Maslow 4 - The next generation of Maslow

Hi, I still have a couple of questions. I installed the 0.73 update but pressing Test says INDEX HTML VERSION 0.71, is this correct?
I ask because I still have trouble releasing the straps backwards.

1 Like

Enea wrote:

Hi, I still have a couple of questions. I installed the 0.73 update but pressing Test says INDEX HTML VERSION 0.71, is this correct?
I ask because I still have trouble releasing the straps backwards.

That means that you didn’t upgrade the index.html.gz file at the same time, you
always need to do both.

David Lang

1 Like

Unfortunately in this case that means I was rushing to get things shipped out and I forgot to bump the version numbers :man_facepalming:

Thanks, maybe I wrote it wrong because I have to translate, the problem is that to extend the straps I have to push in as with previous updates.

That could be an issue with the encoder boards themselves, can I send you some modified ones to try?

So not everyone has this problem. The strange thing is that with the 0.68 update the straps extended correctly, I was even able to do a calibration. The problem appeared with subsequent updates.

1 Like

I noticed that I no longer have the FAVICON.ICO file, is it possible to find it?

If it’s the only solution, that’s fine.

I think it will be a problem to dismantle the encoder boards, because the screws have thread locks and are damaged very easily

1 Like

It seems like there should be a software fix because the issue only appeared after about 0.68, but I haven’t figured out why that could be. I haven’t been able to make it happen myself. The other person who had the issue fixed it with modified encoder boards so that seems to be our only for sure solution at the moment. The threadlocker on the motors is the removable kind so it should come out (it just resists vibration), but it is a hassle to have to disassemble them.

DM me your address and I’ll get some in the mail ASAP so we at least have that option

Hi Bar, thanks for sending me the encoders, I replaced them and the problem was solved.

I noticed that during the vertical calibration, when you tighten the straps, the lower part of the slide rises from the panel, is this normal? The strap anchors are a few cm higher than the work surface.

1 Like

Excellent! That is valuable feedback.

Is this with the z-axis lowered all the way down?

During the calibration test the Z axis was raised by approximately 2 cm. The anchors in the upper part of the frame protrude 9 cm from the work surface.
!

One other idea that I just want to throw out there into the public domain is that I don’t think you actually need current feedback on the motors. I think that we could do calibration by driving the motors with a set torque (voltage) and then monitoring the encoder to see when it stops spinning which would be useful on a bare bones budget version.

Not something I’m pursuing, just wanted to keep the idea out there so nobody else can patent it

2 Likes

This is interesting. One benefit is that it would align everyone, meaning we currently have variability of resistance across arms, across all M4’s in the wild and this approach would take that issue off the table and normalize the results of the read-belt-length function.

1 Like

For a permanent magnet dc servomotor such as you are using, torque is proportional to current, not voltage. This relationship is independent of RPM and depends only on the torque constant of the motor.

If you apply a fixed voltage, the torque is constant only if the motor is stalled. If the motor is turning (say you are retracting or moving the sled) then the voltage is reduced by the back EMF generated by the rotation of the motor.

You have a choice - contant current or constant voltage.

With constant current, torque t = I*Ct, where Ct = the torque constant.

With constant voltage, torque t = Ct*(V - n*Cb)/Ra, where

Ct = torque constant
n = RPM
Cb = back EMF constant
Ra = armature resistance

So if you use constant current, you get constant torque independent of RPM, and motor to motor variation is a function only of one parameter, Ct.

If you use constant voltage, you don’t get constant torque, it decreases with RPM, so the belt that has to move faster for a given move will have less torque and be more likely to fail to follow the desired profile.

In addition, the motor to motor variations in torque you get for a given voltage is a function of three parameters - Ct, Cb, and Ra, plus the effects of RPM.

If you care about torque, controlling current is the only way to go for permanent magnet DC motors.

True, but when we are taking a measurement we are doing it when the motor is stalled which we can verify by watching that the encoder is not rotating so we know that the motor is not moving at that moment

WFD wrote:

If you care about torque, controlling current is the only way to go for permanent magnet DC motors.

In theory you are correct, but in practice, I don’t think it matters.

Not only is there the variation between the motors, there is a huge amount of
variation on the mechanism (how tightly meshed is the idler to the motor?, how
much friction to rotate the spool, including how tightly are the bolts holding
the arm together tightened)

so we aren’t talking about any measurements that can be traced back to
meaningful numbers.

So what I think @bar is suggesting is that instead of applying the PWM and
watching the current spike to determine that the motor has stalled and the belt
is tight, apply the PWM and just watch the encoder for motion to stop to
determine that the belt is tight.

potentially saving a few cents (or probably more useful, a few I/O pins) by
eliminating the current sensor and just relying on the encoders.

David Lang

1 Like

My concern is that using constant voltage there will be much greater variation in actual torque from arm to arm and unit to unit, and so it is more likely that you will have retraction problems in which you will stall out when the belt is actually slack, or at the very least, larger variations in belt stretch and frame flex.

I have two degrees in mechanical engineering and spent 20 years of my career designing and building robots and automation and I am quite confident that moving away from constant current is something you will regret.

1 Like

What is the current status of M4. have the calibration issues been fixed? I dont’ want to buy it till all the bugs are worked out. god bless those that are the first users and finding these issues; i know from past maslow experience I just dont’ have the patience for that.

2 Likes

I’d say if you don’t want to be in the early phase of this, to hold off. There are MANY people having great success, but there are a few issues that remain that are very impacting to some of us more than others.

2 Likes