ELECTRIC TRANSMISSION DELTA /WYE and SERIE/PAPALLEL stator

[Doctorbass]

Doc- This was a post I wrote for Arlo originally on the topic of rewinding a burnt x5305, but as it relates to the topic of this thread so closely (improving crystalyte motor performance), I thought it might be handy to have the info in here as well. The relays help the controller heating, but actually decrease the performance of the motor by just a touch by adding a bit more resistance to the current path. With a re-wind, there is a lot of potential to make some big improvements in motor performance, but of course it also means a TON of time and work to do it. I figured you would be another one of the few folks with the skills and determination to actually manage to do it.

From the factory, a 5305 motor is wound with 4 pieces of 21awg wire wrapping around each tooth 5 times. 21awg wire has a x-section of 0.412mm^2, so the combined 4 strands have a x-section of 1.648mm^2, or roughly equal to a single pass of 15awg wire.

The 5304 motor uses 6 pieces of 21awg wire wrapping around each tooth 4 times. Combined x-section of 2.472mm^2, or roughly equal to a single pass of 13.5awg wire.

The 5303 motor uses 8 pieces of 21awg wire wrapping around each tooth 3 times. Combinded x-section of 3.296mm^2, or roughly equal to a single pass of 12awg wire.


Arlo- I believe you could take square/rectangle wire with a x-section of 12awg (depending on how sloppy the factory windings are, maybe possible with round wire as well), and get the same 5-turns with it that you're using now. This would mean identical torque per amp for your motor, but HALF the winding resistance! This would be a huge improvement in power handling and top speed. It should mean roughly ~50% higher power handling for the motor, and a ~100% increase in continuous torque capability over a factory 5305.

OMG !! rewinding a X5 motor.. yes.. i'm capable.. but i dont have time to do that right now. i'll try that for sure but in the 5303 mode i'm pretty satisfied now..

The problem i had on the controller was simple.. A mosfet trio blown for two possible reasons
:
- The lack of thermal paste on the back of one of the mosfet
- I forgot to re-thight one of the 3 scret holdiong the 4110 mosfet.

one of the mosfet become so hot than the plastic insulator washer melted and decreased the pressure against the aluminum bar.

I discover that puting two layers of kapton tape instead of the grey rubber style band on the alu bar of the controller help more for the heat transfet ( better thermal transfet performances) but the kapton tape is not really compressible as the grey band.. so the pressure from the mosfet to the alu bar is more critical with the kapton. a quater turn of screw make more pressure difference on the kapton than on the grey band.. so the thightening of teh screw MUST be done carefully to let each mosfet transfert the same amount of heat to equalize the current transfer over all mosfet trio.

Now i repaired the controller and it work fine!.. I did some more km tonight and i've reached again 103-104kmh burst on flat... and as well.. this setup is REALLY STABLE and IMPRESSIVE !!

i'm sure this ebike style and performances is just like the ebike everyone are dreaming!!

LIPO ARE SO NICE !!! LFP I 100% AGREE ON EVERYTHING YOU SAID ABOUT THEM !!

I charged them ( the 12 packs) to 4.15V per cell using my meanwell 1700W DC PSU and it REALLY ROCK !!!

Charging these 20C LiPO with 1.3C ( parallel 44.4V 15Ah ( 30Ah)) with 38A current was so intense!!.. The battery just cooled as the SOC raised !!! and at the end they was at ambiant !! :| :wink:

I saw this phenomenon occuring often on my setup... The lithium battery cool down as they get charged !!!.. I perceive that they cool down faster when charging than just leaving them not connected...

Doc

 

[John in CR]

Doc,

So you have batteries with a negative internal resistance. I'd believe 140kph before I'd believe that.

Regarding the bike, I'm waiting for the permanent battery containment before poring on the praise.

 

[karma]

would someone just rewind a x5 already im dying here :wink:

doc did you make any changes to the relays?

 

[Doctorbass]

would someone just rewind a x5 already im dying here :wink:

doc did you make any changes to the relays?

still not.. i m leaving it in the Parallel ( 5303) mode and it work well but the controller heat up but it<s ok from now... no REAL danger..

it<s incredible how fast this ebike can go from 40 to 80 kmh !!!!! The 80 to 106 is still impressive too...

Ahhh.. the 0 to 106 kmh is ALL impressive!!

I really wanna drag race with it and put some Maxxis slick tires!!

i m convinced that i could do under the 18 sec on the 1/4 mile! you agree LFP, John ?

 

[liveforphysics]

i m convinced that i could do under the 18 sec on the 1/4 mile! you agree LFP, John ?

One good way to find out. Take it to the track!

 

[Doctorbass]

i m convinced that i could do under the 18 sec on the 1/4 mile! you agree LFP, John ?

One good way to find out. Take it to the track!

I realized that what limit myu speed is now the current limit and not the voltage. That make sense since i<m using a low count winding on a large diameter wheel.

i<m using the CA-HC and i needed to update it to High current mode because the max current limit in the low mode was 99A and .. that<s what limited my speed!

now i<m in high current mode and i<ll try step my step current increase.

now it is set to 120A and that little 20A seem to make a huge difference!

yesterday i<ve climbed a big hill at 55mph!!! .. but when i arrived on the top, a 1/8 mile later, the motor winding was to 180 celsius!!!

but hey.. 55mph while going up hill in electric bicycle is nice!!!

i<m convinced that if i put some 1.95" x 24 slick, rise the current limit to 150A i could break the 70MPH and enter in the 18sec.

but hey... 88MPH would be nice for a Doc !!!

 

[Arlo1]

i m convinced that i could do under the 18 sec on the 1/4 mile! you agree LFP, John ?

One good way to find out. Take it to the track!

I realized that what limit myu speed is now the current limit and not the voltage. That make sense since i<m using a low count winding on a large diameter wheel.

i<m using the CA-HC and i needed to update it to High current mode because the max current limit in the low mode was 99A and .. that<s what limited my speed!

now i<m in high current mode and i<ll try step my step current increase.

now it is set to 120A and that little 20A seem to make a huge difference!

yesterday i<ve climbed a big hill at 55mph!!! .. but when i arrived on the top, a 1/8 mile later, the motor winding was to 180 celsius!!!

but hey.. 55mph while going up hill in electric bicycle is nice!!!

i<m convinced that if i put some 1.95" x 24 slick, rise the current limit to 150A i could break the 70MPH and enter in the 18sec.

but hey... 88MPH would be nice for a Doc !!!

As the guy in your picture would say! Just make sure the flux capacitor is ok! :lol:

 

[Doctorbass]

For sure !! :wink:

 

[Doctorbass]

Well.. low count winding motor are really tough for controller!!!

I just blown the one on my DH ebike 4th time from now!!!

Each time one different set of mosfet...

I observed that it happen at a REALY PRECISE MOMENT:

-Battery 100% full ( ready to go)
-Controller not warmed up ( cold)
-2-3 first second of each new run ( cycle)

it's like the mosfet that are quiet and sleeping during the standby, REALLY DONT LIKE TO BE WAKED UP FAST !!!..

I blown the controlle all 4 time just at the begining of each run... when i go WOT dand during the 2-3 first second of each use...


I guess that in 5306 mode it would not happen and still push my ebike to 75-80kmh without any problems....

but hey.... frm 80 to 106 in the 5303 mode it's still really fun !! so... :|

Doc

 

[mwkeefer]

Doc,

When you blew the controller up the other 3 times (only one bank of FETs per right), what did you use for replacement FETs?

I'm talking about matching the FETs between phases for evenly distributed / matching phase currents, I've tried to repair just "blown" FETs (or banks) and never with good results when used for "performance setups" like yours.

I've found the only way to really be "carefree" about the longevity of a rebuild controller used in more extreme setups is replacing all the FETs with new units from the same batch of production... in theory I suppose you could test the FETs RDSon and other specs and match them by hand.... Im thinking it would be much more work than I'm interested in doing

-Mike

 

[karma]

Doc,

When you blew the controller up the other 3 times (only one bank of FETs per right), what did you use for replacement FETs?

I'm talking about the between phases for evenly distributed / matching phase currents, I've tried to repair just "blown" FETs (or banks) and never with good results when used for "performance setups" like yours.

I've found the only way to really be "carefree" about the longevity of a rebuild controller used in more extreme setups is replacing all the FETs with new units from the same batch of production... in theory I suppose you could test the FETs RDSon and other specs and match them by hand.... Im thinking it would be much more work than I'm interested in doing

-Mike

matching fets works well for amplifiers. last time i blew my controller i replaced all of them. its open anyway. fets are sensitive to almost everything.

doc do you have a picture of the controller and how there Mounted ?

 

[mwkeefer]

karma - is that a Pot Leaf icon in your Links section? {lol, cough} -Mike

 

[karma]

ya looks like it lol . i just haven't had the time to change it. i should i don't want people to get the idea im promoting drugs or something

 

[Doctorbass]

Doc,

When you blew the controller up the other 3 times (only one bank of FETs per right), what did you use for replacement FETs?

I'm talking about matching the FETs between phases for evenly distributed / matching phase currents, I've tried to repair just "blown" FETs (or banks) and never with good results when used for "performance setups" like yours.

I've found the only way to really be "carefree" about the longevity of a rebuild controller used in more extreme setups is replacing all the FETs with new units from the same batch of production... in theory I suppose you could test the FETs RDSon and other specs and match them by hand.... Im thinking it would be much more work than I'm interested in doing

-Mike

I never matched any fets. When a bank blow i replace all them that are in parallel. on my 18 fets controller i replace them multiple of 3...

matching RDS would be nice... but i think that something i did that help performances could also create trouble:

I used 2 layer of kapton tape instead of the grey strap as fet insulator. The kapton have a better heat transfet than the grey stuff.. but it is less "spring loaded".. I mean the grey stuff is more thick and also equalize better the pressure on all fets i a bank so if one of the screw of one fet is less thightened than the 2 others it dont make a huge difference.. BUT using kapton tape that is less compressible make the thightening of the screw a biger challenge cause it is not "spring loaded" so it can't really compress so one fet with a bit less torque on the screw make a bigger difference on teh heat transfer...

I really think i need something strong!! like a 36 fets!!! 4110

What is limiting my speed is not voltage.. IT'S CURRENT !!

and for a 18 sec run the motor can take it and probably another 5kW



Last night i ran my ebike up a big hill.. and hit 55MPH while climbing!!!!

YES.. 55MPH in up hill !!!

The motor was already at 110 celsius when i started down to the hill... when i arrived up the hill it was at... 182 celsius!!! and... 55mph... and 13kW continuous power for 20sec during the run!!!

Doc

Doc

 

[liveforphysics]

You need to be using belleville washers on your FETs. Then they don't have to be warm to expand and get tensioned, they will always have tension.

http://en.wikipedia.org/wiki/Belleville_washer

Due to the nature of the fet loading being inherently unbalanced in a controller (the high side and low side have different heat loads from the intrensic diodes clamping freewheel currents), you've really got not advantage to matching all fets in a controller. Just matching each parallel group is what counts, and in these controllers where the gates are tied common with no resistance between them, if you were going to match something, the RdsOn would be the least critical, matching the gate voltage plateau (Miller Plateau) would be most critical part to match.

 

[Doctorbass]

You need to be using belleville washers on your FETs. Then they don't have to be warm to expand and get tensioned, they will always have tension.

http://en.wikipedia.org/wiki/Belleville_washer

Due to the nature of the fet loading being inherently unbalanced in a controller (the high side and low side have different heat loads from the intrensic diodes clamping freewheel currents), you've really got not advantage to matching all fets in a controller. Just matching each parallel group is what counts, and in these controllers where the gates are tied common with no resistance between them, if you were going to match something, the RdsOn would be the least critical, matching the gate voltage plateau (Miller Plateau) would be most critical part to match.

Ok I know about these belleville i used some in many project at work for aerospace optics.

Each time my one of my controller blow i recover some 4110 on the first version 18 fets controller i assembled but never used ( not instant start)

They come from IR directly but i admit i never tested the miller plateau.

Could be nice... or maybe just using a 36 fets controller with less lead on the trace of the drain and source to make a kind of shunt to help balance the load... just like i the principle of the good old class AB amolifier output transistor.....

Doc

 

[Arlo1]

Welcome to the wonderfull world of lipo doc!

 

[curious]

Doc -
The problem of paralleling mosfets is the temperature dependence of Rds (to a much larger degree then Rds variation in a batch at room temperature). In a group of parallel mosfets the one with slightly lower Rds will conduct more current and heat up more, then it's Rds further drops relative to others and the cycle repeats until that device is gone. This process happens faster than thermal equalization feedback through the common heatsink. One solution is to mount all mosfets in a parallel group on a small but thick copper bar without isolation (using only silver thermal compound) and then mount that bar using isolating separator of choice on a common heatsink. Doing so will substantially speedup local thermal equalization within a group. Plus the increased area of isolating separator between the spreader copper bar and the heatsink will greatly reduce overall thermal resistance.

 

[Doctorbass]

to mount all mosfets in a parallel group on a small but thick copper bar without isolation (using only silver thermal compound) and then mount that bar using isolating separator of choice on a common heatsink.

That's a good idea.

Now let see how we can make that in that 18 fets box... or 24 fets... or... 36 fets ( if they still exist!)

Doc

 

[karma]

its funny i got into a discussion with ken at its electric.ca over how fets are mounted in controllers. not to long ago i had a few ideas. he told me there trying to keep the controllers small

sounds like a good idea :wink: if you can find a 36fet controller

 

[curious]

I have to correct myself. The dynamic thermal runaway condition is only true for mosfets used in linear mode (power amps etc.) and during switching transient (when Rds > Rdson). Once you turn the device on completely Rds=Rdson increases with temperature - steady state conduction should be runaway safe. So if the gate drive circuit is sufficiently fast and the dominant factor of heat dissipation is device Rdson (not the transient Rds) then there should be another explanation for the failure (perhaps quiescent current spikes or armature saturation).

That said it is still helpful to use heat spreader per group of parallel mosfets. Reduction in overall thermal resistance can be very substantial.

BTW is that the low side or high side mosfets that fail ? If the problem is with the low side (and usually only low side has high-freq PWM) then depending on the gate driver circuit mosfets could be transient limited not Rdson limited. In that case my original hypothesis is valid.

 

[Doctorbass]

BTW is that the low side or high side mosfets that fail ? If the problem is with the low side (and usually only low side has high-freq PWM) then depending on the gate driver circuit mosfets could be transient limited not Rdson limited. In that case my original hypothesis is valid.

It's on the High side ( positive) 80% of the time

 

[steveo]

Hey Doc,

could you tell use which mosfet blew? ... my bet is that everytime you blow them its high side fets!

the most heat i find that comes from the controller is the high side fets in the center of the bus bar... let me know..

i sent you a pm for 24 fet pcb

-steveo

 

[Doctorbass]

Hey Doc,

could you tell use which mosfet blew? ... my bet is that everytime you blow them its high side fets!

the most heat i find that comes from the controller is the high side fets in the center of the bus bar... let me know..

i sent you a pm for 24 fet pcb

-steveo

Yes it's on the High side .. and 50% time on the center or on the end of the bar.

Doc

 

[fechter]

Interesting that the high side tends to blow. Reminds me of the 9 FET controllers with 2 on the high side and one on the low side. Maybe they actually did know what they were doing. I suppose the high side ones are doing most of the freewheel rectification during switching, which could explain why they take more stress.

There could be some inductive ringing that' s zapping the high side too. Core saturation is almost a certainty with Doc's setup. I think the devices have a dissipation limit that's not really thermally determined too. If it tries to dissipate over a certain amount even for a microsecond, I think you can blow them instantly.