Building the Best Controller

[rhitee05]

Those FETs are a "Super-247" package. They're the same size and footprint as regular TO-247, but without the mounting hole or tab. That means they'd require clip mounting, a hold-down strap across the front, or something along those lines.

It's a nice part, but I'm a little skeptical if a $10-$15/ea FET is where this design should be going. I'm not in the 150V market, though, so my opinion may not matter. Since they use the same footprint as any other TO-247 device, there's no reason why they couldn't be a high-end option, and the user would only need to figure out the sink mounting problem. The rest of us peasants could keep on using $3 60V parts.

 

[Kingfish]

RE Luke: Understood. <sigh>

Well – I think we are at a crossroads:

1. We could craft a really awesome 6-FET SMD but likely be capped out at ≤ 100V
2. A powerfully strong 6-FET in the 3-legged TO-247 (or similarly robust packaging)
3. A 12-FET amp monster again with TO-247.

Have I characterized this correctly? If so, I pose the question:

• Which path would you like to take for Version 1?

FWIW - the first two are appealing to me

~KF

 

[liveforphysics]

Identical price for the TO-264 version with identical specs, and a screw-tab for mounting.

$10-15 per fet in a 6-fet design seems like too much for DIY hand-built controller??

The FET stage is like the heart and lungs of the controller. The fet's (and layout) are the single most critical aspect that will determine the performance, reliability, running temp, etc. The rest of the controller is just fluff and gimicks IMO.

But, I'm just using this part as the logical choice for a 150v option if people are so-inclined. It's the best 150v option out there.

I think the rest of us will be able to do a very simple compact 6-fet using the little 1"x1" isolated pucks, and be happy as can be.

 

[liveforphysics]

RE Luke: Understood. <sigh>

Well – I think we are at a crossroads:

1. We could craft a really awesome 6-FET SMD but likely be capped out at ≤ 100V
2. A powerfully strong 6-FET in the 3-legged TO-247 (or similarly robust packaging)
3. A 12-FET amp monster again with TO-247.

Have I characterized this correctly? If so, I pose the question:

• Which path would you like to take for Version 1?

FWIW - the first two are appealing to me

~KF

Well, I'm all about option 1 using the awesome little pucks, but we would be capped at <75v with that path. They don't make a 100v fet in that package.

For >75v applications, it would require going to option 3 to get the same current potential from the 3 leg packages, and require a PCB (with 4 heavy layers), and it would require isolation of the drain tabs with kapton tape during the build, or SMT mounting on an aluminum board. Neither of those options are nearly as fun as just ducking the 75v cap and running the pucks. But that would require folks with high-wind hubmotors to buy lower wind-count or re-wind to get the speed they are used to having back.

 

[Kingfish]

Q: For either package discussed, how tolerant is the 100V limit? Can it be pushed beyond say by 5 or 10%?

I'm thinking of flexibility.
~KF

 

[Kingfish]

Q: For either package discussed, how tolerant is the 100V limit? Can it be pushed beyond say by 5 or 10%?

I'm thinking of flexibility.
~KF

Same question would apply to 75V.

 

[rhitee05]

Q: For either package discussed, how tolerant is the 100V limit? Can it be pushed beyond say by 5 or 10%?

I'm thinking of flexibility.

The flexibility is more like -5% or -10%. You want to have some headroom between the max operating voltage and the FET rating. Otherwise, poof! I'd say 10% headroom is probably the minimum I'd recommend, and more would be nice if possible. The lower the headroom you have, the more careful you need to get with the bypass caps, layout, etc. to minimize the voltage transients during switching.

 

[Kingfish]

Understood.

Well, that given here's the breakdown:

• ≤ 75V, massive amps, SMD, 6-FET
• ≤ 150V, ≤ 150A, TO-247 (or like), 6-FET
• > 75V, massive amps, TO-247 (or like), 12-FET – or external FET

Do I have that correct?
~KF

 

[oldswamm]

Isn't the fet setup Luke's talking about kind of a home made version of the industrial style all in one block 6 fet motor drivers (only more)?
What I mean is, isn't it mounted external to the stack of boards (MCU, PS, fet driver), and an alternative to the giant fets, OR the 3 pin devices ON the driver board? We can still use 3 pin 150v at 'reduced' current. Someone who needs massive current at 150v or more can still use external 'pucks'.

Bob

 

[CamLight]

Q: For either package discussed, how tolerant is the 100V limit? Can it be pushed beyond say by 5 or 10%?

I'm thinking of flexibility.

The flexibility is more like -5% or -10%. You want to have some headroom between the max operating voltage and the FET rating. Otherwise, poof! I'd say 10% headroom is probably the minimum I'd recommend, and more would be nice if possible. The lower the headroom you have, the more careful you need to get with the bypass caps, layout, etc. to minimize the voltage transients during switching.

I agree.
You can't even approach the max voltage rating if you want a reliable controller. Derating to about 80% of the max is the accepted standard for long-term reliability, i.e., there's no concern about the FETs being run at that voltage. Go above that and the reliability starts dropping.

Can you operate a FET above its max rating....yes!!! Are you willing to do the repairs when (not if...when) one or more FETs blow? That's an individual decision to take when designing a controller. You can save some $$ now by exceeding the rating and taking a chance or spend more $$ now and not worry.

There's no "right" decision though (especially for a DIY project), just the one the designer prefers.

 

[CamLight]

Hmm....actually, if the FETs got that hot, the rider should have their controller blow up anyway for thinking they could undersize it so badly!! A thermal switch on the bus bar could completely solve this problem I guess.

I was thinking of how hard it would be to change one when it blew, then realized what a challenge it would be to blow one. :lol:
Bob

Actually, I was thinking about this earlier today and it's a concern.
The FETs can be protected from a slowly rising temp increase to above their rating (or desired operating point) but they're still vulnerable to any number of ways of being destroyed. IMHO, ease-of-repair should still be on the list of things to consider. Especially if voltage ratings are going to be pushed or hand wiring is involved in any install (causing more inductance and, therefore, greater voltage spikes).

Having said this, module replacement can be a repair option. That is, the individual FETs do not need to be easily replaced if the entire module can be removed and sent out for repair. Each user can decide whether or not to have spare FET modules ready for for install while the blown module is repaired.

If the design pushes up against any FET ratings, this approach does not have to be any more expensive than repairing individual FETs as the controller will, hopefully, be less expensive (or smaller, etc.) because of the modular block-of-FETs design. Since the FETs will probably blow at some time in the life of the controller (when pushing the ratings) the user would have to order FETs, pay shipping, etc., anyway so stocking a FET module really isn't any more expensive.

But, this is much more of a systems approach that considers costs over the entire lifetime of the controller and usually isn't considered. Most folks just decide to save the $$ up front, take their chances, and deal with any necessary repair costs when they need to. I still do that plenty of times.

 

[Kingfish]

De-rating: If we assume 80%, a 75V FET, regardless of amperage, is safely operated up to 60V. If this is the case I would suggest dropping SMD and go back to the TO-247 (or like).

Guys, having been delightfully baited by the possibilities I still like the idea of 6-FET, and I like the idea of having a place where we could off-board FETs if this can all work itself out: One FET board that can support the lesser fish in the sea, and yet flexible enough to support the barracudas, sting-rays, and tiger sharks.

Makes sense?

~KF

 

[Alan B]

how about a board with 6 to247, drivers, cpu and a second board for the optional fet setups that is just fets.

 

[liveforphysics]

how about a board with 6 to247, drivers, cpu and a second board for the optional fet setups that is just fets.

That would be great.

 

[oldswamm]

lfp:
I'm still thinking about your idea with the SMT soldered to the bus bars.
Do you have any idea how to change a bad fet, or would we just consider it a bad 6fet unit, which is what we would do with an commercial 6 fet unit after all.
What would be a minimum production run, and what do you think each unit would cost.
I might be interested in one (or 2?) when the time comes, assuming we have the driver for it.

Not asking you to repeat what you've already said in other places, but would you confirm my understanding?
If I put 60V with a 100a limit to an x5302 (ignoring external losses), I should get almost exactly the same torque and power curves as with a an x5304 at 120v and 50a, right?

Thanks,
Bob

 

[liveforphysics]

lfp:
I'm still thinking about your idea with the SMT soldered to the bus bars.
Do you have any idea how to change a bad fet, or would we just consider it a bad 6fet unit, which is what we would do with an commercial 6 fet unit after all.
What would be a minimum production run, and what do you think each unit would cost.
I might be interested in one (or 2?) when the time comes, assuming we have the driver for it.

Hi Bob, I re-work soldered heavy bus bar setups often. It's even easier than putting them together. Takes about 20 seconds of work to do (after 5mins of waiting for the plate to heat up).

Caps and things that are hand soldered on after the FET array is made will need to be removed the same way they were installed. Screw terminal caps fit the bill best for these high current controllers, so it can be as simple as unscrewing a few caps and unbolting a few terminals. Or, unsoldering the accessories if you took that route.

Next, heat up that steel plate you used in assembly to very high temps. Think steel plate sitting on a red oven burner. Touch your array to it, wait 5-10seconds, lift up, and you find the buss bars are all layed out nicely on the plate, the FETs never got hot, and you're free to remove/replace whatever you like. If you don't get the bus array off that heat very quickly, it will oxidize and the solder will oxidize, so have a wet towel handy, and have hot-pads (or welding gloves) on all ready so you can pick the plate up and dump the buss parts onto the wet towel as soon as you set down the FET/sink combo.



I actually find it much easier to re-work anything with flat buss-bars than re-working with 3-leg through-hole FETs, which can be a PITA sometimes.

Not asking you to repeat what you've already said in other places, but would you confirm my understanding?
If I put 60V with a 100a limit to an x5302 (ignoring external losses), I should get almost exactly the same torque and power curves as with a an x5304 at 120v and 50a, right?

Thanks,
Bob

Yep, you will have current density going through the copper in the slots, identical flux density on the tooth, and identical torque/speed etc.

Also, while I fully agree it's a good idea for a professional product to never use right up to the FETs rated maximum voltage (like everyone running the 4110's on 24s LiPo packs is doing, including myself), if the cap design is well layed out, and you've got some killer snubber caps across the FET power rails (even a thyristor wouldn't hurt if you wanted to be bulletproof), you're pretty darn safe to run at the max rated voltage I've got snubbers now with 400vdc max (lowest voltage I could find) with 5.5mOhm ESR at 10khz. These things can offer a lot of protection for FETs being run close to the maximum voltage rating.

Throw one of these little guys on the power bus:
(they make way better ones, just a quick 30 second TSPD search)
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=P0640SDLRP-ND

And have one of these guys screwed on the power bus as well:

 

[Kingfish]

Yep, you will have current density going through the copper in the slots, identical flux density on the tooth, and identical torque/speed etc.

Luke, I was under the impression that higher voltage gives you better speed potential rather than higher current. Adding to that – the reason we seek and prefer voltage over current is to reduce the overall effect of heat from the system resistance e.g. windings.

What have I missed? :?
~KF

 

[rhitee05]

Oldswamm was asking about two different motor windings, which is why what Luke said is correct.

The X5304 has twice as many turns per tooth as the x5302. If you put twice as much current through the '2 as through the '4, then the current*turns product is the same which is what gives the torque. Thus the torque and power curves will be the same (power being torque * speed)

But you're also correct that higher voltage usually is more efficient for a given power output. I'd expect that the higher voltage x5304 case would be somewhat more efficient due to lower copper losses.

 

[liveforphysics]

Yep, you will have current density going through the copper in the slots, identical flux density on the tooth, and identical torque/speed etc.

Luke, I was under the impression that higher voltage gives you better speed potential rather than higher current. Adding to that – the reason we seek and prefer voltage over current is to reduce the overall effect of heat from the system resistance e.g. windings.

What have I missed? :?
~KF

Commonly held fallacy my friend.

Not an ounce of truth in it.



Half the turns = twice the wire cross section, and half the distance. This perfectly balances the x^2 relationship between current and heat in a motors windings.

Half the turns means 1/4th the winding resistance, and half the BEMF growth rate. The enables it to exactly match the performance with half the voltage.

The copper fill determines the continuous torque handling of the motor. The 5302 and 5304 both have identical copper fill (both have 24 pieces in the slot, 2x12 and 4x6 respectively.)

As Bob stated above, you get identical torque/speed/efficiency/heating performance at 60v 100a with a 5302 as you do at 120v 50a with a 5304.
Just in the case of the 60v setup, you have a lot safer battery voltage, easier to charge, much more simple BMS and charger options etc etc.
Until you get to roughly ~400amps or so (>20kw), it really is only an exercise in adding battery/BMS complexity to go higher voltage than 60-70v with an ebike if you get to choose the motor wind yourself. Way way easier to run bigger battery and motor phase leads than it is to add BMS channels IMHO). And that's from a guy who is an absolute performance nut!

 

[liveforphysics]

Oldswamm was asking about two different motor windings, which is why what Luke said is correct.

The X5304 has twice as many turns per tooth as the x5302. If you put twice as much current through the '2 as through the '4, then the current*turns product is the same which is what gives the torque. Thus the torque and power curves will be the same (power being torque * speed)

But you're also correct that higher voltage usually is more efficient for a given power output. I'd expect that the higher voltage x5304 case would be somewhat more efficient due to lower copper losses.

Both have matching copper in the slots, 2x12 and 4x6. When they do the motors with a lot of turns, remember these things are all externally wound, then the pre-wound coils dropped into the slots. They tend to do the many-turn motors with a pretty generous coil size to layer all the end-turns together with the adjacent tooth when winding. This is going to be weighed against the section of copper between the axle to the coils that is the same on both motors (but under twice the current on the 2-turn).
It's going to be a pretty close match IMHO, I think the 2-turn may well have the advantage in efficiency due to the decrease in end-turn mess from the sh*tty way they wind these things, but either way it's an extremely close match.

Ironically, the 5305 with only 20 pieces of copper in the slot has the lowest continuous torque capability, most heating+lowest efficiency (highest copper loss), and lowest continuous power handling of the whole 530x series, yet it has the reputation of the motor to choose for someone seeking torque.

 

[rhitee05]

If I had my 'druthers about the FET board, here's what I would propose:

- Layout for 12x TO-247 FETs. This would accommodate a large number of FET options, from standard parts like the '4110 and '4468 to high-end parts like the FET Luke suggested in Super-247 package.
- Connectors on board for drivers to connect to external FETs
- Include connections for voltage rails, so you could use an external driver "booster" board similar to what Jeremy has posted elsewhere
- For the on-board driver, use a good high-current IC, 4A+ if possible. This should provide sufficient oomph to drive most FET configurations, so only the most obscene external FET arrays would require a "booster"
- Very careful attention to layout and generous use of capacitors, both large and small
- Provisions for protection circuitry, like TVS diodes or MOVs across the FETs, maybe a place for TVS on the gate drive, etc.
- Use a 4-layer board, with 2 oz external and 1 oz internal copper, if economically feasible
- It would also be nice to include provisions for one or more temp sensors which can communicate back to the MCU

 

[Kingfish]

Understood; thank you Eric & Luke both for the clarity.

Generally speaking, slightly off-topic, and with brevity - what would be the effect of high current vs. high-voltage in a coreless motor design, especially in regards to BEMF?

~KF

 

[oldswamm]

IF you use twice as large of wire all the way from the batt to the windings, the efficiency and heat should be identical.
[edit] Not trying to address your question KF.

Even ebikes calculator shows the 5305 as being short of torque.

And, what rhitee05 suggests in the post while I was typing this, is about what I had in mind.
(I need to search out Jeremy's booster now.)

Bob

 

[liveforphysics]

I'm totally fine with all that suggested plan as well rhitee05. Lots of excellent input for the parts the REALLY matter in a controller.

But, if it's going to be a paralleled fet design... I would like to see the 12fets arranged in an inductively and space logical manner. No stupid single long row across the back edge of the board crap if you're going to do a paralleled FET arrangement. We can make the device much smaller and better performance with a proper layout, and who cares if it takes an extra 20-30mins longer to assemble? It's a hand-built kit controller, that's kinda the idea of using your own man-power to enable better designs that a mass production factory bean-counter would green-light.

 

[liveforphysics]

(I need to search out Jeremy's booster now.)

Bob

I knew I liked you Bob