Alan B
100 GW
Now if it was so easy to analyze and predict the thermal performance. 
eBike Master Switch Design
- Thread starter Alan B
- Start date
richmpdx
100 W
Alan, please add me as another individual interested in a couple of boards. I have been following the Fechter design on the "active pre-charge" thread and have been thinking about building Richard's design. Very interested in having a master switch to cut off the battery from the system.
Rich
Rich
Here's a spreadsheet similar to the one provided by mvly in a post above showing parallel FET static ON power dissipation without a heatsink. The snapshots below show the spread configured for 1-4 and 5-8 FETs. IRF datasheet values in the lower left configuration table are used to calculate power and identify acceptable configurations by cell color. Using a heatsink will lower the 'Junction to Ambient' thermal resistance (degC/W) allowing higher currents.
View attachment ParallelFetPowerDissipation_6.zip
EDIT - 2016-10-07: update to correct battery current column header text
- The usual ambient temp of 25degC (77degF) may be a bit conservative for outdoor applications so the ambient value for the snaps is set at 38degC (100degF) - perhaps a more real-world value (in the US anyway).
- The green/yellow/red ranges reflect the per FET power dissipation at the specified temperatures as calculated in lower left configuration table.
- the max allowable junction temp is 175degC so the 'red' cutoff temp is set at 125degC in the table (really hot!),
- a more reasonable operating temp of 85degC is used as the recommended (green) 'safe' limit,
- the yellow values fall in between.
- Changing any value in a gray cell (e.g. "W@degC", "Ambient Temp", or an entire device datasheet row) will cause the tables to re-calculate and re-format.
View attachment ParallelFetPowerDissipation_6.zip
EDIT - 2016-10-07: update to correct battery current column header text
johnrobholmes
10 MW
I used IR3077s, they have the back side drain like a TO-220. So far so good, we have about 30 cycles on the setup. It is a low power bike, 36v 25a.
I like the idea of using a zener for the gate charge. Nice and stiff voltage supply without having as much calculations for the max and min voltage swings present on a voltage divider.
I like the idea of using a zener for the gate charge. Nice and stiff voltage supply without having as much calculations for the max and min voltage swings present on a voltage divider.
What is so difficult about a MOSfet master switch?
Your design puts the MOSfets horizontal to the board, why not put them vertically up with a proper heatsink attached?
What are the other problems that make such a MOSfet master switch difficult to built?
Your design puts the MOSfets horizontal to the board, why not put them vertically up with a proper heatsink attached?
What are the other problems that make such a MOSfet master switch difficult to built?
nicobie
100 MW
Chuechco said:
Nobody has marketed one yet,
give it a go. it's beyond me
Alan B
100 GW
Chuechco said:
Welcome to ES.
The FET heat is not actually the problem here, and we're trying to make something compact and easy to produce.
I'd rather not have a hot switch, and the bulk, sharp corners and airflow requirement of a heatsink (heatsinks don't work unless they are hot and have moving air). I've built switches like this before with very little heat rise, if done properly. It really depends on how much average current you are running and how many FETs you parallel.
Teklektic's spreadsheet shows heat rise for bare FETs. We have a lot more heat spreading area than a bare FET, so temperatures should be even lower. The average current on most ebikes is a lot lower than the peak, otherwise the batteries would discharge before the destination was reached. And it would be easy to extend this board to six FETs, or even use two in parallel which is still fairly compact and would carry a lot of current without generating heat.
As an example, on my work-bound commute I use about 13 amp hours in half an hour (and most ebikes use less current than mine). So my average current is about 26 amps. With four 4110's in parallel the current handling is sufficient for 50 amps (with zero heatsink), so this should handle it with lots of margin in terms of FET heating. The PC board traces carrying high current are another problem, but I've taken steps to control that (and standing FETs up makes that worse).
There is one other electronic switch for sale made by an ES member. Was it kfong? I don't think he shared the schematic, and it used up to six FETs standing up with no heatsink, and not a lot of copper for current carrying. It had a micro and latching relay and I don't know how much current it could really handle, though he did mention 10 amps per 4110 FET being no problem. Yes, it was kfong and it is here:
http://endless-sphere.com/forums/viewtopic.php?f=31&t=32135
I could go ahead and lengthen this board for six FETs but I think four is enough for most ebikes and we can always make a longer version when we know what is actually needed. Who needs more than 50 amps average? The vast majority of ebikes will be fine at that level.
Alan B
100 GW
I received my first OSH Park pc board back (from a non ebike project) and they look quite nice! Definitely a deep purple color.
Alan B
100 GW
Three board prototype run ordered.
Alan B
100 GW
knighty said:
I missed this posting earlier, trying to read them from my phone probably.
How many amps at what voltage are you running?
I've paralleled them by stacking them on top of a board, haven't tried on the back, tricky to do that well.
Alan B said:
So if I understood it correctly, the difficulty is to handle the heat and to have it all compact? This still doesn't sound too difficult to solve for me. Why isn't there a switch like this on the market? The cost should be low, one IRF4110 costs like 2€ or about $2.
This thread seems to be suitable to ask a question I've had in mind for while now: How do I know how much current I can push through a FET? The datasheet of the IRF4110 says something like 120A(package limited) which sounds ridiculous. If you would just consider this value, you would only need one single FET for basically any ebike master-switch. What are the other parameters you need to take into account?
Alan B
100 GW
The goal is to avoid generating the heat, and to dissipate what heat is generated, and to keep it compact, rugged and low cost.
Learning to read and decode datasheets is a deep subject all by itself.
There are a number of factors that determine the current an FET can handle, primarily the thermal environment, but also the heat from interconnects (which can be substantial at these currents), switching losses, etc. You look at the thermal flow from the junction to ambient and determine how much heat flow can be tolerated while keeping junction temperatures within ratings. Of course the junction temperature affects the drain-source on resistance which in turn affects heat generated, so it is an equilibrium. If there are parallel FETs there are considerations to ensure the current divides fairly equally and derating to account for the non-equity. There are many tradeoffs to consider.
If you have a large heatsink with excellent thermal contact and a cool ambient environment with adequate airflow and no switching losses and big solid interconnects one 4110 FET can handle a lot of current, perhaps 50 amps. But if you want to avoid the heatsink and make it small and cool then 10 amps (average) per FET might be closer to the truth.
How much average current are you running? 48V battery 500W motor is just over 10 amps. Four FETs should be adequate for a 2KW motor at 48V with little heatsinking.
Learning to read and decode datasheets is a deep subject all by itself.
There are a number of factors that determine the current an FET can handle, primarily the thermal environment, but also the heat from interconnects (which can be substantial at these currents), switching losses, etc. You look at the thermal flow from the junction to ambient and determine how much heat flow can be tolerated while keeping junction temperatures within ratings. Of course the junction temperature affects the drain-source on resistance which in turn affects heat generated, so it is an equilibrium. If there are parallel FETs there are considerations to ensure the current divides fairly equally and derating to account for the non-equity. There are many tradeoffs to consider.
If you have a large heatsink with excellent thermal contact and a cool ambient environment with adequate airflow and no switching losses and big solid interconnects one 4110 FET can handle a lot of current, perhaps 50 amps. But if you want to avoid the heatsink and make it small and cool then 10 amps (average) per FET might be closer to the truth.
How much average current are you running? 48V battery 500W motor is just over 10 amps. Four FETs should be adequate for a 2KW motor at 48V with little heatsinking.
When looking at teklektik's chart, it all becomes clear to me. There's a constant value which tells you how much temperature above ambient is generated for one watt. R(OJA). The datasheet names it thermal resistance.
So for every watt the MOSfets generates ( P = Id²*Rds(on) ) the MOSfet's junction temperatures rises by 65 degrees.
I wonder by how much you can lower this thermal resistance value (65°C/W) by attaching a heatsink to it? Are values below 30 °C/W achievable? Reducing the thermal resistance by halve would enable the FET to handle 1,41 times more current.
So for every watt the MOSfets generates ( P = Id²*Rds(on) ) the MOSfet's junction temperatures rises by 65 degrees.
I wonder by how much you can lower this thermal resistance value (65°C/W) by attaching a heatsink to it? Are values below 30 °C/W achievable? Reducing the thermal resistance by halve would enable the FET to handle 1,41 times more current.
Alan B
100 GW
24S Cromotor would imply 4115 FETs, and two stacked boards would be about right.
I'm running 18S on my Cromotor now and planning to go to 24S, but plan to limit the current a bit more than you are.
I'm running 18S on my Cromotor now and planning to go to 24S, but plan to limit the current a bit more than you are.
Alan B
100 GW
Chuechco said:
That is a simplified model which is useful. Some more info here:
http://en.wikipedia.org/wiki/Heat_sink
Heatsink manufacturers give some data, but it involves a set of conditions involving moving air. In this switch application the availability of cooling air varies considerably, both in velocity and ambient temperature.
A good example of this is in Radio Control applications like electric powered airplanes. The airflow is so high that they routinely exceed the normal still-air ratings of connectors, wires and electronic components without problem.
It is not hard to lower the thermal resistance below that of a bare device, this is one of the reasons the FETs are mounted to the PC board. The copper on both sides of the PC board has considerable heat conduction and dissipation capability. This paper shows some data for some PC boards used as heatsinks:
http://www.ti.com/lit/an/snva424a/snva424a.pdf
From the above paper it is clearly shown that size matters, the smaller PC boards were not as effective. It also showed that increasing the thickness of the PCB copper didn't help much in terms of heat dissipation (but it would help a lot in terms of electrical resistance and ability to carry current which we could use as this reduces the heat load of the current carrying conductors which is in addition to that of the FET DS junction).
Based on figure four from the paper above the board design here might have a thermal impedance of around 25 degrees C per Watt to static air if it was not wrapped in heatshrink or tape. But with full battery voltage on the TO-220 tabs and mounting bolts (when switch is off) we need to provide some protection and that will likely lower the thermal performance. A box with airflow through it would be great.
One trend shown in the above graphs is that getting down to 30 deg C/W was relatively easy (it didn't take much), getting to 20 deg C/W was much more difficult. Additionally, just a little airflow approximately doubled the thermal performance over still air.
Of course this must be combined with the Rds of the FETs used (due to voltages needed), the RMS current dialed in by the rider, etc. to determine the heatload on the switch.
At the end of the day we need to build something and test it. We may get some toasty fingers or even blow one up, but we will learn what works and what does not.
The PC board design has been panelized and forwarded to the manufacturer.
Alan B
100 GW
Circuit Boards are in the mail. Probably arrive Monday. (They did not arrive Saturday).
powersupply
100 W
Is 1MOhm enough for the (general?) zener diode to start zenering?
Alan B
100 GW
powersupply said:
Depends on the diode, it could be marginal - too low a current to bring the zener to its rated voltage. I changed it to 100K because I had that same concern a week ago, see the parts list near the end of page 1. It will be easy enough to check when a board is tested, and a good idea to verify before putting load on the FETs.
Alan B
100 GW
Boards due tomorrow.
Alan B
100 GW
Correction. Boards arrived a day early.
Alan B
100 GW
Preparing to ship board to tester.
Gregory
100 kW
Good job Alan. I'll buy 5 pcb once you've tested it for the next run. I'll put them together for a couple of locals who im sure would be interested .
And I'll finally be able to use up some of my 150 4110 fets I bulk ordered.
And I'll finally be able to use up some of my 150 4110 fets I bulk ordered.
Alan B
100 GW
Wow, that's quite a collection of FETs. Too bad you are so far away.
Similar threads
