PINNED - Roboteq Controller - developing for powerchairs

[LROBBINS]

25mV is a 1.25 Amp signal (and I assume that this is peak-to-peak, so the average is much less) and I doubt that it will cause any problem at all. Lenny

[Williamclark77]

Many thanks to you three. I might get a few hours at home tonight to mess with it. Maybe. I will check for certain and report back when I get time to. I may still have a few small ferrite doughnuts pirated, out of all things, old joysticks. I have one on my joystick wires in my chair that came on it.

When I checked the sensors the wires were in the bird's nest testing configuration from the cell phone picture I posted a few days ago. The bird's nest from going through the fun procedure of reversing the brushless motors.

[Williamclark77]

Very quick test. I only had about two hours, but we got the wiring 95% complete. Good enough for a quick test. All safety functions tested to work (except unhooking the throttle axis of the joystick) on jackstands and then tied down. Even with no script it moved around very smoothly and turned easily unloaded. Impressive. I loaded in the last working script BM had been using and adjusted for my setup. Not the newer test script. MC set on 10. Basically off. Sommer drove it (not enough time for me to transfer into it) all around the living room with the pot bottomed out. Very smooth. No cogging. Turns at angles very smoothly. It will not quite zero turn. Stalled, but very close. I am now uber excited. They appear to need a little mc, but not much at all. Set at 20% Roborun shows it to be pulling about 35-40 amps

I did not try turning up the mc. I'm not testing that with her in it. That will wait until I can get an expendable test dummy over. Can't test it by rc. My transmitter isn't back from Horizon Hobby. I'm headed back to work. Just wanted to share my excitement.

Oh, current sensors are only bouncing around about 10mv now. 10 total. 5 in each direction.

[Burgerman]

Great. That was at low SpeedPot setting?

Stick to all my settings for now. Don't be tempted to increase turn or add any expo or acc or anything.

Just up motor compensation a bit until it works smoothly. It suddenly starts to help. Then report back! You will be surprised how good it will be.

It sounds exactly like mine. No compensation isn't quite good enough with a body in it. 50 makes it ok, may need a bit more (on mine).

Set at 20% Roborun shows it to be pulling about 35-40 amps

Stalled? Is that motor amps? Don't know what that means on a brushless 3 wire system! Probably = to about 40 to 60 amps on mine? I wait in anticipation! All that evil bad acc steering and deceleration turning will be gone and sensible deceleration rates, back stick braking etc all now works.

[Williamclark77]

Yes, speedpot bottomed out as slow as it would set.. A crawl speed. Motor amps when she tried to zero turn jumped up to 40ish in Roborun and I told her to let off if it didn't turn easily. It wasn't moving, but still trying. I forgot to air my casters back up too. I never aired them back up after separating the rims to polish them. It was a quick 10 minutes of rolling around the living room. I'm tickled pink though!

I won't be back home until 7pm or so my time tomorrow night and won't have but an hour or two. Should get the rest buttoned up for proper testing the next day and then I can help test Lenny's newest version. I do have to make a bracket to hold my relays on. Shouldn't take long.

Oh, and there was nothing wrong with my tilts. The joystick handle is removed for mounting. With it removed it doesn't hold the guts in the neutral position. Screwed the handle on and worked perfect.

[Burgerman]

A crawl speed. Motor amps when she tried to zero turn jumped up to 40ish in Roborun and I told her to let off if it didn't turn easily.

That's quite high for it not to turn. Esp on a 3 wire brushless setup. I would expect it to start to turn at that level. Its eq to about 60A ish on brushed. Your tyres being a bit flat shouldn't make that much difference. Maybe a bit nose heavy or tall geared. You may need to fine tune a few things. You may find its geared for about 20mph or something.

That said MC should still fix it. But its going to waste a lot of battery.

My stock mobility chairs will turn very slowly on the spot at 30-40A on a carpet. When you do it faster at full stick because you don't have all day, it hits 100 (max) easily enough due to the mass. But these are very rear weight biased.

Makes you laugh when you see nose heavy chairs sold with 50A controllers! Doing this experiment with a clamp ammeter on a motor wire, years ago, was one of my reasons for moving c of g back. Makes batteries last longer, and the chairs steer better.

[Burgerman]

You are getting 40A at 9% pulsewidth.
Call it 10...
Based on this,

You will have a 400A stall current at full battery voltage... That's limited to 80A obviously by the brushless roboteq.
Mine works out at about 260A stall doing this limited to 150.

A 50 percent pulsewidth will give 200 stall...
A 25 percent pulsewidth will give 100A...
A 20 percent pulsewidth will max out the roboteq at stall.
You will STILL have full stall torque available at 100% pulsewidth at about 10 mph. So steer shouldn't be a problem!

I hope you didn't go a bit far with the low impedance thing!
Motor compensation will need to be set quite low.

[Williamclark77]

Don't get scientific yet. Haha. Far from a repeatable or accurate test. Just her sat in it and moved the joystick. I will get real numbers later as well as go over everything mechanically soon as possible.

[LROBBINS]

For your (enforced bedtime - John, enforced away-from-project time - Will) reading.

I've finally drafted the first of several messages I want to post at the Roboteq forums. Before I do so, however, I'd appreciate if you'd look this over to see if it makes the case with reasonable clarity.
**********************************************************************************************************************************************************************
Acceleration/deceleration with mixing

I am helping two friends with scripting their Roboteq HDC2450 and HBL2360 controllers. In both cases they are using Throttle and Steering joystick outputs that are mixed by the controller to get Motor 1 and Motor 2 command values and both of them reported a serious controllability problem when decelerating from a forward-moving turn. Indeed, there is a problem and it applies to acceleration while turning as well as deceleration, though it was first noticed during deceleration. It derives from the fact that although the motor commands are the result of mixing the Throttle and Steering values, the mixing algorithm is not in any way applied to acceleration and deceleration rates.

It is easier to see this graphically than to explain it with words, so I have attached several graphs. The first illustrates the "normal" result of centering the stick after being in a steady forward turn. The graph contains four curves; two correspond to Motor 1(red) and Motor 2 (yellow) power as one would see them in the Roborun Run tab, and two show what the vehicle does as the result of these - these are just the result of un-mixing the motor power values to get fore/aft (green) and left/right (blue) motion. (BTW, These graphs were made in a spreadsheet, but I think it would be very useful, and rather trivial, to add the fore/aft and left/right curves to the Roborun graph choices as they give a fairly intuitive picture of the effective results under mixing.)

No mix decel.jpg (19.68 KiB) Viewed 5929 times

At the start of this graph, we are in an 18% turn with M2 power less than M1 power. When the stick is centered, both power curves decline at the same rate, and forward motion decays in parallel. Notice, however, what happens to turning motion. Turning does not stop, nor decay at all, until M2 power has gone to zero - when forward motion has already largely ceased, and fore/aft deceleration rate is halved because one motor is already stopped. If this were an automobile, I think I'd describe this as severe torque oversteer during deceleration. As these are actually high-performance wheelchairs, the result is busted doorways and perhaps busted legs.

The second graph shows that, if instead of just applying constant deceleration values to both motors, we use the un-mixed fore/aft and left/right results to bias the deceleration rates, we can get rid of this problem. Here, deceleration rate of the inside wheel is reduced and that of the outside wheel is increase, but because the turn rate decays over time, so does this bias.

Mixing decel.jpg (17.23 KiB) Viewed 5929 times

Now, the power curves are not parallel straight lines but converge, and fore/aft motion declines linearly even at the end. Most importantly, instead of continuing to turn at a constant rate until forward motion has nearly stopped, the vehicle also straightens out.

As shown in the next two graphs, we can get an even snappier steering response by over-weighting the biasing of deceleration rates.

Mix weight 200.jpg (16.16 KiB) Viewed 5929 times

(Continued in next post because no more than 3 files can be attached per post.)

[LROBBINS]

(continued)

Mix weight 400.jpg (15.76 KiB) Viewed 5927 times

In both of these, initial fore/aft deceleration is not linear, because the deceleration rate of the inside wheel cannot be less than 0, but the net result is that turning stops quickly while fore/aft motion decays in a nicely controllable fashion.

Although I have been able to get this desirable behavior by scripting biased, variable, acceleration and deceleration rates (some apparent bugs in _ACEL, _DECEL, _MAC and _MDEC will be covered in another post), I think that this is really something that should be "built in" to Roboteq's mixing algorithms.
******************************************************************************************************************************************************************
Ciao,
Lenny

P.S. Here's where I think the 10-fold error in _MAC and _MDEC happened - on 10/23/2013 there was a firmware change:
- Increased absolute max motor acceleration from 100ms to 10ms to max RPM

I'll bet the firmware change was accompanied by a change in the profile XML file, but was not accompanied by a change in the GetConfig and SetConfig functions (and I'll bet not in their serial command equivalents either).

[Burgerman]

Well it looks clear to me.

Baffles me that of all the other users with all the other controllers, that nobody ever noticed this.

[Williamclark77]

Very nicely written. They SHOULD be able to picture it pretty clearly with your description. I really don't understand why it was made to operate how it does. I'm just glad you're able to write something to fix it and HOPEFULLY they will do a firmware update to correct it. Probably best for a different post, but I'm also unsure why it will not apply any direction change to a motor until it completely finishes the deceleration routine. Ie - Can't brake using back stick. Motors will not change direction until they come to a stop at the programmed deceleration rate. That's just dangerous.

[LROBBINS]

Well, you won't notice this if you are running a closed loop algorithm, or if you're using a magnetic stripe tracker, or if it's a very slow vehicle, or if your application doesn't use mixing, or if you're positioning solar panels or theater scenery, or moving a strawberry harvester down rows of plants Very few of the user applications shown at the Roboteq site are of a sort that would make this problem rear its very ugly head, and the one wheelchair they show looks like a slow, indoor-only device probably designed by people with the same level of control savvy as they have for hardware design (i.e. their cure for bad control would be delay, delay and more delay).

Ciao,
Lenny

[Burgerman]

But they sold hundreds to robot wars people in tiny school 10 and 20 amp sizes for rc robotics use, and grass cutting or snow clearing shovels etc.

I could tell it had issues after 10 secs in my drive.

I am now wondering what mix algo 3 does.

[LROBBINS]

Were those 10 and 20 amp sizes 1 or 2 channel? Did they have a mixing algorithm, or did the user have to supply that himself? At that level of robot wars maybe ridiculous turn behavior is a good way to confound the enemy.

Neither Woody nor I were ever able to make heads or tails of mixing mode 3. Strangely, although the config tab lets you set separate, mixing mode 1, mixing mode 2, or mixing mode 3, if you search for MXMD (the serial or script command for setting mixing mode), you'll find that it can take only 3 values: 0 - no mixing, 1 = mode 1 or 2 = mode 2. You'll either have to experiment or ask Roboteq for some explanation. Warning though: my script explicitly deals with mixing modes 1 and 2, but may give some weird results with mode 3!

Ciao,
Lenny

[Burgerman]

Were those 10 and 20 amp sizes 1 or 2 channel? Did they have a mixing algorithm, or did the user have to supply that himself? At that level of robot wars maybe ridiculous turn behavior is a good way to confound the enemy.

I am at the wrong PC on my bed but you can load all those small single board dual channel controllers. I thing they all have the same basic setup, options etc, but less inputs/outputs.

I just checked. All the small controllers like these that schools or the tiny battlebots use for eg, exactly the same. http://www.roboteq.com/index.php/robote ... 130-detail

EG, Open the Roborun software, and load a SDC 2130 which is a tiny hobby sized single board, and its got almost exactly the same options as the large controllers. These are used on mini battlebots and other small hobby robots and almost exclusively skid steer. They just don't know there's a problem!

Which frankly amazes me since the powerchairs have no steering as you accelerate, or as you slow! And if deceleration is set to a sensible low rate, you cant brake fast when needed either. It just continues straight, or if turning continues in an ark that tightens up. I noticed it in a few secs, thought I had RC problem! Its not drivable unless you have no clue how it SHOULD respond. Which I expect applies to most people...

[Irving]

Being able to write/modify your own control/mixing algorithms is critical.. which is why I'm going with a homebrew controller. Since there aren't enough hours in a day I'm going to work with a friend to uprate a single channel design he's already built and tested. His is good for 1kW motors, either sensored/encoded brushless 3-phase BLDC/PMSM or brushed DC (for those in the 19thC - his words not mine ). Currently runs 140v @ 15A at 40degC ambient with no extra cooling, but with extra MOSFETs & driver chips will easily do 50 - 90v @ 200A; needs a new PCB layout and uprated current sensors but we'll keep it as three PCBs - dual channel controller, plus two output/current sense for maintainability.

driver-small.jpg (75.56 KiB) Viewed 6068 times

[Williamclark77]

That is some impressive hardware! I look forward to updates on your progress.

[Williamclark77]

Got everything buttoned up, safety tested, and mechanically checked last night. The suspension sag (way too low) was the slight issue with turning when we did the quick test a few nights ago. I still have to put on the back cover after making some spacers for it on the lathe and make a bracket for my relays. The tape holding them now is perfectly suitable for the job, but looks ghetto.

After 30 minutes of butt time in it this morning my first impressions are FANTASTIC! Little motor compensation will be needed. On my wood and tile flooring inside my house, it is not needed at all. With the pot bottomed out and the slightest stick it goes. Literally at the least amount of stick I can physically give it gets smooth motion in any direction. Perfect. No lag. No stalling. No cogging. No jumpiness. Just buttery smoother movement as good as any powerchair. Even zero turning. Caster angle does not matter. Their position has no affect.

Outdoors motor compensation will be needed a wee bit. In thick grass or sand (see video) it will not stall, but hesitates then jumps a bit when zero turning if the casters are in the worst possible position. That is at the least amount of stick I can give it though. Pot bottomed at 20% and only applying maybe 1/4 stick, then it will kind of hesitate and jump as the casters swing around to where they're giving less resistance. Give 1/2 stick and it makes no difference and still going very slow. Being in mud, deep grass, or binded against something may exacerbate this minor issue and require motor compensation. My expendable test dummy (little brother) is supposed to come by tonight and pilot while I adjust settings.

Two major issues though:

1 - The analog safety stop will not work on my throttle axis. Every other input works right and triggers a shutdown when any figure outside the deadbands are reached except it. Analog input 10 is my throttle. As it is right now, if that wire breaks, shorts, etc it goes wide open. Unacceptable. I will test when I can later and see if it will work when not set to a mixed mode or in single channel operation before trying to change input pins. My better half will probably present me with divorce papers if I take that plug apart.

2 - Short circuit protection keeps getting tripped if I make a quick joystick movement. I have it set on slow. If going forward and apply back stick, or turning and quickly change directions, or quick joystick stab from neutral it short circuit faults. I read issues about this and very low resistance motors months ago. Heavy input doesn't cause any issue as long as you add it slowly. It's seemingly not triggered by the amount, but by the speed at which it is applied. I will research this as time allows. Ideas are welcome if anyone has any (besides turning the short circuit protection off).

I will post a video when it is done uploading.

[Burgerman]

Script working good? Balanced?

The short circuit trip is what had me concerned when you went for the extra low impedance motors... It trips due to quick stick movement /acceleration as it pulls more amps from EMF on a stopped motor. As it turns this load drops away.

What is speed like? If its faster than about 15mph you may need to use shorter gearing. Or a lower battery voltage. And / or longer maybe thinner motor cables will also help. You are hitting the 75A limit of the brushless controller too easily. And before the roboteq can drop the current far enough it trips. It doesn't matter if its limiting amps properly, but if you go too far...

Set max amps a bit lower to say 60.

How does it feel when you hit the cutoff? Mine hasn't done this so far set to the medium setting. But 1 cell lower voltage, and a higher 150A max limit probably stops that. And I have it set to 120.

[Williamclark77]

Yes, script seems to work great. Turning is too slow at speed and what not, but nothing that shouldn't be able to adjust to my personal preference. Feels very nice.

I did not butt test top speed. Didn't want to trip the short circuit protection while going 15+ mph. The cutoff feels like you simply turned the controller off. For me, being I have no mechanical brakes, is nothing alarming. It just coasts to a stop - no big deal as long as you are coasting in a safe direction, hence why speed not tested yet. That could be different if using brakes. I'm not sure if the little mechanical brakes are powerful enough to lock up the wheels at speed if activated, but could get your attention.

Gearing is also too tall. As I expected, the motors turn more rpm than advertised. Supposed to spin 3200 @ 48v. They're turning 3475 rpm at 46.5 according to Roborun. I was shooting for and bought gearing for 13.5 to 14 mph. I'm getting a bit over 15 on jackstands.

[Burgerman]

Use 12mph gearing, longer motor cables, and that trip may just go away. It wont trip at high speed, as current is low unless you jam it in reverse.

IS there a setting to disable it?

Says in the manual, a shorted motor cant be detected at low pulsewidths, or with long motor cables...

[LROBBINS]

Hi WIll,

What happens if you decrease the base acceleration? Say from:

M1Accel = 3000
M2Accel = 3000
= 1.4G

to

M1Accel = 1000
M2Accel = 1000
= 0.7G

or even

M1Accel = 500
M2Accel = 500
= 0.35 G

Ciao,
Lenny

[Burgerman]

What affect does this actually have?
And on other settings?

Yes, script seems to work great. Turning is too slow at speed and what not, but nothing that shouldn't be able to adjust to my personal preference. Feels very nice.

Will, mix mode 1 or 2?
2 gives a 10 percent faster speed.
And is the one we should be using with a round joystick.
Meaning you can gear about 18 percent lower, get 14mph still and not trip anything as it will pull about 18 percent less amps in any given situation.

[Williamclark77]

My test dummy should be here in a bit and I will try some different settings. I'm not keen on the idea of me testing with nobody else here. I will keep some notes of what I do and what works and what doesn't.

Short video. Pardon the camera operator. She was a volunteer. I can't complain. Pot low as it will go. Giving it about 1/4 stick. No need to post indoor. Smooth floor didn't add enough resistance to notice. I'm intentionally turning into the wrong direction of the casters. This is with no motor compensation.

[Burgerman]

If you are using my script settings, that is 18 turn scaling, and 50% reduced at lowest speedpot setting.

FULL STICK at lo pot, is a tiny 9 percent pulse width.
I set the low end of the speed pot very slow on purpose to test exactly this sort of thing. If it turns that well at HALF stick, (even ignoring the dead band) at just 4.5 percent pulse width, then a tiny bit of compensation will sort it out completely.

With no compensation a stock mobility device will not even attempt to turn at all at that sort of pulse width. Or even 2 or 3x higher.

Its steer behaves exactly as mine does it seems. Even though its longer and more nose heavy.
You have some creaking. Seat suspension or something I think.
Don't change a thing yet, other than adding m.compensation in small steps with a laptop outdoors till it gets too good! Then go back a bit. It works very affectively... Including high speed steer.

[LROBBINS]

I don't expect lower base acceleration rates to have a very noticeable affect on how it drives unless set way too low. Nor should it affect how other settings work. I am just wondering if slowing the ramping of PWM will reduce the false "short circuit" cutout. Ciao, Lenny

[Williamclark77]

The creaking is the seat pan (not sure what it's called. Flat sheet plate the cushion velcros onto). Another nit picky item on the list of things to correct.

I had the short protection on fast when testing/assemblimg the motor wiring and it was throwing one. Chair behaved the same after changing to the "slow" setting, so I ASSummed it was still it. Turned short protection off this afternoon to see. Still did it with him in it. Hooked the chair up to the computer and had him mess with it until throwing a fault code (I had no way to do this alone earlier). It was not a short code, but an e stop. After much sorting and checking it was the current sensors. Apparently, they give a little more range in the output than amps x the resolution that I set the deadbands to. Extended the deadbands a bit and problem cured.

Adjusted the motor compensation up to 85 (motors are supposed to be 70 mOhm resistance). I didn't try any higher. Never got "jumpy" like it was set too high. It is, however, nearly unstoppable above about 15% input (I didn't see your comment nor check that closely in the script to see that it was set much lower than I thought). So, even better. I have the preload on my casters pretty tight to stop from grocery carting at speed (tighter than they were for that video). If they refuse to swing around, the back tires start to spin or slide them. Same test as above video but with motor compensation and it will drag the casters or tire spin, but it doesn't hesitate then go. I honestly could not ask for it to do better.

I THINK, and it may simply be because it's the result I'm expecting, but they don't feel quite as smooth as a stock powerchair at super low inputs. Like below 15% or so. Given how over geared they are they should not be. I would say it was moreso the gearing than cogging. Nothing you would notice unless looking for it. I'm really being over critical.

One other problem. My right motor, which is wired to run backwards, must not have the best phase order. The wires on it got very hot. The left motor, which I wired A-B-C on both the phase and hall sensors, never got even slightly warm. It sounds smooth and is not drawing more amps than the other, but apparently is not happy with the hall/phase relationship either way. I'll start the fun process of unlooming and figuring out a better order tomorrow.

I did not get to changing any other settings. I quit as soon as realizing the phase issue.

[Burgerman]

Interesting! So you are happy!

So its now good to go! It should (and likely will) be just as smooth as a stock brushed chair once gearing lowered to a sensible level, at which point it should be smooth and controllable everywhere.


My current sensors are:
0 --- calibrated 2438 --- 5000
0 --- calibrated 2442 --- 5000
And all analog inputs are set to relative - or you would get many issues!

And mix B is best for much more speed, allowing shorter gearing to be used for less amps at the same speeds. And less heat in motors, better range, better torque.

[Burgerman]

I have the preload on my casters pretty tight to stop from grocery carting at speed (tighter than they were for that video).

Interestingly I expected this issue. Some of my slow chairs do it. But for whatever reason this one just doesn't shake and they are not preloaded. They tend to do it most when tyres are new if they are going to do it at all, and it literally wears off!

It sounds smooth and is not drawing more amps than the other

Then the heat cannot be coming from wiring incorrectly. I am not sure how you could! Can you do that? It would either pull more amps, turn slower, be less smooth at low speed, or something wouldn't it?

My test dummy should be here in a bit

I don't have one!