Opened My 35a controller and... SIX! x irfb4310 Fets STOCK

[BiGH]

i originally posted this in my build thread, but thought a lot of you might be interested in this (and not so much in the build so i separated them out.)

I opened my controller tongiht for the first time (to plan where the 8ga wire is going to get routed, and found a NEW controller board with 6 (still has room for 12) irfb4310 fets. The irfb4310 fets though eases my mind a lot about my chances of blowing the controller.

The board text is:

RFW4850-V06 07-05-11 indicating a May 07 revision.

To mount another 6 fets would require drilling of the heat sync. (to fit the 8ga will require drilling of the end plates).

now for the pics:


<img src="http://www.hnrglobal.com/ebike/35a%20controller-1.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-2.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-3.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-4.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-6.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-7.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-8.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-9.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-10.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-11.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-12.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-13.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-14.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-15.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-16.jpg" width="512">

<img src="http://www.hnrglobal.com/ebike/35a%20controller-17.jpg" width="512">

 

[fechter]

The gate driver might not be able to handle two 4310's per section. This could be tested with an oscilloscope.

What is the big, many legged chip on the board? Can you read the number?

 

[BiGH]

The gate driver might not be able to handle two 4310's per section. This could be tested with an oscilloscope.

What is the big, many legged chip on the board? Can you read the number?

i'll be opening it up again soon - its back togehter now ready for the test ride later...

i'll let you know- there are two ICs on the board. i'll try to get some very good pics of it tomorrow (time permitting).

At the time i thought "this is a bit strange and doesn't match other's controllers", now i'm fully aware of its newness. so i'll rip it open and get as mcuh info as people wanna know.

Are there any other part numbers i should look for?

 

[xyster]

Here's the FET spec sheet:
http://www.irf.com/product-info/datasheets/data/irfs4310pbf.pdf

100V 130A it looks like. So from what Fechter's saying this new controller may not be as easily mod-able as the 12-FET controller.

Would a IRFB4110 upgrade still be worth it?

How much current could one run with 6 IRFB4110s versus these 6 IRFB4310?

Could this controller safely run more amps by just modifying the shunt and/or current sensor resistors?

 

[cadstarsucks]

The gate driver might not be able to handle two 4310's per section. This could be tested with an oscilloscope.

What is the big, many legged chip on the board? Can you read the number?

My suspicion would be that it was designed either for older FETs or for alternates when the preferred ones were not available.

Dan

 

[Jozzer]

I just took my 20A controller out for a spin at 82v and 35A. It too has 6 4310's (upgraded myself). I understand these fets can handle more current, but have a higher resistance (though still much better than the cheaper fets that have been having problems). I would bet that 45 or 50 amps would be ok, but i'll bet on YOUR controller...you try it first and let me know if it smokes , then i'll try more :lol:
So is this the new controller that Crystalyte has put together at Justin's recommendation, or some old controller that got put together differently?

 

[BiGH]

I just took my 20A controller out for a spin at 82v and 35A. It too has 6 4310's (upgraded myself). I understand these fets can handle more current, but have a higher resistance (though still much better than the cheaper fets that have been having problems). I would bet that 45 or 50 amps would be ok, but i'll bet on YOUR controller...you try it first and let me know if it smokes , then i'll try more :lol:
So is this the new controller that Crystalyte has put together at Justin's recommendation, or some old controller that got put together differently?

not sure if its the one justin reccomended, but its a very recent revision on the board. The text on the board was 4850 - which indictes its use for 48 v 50amps (if you follow the old naming convention of the c-lyte controllers)

I popped it open again. I also noted it only has TWO busbars.

Small IC writing: Low Power Dual Op amplifier
http://www.bcdsemi.com/ASP/productpic/2005471756146518.pdf
AS358m
707-j65

Large IC: Mixed-Signal Array with On-Chip Controller
http://www.datasheet4u.com/html/C/Y/8/CY8C24123A_CypressSemiconductor.pdf.html
CY8C24423A
24PVX10713
B 02 CYP 623 788

- wow!

oh and my controller failed for the first time - the solder joint to the main on off switch broke

 

[fechter]

Would a IRFB4110 upgrade still be worth it?

How much current could one run with 6 IRFB4110s versus these 6 IRFB4310?

Could this controller safely run more amps by just modifying the shunt and/or current sensor resistors?

4110's are only slightly better than 4310's. They do have a lower gate charge, so if you wanted to install 12 of them, it would have a better chance of working.

With a six pack of 4310's I wouldn't go over about 40 amps. They might take more, but I wouldn't bet on it.

Interesing about the CY8C24423A. This is how a controller should be built, as long as the code is good. Hacking might be a bit more difficult.

 

[BiGH]

Would a IRFB4110 upgrade still be worth it?

How much current could one run with 6 IRFB4110s versus these 6 IRFB4310?

Could this controller safely run more amps by just modifying the shunt and/or current sensor resistors?

4110's are only slightly better than 4310's. They do have a lower gate charge, so if you wanted to install 12 of them, it would have a better chance of working.

With a six pack of 4310's I wouldn't go over about 40 amps. They might take more, but I wouldn't bet on it.

Interesing about the CY8C24423A. This is how a controller should be built, as long as the code is good. Hacking might be a bit more difficult.

hmm interesting - in that case i think i'll stick with stock for the moment. will upgrade the wiring anyways, but won't touch current limit (there are only two of the bars to control it insted of 4.

reprogrammable controller would be awesome to have! I'd love to see a custom endless-sphere controller. using a PIC or something with USB and datalogging. The more simple we could produce a controller the better. i'm not sure if its possible to simplify down the actual electronics of the controller. could hall sensors insted of applying to circuitry, be sampled by an IC? and controlled via that?

will have to wait and see too how hot it gets at 72v 35a. - should be more than enough speed for me anyways .

With the CY8C24423A - i was happy but also sad to see the introduction of this. I get the feeling this new controller is going to be more robust, but not as capable of extreme modding as the 12fet old controllers. ie lower max current. it also makes it dificult to see how it works.

Fetcher - any ideas whether it would be possible to put more fets in? you mentioned something about the driver not supporting it... could this still be a possibility?

 

[fechter]

It depends on the driver and the FETs. It could be tested by installing one high side FET and one low side FET, then watching the gate signal on an oscilloscope. Without a o'scope, I don't know how you could check the drive. I suppose the controller might get really hot if the switching gets slow.

 

[cadstarsucks]

It depends on the driver and the FETs. It could be tested by installing one high side FET and one low side FET, then watching the gate signal on an oscilloscope. Without a o'scope, I don't know how you could check the drive. I suppose the controller might get really hot if the switching gets slow.

Very true... Is the thought here on designing a hobbyist buildable controller for a power house ride?

Dan

 

[BiGH]

It depends on the driver and the FETs. It could be tested by installing one high side FET and one low side FET, then watching the gate signal on an oscilloscope. Without a o'scope, I don't know how you could check the drive. I suppose the controller might get really hot if the switching gets slow.

Very true... Is the thought here on designing a hobbyist buildable controller for a power house ride?

Dan

I was thinking a bout that today - having a home built kit available to contorl these motors with the best components (ie 4110's) etc to allow large voltages to be run.

I'm still in the stage now of learning more how the controllers work and the ideas behind electronic circuits. (im a digital man).

However i do wonder if it would be possible to have a PIC run the phases, and adjust the timing etc. would this allow a higher top speed- by advancing the phases to take into account increased rotational speed? would it be possible to make a "digital" contoroller? Ie an output from a pic that switches on and off the power in each? can we minimise the amount of components on the board by doing this?

I'd learn more about it but i have 3 essays (7500words, 2000 words and 2000 words)to write at the moment and some prac reports for uni, plus exam marking is coming up (for work)

 

[cadstarsucks]

I was thinking a bout that today - having a home built kit available to contorl these motors with the best components (ie 4110's) etc to allow large voltages to be run.

I'm still in the stage now of learning more how the controllers work and the ideas behind electronic circuits. (im a digital man).

However i do wonder if it would be possible to have a PIC run the phases, and adjust the timing etc. would this allow a higher top speed- by advancing the phases to take into account increased rotational speed? would it be possible to make a "digital" contoroller? Ie an output from a pic that switches on and off the power in each? can we minimise the amount of components on the board by doing this?

Yes and no. It is possible but impractical with a PIC. I would prefer something with a bit more power (and lower cost) or an external sine generator. Of course BLDC normally runs chunky anyhow, but a sine generator will smooth out the torque at low speeds.

Dan

 

[fechter]

I think a PIC might do it depending on the code. Most of the RC controllers use PICs.
Maxwell built one that had adjustable timing. It did not improve performance of a hub motor much, but for a geared motor running high rpms, it might extend the high end a bit.

A TI DSP processor would be ideal for motor control. They even have sample code on their website. Pretty much everything could be done in the processor leaving just the power stage. The power stage from an existing Xlyte or similar controller should work.

I'm sure we'll see more designs like this in the future. The code part is beyond my skill at this point.

 

[The7]

Maxwell built one that had adjustable timing. It did not improve performance of a hub motor much, but for a geared motor running high rpms, it might extend the high end a bit.

Could find Maxwell's test results on his adjustable timing controller!?

From my understanding, the relative phase shift (spacing) between Hall sensor and motor winding is designed such that the power factor for the 3-phase motor is (nearly) unity.
The motor current will be mininum if the power factor (pf) is unity
If the relative phase shift is increased or decreased ( by electronic means or mechanical spacing means), the power factor will be less than unity. The motor current will increase and has more losses.
Theorectically the rpm will increase at the expense of poor p.f., higher motor current, higher motor losses.

Some simple maths:
pf = 1 gives 0 deg advance.
pf = 0.8 will give (arccosine 0.8) 37 deg advance.
Suppose 100V 10A to run 100 rpm at pf=1.
By phase advance, you may need 80V to run 100 rpm.
The pf of the motor becomes 0.8.
But the motor will take (100X 10)/80/0.8 = 15.6A.
The motor current is increased by 56%.
The motor losses will be greatly increased.
This calculation is only for illustration purpose.
(Factors of 1.73 and actual phase between current, drive voltage and back emf have to determined with known motor inductance and resistance).

In reality, although the spacing of Hall sensor and motor winding is fixed, it is impossible to obtain unity pf at the whole range of rpm due to the motor inductance. The reactance of motor inductance increase with rpm (drive frequency).

 

[oatnet]

Wow, timely info.

I ordered a batch of samples from Kenny a few months ago, and finally popped upen a box with one of these controllers inside. It was alleged to be a 72v40ah controller, but after working on the 72v40ah controller I did the CA install on, I was suprised by the lack of FETs. I also saw the 4850 designation and wondered if this was actually a 48v50a controller...

What v+a have you put through it?

-JD

PS here are pics of mine:

 

[BiGH]

i've only put 36v 35a throuhg it (not 100% sure on the amps tho)

will let you know how it fairs at 72v when my batteries come in

(paying for them today) - they should be shipped soon.

 

[The7]

i've only put 36v 35a throuhg it (not 100% sure on the amps tho)

will let you know how it fairs at 72v when my batteries come in

(paying for them today) - they should be shipped soon.

V

According to the current multiplication, 72V battery will give double the motor current at low rpm and high torque when compared with 36V battery. So the FETs will work double harder.

 

[BiGH]

thinking about running higher voltages on this stock controller - what are your views on running 96v using LiFePO4s? is that too dangerous in terms of the raitings of components?

maybe i should try 84v first...

 

[cadstarsucks]

i've only put 36v 35a throuhg it (not 100% sure on the amps tho)

will let you know how it fairs at 72v when my batteries come in

(paying for them today) - they should be shipped soon.

V

According to the current multiplication, 72V battery will give double the motor current at low rpm and high torque when compared with 36V battery. So the FETs will work double harder.

Actually 4X... power is I²R.

Dan

 

[justin_le]

Interesing about the CY8C24423A. This is how a controller should be built, as long as the code is good. Hacking might be a bit more difficult.

I actually managed to get the proper schematic for these new controller boards which saved a day or two of following PCB traces, but I don't think I have permission to post it.

To summarize though, it is quite a different design:

1) The 15V bus comes from a DC-DC converter rather than a linear regulator, this is certainly a step in the right direction. The buck converter is done entirely with analog circuitry and discrete transisitors, no IC or anything so it's hard to say exactly what it's rated for and if there's enough power to draw bike lights from this bus.

2) The main control is now done with a Cypress Semiconductor microchip rather than an analog PWM regulator and hall effect decoder. Obviously this gives a huge amount of flexibility for modding the controller but it's now in the programmer's domain rather than the soldering / hacking domain. It would be trivial to add things like proportional regen and other intelligence, and a collaborative open-source code project could be a fun pursuit.

3) There appears to be both pulse-by-pulse current limitting as well as average battery current limits. The LM358 dual op-amp U1 senses the shunt current, with the lower op-amp (pins 1,2,3) acting like a comparator and going to a digital I/O on the microchip, while the upper op-amp (pins 7,6,5) configured as non-inverting amplifier circuit and going to an analog input. Both of these limits could still be tweaked or scaled with resistors and pots without getting into the code. If it has a fast enough PBP current limit, then that could mean the END of fried mosfets. Someone would have to check with an oscilloscope to see.

4) The gate drive circuit is now done with discrete transistors and diodes rather than a mosfet driver IC. I suppose this is cheaper if there is no labour cost to solder the dozens of extra parts it requires (23 parts per half bridge driver if I counted right). It also makes it harder to say what the actual gate drive current is like without a pen and paper and some spare time for calculations.

5) The throttle input, ebrake line, etc. are all now simple items which go straight to various inputs of the microchip, there is little if any signal conditioning on these beyond pull-up and pull-down resistors and filter capacitors.

6) The resistance against water damage and the elements appears no better than the previous crystalyte controllers. No conformal coating on the circuitboard, no sealing of the wire grommets or side plates. So those riding in wet weather will still have to keep their controller in the dry.

I'm glad that they've now decided to make the 4310 mosfets stock. I suggested to Kenny that he do this a few months ago and also sent a batch of 4110 mosfets so that he could prepare some controllers with a 50A limit.

Anyways, we'll see how this new controller circuit pans out, the old design was long due for retirement IMO. Also, the extruded enclosure looks suspiciously like the one for the regen controller on the golden motors site, so it might be a board design that's become public domain in china.


Justin

 

[BiGH]

awesome info there!

Thanks Justin

now the question begs to be answered... how do i soldier a Cycleanalyst to it

 

[justin_le]

awesome info there!

Thanks Justin

now the question begs to be answered... how do i soldier a Cycleanalyst to it

Not sure if this was tongue in cheek, but anyways...

The shunt and Vbatt should both be pretty obvious, but naturally you can't use the ebrake line for the throttle over-ride because it's a digital signal in this controller and the bike would simply jerk on and off.

Attached is the circuit mod for where to attach this line. Basically there was an unused resistor (R39) after the throttle signal (SP) which went to an unused pad (XS+) on the PCB, and this can be replaced with a SMT diode and then the pad becomes the new ebrake for the CycleAnalyst.

Most motor controllers have some series resistor after the throttle signal, so it's usually pretty easy to figure out a tidy way to wire in the CA throttle over-ride. But the diode is crucial.

Justin

 

[BiGH]

actually it wasn't - i'm in the market for a Ca - but can't decide if i will get the external shunt or risk my stupidity in wiring up one on this controller.

I'm assuming you'll be selling pre-modified 35a new model controllers soonish? i might just wait till then and get one premodded

Take Care,
Haydon

 

[fechter]

Great info!

Where's the rest of the schematic? I want the whole thing!