Printable View
I am hoping there are still some people on here who have successfully completed a conversion from the Meldas3 controls to a modern PC interface. I read the whole thread on the 4400 which has some good info in it but I am hoping to find some more specific info pertinent to my machine and more up to date than 2008 when that thread was active...
-josh
I know you've been in contact with Duke. I would say that you should follow his efforts closely as he is trying to do the same thing as you. More importantly, he is obviously quite creative and ambitious with his project, so it could be very interesting and inspiring for you. The downside is that the project also involves reworking the spindle, and that is just an added complication that could slow down, stall, or sideline the project too. That is always the risk with retrofits (I can personally attest to that!). It is hard when you have a life/wife/family outside of the shop!
Fundamentally, there is nothing special about the AC servos on an M3 Meldas control. The encoders are also pretty stock except that they have extra tracks to help with the synchronization of the windings. But you don't need to do anything with those tracks, so you can just ignore them if you are using a drive that can take the typical phase quadrature or can use vector feedback to perform the same synchronization function...
But I think your question is a little broader than that.... I am guessing that you are asking if there is anyone that has cracked the bus structure of the Meldas amplifiers (MR-S11, and MR-S12 for the 4400m). The answer to that is "no, not yet, but there are a few of us that are interested in doing so." it shouldn't be that horrible to do. How beneficial is another matter and, frankly, there are other more pressing questions that some of us have. I think there are two here that are especially interested in the M3 control (myself being one) and extending its capabilities. But, to be honest, I'm not sure I want to *replace* mine, which is where I think you are going. Why am I not? Well, because I'm not convinced that it is a lost cause. I mean, let's face it, the thing is pretty capable. I've already gotten mine so that it can handle a 4th axis, which is something that was important to me. And I've got a Renishaw tool setter working now, and ditto for a Renishaw probe. The only real limitation right now is that I don't have the memory. Well, that isn't quite true.... I've managed to double the memory and I'm working on getting full memory. And, in cooperation, we are trying to see if we can come up with a much more economical way to get full memory on all of these using more modern components.
Memory, in my mind, is probably the single biggest issue. The reason for this is primarily macro space. The machine will work fine with DNC, but if you want to use a Renishaw setter, it will take up about 28,000 lines of space (everything is lines based, not KB of RAM). We only have 64,000 to start with, so that is a problem. And the Renishaw touch probe is even worse. It takes about 100,000 lines of space, as I recall, for the full suite. I don't have the exact numbers because the more important point, that you don't have enough space even with the 128,000 lines I have now, is a known fact. So, memory expansion is more important to us right now than figuring out the protocol for the amplifiers.
The biggest issue with the amplifiers is that they are expensive. But, at this point that isn't something that bothers me (personally) all that much. I've managed to fix a few now so I feel better about that. The amps are gigantic, however, and a modern one would do the same thing in literally about 1/20 the space. Yeah, that would be nice, but they are also not repairable. And, the reality is that the amps are stuck in a cabinet... Do I really care that they are big and old? No.... Just change the electrolytic caps and the amps will probably work for another 30 years.
If there is one area of the control that worries me the most, it is probably the spindle drive. I've not dug into that at all, but it is also basically a servo amplifier on steroids. I'm *guessing* that it isn't too difficult to deal with, assuming you give it time.
To me, the biggest frustration I have right now is probably (well, beyond finally cracking the memory thing so I can get 1GB) the tool changer. I have a mixture of BT and CAT. You can use either by adjusting the Z height for the change, but you cannot use both at the same time. I'd love to have a carousel that could be configured with a *mechanical* offset so that you could use both at the same time. But there isn't a whole lot of room in there to do this. And, of course, it is a lower priority.... I haven't even *used* my machine for hobby stuff because I'm too busy mucking with it from a control perspective.
I should also point out that part of my reason for sticking with the control for now is that I've managed to pick up some spare parts. I've had to fix these, but that is OK. It was not only part of the deal but it was also practice and gave me the confidence that I can fix them again in the future. Sure, nothing is guaranteed, but I think the Meldas control actually has some nice features and is probably more nicely integrated than a lot of the home brew stuff we might do. As such, I'm not too eager to replace it just now.... Instead, I'll just hang a DNC off the side of it (make your own, use a PC, or buy one) and I think it will be fine.
I'm getting a DM4500 and am planning on a retrofit as well. I'm having a hard time finding anything about this machine on line. Hoping to see some activity on this thread...
Most people that do "retrofits" rip out the servos and drives and completely replace them (usually with inferior stuff due to cost reasons, making the machine much slower in the process), which in my opinion is a waste, as there's nothing really wrong with the motors or the controls. The problem is the proprietary parallel interface between the M3/M300 control and the servo and spindle drives. When / if I ever get my machine back together, I'd like to work on reverse engineering that parallel interface bus with the goal of being able to replace the M3/M300 controller with something modern, like LinuxCNC and talk to the servo and spindle amps. There already is an interface that works with LinuxCNC for the slightly newer M500 system that does just that, but that control bus is completely different, and serial. I've talked to the author of that work, and he has no interest in expanding it to the older M3/M300 parallel bus (probably mostly because he doesn't have access to any M3/M300 hardware to test with, and he's of the opinion that the M3/M300 hardware is just "too old" to bother with, even though there are thousands of M3/M300 machines still out there).Quote:
Originally Posted by goldenfab
I've talked to and worked with MrMetric extensively about this, and as he said, there has been some effort in how to expand the memory capacity of the M3/M300 with modern parts, but my work on those efforts has been sidelined with some other urgent things that came up recently. Case in point is the expense / stupidity of how to add more memory to the old controls, where Mitsubishi wants to charge hundreds or thousands of dollars for 1MB of sram - literally something you can buy from Digikey for $15. And then you want to add Renishaw spindle probe / toolsetter macros to the thing, and it eats up a substantial amount of the program memory.
There are pros and cons of trying to shoehorn more ram into the existing M3/M300 controller, but it's ultimately a lot more enticing to me to be able to re-use the servos / servo amps and spindle motor / spindle amp and be able to control it with LinuxCNC, where there isn't an upcharge for every single capability (like you get with every commercial controll, or even with Acorn or Mach), or laughably "piddly" amount of program memory you want to add to the machine. It's 2022 - an upper memory limit of 1MB or 2MB is just incomprehensibly stupid these days!
Now, being as this is a hobby project, and I need to get my machine physically back together, don't count of any of this happening in any particular timeframe (if ever). I am about ready to start actually working on the machine again, and hope to have it actually back together and powering on in a few months.
I agree 100% with hwgasdfasdf. Although I understand the allure of a retrofit, I'm not sure I'm really keen on promoting it if you have a working machine. I mean, the M3 is a pretty capable control for most of what we are talking about. And we've been working to crack some of the nuances that are annoying. For instance, we now know how to add that 4th, and possibly even 5th, axis. The big limitation has been memory. We know how to increase the memory from an activation perspective, but the big issue is that the card cages are often too small to support the additional boards. That is where the current work hwgasdfasdf and I are doing comes into play... We've been working hard to reverse engineer the system so that we can add new modern chips. There are two approaches. The preferred one is to replace the MC852 daughter board that is on the CPU board with something that has far more memory. An alternate approach that I've been tinkering with (because the signals are easily accessible) is to make a board that slips onto the side of the control where an expansion connector resides. But that is likely a temporary approach.... I've mapped more of those signals than the MC852 daughter board... Bear in mind that both hwgasdfasdf and I are hobbyists, so we have competing projects and obligations.
What is the biggest reason I personally want more memory? I get this question asked of me because there are so many DNC solutions that work really well now.... Put bluntly, there is only one *real* reason I can say why I want more... macro space. I want this specifically because I don't have the room for the Renishaw macros. This all started because the native control only has 64000 "lines" of g code available (just how they measure it). The TS27R macros take about 40000, if I remember correctly. OK, that will work... but I want to add a touch probe too. Hey, I'm a hobbyist and at geek... I want to play with both! The MP macros are about 90000 lines, I think (some of these number are a bit contrived, but you get the idea). Yes, I might be able to trim the macros down, but I really don't want to do this. So, no space. I've managed to get about 130,000 lines of space now, but I still don't have enough room for all of the stock macros and the extra stuff for the embedded functions that the M3 has (those take about 20,000 lines I think). Hence the desire to add more RAM.
I think we are getting pretty close to success... But let's face it... 90% of the time is spent doing 10% of the work, so we may be a long ways off for all I know. And yes... I have an extra control, boards strewn all over the place, as well as amplifier parts here and there too. My garage is a friggin mess!
I can say that the current control I have in my machine is probably a lot better than the LinuxCNC I put into my old Hurco. That isn't to say that you couldn't make a better LinuxCNC setup, but it *is* a lot of work. To that end, I have no immediate plans to trash the M3. Like hwgasdfasdf said, though, I think one of the next tasks will be to better understand the parallel interface that is use for the control of the servo and spindle drives
I'm still waiting on a rigger to get the machine here. If I'm able to get it going ok you're probably right its best to keep it with the stock controller. I guess it just depends on if I can get it going.
Is this the machine that was on fleaBay a while ago? I think it was in Florida?
I got the machine from a small shop in Casa Grande, Arizona. I'm in central Arizona. I'm not sure where he got it or how long ago. Could have came from FL for all I know. He said he was close to having it running, he said it was all powered up but just missing one thing. Whatever that means. I called the guy at Dyna Metronics to ask him some questions about rigging the machine and he remembered the guy in Casa Grande and said their problem was a a fault with on of the drives.