BGA vision/placement demo

[smdude]

For the added weight and the fact that you would still need to come up with a way to get vacuum to the nozzle, I think the geared solution would be much better and allow very fine rotation control. I see on the openpnp group there is a nice dual nozzle carriage coming along. There's just so many options around and each one just gets better and better!!!
I need a 3d printer And about 60 hours in a day

[JuKu]

> Not sure about the Juha-original rotational axis; it's a 1.8-degree stepper but it has some gearing.

The original z motor is 0.9 deg.

[WayOutWest]

> Not sure about the Juha-original rotational axis; it's a 1.8-degree stepper but it has some gearing.

The original z motor is 0.9 deg.

Hi Juha,

(I think you meant "a motor").

Yes, my mistake, the stock A-axis motor is 0.9-degree.

So it can definitely handle 0.5mm BGAs, even very large ones, easily. It's more than twice the accuracy (2x from the 0.9-degree stepper, and a bit more from the gearing) than the direct-drive 1.8-degree stepper I'm using to do 0.8mm BGAs.

[mawa]

@wayoutwest:

have you thought about using 1/2, 1/4, 1/8 or 1/16 step drivers like the trinamic pan.drive for the A axis stepper?

Due to the fact that there is almost no torque load the positioning should be very accurate and with very little slew.

Here's a link to trinamic.

http://www.trinamic.com/products/integr ... per-driver

I will take a try using this board I am using in an other project, which utilizes that kind of stepper controller. It easy to control via TWI/I2C and does position count, ramping and a bunch of safety measures.

http://www.elv.de/intelligentes-schritt ... usatz.html

Sorry the page is in German but you probably get the idea .

[JuKu]

Oops; yes, I meant a motor. TinyG can do 1/8 microstepping.

[WayOutWest]

@wayoutwest:
have you thought about using 1/2, 1/4, 1/8 or 1/16 step drivers like the trinamic pan.drive for the A axis stepper?

I'm already using the TinyG's 1/8th microstepping. Upgrading to ST's chips is high on my to-do list; they offer 1/32 microstepping but that's not my main reason for liking them.

Due to the fact that there is almost no torque load the positioning should be very accurate and with very little slew.

Open-loop microstepping isn't a cure-all for positional accuracy; doubling the number of microsteps will not help accuracy nearly as much as doubling the gearing or halving the motors' step angle.

At rest a microstep is the result of making the motors' two phases fight against each other, with some fixed ratio (1:2, 1:4, 1:8, etc) between the drive strength applied to the two H-bridges. But just setting a particular ratio in drive strength doesn't guarantee you get that ratio of position, and it radically increases the number of factors involved in accuracy.

Think of it this way: with whole-stepping you don't have to exactly match the resistance of the wires leading to the two phases of the motor. If they're mismatched the machine will run a bit choppier but still come to a stop at exactly the same position. If you're relying on microstepped positions, your positioning within a full step now relies on stuff like this! Seriously, now a slightly-weaker solder joint can affect the final resting position of the machine. Or minor ESD stress that slightly weakened one branch of your steppers' H-bridges. Ugh. For 1:2 and 1:4 microstepping this is still an obvious win, since the uncertainty added by all that other gunk is not likely to be more than 12.5%. But as you start pushing towards 1:8, 1:16, etc you have to ask yourself if all of these (likely uncontrolled) electrical properties are more trustworthy than the mechanical properties. Can you make solder joints with 3.125% consistency in resistance? Are you sure the heatsinking of the motors' two coils is identical?

I dunno, part of my aversion here may be as a digital IC design guy. It's been drilled into my head that the relative drive strengths of components are not something to be trusted.

On the other hand ultrafine microstepping would be a very nifty tool for integrating a linear encoder. It can allow really really cheap mechanics to be as precise as the linear encoder if you mainly care about the final resting position (and not the path). But good linear encoders seem to be more expensive than good steppers right now. Trying to figure out some trick for getting a high-resolution linear encoder is also on my to-do list, but further down. Probably somehow leveraging cheap PCB manufacturing. Going closed-loop would make a lot of mechanical stuff matter a lot less.

[Mark Harris]

Check out TI's Inductance to Digital converters, they have some neat demos of absolute positioning of a shaft based on induction, and ridiculously accurately too - the biggest bonus is that both the IC's and the method are super cheap.

http://www.ti.com/lit/ml/slyb212/slyb212.pdf - see page 4 for a diagram.

[JuKu]

Also AMS. Magnet strips <10$/300mm, ICs <6$/1k, sub-um accuracy. http://ams.com/eng/Products/Position-Se ... on-Sensors

Lots of other players, too. Google keywords are linear magnetic position encoder

[JuKu]

[quote="WayOutWest"]
Open-loop microstepping isn't a cure-all for positional accuracy; doubling the number of microsteps will not help accuracy nearly as much as doubling the gearing or halving the motors' step angle.

At rest a microstep is the result of making the motors' two phases fight against each other, with some fixed ratio (1:2, 1:4, 1:8, etc) between the drive strength applied to the two H-bridges. But just setting a particular ratio in drive strength doesn't guarantee you get that ratio of position, and it radically increases the number of factors involved in accuracy.
[quote]
True. Microstepping has two benefits: It makes the machine run smoother, and it does increase accuracy. Both benefits start to diminish at 1:8 and down. The more you load a stepper motor, the less accurate it will be, up to the point of losing steps(*). I would be confident that in our application, microstepping will get you to the right quadrant and maybe in the right 1/8 region. I'm not sure about the latter and I wouldn't bet that 1:16 would be any more accurate, even if it would be available.

(*: Losing single steps is a problem for CNC machines, not for a pnp machine. A pnp machine doesn't have (at least shouldn't) much stuff resisting movement. We don't lose a step or two resulting to a small error. If this machine is driven too hard (=too fast) the motor will stall and lose a big portion of the whole move, going _way_ off. Next move will likely trip a limit switch.

[mrandt]

Inductance to Digital converters

How cool is that?

I think that might also be an answer to the Z-limit-switch with Juki nozzle quest... I need to check how I could connect to that spring at the tip. Or just integrate a coil into the nozzle holder; it's inductance would also change when the inner nozzle tube moves upwards.

Thanks for this hint!