Arlo's power stage Leaf controller runs and drives page 103
- Thread starter Arlo1
- Start date
Lebowski
10 MW
Arlo1 said:
depends also on the current sensors. But lets say 24 samples per e-rpm (so every e-rotation is sub-divided in 24 pieces, pretty
high compared to the 6 for hall sensors), then you end up with 78k * 60 / 24 = 200 000 erpm .... But this uses the assumption
of 24 and also needs a PWM frequency higher than 78k/2 = 39 kHz
Ok so I replaced the gate resistors but now I got a new problem. The motor does not sound right when running but I don't think its the motors fault. I tested all three phases and there is no shorts to ground (the stator) and the resistance is all the same and the output on the scope from all three phases looks perfect. As well when testing the wires from the motor to the motor were cool. But the wires going to the power
stage were hot.
And I think its time to upgrade my scope... I was scoping the voltage drop of the gate resistors and saw what looked to be a bounce but I now think its the miller plateau.
I am now wondering if the controller is firing two outputs at once. I will run some more tests tomorow.
[youtube]zl1eAY-v8WQ[/youtube]
One thing is I am testing at 12v for the driver stage controll but I think it should still work and I think thats what I was using when I run it yesterday!
stage were hot.
And I think its time to upgrade my scope... I was scoping the voltage drop of the gate resistors and saw what looked to be a bounce but I now think its the miller plateau.
I am now wondering if the controller is firing two outputs at once. I will run some more tests tomorow.
[youtube]zl1eAY-v8WQ[/youtube]
One thing is I am testing at 12v for the driver stage controll but I think it should still work and I think thats what I was using when I run it yesterday!
Attachments
Lebowski
10 MW
If it's running rough in drive_3 there's something wrong with either
the output stage and its drivers OR with the current sensors. If your output
stage etc is OK, check the current sensors are still working. I once had
a blown ACS768 - 150A sensor (output stuck at 0 V), the motor was
still running but very rough and FETs were blowing left and right...
What you could also try is to reset the calibration of the current sensors
(menu d, option k and then save) but I seriously suspect the current sensors.
Also check you accidently set 3 current sensors in the menu while
you only have 2 connected !
about a better scoop, Kingfish just bought a new one, maybe ask him
how he likes it and get the same one ?
http://www.endless-sphere.com/forums/viewtopic.php?f=30&t=42666&start=105#p627581
the output stage and its drivers OR with the current sensors. If your output
stage etc is OK, check the current sensors are still working. I once had
a blown ACS768 - 150A sensor (output stuck at 0 V), the motor was
still running but very rough and FETs were blowing left and right...
What you could also try is to reset the calibration of the current sensors
(menu d, option k and then save) but I seriously suspect the current sensors.
Also check you accidently set 3 current sensors in the menu while
you only have 2 connected !
about a better scoop, Kingfish just bought a new one, maybe ask him
how he likes it and get the same one ?
http://www.endless-sphere.com/forums/viewtopic.php?f=30&t=42666&start=105#p627581
I just tested the current sensors and they all have 4.88v at the positive and negative pins and ~2.44 back to the dspic30 chip. So i think they are ok. I tested the driver stage mosfets and they seem ok its realy funny it works great testing on pwm test mode but in drive 1,2, and 3 its pulling huge amps!
I got it running again.... It was the fet driver for the same phase that blew the two mosfets. It seems it is turning the hi and low on at the same time.... Now the question is were the limits of the ir2113 exceeded by going to 110,000 e,rpm at 40khz pwm?
I checked the dead time and it looks good. 500nS -1500nS are .1 amp use on the power stage. 400 = .2 300 = .8 etc
Time to read the data sheet for the ir2113. It is my guess that the driver was the cause of the failure.
The pic is 2 different tests of dead time. One was yesterday. Column A and B are the same because I was to lazy to clear one lol
I checked the dead time and it looks good. 500nS -1500nS are .1 amp use on the power stage. 400 = .2 300 = .8 etc
Time to read the data sheet for the ir2113. It is my guess that the driver was the cause of the failure.
The pic is 2 different tests of dead time. One was yesterday. Column A and B are the same because I was to lazy to clear one lol
Made some great progress today. Man its quiet.
Here is the first run ever!
[youtube]5QGI8JjOKCs[/youtube]
And our best run so far (remember its only 6 fets)
[youtube]lWSJaghYnOk[/youtube]
And blowing up at 32s got another pass through but it might have been a mechanical failure due to a peice of something loose rolling where it wasnt suposed to be.
[youtube]5TsqzTcFicQ[/youtube]
Ok so I have the dead time set at 666 lol it wont alow 699 and I scoped my fets and they are turning on in 1.2uS and off in 1.2uS I am using 12v gate drive and 22ohm gate resistors what would you guys sugest?? I am thinking 15 ohm gate resistors and maybe 15v? Shorten up the switching time to about 500-700 us?
Here is the first run ever!
[youtube]5QGI8JjOKCs[/youtube]
And our best run so far (remember its only 6 fets)
[youtube]lWSJaghYnOk[/youtube]
And blowing up at 32s got another pass through but it might have been a mechanical failure due to a peice of something loose rolling where it wasnt suposed to be.
[youtube]5TsqzTcFicQ[/youtube]
Ok so I have the dead time set at 666 lol it wont alow 699 and I scoped my fets and they are turning on in 1.2uS and off in 1.2uS I am using 12v gate drive and 22ohm gate resistors what would you guys sugest?? I am thinking 15 ohm gate resistors and maybe 15v? Shorten up the switching time to about 500-700 us?
OK so I am sucsessfull up to 100v battery (24s) But once I go over 100 I have problems. I thought the diodes would solve it then today I burned down big time! But good news I think I know why, With 32s ~130v you have 260v differnce peak to peak and that was running through a ribbon cable where it was fed to a seprate brain board and was shunted down to 5v though resistors and diodes this again putting a rather large amount of amps flowing though the ribbon wires on the negative and all 3 phase voltage wires. So... here is the carnage. And any imput that will help progress is welcome.
[youtube]7Cyo2hYEEhQ[/youtube]
[youtube]7Cyo2hYEEhQ[/youtube]
Attachments
What is your resistor divider between the motor and the PIC?
Maybe you should increase the resistor facing the motor when you add more voltage, to make the voltage at the ADC the same?
How fast is the motor spinning at 130V? is the components in the RC-filter small enough to let the motor frequency pass?
Does the chip handle the frequency/is it able to process in time?
Maybe you should increase the resistor facing the motor when you add more voltage, to make the voltage at the ADC the same?
How fast is the motor spinning at 130V? is the components in the RC-filter small enough to let the motor frequency pass?
Does the chip handle the frequency/is it able to process in time?
Ok so as seen here http://endless-sphere.com/forums/viewtopic.php?f=30&t=36602&start=15bearing said:
I asked and used 10k 3 watt resistors. But remember if the resistors need to remove 3 watts then the little ribbon wire feeding it will melt and because its ~the speed of ~110v when it melted then there is 110v+ in a wire next to 110- next to the ground that also has to flow the amps fetting the ground for the watt resistor.
And if the amps flowing through the ribbon cable were not enought well then the 220v peak to peak with less then .5mm between them on the headder pin solder connectrions is enough for a failure!
I just double checked with lebowski but I think the max erpm is ~200000 erpm so 20,000 rpm for colossus. 65mph = 351000 feet per hour = 5850 fett per minute the wheel dia is 18.5 inches so 58.1 inches per rev =4.84 feet per rev 5850/4.84= 1208 wheel rpm at 65 mph x 4:1 (which I will tripple check) reduction in gearing so 4834 at the motor when you saw it melt down.
http://endless-sphere.com/forums/download/file.php?id=78273&mode=view
Hm, so it's this schematic, but with 10k R_Bs. I assumed there would be an RC-filter, but it seems like there is no RC-filter, so I guess it's not needed.
With R_A = R_B, the EMF will be capped at a very low voltage. I guess the chip is only interested in voltages around of a few volts. I think the circuit looks a little strange though, because the capped voltages will mess with the "zero point" (where all R_Bs are connected). (But I'm not the expert here.)
If the current through the diodes gets higher than what the total circuit consumes, then the 5V rail will increase. But thats probably (and hopefully) not your problem.
Hm, so it's this schematic, but with 10k R_Bs. I assumed there would be an RC-filter, but it seems like there is no RC-filter, so I guess it's not needed.
With R_A = R_B, the EMF will be capped at a very low voltage. I guess the chip is only interested in voltages around of a few volts. I think the circuit looks a little strange though, because the capped voltages will mess with the "zero point" (where all R_Bs are connected). (But I'm not the expert here.)
If the current through the diodes gets higher than what the total circuit consumes, then the 5V rail will increase. But thats probably (and hopefully) not your problem.
Ok after looking at my ribbon cable again and talking to lebowski. I think I know whats up. The power to drive the fet drivers was coming in from the brain board from a 3s lipo pack powering the brain and drivers. This was fed though the 28awg ribbon cable I went and inspected again yesterday and saw the negative on the ribbon cable was burnt most from one end to the other (wire inside got to hot) then the second most damage was the 5v (the other feed to the drivers) SO as rpm increases driver current usage increases this explains the problems with going over x rpm or x voltage but the failure was because of rpm of the motor and the negative is a common for a few other things so it would have the most load on it! Now I will also move the 10k 3w resistor for the phase voltage monitoring to the driver boards to lower the voltage in the ribbon cable to 0-5v and help keep the chance of arcing on the header pins and everything else down.
One thing that would help me prevent this and debug without blow ups is a thermal imaging camera I have a shitty iPhone and my gf has a good one we could do the hack on but does anyone have a idea where you can find a thermal imaging camera for a reasonable price??
One thing that would help me prevent this and debug without blow ups is a thermal imaging camera I have a shitty iPhone and my gf has a good one we could do the hack on but does anyone have a idea where you can find a thermal imaging camera for a reasonable price??
MitchJi
10 MW
Hi Arlo,'
http://www.sheldonbrown.com/gears/
I ran the following:
It will be simpler to use this calculator, plug in the correct wheel size (doing a little interpolation will be easier than doing the calculations yourself) and choose MPH at 100 RPM in the "Gear units" pulldown:Arlo1 said:
http://www.sheldonbrown.com/gears/
I ran the following:
salty9
1 kW
Arlo1 said:
IR thermometers could be useful for finding hotspots. Slower than an IR camera but orders of magnitude cheaper.
Yeh I already have one. And I was checkign the fet, motor, and fet drivers every run but Never went over every single wire.salty9 said:
I will get it sorted out soon. I just received 70 more ixfk230n20t fets and digikey matched mousers price. I also got some of methods favorite caps. I'm going to Alberta to check out my good friends motocross track for a long weekend then back to this! I have one more driver board I should be able to fix everything and replace the really bad board and be up and running on 6 fets again in a few hours of work. Then I will run to the limits (40s 170v off the charger) to prove the concept and find the limits of the fets in my design. Then.... 48fets
Make friends with your local fire department, and see if you can borrow one from them.Arlo1 said:
Thanks AW. I do have a firedept hang around who is a customer who likes to watch my experiments.
Now I was doing some quick reading to let my self think while I drive for 12hr tonight and I came to the conculsion I need to learn how to calculate the amps used to charge each fet gate.
I tried one method I found a Nanocolumb to AH calculator and I used 50,000 (pwm) x 378 (ixfk230n20t nanocolumb spec from sheet) x 60 x 60 to get 68040000000 (nanocolumbs used per hour) then put that into a calculator and got 0.01890000000001512 per fet so .1134 for all 6.
Power transfer for 28 awg is max at .226 amps so giving that I would have been ok. But I think I need to find a better way to calculate mosfet gate charge amp draw at x pwm. Its not just the fets using the power it would be the drivers and other items on the driver boards too.
Now I was doing some quick reading to let my self think while I drive for 12hr tonight and I came to the conculsion I need to learn how to calculate the amps used to charge each fet gate.
I tried one method I found a Nanocolumb to AH calculator and I used 50,000 (pwm) x 378 (ixfk230n20t nanocolumb spec from sheet) x 60 x 60 to get 68040000000 (nanocolumbs used per hour) then put that into a calculator and got 0.01890000000001512 per fet so .1134 for all 6.
Power transfer for 28 awg is max at .226 amps so giving that I would have been ok. But I think I need to find a better way to calculate mosfet gate charge amp draw at x pwm. Its not just the fets using the power it would be the drivers and other items on the driver boards too.
bigmoose
1 MW
Hack your phone or a cheap camera with the right sensor and a visible filter. I picked up a FLIR i7 this year. It is extremly useful, but has no focus capability, nor can you set the temperature range on the view. It autoranges whatever it sees. I can get a decent idea of board temperature maps, and a whole lot better than I did before. I also doesn't have visible picture overlay capability like the nice Flukes, but then again it was around $1600 US on eBay. It does save the picture and has an incredible battery life. I am always sticking my finger into the view to ensure I know what is the hot component.
The things I see is when I buried TO92 regulators close to the filter caps and ran them hard, I was heating the side of the caps. Not good for life. Another neat, fun tool.
BatteryMooch
10 kW
Ooooohhhhh....very jealous!!bigmoose said:
Thanks for the feedback on it. I've been looking at the Flukes recently (getting ever so tired of IR thermometers!) but haven't decided on what my budget is going to be. Sooooo tempted by the high end models! Just not sure that I want to sell my apartment to buy one.
Lebowski
10 MW
Arlo1 said:
I also arrived at around 0.11 A for the 6 FETs, plus in addition there´ll be some return current
from the 10K resistors. What I would do, measure the return current in the ground line in the ribbon
cable (and all other currents associated with burned out wires) and see whether it makes sense...
My battery that run all this was a 1000 mah 3s lipo pack and it was hooked to the brain board so all the power for the driver boards come in on the 28awg ribbon cable and that battery would not last long for testing like someware ~10-30 min so that meens something is using aproxamitly 2-10 amps and if there is only 1 ground wire thats 28 awg ( the one thats very burnt) than its definatly over loaded!!!
Lebowski
10 MW
Still, it sounds rediculously high... I would measure all currents and look for
abnormalities. Maybe there´s shoot through or some fundamental error in the
extra driver transistors. 2 to 10A sounds way too high to me. On my ´big lebowski´
controller the power budget for the complete 15V and 5V subsystem is 15W
(1A max at 15V) but in practise I think it´s only 20% of that.
abnormalities. Maybe there´s shoot through or some fundamental error in the
extra driver transistors. 2 to 10A sounds way too high to me. On my ´big lebowski´
controller the power budget for the complete 15V and 5V subsystem is 15W
(1A max at 15V) but in practise I think it´s only 20% of that.
Similar threads
