Compact Field Oriented Controller, ASI + Grin, limited run
- Thread starter justin_le
- Start date
It is not that simple. The elastic modulus must be high enough to provide mechanical protection but low enough to absorb strains. Also the coefficient of thermal expansion is a critical parameter. Epoxies actual suck for all weather apps as they have a high glass transition temperature which means they become so hard at low temps that they break solder connections. Its is not the fact they become hard that breaks the connection. It is the fact the epoxy is shrinking faster than the pcb and applies pressure to the components. If the CTE was matched the epoxy getting harder would not matter. The CTE of potting compounds is almost always higher than the PCB (no matter what kind it is) so using a soft silicone with a lower Tg allows for a CTE mismatch. The reason you dont see silicone potting as often is it is more expensive. You should also throw the data sheet away.
Here is the what military says. It is really overkill for an consumer grade controller but it is interesting none the less. Historically the military actually has more problems caused by potted electronics in some applications due to low cycle fatigue stress caused by daily ambient temp changes. This is because someone had the simple idea harder is better. The problem can take years to show up. Silicone makes most of these issues go away so you dont have to do a full analysis really
"""A word of caution is warranted here - in ANY design the potting materials need to be well- characterized over temperature and strain rates. This does not mean to simply rely on the information from a manufacturer's data sheet. This is usually woefully insufficient and often inaccurate. The material has to be tested for mechanical strength, thermal expansion characteristics, curing shrinkage, adhesion to surfaces, and also creep over the full temperature range and at different strain rates. Without this data a potted design should not be attempted.
Here is the what military says. It is really overkill for an consumer grade controller but it is interesting none the less. Historically the military actually has more problems caused by potted electronics in some applications due to low cycle fatigue stress caused by daily ambient temp changes. This is because someone had the simple idea harder is better. The problem can take years to show up. Silicone makes most of these issues go away so you dont have to do a full analysis really
"""A word of caution is warranted here - in ANY design the potting materials need to be well- characterized over temperature and strain rates. This does not mean to simply rely on the information from a manufacturer's data sheet. This is usually woefully insufficient and often inaccurate. The material has to be tested for mechanical strength, thermal expansion characteristics, curing shrinkage, adhesion to surfaces, and also creep over the full temperature range and at different strain rates. Without this data a potted design should not be attempted.
speedmd
10 MW
Flathill is spot on with regards to CTE issues and hard potting compounds. We saw this back in the 1980s on some flight hardware I was involved with. The larger and more flex / stress in the assemblies the bigger the problem got. May be worth trying a few different types, but this little cube should be a reasonably good reliable shape /size for most any medium /soft potting compound.
Read what I wrote again
"""Its is not the fact they become hard that breaks the connection. It is the fact the epoxy is shrinking faster than the pcb and applies pressure to the components. If the CTE was matched the epoxy getting harder would not matter. The CTE of potting compounds is almost always higher than the PCB (no matter what kind it is) so using a soft silicone with a lower Tg allows for a CTE mismatch.
"""Its is not the fact they become hard that breaks the connection. It is the fact the epoxy is shrinking faster than the pcb and applies pressure to the components. If the CTE was matched the epoxy getting harder would not matter. The CTE of potting compounds is almost always higher than the PCB (no matter what kind it is) so using a soft silicone with a lower Tg allows for a CTE mismatch.
voicecoils
1 MW
How about starting a separate thread to discuss potting further?
What's funny about all this discussion is that I didn't really detail what potting compounds we were planning to use, be it silicone, poly-urethane, or epoxy based, and among those there are a gazillion varieties. I just mentioned the word "potted" and it set off this firestorm!
Flathill raises a number of excellent points and challenges in getting this right in his last post, including the risks of very cold weather usage. I'm not so concerned of the thermal cycling when the controller gets hot in use because it's easy to find compounds that are soft enough at elevated temps not to cause much component stress even with mismatched CTE's. But when people have their bikes out in -20 oC eastern Canada winters, then even softer compounds can become hard as glass.
Of several dozen different PU, Epoxy, and yes, even Silicone potting resins that we've sampled, the ones that so far stand out the most have been two products from epoxies etc. First is the 50-3170 series which is an epoxy formulation that has a rubber like consistency and has a large amount of fillers to enhance thermal conductivity:
https://www.epoxies.com/_resources/common/userfiles/file/50-3170R.pdf
And the second is the 20-3001 series for a lot of general purpose usage, since it pours with very low viscosity to avoid degassing, and the 65D hardness is a lot less than most (80-90D us typical) allowing it to still feel tough without being rigid and brittle.
https://www.epoxies.com/_resources/common/userfiles/file/20-3001NC.pdf
The 50-3170 product is totally inappropriate for the full pour, since it's much to viscous to fill all the voids and the resulting rubber isn't strong enough to hold the wire assembly. But it should work as an initial layer that we apply first to the bare circuitboards almost like a thick conformal coating, prior to putting it in the mold and potting the full shape with the most standard epoxy product. That way there is a soft rubber-like intermediary between the PCB and the outer shell, which I'm hoping would really reduce the shear stress resulting from mismatched thermal expansion coefficients.
Yes exactly. That's why I thought your use of the word 'copout' for our potting plans wasn't totally fair!
Flathill raises a number of excellent points and challenges in getting this right in his last post, including the risks of very cold weather usage. I'm not so concerned of the thermal cycling when the controller gets hot in use because it's easy to find compounds that are soft enough at elevated temps not to cause much component stress even with mismatched CTE's. But when people have their bikes out in -20 oC eastern Canada winters, then even softer compounds can become hard as glass.
Of several dozen different PU, Epoxy, and yes, even Silicone potting resins that we've sampled, the ones that so far stand out the most have been two products from epoxies etc. First is the 50-3170 series which is an epoxy formulation that has a rubber like consistency and has a large amount of fillers to enhance thermal conductivity:
https://www.epoxies.com/_resources/common/userfiles/file/50-3170R.pdf
And the second is the 20-3001 series for a lot of general purpose usage, since it pours with very low viscosity to avoid degassing, and the 65D hardness is a lot less than most (80-90D us typical) allowing it to still feel tough without being rigid and brittle.
https://www.epoxies.com/_resources/common/userfiles/file/20-3001NC.pdf
The 50-3170 product is totally inappropriate for the full pour, since it's much to viscous to fill all the voids and the resulting rubber isn't strong enough to hold the wire assembly. But it should work as an initial layer that we apply first to the bare circuitboards almost like a thick conformal coating, prior to putting it in the mold and potting the full shape with the most standard epoxy product. That way there is a soft rubber-like intermediary between the PCB and the outer shell, which I'm hoping would really reduce the shear stress resulting from mismatched thermal expansion coefficients.
flathill said:
Yes exactly. That's why I thought your use of the word 'copout' for our potting plans wasn't totally fair!
MrDude_1
100 kW
man... these guys would really be mad at me if they saw the potted electronics I used to sell...
I used to make both the outer plug/shell and pot the electronics in one shot using hard polyurethane... :lol:
I used to make both the outer plug/shell and pot the electronics in one shot using hard polyurethane... :lol:
justin_le said:
Good to hear you are aware of all the issues. You mentioned thermal epoxy and Robbie mentioned epoxy and non-serviceable in the same sentence which sent me off. I just know it gets cold up there in Canada and epoxy might not be the best choice no matter what the data sheet says. In any case you mentioned the cast design is only for the first run so you guys will have more options when you get a case later. I apologize for calling it a copout but It was intended to challenge you to make it serviceable and potted.
Are you guys going to hack the cycle analyst to communicate with the controller over ModBus/232? I wish they just used CAN bus so you guys could start working towards EnergyBus integration in a couple years
Also I forget does this controller offer variable regen? If so is it down to zero speed?
Thanks for all your hard work!
mrbill
10 kW
justin_le said:
I am looking forward to trying this style of controller on my ebikes.
A couple of questions:
1) Can the two larger controllers function or be modified to function at lower voltages (e.g. down to 20 volts)? Ideally I'd like something I could run as low as 20 volts at 1000 watts.
2) When might beta units be available?
Thanks.
Alan B
100 GW
flathill said:
I just reviewed post number one, it states smooth variable regen braking. Yes!
My experience with the Sabvoton's eBraking is that it works to a very low speed, but not quite zero. At very low speeds regen doesn't work, they have to apply reverse power to get braking, so it consumes some power (but not much). So you may not want it to keep drawing power when it is nearly stopped.
I haven't tested mine yet so I can't report on it. Actually, come to think of it, I'm going to put it on a geared hubmotor, so no regen testing for that one. The voltage of this model is too low for my other hubmotor (which uses 18S), so I'll wait for the BAC800 to try a bike that has regen. Looking forward to it.
Sweet! Also might be interesting to compare the regen braking performance sensored vs sensorless.
I think true zero speed braking would be a cool feature for eboards and even regular ebikes even if it does waste a little more power than regen only braking. You could even have a hill hold feature but that would really waste power.
I think true zero speed braking would be a cool feature for eboards and even regular ebikes even if it does waste a little more power than regen only braking. You could even have a hill hold feature but that would really waste power.
Alan B
100 GW
To eBrake at low speeds you need a rotational encoder that is finer than digital hall sensors. FOC might be able to do it, but the signal to noise ratio of such low velocity angular position may be inadequate.
robbie
1 W
- Joined
- Sep 28, 2012
- Messages
- 50
Thanks for the push - in many ways I agree with your standpoint on designing for product end of life.
It would be very neat to get serialized control working with the CA, it would provide some very interesting options with this controller!
It does indeed offer variable regen - at the moment we've configured them to have the analog braking (ebrake) connected in parallel with the throttle, so that a CA3 can take advantage of the 0-0.8V unused portion of a standard throttle. You can configure the controller in software to have a percentage of phase current as regen as well as setting regen battery current limits and battery high voltage limits. There is also a minimum brake speed setting. The controller will do very close to zero speed braking, with hall sensors. If one is running sensorless, the throttle typically must be applied before regen braking is allowed.
Alan B
100 GW
I have a separate brake lever with analog output voltage as well as a contact closure. I don't believe I want the ebrake/regen tied up with the throttle. Is that supported as well?
MrDude_1
100 kW
I have the same question. I want a regeneration brake lever.Alan B said:
Alan B
100 GW
robbie said:
Excellent!
Thanks,
Jason3
10 mW
I have a problem with my BAC500 not connecting to Bacdoor. It was working fine until I inadvertently set the motor rated RPM to 50 instead of 500. The motor made some funny noises, the connection dropped out and since then I've been unable to reconnect. The drive LED is solid red, not showing any errors.
Anyone else had a similar problem, or have any ideas for a solution? The cable is ok, I can still connect to my Satiator fine. Any ideas appreciated!
Jason
UPDATE: Problem solved - it was not the fault of the BAC500 but the battery, which was showing 53v but dropping to <5v under any sort of load.
Anyone else had a similar problem, or have any ideas for a solution? The cable is ok, I can still connect to my Satiator fine. Any ideas appreciated!
Jason
UPDATE: Problem solved - it was not the fault of the BAC500 but the battery, which was showing 53v but dropping to <5v under any sort of load.
johnrobholmes
10 MW
In regards to potting and product long term value, I have also found that potting advantages far exceed the disadvantages of a less serviceable product. I generally move a few thousand new controllers per year with more than 1000 of them being potted in-house. We also pot and conformally coat an additional 500ish products per year for customers. Failure rates drop significantly when water is involved, and the bain of ebikes is nasty road water. The big hurdle will be the wide temperature variation, but I think the two product strategy will work wonderfully. Sure, it is a cost savings for the supplier to pot a controller VS a sealed housing. It is a cost savings because the end product is superior and the customer has fewer failures. Even sealed housings don't solve water egress through the wires, so you still end up needing to coat the board with something that wicks into the wires. That is every bit as non servicable as any thickess of coatings, so what is gained by a sealed housing over a potted device? Bragging rights? An expensive case? A lot of effort wasted to watch water still get in?
For effectiveness, I would suggest an epoxy or epoxy over a silicon. Silicons will routinely separate from the wire jackets and allow water to wick in. A base silicon layer would give a great CTE cushion, and an epoxy outer will give more durability and protection from delamination from the wire jacket. However, I don't know how well epoxy will hold to silicon as I have never tried it. The best strategy may indeed be a softer epoxy conformal with a harder epoxy outer. At any rate, a two part epoxy will cure much better in a mold than a silicon that requires offgassing. It could take days for offgassing to complete, maybe even a week unless you pulled a vacuum on it too. Have a vac chamber that will fit 50 controllers? :lol:
Justin, if you can get failure rates below 1% you will be my hero! Even with extensive pretesting I still see about 2.5% infantile failure rate within 30 minutes of customer installation. Failure modes vary, but the most common is a FET or driver failure. You will hopefully have a lower rate because your FET doesn't rely on a soldered drain for heat dissipation. I also saw just under 10% infantile failure with Xie Chang controllers, also seemed to be FET and drivers as the most common mode but I only have a few hundred controllers to pull experience from.
For effectiveness, I would suggest an epoxy or epoxy over a silicon. Silicons will routinely separate from the wire jackets and allow water to wick in. A base silicon layer would give a great CTE cushion, and an epoxy outer will give more durability and protection from delamination from the wire jacket. However, I don't know how well epoxy will hold to silicon as I have never tried it. The best strategy may indeed be a softer epoxy conformal with a harder epoxy outer. At any rate, a two part epoxy will cure much better in a mold than a silicon that requires offgassing. It could take days for offgassing to complete, maybe even a week unless you pulled a vacuum on it too. Have a vac chamber that will fit 50 controllers? :lol:
Justin, if you can get failure rates below 1% you will be my hero! Even with extensive pretesting I still see about 2.5% infantile failure rate within 30 minutes of customer installation. Failure modes vary, but the most common is a FET or driver failure. You will hopefully have a lower rate because your FET doesn't rely on a soldered drain for heat dissipation. I also saw just under 10% infantile failure with Xie Chang controllers, also seemed to be FET and drivers as the most common mode but I only have a few hundred controllers to pull experience from.
Hi JRH and thanks for sharing your own real life experience on the matter here, it's nice to have feedback from people who have firsthand expertise of the pros and cons in an almost identical application.
Well we got our first trial experience in last week after the machined heatsink parts came back from hard coat anodizing. It turns out that the epoxy rubber product which I thought would make a great first coat for CTE isolation was almost impossible to keep on the raw board in any kind of thickness. Even though it is quite viscous and goopy, when sitting it would slowly drain off until just the thinnest of films was remaining so I ended up putting the controller in a lathe chuck at 100rpm while the epoxy cured in an attempt to keep the film distributed. But even then it was still pretty thin in some sections, and almost for sure we'd want to switch to another product for this step:
You can see the small square of acrylic glued over the surface mount LED on the bottom right of the PCB.
The 2nd stage pour though worked out just as hoped. I needed a bit of sideways clamping pressure on the silicone mold to prevent leakage, but pouring in the resin from the top at the wire exit didn't cause problems.
And after popping out it looks identical to the 3D printed prototype, including the hexagonal fill marks on the top plane and all the other small visual defects of the model. And the LED shines through great.
I had this installed on my bike and on the way home last night got hit by a surprise rainstorm, and I've never been so gleeful to watched my motor controller get splattered with water!
From my experience here's pretty much zero bonding between cured silicone and anything other than silicone (though I see a patent on the topic here
http://www.google.com/patents/US3519465.) But I'm not sure if a bonding between the two layers is all that important. They're more or less mechanically entrapped in each other and would water get between the two it's not exposing the circuitry or wiring.
Trying to avoid all usage of both our pressure pot and vacuum chambers in this process
We were at the DEX expo today and had a great discussion with a local distributor of electrical resins and adhesives about this project. He had some good things to day about some marine grade urethane rubber potting compounds, which we'll likely try as well since then it could be a single stage pour. Tg is -16oC and durometer at ~80A is firm enough for the finished part yet soft enough not to transfer much thermal stress. I'd generally ruled out PU over epoxy for long term water and UV resistance, but if it's meant for marine...
Well it would mostly be ASI who'd deserve the credit for that since they did all the electronics design, we're just trying to package it nicely for end users. But yeah, <1% is the goal. Realistically, 2-3% would still be pretty decent, that's about what we have with the Cycle Analyst and it's manageable although in those cases it's almost always an easy repair. With potted that won't be an option.
johnrobholmes said:
Well we got our first trial experience in last week after the machined heatsink parts came back from hard coat anodizing. It turns out that the epoxy rubber product which I thought would make a great first coat for CTE isolation was almost impossible to keep on the raw board in any kind of thickness. Even though it is quite viscous and goopy, when sitting it would slowly drain off until just the thinnest of films was remaining so I ended up putting the controller in a lathe chuck at 100rpm while the epoxy cured in an attempt to keep the film distributed. But even then it was still pretty thin in some sections, and almost for sure we'd want to switch to another product for this step:
You can see the small square of acrylic glued over the surface mount LED on the bottom right of the PCB.
The 2nd stage pour though worked out just as hoped. I needed a bit of sideways clamping pressure on the silicone mold to prevent leakage, but pouring in the resin from the top at the wire exit didn't cause problems.
And after popping out it looks identical to the 3D printed prototype, including the hexagonal fill marks on the top plane and all the other small visual defects of the model. And the LED shines through great.
I had this installed on my bike and on the way home last night got hit by a surprise rainstorm, and I've never been so gleeful to watched my motor controller get splattered with water!
From my experience here's pretty much zero bonding between cured silicone and anything other than silicone (though I see a patent on the topic here
http://www.google.com/patents/US3519465.) But I'm not sure if a bonding between the two layers is all that important. They're more or less mechanically entrapped in each other and would water get between the two it's not exposing the circuitry or wiring.
Trying to avoid all usage of both our pressure pot and vacuum chambers in this process
We were at the DEX expo today and had a great discussion with a local distributor of electrical resins and adhesives about this project. He had some good things to day about some marine grade urethane rubber potting compounds, which we'll likely try as well since then it could be a single stage pour. Tg is -16oC and durometer at ~80A is firm enough for the finished part yet soft enough not to transfer much thermal stress. I'd generally ruled out PU over epoxy for long term water and UV resistance, but if it's meant for marine...
Well it would mostly be ASI who'd deserve the credit for that since they did all the electronics design, we're just trying to package it nicely for end users. But yeah, <1% is the goal. Realistically, 2-3% would still be pretty decent, that's about what we have with the Cycle Analyst and it's manageable although in those cases it's almost always an easy repair. With potted that won't be an option.
Looks cool. I agree getting silicone potting compunds to stick is tricky and dont recommend useage without a hard shell, but if you ever do try them again most need to be used with a primer (adhesion promoter) after cleaning. There are primerless versions but I have not had good experience with them.
MrDude_1
100 kW
So what is the absolute max power of this little controller once potted? Is it any higher or lower?
speedmd
10 MW
The marine grade urethane rubber sounds like a perfect potting material to me. 80A is very similar to a road skate wheel in hardness so it is easy to find a part to play with to see how strong and durable this stuff is. Extremely Tough, bonds well and very good at vibration dampening. Some types setup reasonably quick also. As long as your materials / assemblies are good and dry there should be very little air bubble issues with this material.
Similar threads
-
- Locked
