How I fixed my Chinese TB6560 controller (updated)
I purchased a 4-axis controller base on the Toshiba TB6560 Driver IC from a Hong Kong vendor thru e.b.a.y. Although the board could be made to function by setting all parameters to low / relaxed / lame, I decided to find out why the controller greatly underperformed as compared to what the TB6560 IC specification implied.
Performance issues of MY board:
- required maximum (slow) timing settings in EMC2 to work without missing steps
- could be made to work at 12v but not 24v without missing steps
- would work with drive current set to minimum (25%) otherwise motors would miss steps
- at working controller settings, rapids weren't rapid...
All of these factors suggested a noise problem in the circuit (faster timings = less noise margin, higher drive voltage = more switching noise, higher drive current = more switching noise, rapid rapids = more switching noise).
Disclaimer: The information here pertains to MY TB6560 board which may not be the same as YOUR TB6560 board. Use the information here in at your own risk. I accept no responsibility as to the application of this information to YOUR board. However, the information is believed to be correct and does seem to have completely fixed MY board.
A schematic did not ship with my board. Another e.b.a.y. customer had requested the schematic from the my vendor, but the vendor response was "we dont do that". So I developed my own schematic based on inspecting and measuring MY board. See attached.
Seven issues were identified with the design on MY board:
1) Driver IC CLOCK input improperly driven: The TB6560 uses the rising edge of the CLOCK input (not explicitly called out in IC spec except for “rising edge” symbol shown in one table). The slow turn off time of the opto-coupler combined with the passive pull-up resistor are insufficient and allow system noise to cause unstable driver circuit operation. ( I measured a rise time with and oscilliscope of app 50uSec on my board.) Solution: Actively drive the CLOCK input using added 74HC14 IC as buffer.
2) Driver IC ENABLE input improperly driven: The opto-coupler for the ENABLE input is wired incorrectly having the npn emitter connected to supply. Although the circuit does function, the improper wiring of the npn will cause very low transistor gain opening the possibility of system noise corrupting the ENABLE signal. Solution: Rewire the npn of the opto-coupler with collector to supply and emitter to output.
3) Drive Current Manipulated by CLOCK/STEP signal: It appears that an attempt was made to implement a feature where the driver would reduce current to the stepper motors when the motors were idle. This was accomplished by pulling the drive current set input low when the CLOCK signal was active high. And although the feature does work for drive current settings of 25% and 50%, the implementation has negative side effects that outweigh the feature usefulness:
- the feature departs from the reference design of the TB6560 spec resulting in the spec no longer accurately describing operation
- motors may be overdriven with higher than programmed drive current
- artificially raises the lowest current motor that the controller board can safely drive
Solution: Disconnect the non-spec circuit.
4) Activity LED improper function: The activity LED circuit works properly when the STEP signal from the PC has high duty cycle (active low orientation). However, the TB6560 spec directly states that proper operation requires low CLOCK (e.g., STEP) signal duty cycle (or active high orientation). With active high orientation signal the activity LED remains lit even when the driver is not being ‘stepped’. Solution: None. Ignore the LEDs.
5) Potential for Overheating of 12V Voltage Regulator: The cooling fan is fed directly from the 12V Vreg output adding significantly to the loading of the device. The 12V Vreg will shut down if it overheats protecting itself; such a shutdown would result in lost steps of an active motor. Solution: Power the cooling fan with an external supply or (with suitable step down circuit) from the stepper motor supply voltage directly. With the fan connected to the 12v Vreg the heatsink temperature was 155 F, disconnecting the fan and waiting 5 minutes, the 12V Vreg heatsink temp dropped to 125 F.
6) System Not Optically Isolated: Despite having opto-couplers on all PC inputs and output, the controller does not properly implement optical isolation of the PC from the driver IC’s nor from the stepper motor power supply. This is due to sharing of ground among all circuits. Solution: None. However, do assure PC and stepper motor power supply externally share the same earth ground.
7) Operating Voltage Mis-information: The board is marked for input voltage “12v to 36v” but the IC spec indicates maximum operating voltage of 34V. But even that is probably not a safe operating point due to potential stepper motor generated voltage spikes. The controller board output diodes will clamp large overvoltage spikes when the diode breakdown voltage is reached, but the breakdown voltage is likely above the maximum voltage the driver IC can sustain without damage. Solution: Recommend limiting input voltage to 30v or less to provide margin of safety for operation.
Update: Immediately after posting the fixes, it occurred to me that it would be better to buffer the CLOCK/STEP signals with 1 74HC14 inverter instead of 2 :
- allows the 'activity LEDs' to work (using 1 inverter and setting software to invert the clock)
- allows 1 IC to 'fix' a 4 or 5 axis board instead of needing 2 IC's
- simpler to wire
See attachment slide10ver2.jpg for updated fix. I tested this morning and works fine.
Similar Threads:
Last edited by lasik2025; 08-28-2010 at 01:42 PM. Reason: better (simpler, more complete) fix described
Bravo on making the effort to examine in detail how the driver board is actually wired, and what design decisions (and possible mistakes) were made by some anonymous board designer on the other side of the world.
I'm considering getting a board that looks suspiciously like the one that is shown in your photos, and so I'll have to take a closer look at the TB6560 spec sheet and reference design and compare it with your reverse-engineered info. Thanks...
Very well thought out!
I do not own one of the boards you mention but you have done everyone that owns one a great service.
great!. Many thanks for share. I have same issues on my 3 axis board, with 1 motor seems to work fine, but when try with 2 o 3 motors the steppers do not work fine (some step stalls on all motors). I replace some parts (tb6560, capacitors, added pullup without good results). I am going to try your fix.
What do you think about the capacitor used for clock? I think that could be lower value for decrease the needled pulse time.
(Sorry my Spanglish
villamany,Originally Posted by villamany
great!. Many thanks for share. I have same issues on my 3 axis board, with 1 motor seems to work fine, but when try with 2 o 3 motors the steppers do not work fine (some step stalls on all motors). I replace some parts (tb6560, capacitors, added pullup without good results). I am going to try your fix
What do you think about the capacitor used for clock? I think that could be lower value for decrease the needled pulse time.
(Sorry my Spanglish
The capacitor on OSC does set the step pulse time as you indicate. I was not able to determine the value of that capicitor on MY board and I could not get my half-working oscilliscope to trigger properly so that I could measure the frequency on the OSC line. The silk screen on MY board says "102" which usually stands for "1000" and since the TB6560 IC spec says acceptable values are 100pF to 1000pF one would think the cap is 1000pF. But there are several parts on the board that whose values do not match the silk screen...
With that said, I am unsure if changing the OSC cap will help or not. On MY board the rise time on the step/CLOCK was simply too long. I did try changing
the step/CLOCK pull-up resistor from 4.7k to 1.0k which changed the rise time from 50uS to app. 20uS as measured on my o-scope; this change seemed to reduce the missed steps by only a small amount; only by adding the buffer with its very fast rise time was I able to eliminate missed steps (the 74HC14 biffer IC spec says rise time is less than 0.020uS or 1000 times faster than without the buffer, e.g. 20nS)
At what voltage are you trying to run your controller ? Without the buffer, I could get 12v to work but never found any settings that allowed 24v to run properly.
at 13.8v i get some step stall, at 24v are very more step stall.
The stall are increase with:
- more steppers connected.
- with more torque settings.
- with more voltage.
- with less time decay settings.
- with low microstep settings.
Seems to be also noise problems.
What are the added resistors that i can see on "cntr_bot.jpg" picture?.
I only need 4x10k resistors, 1x74hc14 ic and 1x decouplig capacitor for these mods is ok?.
I has my controller working for about 5hours without problems on 12v regulator (very hot for my hand but not thermal shutdown ocurred).
A buffer board schematic for your consideration
villamany,
The 4 leaded resistors that can be seen on "cnr_bot.jpg" were my attempt to fix the controller by changing the STEP/CLOCK pull-up from 4.7k to 1.0k. I did this by leaving the 4.7k in place and adding 1k resistors in parallel between the +5v supply and the Step signal ; the resulting effective pull-up was 4.7k || 1.0k = 0.8k. On my board, the chip resistors and capacitors are glued and soldered down making removal of those parts a little difficult and I did not want to risk damaging any printed circuit board traces by applying enough heat to cause the glue to release.
For the mod's, you may want 6 resistors as you will not want to leave the unused 74HC14 inputs unconnected; you can tie them to ground or +5v or, as I would suggest, to pull-up resistors. That way if you come across another signal that needs buffering the you will have spare buffers in place.
To hold my (ver 1) buffer board to the controller I used double sided foam tape (see "bufbrdmntd.jpg - buffer board mounted").
I have attached a schematic for a ver 2 buffer board for your consideration.
many thanks, finally i am going to use 3 buffers for the 3 step pins and the another 3 buffers for apply the same fix to dir pins (my controller only has 3 axis
If i can measure the real osc capacitors value i let you know.
I was considering buying one of these boards, guess I wont bother.
John
Hi folks,
I'm very new to CNC and have a long way to go no doubt. I have also bought the TB6560 board and have not hooked it up yet and I'm glad I ran across your board fix post first before I do hook it up. I have a question. Should a breakout-board be used with this TB6560 board ? ? ?
brutusgraphics,
MY TB6560 board, as shown in the previous pictures, includes the circuitry that a breakout-board would provide; so the answer is "no you dont need a [separate] breakout-board" if YOUR TB6560 board is the same as mine.
However, there is one caveat. The optical isolation provided by MY TB6560 board was poorly implemented and does not provide all the protection a properly implemented breadout-board could. But as long as you plug in your PC to the same wall outlet (or the same power strip) as the powersupply for your CNC machine then you should be ok. Trying to use a separate breakout board with MY TB6560 would be extremely difficult.
Hi lasik2025,
I was busily reversing engineering a 3-axis controller when another user 'doorknob' pointed out your existence over on this forum page.
Here's some more flaws for the list.
- There is a FR301 diode in the motor voltage supply to each TB6560. When back EMF hits, that current can only dump into the 3300uF decoupling cap. The volts increase and folks on this forum start practicing their French. >> replace the 3 culprit diodes with fuses.
- No back EMF diode on the coil of the relay >> add diode
- Pins 25 of the TB6560 extends over the trace from pin 23 >> trim back
- Heatsink not grounded. The resistance from heatsink to ground is small but non-zero. When failure occurs substrate currents will help the remaining TB6560s to join the party >> according to the datasheet isolate each device from the heatsink or simply ground the heatsink
On the OSC pin issue, I observed a nice sawtooth waveform, ranging from 0.5V to 2.5V, with a frequency spread from 38kHz to just slightly over 40kHz.
The max step rate for the TB6560 is 1/4 of this.
Last edited by circuitz; 09-06-2010 at 02:22 AM.
missing steps and low speed
Hi Everyone.
I have the same issue with 3 axis TB6560 board missing steps and low speed
what ever i try on mach3 to eliminate the problem wont work.
the best working almost no missing steps is reduce the current to 25% and micro-steps to half but then the speed is not there, Increasing the speed result to stall the steppers.
If i preform the buffer will it increase the speed too
specs are Nema 34 890oz-in with 24 volt power supply on mach3. settings are 15000 per steps, 20 velocity /mm, acceleration 1 mm/sec, step puls 4, dir puls 4.
Regards,
Aaron
A word of thanks
Hi there.
This is my second visit here.
I am too trying to make A CNC.
Have bought the motors and the preferred TB6560.
I got these info from www.allaboutcircuit.com members which I am very grateful to.
I have the same user name their and have also started a thread as My mini engraver which is quite old. But has all the info updated. I am a regular member there.
One of them suggested this site and ever since I have been checking for what I need instead of asking.
Better to do my part in searching then ask others to do it.right?
I have yet to buy the parts such as screws and railings.
Since I have the motor and a LPT driver board, I bought the TB6560.
But had doubts in making a circuit considering all the design features of this chip.
And today the OP gave me Just what I needed. A schema for the chip I have and the details on how to go abt it.
I just would like to thank the OP for his hard work on reverse engineering the driver bards. I got what I want without buying any other.
Hope you don't mind me using the schema's and like to ask you if I can post them their if needed for reference.
I will be making the changes that suites me. Cause I will be using a LPT ready made one I got with the motors. I will build the Driver part only.
Thanks again.
Thank you for the members who makes this site for what it is.
keep up the good work.
Rifaa
circuitz,
Thanks for the additional information & investigation.
I mistakenly assumed the SOT23 device on the backside behind the relay was a relay coil back emf protection diode; but it is instead an PNP transistor driver. I have updated the slides (atttached to this post).
On MY TB6560 board (4 axis version), the pin23 to pin25 short does not exist. Everyone should check THIER boards for this issue however.
I did not see the datasheet reference to grounding the heatsink. The pdf I found on the web somewhere has header "TOSHIBA TB6560AHQ/AFG" and footer "2009-07-10".
Thanks for the OSC pin info on observed frequency. May I ask what the capacitance of the probe you used to measure it ?
Lasik2025
Aaron,Hi Everyone.
I have the same issue with 3 axis TB6560 board missing steps and low speed
what ever i try on mach3 to eliminate the problem wont work.
the best working almost no missing steps is reduce the current to 25% and micro-steps to half but then the speed is not there, Increasing the speed result to stall the steppers.
If i preform the buffer will it increase the speed too
specs are Nema 34 890oz-in with 24 volt power supply on mach3. settings are 15000 per steps, 20 velocity /mm, acceleration 1 mm/sec, step puls 4, dir puls 4.
Regards,
Aaron
For me, buffering the CLOCK (eg STEP) line for each driver IC fixed the missing step issues. But before I committed to change the entire board, I did try out the buffer fix on a single driver. For a second trial, I added connections to a 2nd driver. See attached picture (the picture shows the buffer temporarily wired to 2 drivers). I just used breadboard of buffer circuit, then connected wires to +5v, gnd, TB6560 pin 3 CLOCK input, the corresponding optocoupler output for CLOCK (by following the printed circuit board trace from pin3 backward), then finally cut the trace with small knife. I figured that if the buffer did not work, I could just repair the cut trace with a small wire to return the board to its initial state.
Lasik2025
Good call, forgot about the cable loading and so re-measured.
I used an oscilloscope probe. On 24VDC, the x1 setting on the probe showed 45.62kHz, the x10 setting showed 40.38kHz. Of particular note the readings varied by several hundred hertz from second to second despite being careful with the probe grounding. There was no motors attached. The cap silk screen is marked 102 aka 1nF.
The heatsink reference occurs on page 28 or page 26 depending of the date of issue of the datasheet.
"Treatment of Heat-Radiating Fin
The heat-radiating fin pins of the TB6560AHQ/AFG (backside) are electrically connected to the backside of
the die. Thus, if a current flows to the fin, the IC may malfunction. If there is any possibility of a voltage
being generated between grounds and the fin, the fin pins should either be connected to ground or
insulated."
Also attached is an image of the pin 23 to pin 25 overlap on the 3-axis driver before trimming back with side cutters.
Last edited by circuitz; 09-06-2010 at 03:20 AM.
Hi,
I have a 3 axis version of the same driver board and after going through some nightmares I was about to order a more expensive board but I will try your workaround before.
While I am waiting for some 74HC14 I figured i would start cutting traces and getting ready.
See attached:
A - I removed the 10k resistor coming from pin 3 of TB6560
B - I cut the trace and will use both sides to connect 74HC14/resistor connections
C - Coming from pin 4 of TB6560, my board appears to reflect your changes already so I just wanted to make sure you did not meant to change this around.
This is how my board is configured (too bad there is no version number):
From pin 4 of TB6560 going through 1k resistor then going to pin 3 (emitter) of PC817 and pin 4 (collector) of PC817 is going to +5V
Thanks
Mike
lilimike,
Your A, B look good.
And you are correct on C that the board does not have the error I described; actually my board does not the error I described. And so I must [embarrased] admit my reporting incorrect information. Find updated slides attached.
lasik2025
So what is the verdict on this board. If we purchase one and do the mods listed in this thread will it work as advertised? It is still a good deal?
Posting Permissions
Reply with Quote
