Just cranked it up to 350 phase amps and did a video on 79.6 v starting battery voltage. I got 2 burnouts and over 10 min of mostly wide open run time untill I blew a hi side. Video to come. It really didnt feel faster then 300 phase amps. My wires might be to long and small.
Arlo's power stage Leaf controller runs and drives page 103
- Thread starter Arlo1
- Start date
Lebowski
10 MW
I've been playing around with this RC-series network between the gate-drain and it works amazingly well. I cut out the
snubbers on one of the 3 output stages, with the proper gate-drain RC the spikes on the output look much
better than on the 2 snubber outputs. And the nice thing is that snubbers dissipate a few Watts of power, while
the gate-drain RC does not... i think... cause the snubber is driven by a voltage source while the gate-drain RC is driven by a current source..
I still have 10 Ohm gate resistor, and now use a 10 Ohm in series with 560 pF (same as C_rss at low voltage) between gate-drain. I want to
try to reduce the resistor in the gate-drain further, but now 'm going to watch a movie so it'll have to wait.
I'll take some more measurements over the next few days and update my controller PCB for this change
snubbers on one of the 3 output stages, with the proper gate-drain RC the spikes on the output look much
better than on the 2 snubber outputs. And the nice thing is that snubbers dissipate a few Watts of power, while
the gate-drain RC does not... i think... cause the snubber is driven by a voltage source while the gate-drain RC is driven by a current source..
I still have 10 Ohm gate resistor, and now use a 10 Ohm in series with 560 pF (same as C_rss at low voltage) between gate-drain. I want to
try to reduce the resistor in the gate-drain further, but now 'm going to watch a movie so it'll have to wait.
I'll take some more measurements over the next few days and update my controller PCB for this change
Lebowski
10 MW
Arlo1 said:
you blew a high side, where you still got the diode in the gate line...
Do you think it blew because of a voltage spike ? Or maybe it overheated 'cause 10 minutes of full throttle at 350 phase amps, that's some serious power
I wouldn't expect it to be faster at 350 phase Amps, looking at your dyno plots your current is limited by the low battery voltage above around 30 kmh.
350 phase amps should get you 17% more torque wrt 300 (350/300=1.17), but only upto 25 to 30kmh...
I'm using 10awg phase wires and they are long.
I need to put a temp sensor on my fets. Anyone have some good sensors to link me to.
I think I can cool the fets better and switch them a bit faster and up the phase wire size and parallel more Lipo and relay I might be able to see 15-20rwhp
I need to put a temp sensor on my fets. Anyone have some good sensors to link me to.
I think I can cool the fets better and switch them a bit faster and up the phase wire size and parallel more Lipo and relay I might be able to see 15-20rwhp
Farfle
100 kW
That little bugger hauls!!!
Lebowski
10 MW
bloody hell
zombiess
10 MW
Arlo1 said:
This is what I use, it data logs 4 channels and can be downloaded over USB for looking at later. I've successfully soldered part of a thermocouple to a piece of copper to make it stay put and it's worked fine in my controller.
http://www.amazon.com/UEi-DT304-Digital-Logging-Thermometer/dp/B0039LAVEM/ref=sr_1_1?ie=UTF8&qid=1365151155&sr=8-1&keywords=dt304
Then I bought something like this, 5 thermocouples for $17
http://www.ebay.com/itm/K-Type-Thermocouple-Wire-for-Digital-Thermometer-Temperature-Sensor-Probe-TC1-5p-/310349791911?pt=LH_DefaultDomain_0&hash=item48424a02a7
You can also get many different lengths and better quality wire protection on them like this (but can be found cheaper)
http://www.ebay.com/itm/K-Type-Thermocouple-Wire-for-Digital-Thermometer-High-Temperature-Sensor-PK1000L-/221128239868?pt=LH_DefaultDomain_0&hash=item337c452afc
patallen
1 W
it might been asked million time already, but what is the motor you are using ?
spec or data sheet out of it ?
thanks.
spec or data sheet out of it ?
thanks.
Lebowski
10 MW
http://endless-sphere.com/forums/viewtopic.php?f=30&t=16728
I pulled the controller off the bike last night and It looks like I blew a low side. I have not tested the diodes on the other 5 stages yet but one low side is blown for sure.Lebowski said:Arlo1 said:
you blew a high side, where you still got the diode in the gate line...
Do you think it blew because of a voltage spike ? Or maybe it overheated 'cause 10 minutes of full throttle at 350 phase amps, that's some serious power
I wouldn't expect it to be faster at 350 phase Amps, looking at your dyno plots your current is limited by the low battery voltage above around 30 kmh.
350 phase amps should get you 17% more torque wrt 300 (350/300=1.17), but only upto 25 to 30kmh...
So I have a few places I can make up the losses to help correct the problem.
1 I can switch faster
2 1 can cool them better
3 I can lower the power but work on makeing the system more efficicent.
As for the sag if it is in the batteries them selfs I will not worry to much because I plan to run a pack 4 times the size for the range I want on a charge.
From the dyno I seen a max of ~10hp with 15000 watts going into the system so that ~20hp in with 10hp out but thats 10 hp of losses in the wires, conectors, controller losses, motor losses, chain losses, tire flex losses, wheel bearing losses, dyno roller bearing losses, windage losses on the motor and the wheel, it all adds up. But if I can cut down a little bit on a couple of them then I could get the same rwhp with a lower settign and stress the controller less for the same amount of "FUN"
Lebowski
10 MW
I would
- take out the diodes and use a fixed gate resistor, make high / low side equal
- the biggest change: take out the 4 fets/100 ohm resistor and drive the gate from the IRS2186 directly. This will most definitely reduce your deadtime
- see whether the R in the RC gate-drain can be lowered (baby steps !)
- go to more than 100% amplitude
just to illustrate why it is so important to reduce deadtime: you lose about 4 times the deadtime.
So, assuming 2 usec deadtime, this means 8 usec is lost, out of 50usec (20 kHz PWM), this is 16% !
It's like having a 16% lower battery voltage !
I just realised you actually lose way more than 16%
I added resistance measurement next to the inductor measurment in the controller IC. Even though I know 100% sure
my motor is about 370 milli-ohm, the controller reports about 930 milli-Ohm. Adding a 0.33 ohm resistor to each phase
make the controller report 1.26 Ohm, so this fits, the measurment is correct. I messed around with the settings and
found, interestingly enough:
- it seems to stay at 930 milli Ohm, independent of measurement current
- it is heavely dependent on the % the deadtime is with regard to the 1/pwm_frequency
by reducing deadtime I can get it down to below 600 milli-ohm
I'm still thinking about the mechanism at work and whether I can correct the measured value for this. But
it seems to be like a 'virtual resistor' is added by the deadtime. This resistor does not dissipate any power,
but it does come into play when trying to push loads of current into the motor.
- take out the diodes and use a fixed gate resistor, make high / low side equal
- the biggest change: take out the 4 fets/100 ohm resistor and drive the gate from the IRS2186 directly. This will most definitely reduce your deadtime
- see whether the R in the RC gate-drain can be lowered (baby steps !)
- go to more than 100% amplitude
just to illustrate why it is so important to reduce deadtime: you lose about 4 times the deadtime.
So, assuming 2 usec deadtime, this means 8 usec is lost, out of 50usec (20 kHz PWM), this is 16% !
It's like having a 16% lower battery voltage !
I just realised you actually lose way more than 16%
I added resistance measurement next to the inductor measurment in the controller IC. Even though I know 100% sure
my motor is about 370 milli-ohm, the controller reports about 930 milli-Ohm. Adding a 0.33 ohm resistor to each phase
make the controller report 1.26 Ohm, so this fits, the measurment is correct. I messed around with the settings and
found, interestingly enough:
- it seems to stay at 930 milli Ohm, independent of measurement current
- it is heavely dependent on the % the deadtime is with regard to the 1/pwm_frequency
by reducing deadtime I can get it down to below 600 milli-ohm
I'm still thinking about the mechanism at work and whether I can correct the measured value for this. But
it seems to be like a 'virtual resistor' is added by the deadtime. This resistor does not dissipate any power,
but it does come into play when trying to push loads of current into the motor.
Thanks... But ive been put onto the design using irfb9110 and irfb110 to stop rining and drive hi power.Lebowski said:
Im also using a ir2113 gate driver.
I can lower the resistor value to switch the fets faster.
I 100% am keeping the diode on the hi side at this point because it keeps the hi side off preventing me from having a pass through.
Lebowski
10 MW
Very understandable reaction, it works so hands off, no touching ! I would do the same in your case
Get some more transistors , 3 extra power stage pcb's and take my advise in consideration, and go for 150V.
Then you can just do an output stage swap for testing, and you can use what you have now on the road.
I really like the current sensors you have by the way, they are 400A but by looping the wire through it 1, 2, 3, 4 etc times it in effect becomes a 400A,200A,133A etc current sensor, so configurable
Get some more transistors , 3 extra power stage pcb's and take my advise in consideration, and go for 150V.
Then you can just do an output stage swap for testing, and you can use what you have now on the road.
I really like the current sensors you have by the way, they are 400A but by looping the wire through it 1, 2, 3, 4 etc times it in effect becomes a 400A,200A,133A etc current sensor, so configurable
patallen
1 W
Arlo1 said:
where /who do you order pwoer stage from ?
MitchJi
10 MW
Hi,
Great job!
Maybe you guys should set up a partnership?
Great job!
Arlo1 said:
How about kits and completeded controllers for real dummies?Arlo1 said:
Maybe you guys should set up a partnership?
patallen
1 W
MitchJi said:
as i am building the controller for my ETEK motor, i am thinking about making a generic PCB for the controller/low power side, power side beign relatively adaptable for the needs...duno if there would be demands for such thing. fwiw i am making very good progress with this and it works very well for now on a veroboard.
I found one thing to make it better. My fans were spinning the wrong direction... lol I assumed they turned a certain way but I was wrong so I flipped them. I figure if they blow fresh air on the bottom of the heat sink then it lets it flow sideways out the ends it will cool best. I also ground some groves in the heat sink and tomorrow I will bead blast it so the fans will have more surface area to blow on.
We might be able to do something but Chances are you would not be interested when you see the price. IE this controller has $150 of current sensor alone. I did a quick price in my head and I think the parts cost alone was 500-600$ plus the time to hand assemble it all and then the time to set it up and test it.MitchJi said:
Once I get the free time I will post all the files I have but I need to go over them and try to fix the mistakes. I have a brain board separate so I can use one design for the Brian to run any power level of powerstage. If anyone wants to have me build them a brain I will do it for the price of parts and I will ship it out. I would like to make sure its good before trying to go to far. Two things are slowing me right now 1 is time 2 is money. So... If people want to help let me know. And to make this clear I'm not asking for handouts but simply someone who wants to spend money on DIY stuff or use my files and see if they can make it all come alive as well.patallen said:
Lebowski
10 MW
Arlo1 said:
Well, to be fair, that is for a 15kW controller... Last year I made a PCB for my controller (which I want to update now that
we have a new spike suppression scheme). This is for a 100 or 150V 6 FET, so good for a few kW. I'm guessing, including the PCB and controller IC, you're looking at around $200. Arlins $50 current sensors are for 400A, simple 50 or 100A ones are about $7 if I remember correctly... have a look at 'the big lebowski controller' thread
MitchJi
10 MW
Hi,
Not only will the vast majority of users need a less powerful controller but how much would you have to pay for a 15kW controller of similar quality if you want one? For this to be a viable business it can't be priced in the stratosphere although many people will pay a premium for premium quality. The reason I suggested this is if you can figure out a way to do this that's financially viable (and I'm not sure you can) it makes sense in that you are both accumulating a lot of valuable and unique knowledge and experience and you both love doing this. How many people in our society can earn money doing something they love? A large part of your enjoyment would come from producing a product that your customers would love.
Arlo1 said:
Yes. I'd be happy to get one that could handle 75V 80-100A.Lebowski said:
Not only will the vast majority of users need a less powerful controller but how much would you have to pay for a 15kW controller of similar quality if you want one? For this to be a viable business it can't be priced in the stratosphere although many people will pay a premium for premium quality. The reason I suggested this is if you can figure out a way to do this that's financially viable (and I'm not sure you can) it makes sense in that you are both accumulating a lot of valuable and unique knowledge and experience and you both love doing this. How many people in our society can earn money doing something they love? A large part of your enjoyment would come from producing a product that your customers would love.
HighHopes
10 kW
- Joined
- Mar 28, 2013
- Messages
- 930
$500-600 is about right for this size controller. but take your time making kits, you need a stable & proven circuit.
you were talking about switching frequency before.. just thought i'd share some thoughts ..
1. minimum switching frequency should be 10x greater than your maximum modulated frequency, typically 60Hz. so your switching frequency minimum is 6kHz. this is to avoid aliasing in harmonic spectrum, i.e. make sure the harmonics never overlap giving high spikes
2. if you had an output LC filter between inverter & motor drive (useful for long cable run over 50ft) then your switching frequency would have to be 10x less than the cut-off frequency of that. but in your case you don't have a filter i think, so this doesn't apply.
3. for practical reasons the upper switching frequency is limited due to temperature. the more switching the more switching losses for same conduction. it is nice when your switching losses are the same as your conduction losses at max load. you can actually calculate this, but it is a bit harder to calculate for a mosfet as compared to IGBT.
4. for practical reason the upper switching frequency is limited due to your deadtime. your deadtime is there because of propagation delay(s) in your system and its variation over temperature. this value does not change because of switching frequency.. but the deadtime % of switching period is important. remember, you don't want to be higher than 5%, if you need higher you have a bad design (but bad designs can still be functional). the highest switching i ever did for power levels over 50kW was 40kHz, and to do that took very special attention to near elimination of deadtime (actually was around 100ns) and very special attention to switch design at the die level.
5. finally, and perhaps more most importantly, your switching speed is related to the bandwidth of control of current regulation. i know what the text books say about this, but i will give you something that your text books do not say.. the bandwidth of your controller must be set for no-load operation of the motor incase of broken shaft. you can not go out of control during a failure - this is when you need the control the most! this is basically electrical time constant L/R of windings at no-load. also, you must regulate the current DURING acceleration/deceleration .. not during constant speed. i don't know why the text books miss this. it is NOT OK to have your PI(D) current regulator hit the rail during acceleration. it is unacceptable to not be in control during acceleration. the bandwdith requirement of current regulation during acceleration is much higher and so your switching frequency which is your control frequency related here. now that you see how high bandwidth is really required for a true high performance controller, now you see why it has been only recently that true high performance drives have done the current regulation in a DSP. you need a lot of MIPS! that has only happened in the last 3 years.. one of my controllers developed 5 years ago had to change 1000rpm in 10ms .. crazy accelration, and to regulate that current it had to be done with high slew rate op-amps. actually worked beautifully, but was a pain in the ass to tune (lots of soldering).
6. this probably doesn't apply to you too much, but for constant speed applications the switching frequency divided by modulation frequency can also be a multiple of 3 for 3-phase system. this will help cancel some harmonics. well, guess it can't hurt to through this requirement in your controller anyway.
too much info? am i annoying you yet?
you were talking about switching frequency before.. just thought i'd share some thoughts ..
1. minimum switching frequency should be 10x greater than your maximum modulated frequency, typically 60Hz. so your switching frequency minimum is 6kHz. this is to avoid aliasing in harmonic spectrum, i.e. make sure the harmonics never overlap giving high spikes
2. if you had an output LC filter between inverter & motor drive (useful for long cable run over 50ft) then your switching frequency would have to be 10x less than the cut-off frequency of that. but in your case you don't have a filter i think, so this doesn't apply.
3. for practical reasons the upper switching frequency is limited due to temperature. the more switching the more switching losses for same conduction. it is nice when your switching losses are the same as your conduction losses at max load. you can actually calculate this, but it is a bit harder to calculate for a mosfet as compared to IGBT.
4. for practical reason the upper switching frequency is limited due to your deadtime. your deadtime is there because of propagation delay(s) in your system and its variation over temperature. this value does not change because of switching frequency.. but the deadtime % of switching period is important. remember, you don't want to be higher than 5%, if you need higher you have a bad design (but bad designs can still be functional). the highest switching i ever did for power levels over 50kW was 40kHz, and to do that took very special attention to near elimination of deadtime (actually was around 100ns) and very special attention to switch design at the die level.
5. finally, and perhaps more most importantly, your switching speed is related to the bandwidth of control of current regulation. i know what the text books say about this, but i will give you something that your text books do not say.. the bandwidth of your controller must be set for no-load operation of the motor incase of broken shaft. you can not go out of control during a failure - this is when you need the control the most! this is basically electrical time constant L/R of windings at no-load. also, you must regulate the current DURING acceleration/deceleration .. not during constant speed. i don't know why the text books miss this. it is NOT OK to have your PI(D) current regulator hit the rail during acceleration. it is unacceptable to not be in control during acceleration. the bandwdith requirement of current regulation during acceleration is much higher and so your switching frequency which is your control frequency related here. now that you see how high bandwidth is really required for a true high performance controller, now you see why it has been only recently that true high performance drives have done the current regulation in a DSP. you need a lot of MIPS! that has only happened in the last 3 years.. one of my controllers developed 5 years ago had to change 1000rpm in 10ms .. crazy accelration, and to regulate that current it had to be done with high slew rate op-amps. actually worked beautifully, but was a pain in the ass to tune (lots of soldering).
6. this probably doesn't apply to you too much, but for constant speed applications the switching frequency divided by modulation frequency can also be a multiple of 3 for 3-phase system. this will help cancel some harmonics. well, guess it can't hurt to through this requirement in your controller anyway.
too much info? am i annoying you yet?
Similar threads
