tuxcnc wrote:

timo wrote:

There are many reasons why you should not use MPG for turning, but I will give you just one - you can turn the handle faster than the axis is moving, in other words, when you are done turning, the axis will continue to move until it reaches the commanded position.
 

Thats not correct. The user can choose how he wants his MPG to operate when he sets it up. Now I am the old crank!

I don't know why this conversation got down to swap drives and building LinuxCNC from source. None of that is necessar. LinuxCNC only requires 4-8 gb RAM but I think I'd lean towards 8 gb today.
Just follow Andy's directions, accept the default partitions and have fun. After all he is the Release Manager so should know a thing or two about how to install his releases!

@tuxcnc: Are you a nay sayer? image missing /media/kunena/emoticons/cool.png 

Being pointed to weak aspects of something is not a bad thing. Many of my ideas where stupid and one work colleague used to call me out on it. And every time I learned something, sometimes he did as well (i think). 

"You can turn the handle faster than the axis is moving" I think that is plain wrong. I would like to see you hand crank any commercial lathe at rapid speed; plus they stop as soon as you let go of the wheel.
To make it "feel like a direct coupling" is "worth doing" regardless, if you only use it for positioning. Admittedly a stupid idea being happy with untuned overshooting hand wheel. (I am guilty of it as well, and it did end up in tears once or twice)
I prefer not to crash the machine into the chuck, or hitting E-stop buttons when I was hand cranking too fast, so I suggest to tune that. 

"People" can crash an engine lathe using the hand wheel quite reliable. (ask any trade school teacher )
If that makes a lot of sense is another question, I agree to "doesn't mean it's worth doing".

Examples:
Fine adjust for the z-axis of the surface grinder works a treat with an encoder wheel and stepper motor, because the motor does not care and over powers the stick slip without overshooting. (I klick once, it moves 0.005 mm) 
If turn an M2.5 screw on my engine lathe there is not much "feedback signal" in the sense of force feedback as well.
All sorts of machinery is using encoders to cause movement. 

Pretty much I agree, that some things might not be worth the effort.
If something is worth doing for someone else is not for me to decide.

Sure you can delete your swap space and most stuff will just work. Some apps won't. SSD space is so much cheaper than 10yr. ago, do your pc a favor and allow at least 1GB swap space if you have to cut it short.

Long ago I read swap kills SSDs so I carefully removed swap from all my linux os. After a long time I learned some of my mysterious linux troubles were swap related (because I used no swap space). It seems most stuff just works without swap, but it hurts stability a little, even if you have more RAM than you'll ever use. Also, nowadays the OS and SSDs are designed in ways so swap is not so hard on SSDs.

Nowadays I  make sure to provide swap space bigger than my RAM. Like 64GB RAM and 90GB swap. This allows hibernation to work. Then I completely disable hibernation because I don't like it. go figure. image missing /media/kunena/emoticons/wink.png

tuxcnc post=348156

If you really need swap space, you can use a file instead of a separate partition. 
 

That works, also. I recall seeing that automatically created on one of my Linux boxes a while back upon install. I don't remember which distro it was. Might have been Ubuntu.

Need a bit of help from PCW here.

I need to read two Yaskawa absolute-encoder serial position outputs into LinuxCNC — RS-422, async 9600 baud, 7 data / even parity / 1 stop, continuous (~17-byte frame every 40 ms), one per drive. So two independent, always-on UART RX lines.
  
Stack is a 7I97T with a 7I85S on P1 (running the 7i97t_7i85sd bitfile). The snag: every encoder input is a MuxedQCount (2:1 muxed), which can't carry a continuous async stream, and the only free non-muxed RS-422 receiver I can find is the unused sserial RXData1 (7I85 TB1-19/20) — RXData0 (7I97 TB4) already drives a 7I84.
  
  Can a custom bitfile place two PktUART RX on RS-422 receiver pins on this hardware? If only that one sserial RX is usable, what's the right way to get the second stream — a different daughtercard, or decode it off-board?

Thank you
EDIT: Solution plus tips:

I use Yaskawa drives. With my hardware, I ended up with a strange situation where I was one UART receiver short (I didnt want to use the Pi UART, or add any more hardware). The reading I need from Yaskawa is the absolute encoder position (needs a battery to survive power off, and an additional signal to report the servo absolute position back to mesa). Since a swivel head has 2 axis, I am one UART receiver short. However, I only need the absolute position as a single datapoint after which, the incremental position record takes over. My solution was to wire the UART via a 4pole relay so that when energized, i get UART_AAxis, ad when not energized, i get UART_CAxis. I simply reused a harmless relay. So in my case, when the chip blower is on, I read C and when its off, I read A. So with that, my 7i85 is now able to read 2 signals on 1 single UART.

I had to redo the bit file. 

The rest are standard establishing of signals.

tommylight wrote:

After you are done counting the CPU's, see lower on the right side, 16GB of swap space on a PC with 96GB of RAM memory. On laptops with 16GB of RAM i do 32GB of swap always.
Although you are correct that usually it is not needed, if you do need it and do not have it the PC will slow to a crawl.

It's not like that...
The more memory, the more Linux reserves for disk buffers, which does not mean that it needs that much, or that it actually uses that much.
With 8 GB RAM, the Linux installer creates a 1 GB swap partition, which won't help anyway.
If you really need swap space, you can use a file instead of a separate partition.
The problem is basically that modern Linux stores the existence of the swap partition in multiple places, not just in the /etc/fstab file as it used to be, which results in the system failing to boot when the swap partition is deleted...

thanx.

John had some in stock at mesaus.com

what have you guys used to interface from P1 and P2 to the switches, etc?
with the new quick turn PCB houses, I wonder if someone has already made a plug in PCB adaptor to put the same screw terminals on it?

more update. i haven't posted here must because the discord forum usually helps me with understanding something claude is making up and I'd say I've gotten to the point where the machine is commissioned. half of the problems was just about electrical connections so, note to future self, don't use splices. just wait a day for a new wire to show up and spend that extra 20 bucks. 

so a 5axis retrofit is very doable, but:
(a) pay close attention to the physical properties of the machine. It's very use dependent. If you are carving foam, a direct drive or belt reduction is probably ok, and this actually makes a DIY solution for the physical build quite tenable. I think wood would be fine too if you are carving. For joinery, the precision will most certainly call for a harmonic drive (as in my use case). this means (A) weight, and ( the weight is about 300mm away from the Z axis guide. There is really no away around this. This then means that your gantry will either need to be fixed, or the there is some counterweight, or, "built like they used to be" (as in my case). I also think Z axis clearance is crucial. Have at least 18in.

That's it really. The rest is kinda sorta a load of wiring because the extra axis adds a lot of wires. Happy to take questions if there are any, but I think I'm safely migrating towards calibration, gui setup, and hopefully, making parts and prospering soon after. so glad I chose this over the other controllers

Yes, they are at assembly currently.

7i73 boards are not in stock at Mesa, so i'm just checkin if its still a viable board and will be back in stock at some point.
dont wanna spend too much time designing it in if not.
 

Yup! It may take some fiddling to get the speed you asked for to match the speed you get, but it should at least "just work"

gotcha, so I would just adjust it in LinuxCNC?

The 7i96S board has an analog spindle out which should work for that purpose - it's implemented as an isolated digital potentiometer.

Different question: if I want to control my spindle's speed, I do this through software, right? My Sherline mill's speed control uses a 0-7V range to control its speed using a potentiometer.