My musings on the topic are here:
And @madgrizzle being the amazing being that he is is building one here:
I donāt think the RPI has enough PWM pins to run a motor controller directly (not even sure it has enough GPIO for all the encoder pins needed for a four motor design) so if you have an RPI as part of the design, it needs some sort of āintelligentā controller it can talk to since it wonāt be able to talk directly to the TLE5206/whatever chips and motor encoders. Thatās why I like the idea of incorporating a microcontroller directly on the motor controller (4 dc brushed motors + stepper) and just making that microcontroller do really basic work like managing the PID loop and reporting errors. Leave all the complicated stuff to the RPI (e.g., processing gcode, UI, etc.) so you seldom, if ever, have to update the firmware of the microcontroller.
Here is some detail on the Pi GPIO pins: https://www.raspberrypi.org/documentation/usage/gpio/
So, this would look something like the Einsy Rambo with a Rasberry Pi Zero W, used in the 3D printing world?
[Edit - photo courtesy of Prusa Printers blog The Rpi zero W is used to run OctoPrint]
Iām using 10 pwm pins an 10 gpio at the moment in my four motor design. I guess you might be able to rework the pwm with the tle5206 to just use one pwm pins and do something different for the z axis or maybe use softpwm but not sure if that would work for the application.
+1 on that.
While I understand the market concepts/dynamics that @JWoody18 and others were referring to.
Even near $500 is a big ask for many. That can be hard to justify for one tool no matter how useful and amazing.
Not everyone is looking to use this for business/income. For many of us it would be just for hobby.
I was on the fence for some time and missed the opportunity for the first one. Still trying to justify the price on the current offerings.
If it was $700+ I would not even consider it much like traditional CNC router machines.
In the very beginning when it was introduced to me, I might not have either, honestly. I will say that it feels like a majority of people, at least the ones active on this forum, seem to tinker enough over time and add upgrades slowly, pushing their all-in costs way past the advertised price point. EDIT: myself included!! 
Maybe we can have the best of both worlds by having easy upgrade options, as bar said above. If that continues to be the mission, I do feel that communication to new users, who are in the buying phase, about the existing āstockā options and āupgradeā options, could be improved.
Good point, on lower price entries into tech such as this. Overall costs can well exceed the initial price.
The tough sell is that first price.
As @bar alluded to above, I think itās important to keep the initial ākitā price low to get people into these that may otherwise avoid the tech and miss out on the joys and benefits.
3D printing was like that for me, It wasnāt until good printers were in the sub $300 range that I took the plunge and got one.
Now that I like it and understand the value of that tool, I can see and rationalize the benefit of adding a quality extruder that is near the initial price of the whole printer.
Exactly, for many (me included) its better to keep the initial cost down even if much more is spent over time to upgrade. It lessens any single impact on the pocketbook. For others who know that they want it and can afford it, I can see the benefit of getting a completely upgraded unit with all the bells and whistles.
This is part of my point about improving communication on options. Iām not sure if new, inexperienced users could know if a certain kit would be right for them, if āthisā piece of hardware is better than āthatā piece of hardware and why.
What Iām thinking about is: how to inform the new user, who has the money, that there are upgrades that could be done straight away, instead of buying the āstockā option and then saying āwell āthatā part doesnāt work well, wish I would have known about āthe other partā? My felling is that this scenario is frustrating and off putting to the āMaslow brandā.
before we get too worked up in different options, Iāll point out that we donāt
have anyone who is producing a ultra reliable/precision kit yet. I think the
problems are more on the software side now than the hardware side, but in a year
we have not put together a good comparison of the different triangulation kits,
let alone any of the other options.
We do know the Z axis can be improved significantly, but thereās no comparison
between the different Z options (other than they are all better, and mostly more
expensive or harder to build, than the stock Z option)
frame angle, we know that 20 degrees is too much and 5 degrees to little, is 15
degrees right or would be be better going closer to vertical?
a 12ā top beam seems better and it may be good to define that as the default,
ith 10ā or shorter being a fallback (including just a little more chain in the
kits). this would be close to zero cost
it seems pretty clear now that weights instead of bungees improve the machine,
this is pretty much a zero cost upgrade
itās only the more drastic differences (4-motor, gantry) that add a really
significant extra cost. Even more expensive electronics would not be a huge
cost.
David Lang
@dlang, @Metalmaslow is producing a very nice laser cut metal kit, although the first unit arrived today and we donāt have ādataā on its performance, by watching the video of it cutting, Iām pretty optimistic and confident in its performance. In my opinion, their kit options are a step improvement for the community.
The best way to keep costs down is to make all parts available for people to buy themselves. Right now the motors used are realistically only obtainable in large quantities easily and unless you are a skilled electrical person same thing for motor controller.
There are other Chinese motor companies. Aslong is one that make similar motors and sell them in small quantities.
Iām planning to use SPI to talk to both the encoders and the motor controllers. That would give us more information to work with (like real time current measurements) and should ultimately use fewer GPIO pins.
Based on everyoneās discussions about the price point, I agree with the idea that keeping the cost to get a machine as low as possible should be the goal, with optional upgrades. Iāve been playing around with the esp32 and it seems like a great replacement for the Arduino at a lower price. The point about having a clear list of what is available and what the benefits are is a good one. It could be a little bit difficult with multiple people selling different things, but I will work on that.
Great point about the motors. Using a motor which can be bought in smaller quantities would be a big step towards making the kit more open. Do you have any recommended motors which are available in smaller quantities? For the next version I am planning to use an external encoder so the motor does not need to have an integrated encoder.
That looks like itās readily available, how would you calibrate it? The
non-linerity is prety significant (0.8 to 1.2 degrees depending on temp and
magnet placement being off by 0.5mm) Iām thinking of the problems you had with
the makesmith magnetic sensor.
David Lang
Iāve been wanting to use the AMT series of encoders from CUI, they apparently
just released SPI versions (prior versions were rs422/rs485 based). they are
available in 12 and 14 bit absolute versions, but list their accuracy as 0.2
degrees (which with a 25 tooth gear would be ~0.1mm vs ~0.5mm of the magnetic
sensor listed above)
the problem is their price, ~$50 each.
This may be a case for the high-low mix, with a 10 tooth sprocket, the magnetic
version claims to give accuracy about the same as we can get with the AMT
version on a 25 tooth sprocket for probably <$15/sensor ($10 for the chip, +
board, connector, assembly, magnet etc). So support both, the cheap sensors for
the low-end kit and the expensive, but more accurate sensors for an accuracy
upgrade (especially when combined with larger sprockets and faster motors)
David Lang
P.S. one nice thing about going to a spi interface is that you donāt have to
keep up with the encoder pulses, and this drastically relaxes the real-time
requirements. The computer polls the encoders and records when it did so, if it
gets delayed a bit before the next poll, itās not a big deal. This is one of the
things that makes using something like a Pi more realistic.
Exactly my thinking 
My reading of the datasheet is that the non-linearity has much more to do with magnet placement than temperature so it should be pretty constant. Iām hoping that by running the motor at a constant voltage and constant load we can measure the non-linearity of the encoder, but Iām working on testing that right now so it is to be determined. If the non-linearity canāt easily be compensated for weāll have to rethink some things.
I got some sample motors from company below but have not tested them.
There is also another company that makes stepper worm gear motors that sent a catalog, as well as spindles.
I guess a big benefit of removing the encoder from the motor is that as long as we have a standard bolt pattern pretty much any motor would work. If the motor was WAY different from the original one some PID tuning might need to take place, but for the most part they should be swappable.