Inside a 400amp 120v BLDC Kelly

[rhitee05]

You really need to be part EE and part ME to design a good high-power controller. I might even argue (against personal bias) that ME is a little more important - marginal heat transfer design is probably more likely to turn your controller into smoke than marginal circuit design. But, the details of the electrical design are pretty technical, too. So you need both skill sets.

Problem is, most EE's don't learn the necessary ME skills in school and I'm fairly certain the likewise is true for ME's. I even had the benefit of a heat transfer/fluid dynamics course, but it still didn't teach me anything useful about cooling electronics. I had to learn that stuff on my own and on the job.

 

[bigmoose]

Luke since you disassembled a Sevcon, I'll bet you eventually reassembled it... can you share the installatin procedure for the FET pressure springs?

 

[liveforphysics]

Luke since you disassembled a Sevcon, I'll bet you eventually reassembled it... can you share the installatin procedure for the FET pressure springs?

I imagined you would spot that Dave.

I used 12 flat head screwdrivers. One on each side of the spring clips slid in, then twisted to hold the spring open, then set the PCB with fets in place, located properly, and untwisted and slid out the screwdrivers one by one to set the clamps. I imagine they have a jig/tool that does it quickly and easily, for myself it took a bit figure a solution and round up enough of the right width screwdrivers. Lol

 

[localbroadcast]

I was wondering what the large capacitor banks are for? I'm at the beginning stages of making my own controller. I just figured out why anyone would need more than 6 mosfets... because in parallel you can handle higher current! So are the large cap banks there to handle voltage spikes caused by switching the stator coils on and off rapidly with PWM? Voltage spikes caused by collapsing magnetic field? If so, does anyone have a circuit diagram that shows where these capacitors are wired in the circuit? Parallel with the coils? What do you use to calculate the required voltage and farad requirements for these caps?

Thanks in advance.

 

[Njay]

Check this out localb,

http://www.4qdtec.com/pwm-01.html

 

[grindz145]

night and day.

 

[Doctorbass]

Luke,

With your higher experience with the Kelly than me, what is your opinion about using my KBL12221H rated 120V(max136V) and 220A phase current?

Actually i'm using it at 75% power and 100V fresh of the charger ( 24s) on a 5303 motor. I see 233A max recorded by the CA at around 30% of the accelleration ramp ( My shunt is calibrated with trade off value between my Fluke 337 DC clamp meter and also with a fluke 179 for the voltage drop.

Apparently, accordig to Kelly, the % of power in the programmer is from the max RATED motor phase current of the cotroller wich is 220A.. so i should get 165A...

Test 1: I would like to try that same setup but at 100% power... would you recommand that? .. since i am still 36V under the max rating for the voltage?

Test 2: Also i plan on using the 3 new lipo i buight to increase from 24s to 30s.

This will add 25.2V fresh of the charger leaving a margin of 10 volts from the max rating of that cotroller wich is also 92% of the max voltage rating

Test 3: Both 30s and 100% power rating.

Thanks for your opinion Luke

Doc

 

[liveforphysics]

Give it a try and keep your hand on it to see if it gets burning hot, if it does, then back it down a bit until it doesn't.

We got the 800amp 144vdc Kelly skin-searing burning hot yesterday on the TTXGP bike, and it was only averaging around 550amps...

 

[Arlo1]

Give it a try and keep your hand on it to see if it gets burning hot, if it does, then back it down a bit until it doesn't.

We got the 800amp 144vdc Kelly skin-searing burning hot yesterday on the TTXGP bike, and it was only averaging around 550amps...

What do the kelly controllers use to hold the fets?

 

[liveforphysics]

Give it a try and keep your hand on it to see if it gets burning hot, if it does, then back it down a bit until it doesn't.

We got the 800amp 144vdc Kelly skin-searing burning hot yesterday on the TTXGP bike, and it was only averaging around 550amps...

What do the kelly controllers use to hold the fets?

Screws. The beginning of this thread has a bunch of pics of them.

 

[Arlo1]

Sorry luke what i was wondering is if they use the little plastic washers to isolate the fets? It kinda looks like the fet is conected to the buss bar it self?

 

[markcycle]

Luke
I am wondering why you continue to use the Kelly controllers

Is it that in the finale analyst it has the best price performance ratio when derated, so as to work reliably.
Is it that you just get them as free samples.

Seems money is no issue in your current job so there must be a technical justification.

 

[liveforphysics]

Luke
I am wondering why you continue to use the Kelly controllers

Is it that in the finale analyst it has the best price performance ratio when derated, so as to work reliably.
Is it that you just get them as free samples.

Seems money is no issue in your current job so there must be a technical justification.

All bought and paid for at full retail from Kelly, they are such a tiny bussiness, they can't even meet OEM level volune demands to be offering samples.

Why do we run them when we have the best from Sevcon and others sitting around? Because its hundreds of hours on the dyno to setup other controllers, which is fine for an OEM application but for 1-off race bikes, the plug-n-play nature of a Kelly is just too convienient.

 

[liveforphysics]

Also, I find best practice when setting up something to use a Kelly is to buy 3 of them.

One sacrificed to find the limits in tuning. One that you intend to use. And one as a spare, because its not like you can overnight a new one at a race or whatever when they blow, and if you've got say $20-30k wrapped up in logistics and setup for a single event, it really sucks to DNF over not having a spare.

Buying 3 Kelly's is still loads cheaper than setting up a proper controller (because of the hundred hours of skilled labor and dyno time involved).

 

[markcycle]

I've setup control loops and PID for 25 years using Galil Motion control, Copley Controls, kollmorgen and many others. Today with the auto tune algorithms out there it takes but a few hours to get it 99.9% right. In the old days it was a difficult task, but today even loose systems with terrible backlash can be controlled.

I don't understand Sevcons problem. If Kelly can make a good self learning controller that gets it mostly right, I don't see why Sevcon can't do the same and still give the OEM the ability to change the tuning to their hearts content.

I get asked about using the Sevcon all the time but no one can invest the time in getting it tuned, and Sevcon won't give me the time of day.

 

[liveforphysics]

We've flown out the Sevcon techs from Boston. We've flown out reps and techs from Arizona. We hired an outside EV controller setup specialist with 10years of experience and training. Then we sent the whole setup over to Sevcon in the UK.

And still we aren't able to use themin the ways we wanted or get full performance that we get from a Kelly.

Also, Kelly's dont self-learn. Its just a general best-guess 1 size-fits-all algorythm that the user can adjust for some parameters. They sometimes randomly do full regen when you hit the throttle. They sometimes just explode into plasma (even when at cold temps and light throttle).

But they aren't faulting cry-babies like the other options out there, so it works out OK for racebikes.

 

[flathill]

We've flown out the Sevcon techs from Boston. We've flown out reps and techs from Arizona. We hired an outside EV controller setup specialist with 10years of experience and training. Then we sent the whole setup over to Sevcon in the UK.

And still we aren't able to use themin the ways we wanted or get full performance that we get from a Kelly.

Also, Kelly's dont self-learn. Its just a general best-guess 1 size-fits-all algorythm that the user can adjust for some parameters. They sometimes randomly do full regen when you hit the throttle. They sometimes just explode into plasma (even when at cold temps and light throttle).

But they aren't faulting cry-babies like the other options out there, so it works out OK for racebikes.

Did the Sevcon fault in Brammo's race bike at REFUEL? Supposedly they got the kinks worked out last weekend at Thunderhill. We'll see this weekend

Does anyone know what batteries Brammo and Mission Motors are using in their race bikes? I think Brammo is using Leyden Energy in their street bike.

 

[HumboldtRc]

We've flown out the Sevcon techs from Boston. We've flown out reps and techs from Arizona. We hired an outside EV controller setup specialist with 10years of experience and training. Then we sent the whole setup over to Sevcon in the UK.

And still we aren't able to use themin the ways we wanted or get full performance that we get from a Kelly.

Also, Kelly's dont self-learn. Its just a general best-guess 1 size-fits-all algorythm that the user can adjust for some parameters. They sometimes randomly do full regen when you hit the throttle. They sometimes just explode into plasma (even when at cold temps and light throttle).

But they aren't faulting cry-babies like the other options out there, so it works out OK for racebikes.

I know this is way over my head, but what are the algorithms the sevecon controllers have to learn, used for??? To run the motor or what I don't get it..... Why can't you just plug them in like an rc esc and give it throttle and have the motor run...

 

[Arlo1]

Did you give up on sevcon then???

 

[liveforphysics]

Did you give up on sevcon then???

The hardware is awesome. Once you get one properly dialed in it's awesome. If you can't throw 10,000-20,000usd at getting one dialed-in then it's a paper-weight.

Perfect for OEM needs. Not-perfect for DIY'ers and 1-off builds etc.

 

[rhitee05]

I have to admit I'm surprised too. In most of the stuff I've read on motor control, most of the parameters are derived directly from the electrical and mechanical properties of the system - L, R, inertia, friction, etc - which is all stuff I'm quite certain you have the equipment/know-how to find quite easily. A little bit of dyno time to tweak and that's it.

Sevcon must be using a pretty weird control setup if you have to determine everything experimentally. They don't really have any documentation of this on the web - I'd be curious how it's set up.

 

[Arlo1]

Did you give up on sevcon then???

The hardware is awesome. Once you get one properly dialed in it's awesome. If you can't throw 10,000-20,000usd at getting one dialed-in then it's a paper-weight.

Perfect for OEM needs. Not-perfect for DIY'ers and 1-off builds etc.

What about someone who can spend the time lerning to set it up?

 

[Andje]

Hey luke, did you ever receive your 800 amp kelly? If you ever have time to review it's insides that would be appreciated. I am considering getting one of the 300 amp burst 144v KBH's and would love to here your opinion of them.

 

[Doctorbass]

My friend Jean Luke just received his 144VDC 600A Kelly a month ago. It's heavy and black anodized color and seem stronger than the lower power model.

i'll post my opinion about it probably this summer 2012 when we will install it on his GRX-R1100


Doc

 

[daveyjones97]

ah fuggit, every time i find a controller (or other component) i think will do for my proposed 1st build i find an es thread telling me its junk!
what are the chances of this lasting @ 70v 35A?:http://kellycontroller.com/kds72050e50a24v-72v-mini-brushed-controller-p-760.html
when the answer is "none" what controller can you guys reccomend?