Infineon transistor modification

[geoff57]

Hi
The transistor modification has gone through several stages of evolution, starting with a combination of an ideal controller that Knuckles wanted and a circuit design to achieve that by Fechter.

Original circuit diagram by Fechter

The Darlington transistor in this was found to be overkill so a revised design was done, made and tested by Knuckles using a 3055 transistor

This worked fine up to a point, it had one major drawback in red HEAT the transistor got hot, within working temperatures of the transistor. The temperature was above 60 degrees C once it was doing its job. The Infineon’s sent out with this modification worked but we still did not like the heat the transistor was giving off.
This caused Knuckles to brainstorm from shortly before Christmas and come up with a new circuit design that after a few revisions became the circuit below.

When it comes to making these modifications knuckles and I differ he has chosen the approach of joining the components together off the board and then mount them on the board.

I decided to take the modular approach to this idea and make a small sub board that with legs that slot straight onto the main board where the power resistors go.


Geoff

 

[geoff57]

Hi
The module circuit board is a small piece of prototyping board with 0.1” pitch holes called tripad board that a UK electronics company called Maplins supply, this is like stripboard but the copper pads are split every three holes making a board with holes in with copper joining just three holes instead of the normal row. I have searched high and low for anyone else that makes this type of board no one does, not to worry the same thing can be done with strip board or any prototyping board with 0.1” pitch holes as long as it has copper pads on.


Parts list
1x 100 ohm 2 W resistor
1x TIP122 Darlington transistor
1x 36v 5W zener diode
1x 15v 5W zener diode
1x high speed switching diode 1N4148
1x 15K ohm 1/2 W resistor
1x jumper wire
1x grounding wire
2x 220 ohm 2W resistors
2x 470 ohm 2W resistors
The last 4 resistors are placed on the module in pairs that will become obvious later Knuckles uses 200 ohm and 500 ohm making 2 700 ohm 4W resistor bridges, but these values are unavailable to me so I used 220 and 470 making 2 690 ohm 4W bridges. The values are within tolerance levels and these values are more readily available whereas 200 and 500 is not so common.

 

[geoff57]

Hi
Tomorrow I will build a module one step at a time taking pictures as I go

Geoff

 

[philf]

OK.

I'm a little perplexed. The first diagram in that series is almost EXACTLY what I was about to try (Using either a 2N6045 or TIP122 - the value of the resistor is where the "almost" comes in. I was going to start with 10K, but wasn't sure if that would be the final value).

It overheats?

That's surprising...

What might happen if a zener fails is a big concern of mine. Less is more, in this case.

 

[fechter]

I think I used 20k. If the pass transistor is a darlington, the zener current can be very low.
The zener and pull up resistor shouldn't be getting very hot.

It's the pass transistor that will get hot. Very hot. A heatsink would be the best solution, but that adds some mechanical issues. Space is limited and whatever you do needs to withstand a lot of vibration.
With an adequate heat sink, nothing will get very hot.

By adding some resistors and a second stage, Geoff is spreading out the heat dissipation over a larger surface area to avoid needing a heat sink.

 

[geoff57]

Hi
you did say 20K fecher yours is the first circuit diagram.
we used 10K with a 3055.
this idea is NOT mine it is Knuckles circuit design credit where credit is due, my contribution was to design a self contained module from his idea that will be easy to fit once the power resistors have been removed and the holes cleaned up of solder no track cutting drilling the board or anything.

Geoff

 

[philf]

Admittedly, last time I used such an arrangement as a power supply was about 30 years ago. The Darlington was in a T0-3 case, mounted to the outside of the metal project enclosure (with a mica insulator). It didn't even occur to me that it might have run hot otherwise. I just had the transistor and mounting hardware and just used 'em.

Happy accident, I guess

The alternative, here, does look like a bit of a mass to have bobbling about on top of the board, though. Some supporting silicone or hot glue might be in order...

 

[geoff57]

Hi
The module is on 5 legs it would have been 6 but that would require a track cut to stop a short circuit, the legs are the wires at the end of the 2W resistors nice and thick not the thin wire of a high speed diode.
The module stands proud of the main board by enough to stop the solder connections on the module touching the main board about 2 to 3 mm clearence is overkill but allows for the exact match in hole spacing as well (approx 0.5mm), the overall hight of the module is the same as the 2 big capasitors.
The components on this module are as sturdy as any conponenent soldered directly to the main board.

Geoff

 

[ZapPat]

It's the pass transistor that will get hot. Very hot. A heatsink would be the best solution, but that adds some mechanical issues. Space is limited and whatever you do needs to withstand a lot of vibration.
With an adequate heat sink, nothing will get very hot.
By adding some resistors and a second stage, Geoff is spreading out the heat dissipation over a larger surface area to avoid needing a heat sink.

Let's have a bit of fun analysing a controller's different heat dissipation factors.

First is to figure out the power dissipated by the linear voltage regulator setup (counting all parts) to go from maybe 90V down to 15V.
Linear regulator: (90V-15V) * 60mA = 4.5Watts of heat to be dissipated.
This is to provide the circuitry with this much power:
Controller's logic section: 15V * 60mA = 0.9 Watts


Now just for fun let's compare this 4.5W of constant waste heat generated to the controller's own power losses in the FETs. I will consider here a 12 FET controller version using IR4110's (at about 2.5mohms paralleled Rds when hot), runing at 90V with an average of 30A battery current. We'll look at both the full duty cycle and then the partial duty cycle functioning modes of the controller.
EDIT: IRF4310 FET based dissipation has been added in brakets after the 4110's... (3.5mOhms used for these @ ~60oC)

Full throttle, non-current limiting mode - FET heat dissipation
(when battery current is under controller cutoff current, at high speeds and wide open throttle)
--> 100% Duty, Ibatt = Imot = 30A, Vmot = Vbatt = 90V
FET conduction losses (two phase FETs full time): I^2 * (Rds * 2) = 30A^2 * (0.0025Ohms * 2) = 9 Watts
FET switching losses (commutation frequency ~ 300Hz): < 1W
Total losses: 9W + 1W = 10W [13.5W - IRFB4310's]

Partial throttle or current limiting mode - FET heat dissipation
--> 50% Duty, Ibatt = 30A, Vbatt = 90V, Vmot =~ 45V, Imot =~ 60A, fsw = 10kHz (assumed), tr = tf = 400ns(assumed)
FET conduction losses (unswitched phase, LS FET): Imot^2 * Rds = 60A^2 * 0.0025Ohms = 9 Watts [12.5W]
FET conduction losses (switched FET, HS): Imot^2 * Rds * Duty = 60A^2 * 0.0025Ohms * 0.50 = 4.5 Watts [6.25W]
FET conduction losses (switched FET, LS): Imot^2 * Rds * (1 - Duty) = 60A^2 * 0.0025Ohms * (1 - 0.50) = 4.5 Watts [6.25W]
FET switching losses (switched phase): 0.5 * Imot * Vbatt * (tr+tf) * fsw = 0.5 * 60A * 90V * (400ns + 400ns) * 10 000Hz = 22 Watts
Total losses: 9W + 4.5W + 4.5W + 22W = 40W [47W - IRFB4310's]

From this we can see that high power use at full throttle is much easier on the controller than when using that same power at partial throttle, as long as we are working in the non-current limiting region of the controller. 4 times more at 30Amps! Lower duty cycles and higher amp draw would make the 40W figure go up quite fast, just as conservative driving will lower it.

So if we have maybe 25W average dissipation in regular controller functions, the 4.5W in the linear regulator is about 15% of the total controller's power dissipation. When cruising at full throttle (and are using less current than the controller is limited to), that 4.5W gets to around 30% of the total heat dissipation in the controller in this case! Maybe I'll strap a few special power resistors I have to my infinion's heat sink to see how hot 25W gets the controller housing, since I have no idea what the thermal resistance of the infinion's case is.

A PWM switching regulator would lower that 4.5W constant to about 0.7W (at 82% efficiency) - quite a difference, but is it worth it? One of the common arguments against a switcher is complexity, but seeing how complex this linear circuit has become just to cope with it's own waste heat, I'm not sure this is the case anymore. The PWM chips I've looked at are pretty simple to use. The second argument is cost of the PWM curcuit. I've checked and found that it would cost about three to four dollars of parts for the switcher in quantity - more cash than the linear solution here I'm sure, but how much more after heatsinking etc? From these musings, I think I'm going to be going the PWM regulator route for my own controller, since I've found the heat generated by the linear parts when working at high voltages gets very hard to manage.

Just some of my thoughts on the subject of voltage regulators in ebike controllers. Probably this whole post sounds martian to most ES'ers... but hopefully some might appreciate!

Pat

 

[geoff57]

Hi
Pat thanks for the input as I posted earlier I'm not the designer of the circuit, just the builder of the board/module.
That out of the way you have a point, space is limited however to the footprint of "the board" transistor1.jpg in post 2 and a max height of the top of the capacitors, the requirements is a method of dropping the voltage down from anything up to 90v down to about 15v with a current draw of a maximum of 80 mA, one additional problem is the 15v is a BUS so we want it kept to 15v or close, Knuckles was finding the bus getting as high as 18v, there used to be a 12v zener diode on earlier boards but the new Infineon board has no such diode, so a 15v was built into the circuit design as a replacement by Knuckles.
So if there is another way to achieve the same thing as this module does I am open to ideas.

Geoff

 

[Knuckles]

ZapPat's analysis (his theory) sounds dead on correct.

The only missing element (for the VR circuit) is when the controller is connected but the peripherals are not (ie throttle & motor halls not connected).

The “typical” ignition circuit draws about 30 ma with no peripherals connected.
The “typical” ignition circuit draws about 60 ma with all peripherals connected.
At half throttle (on-road), the “typical” ignition circuit draws more than 65 ma.

My goal was to design a linear VR circuit that was stable (at any battery voltage) under all these conditions and also to spread the heat around using by-pass resistors AND keep heat away from the transistor and on-board LM317 regulator. The circuit must also "survive" an intial current "spike" when the ignition wire (key) switch is intially turned on (charging the VR capacitors).

Two additional criteria were ...
1) (Vin-Vout) on the LM317 not to exceed 40V limit
2) 12V bus not to exceed 15.5V limit

also ... It may not be so easy to get a cheap (100 ma) switching regulator that can accept 20V to 90V with constant 12V(15V) output.
Keywin and I have been looking at switching regulators for a while now. No such "off the shelf" luck so far.

meanwhile ... all this "stuff" has to fit in a small little spot on the pcb ...
(but once soldered in it is nice and stiff and ain't bouncin' around)



PS ... This is what it looks like (72V50A stock Infineon LVC=60V) before I "Any Voltage" transistor modify the controller.

 

[Knuckles]

The core Voltage Regulator (VR) on the Infineon is the LM317 ...

View attachment 1
The V(out) pin connects to the 12V(15V) bus.
But a nifty R6 By-Pass resistor is used to take up extra heat.
This takes some of the heat load away from the LM317.
Without R6, the LM317 will "take the heat"
But the R6 value will also increase the 12V bus voltage if there is very small current.


Example: V(in) = 48V, V(out) = 12V, R6 = 600 ohm
The current thru R6 = I = V / R = (48-12)/600 = 0.060 amps
This is exactly want the "ignition system" wants to run the MCU and peripherals.
So the heat (current thru) the LM317 is zero.

But ...
At ... V(in) = 20V, V(out) = 12V, R6 = 600 ohm
The current thru R6 = I = V / R = (20-12)/600 = 0.013 amps
So now the LM317 is putting out 60-13 = 47 ma to make up the difference.
So the heat from the LM317 is now W = V x A = (20-12) x (0.047) = 0.376 watts.
No problem at all for an LM317 without a heat sink.

HOWEVER ...
If we connect the controller WITHOUT the peripherals ... then the current is small (30 ma)
V(in) = 48V, V(out) = 12V, R6 = 600 ohm, I = 0.030
V = I x R = 0.030 x 600 = 18 ... (48 - 18) = 30V YIKES! The 12V bus is now 30 volts! IMPOSSIBLE!
Actually what happens is the current increases as the 12V bus begins to rise above 12V.
Current is being wasted as the 12V bus rises to 13V, 14V, 15V, 16V, 17V etc.
This is why I use the 15V zener (MOD) to cap this upper limit.

The 15V zener does get warm when I test the controller with no peripherals attached.
(Do you really connect a controller to power with no motor connected? But ... it can happen)
A good design accounts for all possible client contingencies.

Anyway ... This "By-Pass" idea is a cool design so I tried the same "By-Pass" approach to a transistor.
But to play it safe I added a diode to the transistor emitter just to make sure it would not pass current backwards.
And I added a 100 ohm resistor in front of the entire circuit to cap the max current "spike" thru the circuit.

I chose 51V as the mid point between 90V and 15V.
90-15=75 ... 75/2=37.5 ... 37.5 + 15 = 52.5V ... Say 51V.

I decided on 700 ohm by-pass values (not 600) because I have plenty of 200 and 500 (2W) resistors laying around.

btw for the TIP122 transistor ... 10k - 20k seems OK. I used 15k (because I had those around also).

BOTTOM LINE ... The circuit passes 60 ma all the time (at higher voltages) whether peripherals are attached or not.
It also passes 60 ma at any voltage 20V to 90V and the minimum 12V bus is maintained and capped at 15.5V.
At lower voltages none of these components get hot at all. 90V was the goal, however, and 90V dominated my design concerns.

Peer review of this circuit is very much welcome.

 

[geoff57]

Hi
Here it is the post of how to make a transistor module to Knuckles circuit design, this will just slot in, in the place of R01 A, B and R6.

The whole thing is built on a small piece of stripboard with the tracks cut to make it work I use a Dremel with a fine tip to cut the track but not the board see board below.





The components are as follows
1x 100 ohm 2 W resistor
1x TIP122 Darlington transistor
1x 36v 5W zener diode
1x 15v 5W zener diode
1x high speed switching diode 1N4148
1x 15K ohm 1/2 W resistor
1x jumper wire
1x grounding wire
2x 220 ohm 2W resistors
2x 470 ohm 2W resistors

1.
Shape the diodes to fit the board, make sure to get them the right way round.






Install all the low components first, the high speed diode, 15K resistor and the jumper wire.



Next install the zener diodes.



First power resistor to install is the 100 ohm spike protector.


Only cut off the wire not over the resistor, the wire left uncut will attach to the main board in pad 1 in the last picture of this post.




The blurred spot on the right corner of the board is the leg pointing straight at the camera.
The rest of the power resistors are fitted next.



Now fit the transistor.


Once the transistor is soldered in place and the excess cut off bend over the rest of the resistor wires to form bridges cut them at this point so it looks like in the first picture below then run a bead of solder down the 2 bridges and the top is finished see lower picture.
more to follow

 

[geoff57]

Hi
here is the rest of the build.

Once the transistor is soldered in place and the excess cut off bend over the rest of the resistor wires to form bridges cut them at this point so it looks like in the first picture below then run a bead of solder down the 2 bridges and the top is finished see lower picture.



Finally fit the grounding wire

The orientation of the module is as below.
View attachment 1
The grounding wire is fed through the hole in the board marked blue below then soldered to GND.


Pad 1 ignition pad this is the pad the power first hits the circuit on.

Pad 2 this has no use other than strength it could be the leg could be cut off if wanted.

Pad 3 Vin to the LM317.

Pad 4 12/15v bus and Vout from the LM317.

Pad 5 at this point the voltage should be clamped to about 51v.

Pad 6 is not used.

As you can see there are tracks from pads 6 to 2 and 5 to 3, the track from 5 to 3 is used and is part of the circuit, to start with and the way knuckles had designed his transistor modification was to use pad 6 for the high speed diode of the circuit then cut the track to prevent a short circuit, I decided that since I was working on a sub board pad 6 was not needed so I designed the module without it and saved on having to cut tracks.

 

[philf]

A wee bit of a head-scratcher, but... HELL - if it WERKS!

Actually, the most "key" thing is that it fits the space available. The switching regulator schemes I've tried cramming in require a bit more real estate - and that leads to more fooling around with the positioning of stuff peripheral to the power supply. And then there's the additional noise...

Sometimes less is more...

Thanks for sharing the details of what you've done, Geoff!

 

[ZapPat]

Knuckles - I've revised my post above, changing mOhms to Ohms (thanks for spoting that), and also adding dissipation for a 12 FET 4310 based unit.

As for the small footprint switching supply, I am quite sure I've spotted a good solution that would use only a very small surface area, easily fitting into that space (about 20mm X 15mm is the available space by the looks of it - I'll have to open up my infinion to measure though). However, I have not done the PCB and tested it yet, since the solution involves surface mount parts of course - not easy to prototype. I will be using this switchter in my next controller PCB revison, and I'll let you know how it turns out. 5-6 times less heat to get rid of makes it possible for much smaller parts to be used. Also, being surface mount, there's less chance of vibration related problems.

Switching noise I don't think will be a problem at all, since there's already lots of noise in this kind of circuit anyways. Also, the 5V linear regulator still stays for the logic bus, and thus any noise present from the 12V switcher would be eliminated by this.

 

[fechter]

Yes, a swticher would solve the heat problem. I'll have to see if I can come up with something that's not too expensive. The stupid inductors tend to be very pricey. The switcher in the Crystalyte v.2 controllers worked (sort of). I think there might be a simpler approach though.

 

[ZapPat]

Yes, a swticher would solve the heat problem. I'll have to see if I can come up with something that's not too expensive. The stupid inductors tend to be very pricey. The switcher in the Crystalyte v.2 controllers worked (sort of). I think there might be a simpler approach though.

What where the problems with the crystalyte's switcher? Noise or something else?

As for inductors, I've found an appropriate one for about 1$ in large quantities (1.70$ in units though); or a cheaper, non-shielded, and larger one for 1$ in units, but not much difference in larger quantities (0.80$).

 

[geoff57]

Yes, a swticher would solve the heat problem. I'll have to see if I can come up with something that's not too expensive. The stupid inductors tend to be very pricey. The switcher in the Crystalyte v.2 controllers worked (sort of). I think there might be a simpler approach though.

What where the problems with the crystalyte's switcher? Noise or something else?

As for inductors, I've found an appropriate one for about 1$ in large quantities (1.70$ in units though); or a cheaper, non-shielded, and larger one for 1$ in units, but not much difference in larger quantities (0.80$).

hi
Pat have you any more information on that, pdf spec files a part number i'd like to have a look at an alternative way of doing it as i said I just built from the circuit design I'm open to improvment.

Geoff

 

[fechter]

What where the problems with the crystalyte's switcher? Noise or something else?

As for inductors, I've found an appropriate one for about 1$ in large quantities (1.70$ in units though); or a cheaper, non-shielded, and larger one for 1$ in units, but not much difference in larger quantities (0.80$).

No noise problems, but I remember a couple of them smoked.

Yes, let's hear more about cheap inductors... We can design the rest around that.

 

[geoff57]

hi
for thoese faced with the old style Infineon with R01 ABCD and R6 for power resistors here is how to fit the module to the board.

this is what we are faced with, now on the original I said PAD/PIN 2 is not used

so this pin on the module is cut. jumper the pads shown

the module then fits on the board ,the grounding wire still goes through the hole, make sure of two things one no shorts between the module and the board, two there is clearence male sure the transistor is not touching the case.

Geoff

 

[philf]

I like the way you've arranged this thing to fit the holes on the board. Looks mechnically very good.

But after being brought to realize the heat problem with the "simple" linear solution that the thread started with, my interest has been rekindled in "round two" of the switching supply module. Last summer, I *did* get a switcher working just swell on one of my old GM controllers, and using less parts than this new linear project is up to now. However, the LM2575HV I used as the core of the thing has a rated maximum input voltage of 60V. This did what I needed it to do - allowing me to easily switch between 2 and 3s Yardworks batteries (with 3 in series, I was over the 60V a bit, but the thing never complained). The beauty was that the whole circuit was one T0-220-5 case, plus a cap, a diode, and a choke.

There are other switchers out there, as ZapPat has said, and they'll handle 100V. Surface mounts all (MSOP-8's - REALLY tiny), and requiring more discrete components for support - but I've become reasonably proficient at making PCBs and hand-soldering SMDs. The chip I have in mind is only good for about 350mA in the best conditions - but that should leave plenty of margin for this application.

And the small surface mount 220uH inductor required is cheap, too! Will share the result (or magic smoke) when I get something together.

 

[ZapPat]

Humm... methinks philf is talking about the LM5008 for 5$ (CAN) each. But you might want to save a whole buck and go with the LM5009 instead for our power levels.
I was looking at the 811-1304-ND inductor as a cheap solution (0.70$ CAN). The model I linked to is a 150uH, which is what I am going to use.

 

[philf]

AHAH!

ZapPat!

The game is afoot!

Well, not really. I'm not a gamer - but I'm betting that we're on the same page National's design notes are going to have to suffice as a starting point, though, unless you have some warnings or other empirical data that could prove this to be a waste of time.

I can definitely say, from my experience with the switching mod that I made to to the GM (Shenzhen) controller, that the power supply was way cooler in operation, and that the thing was way more efficient in standby mode than the original multi-tiered linear regulator scheme that the controller shipped with.

These mA may not mean much in the grand scheme of things, but they add up in a big way - and even psychologically - knowing that my rig isn't running the batteries down any more than it must while I'm stopped or otherwise just pedaling along.

 

[Knuckles]

ZP, philf and fech ... Get the BUCK (converter) Outta Here!
This is a DOPE (NPN) Thread!

Yet I am Induced to Switch from my Warm Linear thinking. I do Oscillate however.
What da BUCK!

Knuckles

:lol:

Shout out to da noobs ... http://en.wikipedia.org/wiki/Buck_converter