eBike Master Switch Design

[izeman]

so where should i buy it? conrad.at is one of the biggest supplier for private customers in austria. all others only
sell to companies only. and ebay doesn't show any results.

 

[Alan B]

A large supplier that has NO p channel mosfets?

I selected the one I recommended at Mouser. They had hundreds if not thousands of P channel FETs, of course not so many with the low gate threshold, package, etc.

I'm not familiar with parts sourcing in your area. Perhaps others will chime in with advice.

 

[izeman]

yes strange i must admit. mouser has a minimum amount to order and high
shipping costs. paying 10x the cost of the product for shipping is a no-go i'll try to find a source. thanks

 

[dmwahl]

I haven't looked over the schematic recently, but why not use a PNP instead of a P-FET to dump the gate charge? They're all over the place and cheap.

 

[2moto]

You might want to try Farnell. They have distribution in Austria. Also, Ebay is worth a try.

 

[izeman]

Farnell doesn't have it, and Mouser charges 20.- for a 54c article.
Maybe there are alternative types available that are easier to buy locally? An I the only one having troubles buying those?
I guess Europe is different from USA as we tend to go to a local shop and buy things, as mail order is not THAT popular here because of higher shipping costs.

 

[2moto]

I buy most of my stuff from Farnell's (UK) and shipping is very reasonable. I'd be very surprised if they didn't have an equivalent type of P-FET. Have you tried searching their website by parameter as needed?

 

[izeman]

hmmm i'm no way into electronics, but what about a bc557? would this work? and you can buy them anywhere.
sorry if this is a dumb question, but i guess i can't really follow the schematics

 

[dmwahl]

hmmm i'm no way into electronics, but what about a bc557? would this work? and you can buy them anywhere.
sorry if this is a dumb question, but i guess i can't really follow the schematics

Only good to 45V, but that was my point about using a PNP instead of PFET since there are tons of them out there. If this were my design I would use a cheap, >100V PNP to dump the gate charge. There are tons of them available: http://www.digikey.com/short/qb4cz

 

[izeman]

yes. please don't take it as offense alan, but i think using parts that are available easily would make your board a much bigger success outside of usa. i really searched a lot, and of course my supplier has p-types. but those are smd types or to-220 or some other case form.
could you think of any other fet type that could fit?

 

[Alan B]

As I recall, I chose the P-FET to make the most robust design. There were plenty of parts available in Mouser so I didn't worry about it being too special.

There was a suggestion for a PNP circuit that should work (from Richard), though the circuit was different, and there was a transient failure mode in my analysis of that design that I didn't care for.

We have an investment in design, boards, etc., so we should probably try to find a solution to obtaining the parts rather than changing the design so quickly.

The voltage this part sees is quite low as I recall. Need to review the requirements for the FET. I updated the requirements on page 7 slightly, the FET only sees 15V max so a 20V FET should be fine. We could do a partial parts kit to solve the availability problem if needed.

 

[Alan B]

I added another part number to the page 7 posting.

A random P channel FET might work, it depends on your main FETs. It needs to turn on at a lower gate voltage than your main FETs.

Best to choose a P channel FET that turns on at 2.0V MAX Vgs.

If we use a PNP the biasing gets tricky.

 

[Alan B]

Sorry about the component sourcing. The first version was designed to be super simple, but the increasing requirements pushed us into a more specific design with more difficult component requirements. The PNP approach had some problems that the FET solved nicely, and I quickly found parts that met requirements, so I didn't worry about it. I just checked Digikey and found 8 parts that met requirements though not all are stocked there, they have thousands of the ones that are stocked.

If we limit ourselves to what we can buy locally then we are doomed. Locally stocked electronics parts are a miniscule fraction of what is available, I probably could not even buy a 4110 FET locally, hardly anything I use could be sourced locally anymore. Local sources no longer even try to carry a fraction of the huge and expanding parts range available. We must develop other source channels to have any serious construction capability.

Another problem is through-hole, most parts today are surface mount. I haven't looked but suspect there are lots more P channel FETs available in SMT than through hole.

If we have to make partial kits we can do that. Perhaps just the hard to find parts. Shipping cost may go way up however, compared to shipping a PC board only.

We could also just skip the quick turn-off requirement and simplify the board again. How important is the rapid turn-off requirement?

I can look again at the PNP approach, but I think that will make the design voltage dependent, with different resistors required for different battery voltages. Right now the one set of values covers the whole voltage range, which seems like a valuable feature.

 

[izeman]

alan, np i just would like to build this myself and can't atm.
the mentioned part "ZPV4105A" is even worse to get. if i search google for it it shows NO result at all. checking mouser etc there is no part either. is this a typo?
maybe it would be a good idea (if you still think about selling those boards) to sell kits. if you ordered bigger quantities from eg mouser shipping to your place will be neglectable and all parts still fit in a small letter envelope.

 

[Alan B]

ZVP4105A is a Digikey and Newark available part. Newark showed 46 parts with 2V gates but only 1 was TO92, most were SMT.

edit - fix typo

 

[izeman]

ZPV4105A is a Digikey and Newark available part. Newark showed 46 parts with 2V gates but only 1 was TO92, most were SMT.

am i stupid? what am i doing wrong??

edit: it still would be €18.-/$25 for shipping. this makes this little FET the most expensive part of the whole switch.
once again: sorry for all my posts - i don't want to whine around, but am looking for a supplier

 

[Alan B]

Sorry, typo. Fixed my postings. ZVP... I looked at that several times and still missed it.

The ZVP4105A is $0.92 in unit quantity at Digikey and they have over 6,000 in stock.
The TP2104NG-3 is $0.68 at Mouser and they have 18,000 in stock.
Newark has the ZVP at $1.14, 2,000 in stock, so Farnells might as well?

I usually combine some parts for other projects in the same order to spread out the shipping cost, ordering one low cost part in small quantity is not efficient from any supplier.

It looks to me like a PNP might work there with this design, I haven't had time to analyze it properly or test it. You might try a 2N3906 but I make no guarantees it will not have some issue. The leakage currents may require adjusting the resistor values, perhaps use 100K for R1 and R2.

One other choice is to skip the fast turnoff, use R1/R2 values from the V1 spec, short D1 and D2 and put a jumper from gate to drain on Q5. Essentially return to the schematic of V1.

Has anyone checked the fit in the Hammond box? I don't have one yet to check it.

 

[izeman]

... You might try a 2N3906 but I make no guarantees it will not have some issue. ...

now THAT'S a GOOD part. plenty supply at my local shop *ggg*
i don't have any problems doing some real life testing to see if it works. that's what beta testers are made for. i will also buy a hammond box (YES those are available locally as well *ggg*) and see if it fits.
will report probably at the end of next week.

 

[dmwahl]

... You might try a 2N3906 but I make no guarantees it will not have some issue. ...

now THAT'S a GOOD part. plenty supply at my local shop *ggg*
i don't have any problems doing some real life testing to see if it works. that's what beta testers are made for. i will also buy a hammond box (YES those are available locally as well *ggg*) and see if it fits.
will report probably at the end of next week.

Should probably figure out what the worst case Vbe will be first, I think that part is only good to 5V.

 

[Alan B]

I looked at that, looks like 0.7V worst case in this design. So 5V should do it.

 

[dmwahl]

Nice. You might even be able to use a PNP with the existing layout by moving pins around or extending the leads with wires as a test. For such a short pulse you probably don't even need limiting resistors on the base. I suspect it's going to make those FETs snap off pretty fast.

 

[Alan B]

The schematic is near the bottom of page 7 in this thread.

My quick "in my head" estimate while I was driving to work this morning was 120 V/S with 1 meg R1/R2 and 1uF cap assuming Q5 PNP with low current gain of 60, and 1200 V/S with 100K values.

The base current is limited with R1/R2 so that should be safe. The collector current is mostly limited by R3 (and R4,5,6,7), not much energy there anyway.

Reverse Vbe is limited by D2, so less than 1V
Vce limited by the zener, so about 12V max

E = S
B = G
C = D

Pinout looks okay.

Seems worth a try. The FET would be much faster, but if the PNP works also, even a bit slower, that will give folks options.

Hopefully I didn't miss anything here, but testing will tell.

 

[izeman]

...Hopefully I didn't miss anything here, but testing will tell.

i guess i can risk those 14c for communities sake.
the caps are all full voltage, right?

 

[Alan B]

Yes, the caps should all be full battery voltage plus safety factor of 20% or more.

 

[richmpdx]

I have completed the 2nd beta version (v1.1) of the master switch board. As I am in the US it was no problem to get parts from Mouser. My bike has an 18S battery and I choose a 1 uF MLCC leaded, 100V TDK capacitor rather than using SMT caps. As the board has few parts, is can be assembled in a short amount of time. I did run into problems soldering the power FETs and the 12 awg leads to the board. My wimpy 20w fine tip soldering iron simply wasn't up to the job, which should have been obvious to me from the start. The big copper traces did just what Alan designed, they rapidly dissipate heat. So either no solder joint or a really ugly, cold solder joint was obtained. To save myself from great humiliation, I won't reveal how long I persisted with the undersize soldering iron before I realized the error of my ways. In any case, I bought a Hakko FX-601 by mail order which did take some time. But it seems like a quality tool with adjustable temp (464 - 1000 F), interchangeable tips, etc. And I am happy to report it makes soldering the big leads to large copper pads relatively easy.

In initial tests the board seems to function as expected. When connected to the battery minus and with output controlled through J1 by a switch and measured by connecting a voltmeter across the controller minus output lead and battery plus lead, there is an initial voltage upon connection to that battery that declines towards something like 1V. Closing the switch at J1 then increases the voltage to the battery voltage, in this case 70.2V in something like 1-2 sec. Opening the switch decreases the voltage, falling to about 1V over something like 20-30 sec.

The time course of voltage changes are much different after connecting the switch to a controller. The controller is a 12 FET from Ebikes Canada which has it's own on/off switch. For the test of the master switch, the controller switch was turned on. With the controller connected to the master switch, it seemed like the on time was similar, probably something like 1-2 sec to reach full voltage of 70.2V. When turning off the master switch, at least as measured by the time for the Cycle Analyst to shut off it's screen, it seemed like voltage fall was almost instantaneous. However, upon shut down there was then a surprising, regular, periodic flashing of the CA screen. This flashing of screen can be eliminated by turning off the controller switch.

When I figure out how to hold down the completed master switch board for a photograph, I will post some pictures.

Rich