Doddy
06-07-2017, 09:18 PM
For better, for worse, I've been trying to sort problems with my mill (heavily modded Denford Starmill), lately rebuilding the control cabinet from scratch. For reference, I'm using a cheap Chinese Ethernet motion controller (NVEM 5-axis) and the TB6600-based HY-DIV268N-5A controllers, with 40V/600VA drive supply.
Part of the reason for the rebuild was cosmetic, but I was concious that pre-rebuild I suffered occasional skipped steps; post-rebuild (and the introduction of the HY-DIVs) I was getting substantially more missed steps.
The obvious solutions - replace the PSU and stepper drivers, but that's £300 in anyone's money for 4-axis so I thought it worth having a look at what's actually going on.
Firstly, the NVEM controller manual suggests feeding the stepper driver from a constant +5V and switched pulse/direction lines (-ve going). The steppers were losing about 50% of all pulses and the steppers sounded awful.
Measuring the actual signalling the pulse/direction lines were positive-going pulses vs common ground. Rewired for this and the situation was much improved, but still losing a random number of steps (example was on 4th axis, running 0-90 degrees I'd lose 3-8 degrees, clearly still useless). The improvement, however, got me wondering about the stepper driver's opto-isolated inputs and whether these were struggling with saturation of the photo transistor and/or switching speeds.
Now, scoping the signals from the NVEM controller, the step pulses were 2.25us in length. Tearing one HY-DIV apart and examining the circuit revealed the inputs for step/direction were based on 4N25 opto-isolators. Now, these are not exactly fast devices, with the datasheet quoting an optimal rise/fall times in the order of 2-3us (depending on configuration and manufacturer) and supply voltage across the phototransistor. Throwing a scope on the transistor-side of the opto-isolator showed poor transfer performance (and, accidentally touching the base-connection of the transistor miraculously improved the stepper performance greatly. That was enough to make me think hard about whether my problems were due to the short pulse period from the motion controller.
Now, with a normal parallel BOB you can tweak the pulse width, but not with this cheap motion controller.
So, one idea - if people are losing steps with these drivers, try increasing the pulse-width to 5us to see if this improves your situation.
For me, though, I was starting to think of solutions such as pulse stretcher (hint: Do NOT search for that term on eBay, especially if at work. Sometimes I wonder how innocent and naive I am).
However, another solution presented itself. The TB6600 is capable of stepping with 2us pulses, so all I needed was a faster opto-isolator. Problem is, they're like hens-teeth to find.
Instead, I came across the 6N137 device. This is a logic-level high speed buffer device. It's 8 pin DIL, compared to the 6 pin DIL 4N25, but pin compatible if you sit the device 2-1, 3-2, 4-3, 5-4, 6-5, 7-5 on the pads for the 4N25, and you then only have to source a +supply for the pin 8. Don't worry, a picture later will make that clear.
One other problem exists, that the driver had a 51k resistor from the opto-isolator's transistor's base to 0V. This pin on the 6N137 is the enable input, and needs to be floating, or pulled high for the 6N137 to work properly. Simple solution - remove the resistor.
So, for example,
22107
Here, you can see that I've remove the 2 6-pin Opto Isolators (lower-right) and the two base-pulldown resistors (just above each opto isolator). I've soldered in 2x3 pin SIL headers for each opto-isolator to accept the replacement 6N137s.
22108
Here, I've inserted the 6N137's, bending pins 1&8 away from the PCB (no holes for these). Pin 1 is N/C, pin 8 is +5V. Also, according to the datasheet the 6N137 requires a 100nF capacitor between VCC and ground (pin 5). So...
22109
I've strapped capacitors across the supply to the 6N137 and sourced a +5V supply from the SMD resistor near the TB6600.
Reassembled the driver, and tested. A-Axis was perfect - responding to much quicker acceleration and max speed than previous and faultless in terms of lost steps. Tried one of the remaining drivers - back to the usual poor performance. So, an evening in the shed later and I've upgraded all 4 of the drivers and reassembled the cabinet. Operation is now perfect, as far as I've been able to test.
So, lessons:
1) Buy different stepper drivers.
2) Or, if that's too late, start looking at just WHAT your problem is. Don't discount the pulse width from your controller.
3) If you think you are losing steps and increasing the pulse-width doesn't help, have a look inside the stepper driver, and read the spec on the opto isolators on the step/dir inputs. If these are slower than your preset-able pulse-width then think about replacing them.
I can't promise my solution for my problem fixes anyone else's but the solution above has given me what I think is solid performance for very little money.
Happy to discuss further with anyone interested, this was a very brief description of my problem/solution.
Mike
Part of the reason for the rebuild was cosmetic, but I was concious that pre-rebuild I suffered occasional skipped steps; post-rebuild (and the introduction of the HY-DIVs) I was getting substantially more missed steps.
The obvious solutions - replace the PSU and stepper drivers, but that's £300 in anyone's money for 4-axis so I thought it worth having a look at what's actually going on.
Firstly, the NVEM controller manual suggests feeding the stepper driver from a constant +5V and switched pulse/direction lines (-ve going). The steppers were losing about 50% of all pulses and the steppers sounded awful.
Measuring the actual signalling the pulse/direction lines were positive-going pulses vs common ground. Rewired for this and the situation was much improved, but still losing a random number of steps (example was on 4th axis, running 0-90 degrees I'd lose 3-8 degrees, clearly still useless). The improvement, however, got me wondering about the stepper driver's opto-isolated inputs and whether these were struggling with saturation of the photo transistor and/or switching speeds.
Now, scoping the signals from the NVEM controller, the step pulses were 2.25us in length. Tearing one HY-DIV apart and examining the circuit revealed the inputs for step/direction were based on 4N25 opto-isolators. Now, these are not exactly fast devices, with the datasheet quoting an optimal rise/fall times in the order of 2-3us (depending on configuration and manufacturer) and supply voltage across the phototransistor. Throwing a scope on the transistor-side of the opto-isolator showed poor transfer performance (and, accidentally touching the base-connection of the transistor miraculously improved the stepper performance greatly. That was enough to make me think hard about whether my problems were due to the short pulse period from the motion controller.
Now, with a normal parallel BOB you can tweak the pulse width, but not with this cheap motion controller.
So, one idea - if people are losing steps with these drivers, try increasing the pulse-width to 5us to see if this improves your situation.
For me, though, I was starting to think of solutions such as pulse stretcher (hint: Do NOT search for that term on eBay, especially if at work. Sometimes I wonder how innocent and naive I am).
However, another solution presented itself. The TB6600 is capable of stepping with 2us pulses, so all I needed was a faster opto-isolator. Problem is, they're like hens-teeth to find.
Instead, I came across the 6N137 device. This is a logic-level high speed buffer device. It's 8 pin DIL, compared to the 6 pin DIL 4N25, but pin compatible if you sit the device 2-1, 3-2, 4-3, 5-4, 6-5, 7-5 on the pads for the 4N25, and you then only have to source a +supply for the pin 8. Don't worry, a picture later will make that clear.
One other problem exists, that the driver had a 51k resistor from the opto-isolator's transistor's base to 0V. This pin on the 6N137 is the enable input, and needs to be floating, or pulled high for the 6N137 to work properly. Simple solution - remove the resistor.
So, for example,
22107
Here, you can see that I've remove the 2 6-pin Opto Isolators (lower-right) and the two base-pulldown resistors (just above each opto isolator). I've soldered in 2x3 pin SIL headers for each opto-isolator to accept the replacement 6N137s.
22108
Here, I've inserted the 6N137's, bending pins 1&8 away from the PCB (no holes for these). Pin 1 is N/C, pin 8 is +5V. Also, according to the datasheet the 6N137 requires a 100nF capacitor between VCC and ground (pin 5). So...
22109
I've strapped capacitors across the supply to the 6N137 and sourced a +5V supply from the SMD resistor near the TB6600.
Reassembled the driver, and tested. A-Axis was perfect - responding to much quicker acceleration and max speed than previous and faultless in terms of lost steps. Tried one of the remaining drivers - back to the usual poor performance. So, an evening in the shed later and I've upgraded all 4 of the drivers and reassembled the cabinet. Operation is now perfect, as far as I've been able to test.
So, lessons:
1) Buy different stepper drivers.
2) Or, if that's too late, start looking at just WHAT your problem is. Don't discount the pulse width from your controller.
3) If you think you are losing steps and increasing the pulse-width doesn't help, have a look inside the stepper driver, and read the spec on the opto isolators on the step/dir inputs. If these are slower than your preset-able pulse-width then think about replacing them.
I can't promise my solution for my problem fixes anyone else's but the solution above has given me what I think is solid performance for very little money.
Happy to discuss further with anyone interested, this was a very brief description of my problem/solution.
Mike