures floating around so it should be fairly straight forward to check! It may be that we have to AC couple the signals in order to view but we'll let you know as soon as we have anything to share.

Thanks again and look forward to hearing more from people trying out the Martins' Method!

Kind regards

Ben

ben.martins
Pico Staff Member
Posts: 317
Joined: Tue May 16, 2017 1:02 pm

### Re: The Martins' Method

Just a quick follow up after some testing I carried out on a 2016 e-Golf.

The resolver output for this vehicle appears to already have the math applied leaving you with 2 sine waves. This means we can watch these in real time using the Martins' Method and watch the direction of the motor change depending on if you're travelling forwards or reverse.
egolf motor position XY.mp4