[lucas]

The drive has arrived yesterday, I ordered a TB1H from First-Supply to have it fast (they deliver from the UK).
The one I received is a HY-DIV268N-5A wich looks exactly the same.

I couldn't wait to have a look at the inside and removed the cover.
The board is floating inside, it has no screws holding it in place. The chip is fixed to the heatsink and the board is held in place horizontally by the cover screws wich go through the holes but these have some play. See picture.
The chip pins experience all the mechanical stress when the plugs are installed or removed.

Next I removed the chip to have a look a the bottom side of the PCB. There's no thermal paste or pad..... and the mounting holes have some burrs, the chip will not make good contact with the heatsink.
Deburred the holes and remounted with thermal paste for testing.

I use a 33V supply, 4Amp Sanyo Denki bipolar motor and did set the drive at 2.5A and 1/8 microstepping.
The drive works but the stepping rate seems wrong, checked the swicth settings again .... Short story: the microstepping table is wrong, looks like 2 colums need to be swapped.

To be continued...

[hollisin]

Lucas,
What is your goal for purchasing and analyzing this? Just curious.
Jack

[lucas]

CNC is a hobby for me and designing stepper drives also.
I have a 30 years career in electronics repair and know where the weak points are, where they save $$'s using cheap components resulting in premature failures.
Good electrolytic capacitors and trimmers are expensive, my employees and me replaced several thousands of these, no kidding.
We had to give warranty after repair and for some models of computer terminals we simply changed approx 20 components to be sure they wouldn't come back.
Big problem is that a simple electrolytic capacitor fails and sometimes causes 10+ others components to blow at the same time, avalanche effect and thus an expensive repair due to a 2$ cent saving.

Back to the topic:
I have designed several THB6064 drives for me and MassMind.org, a TB6600 drive is coming soon.
For questions like the one below it's best to know exactly what the possible problems are with the cheap Ebay drives.
This is the reason I bought one to analyse and compare.

Would one of the electronics experts on this thread compare the two stepper drivers below and tell me which is better and why. I am in the process of upgrading my CNC controller to discrete components.

Thanks

THB6064AH MassMind.org

CNC THB6064AH Two 2 Axis 64 Micro-step 4A Stepper Motor Driver MACH3 No Heat #A559A on eBay.

PS: I'm now working on the schematic of this drive and it looks like I bought a heatsink, some connectors and 2 optocouplers in a socket.

[hollisin]

CNC is a hobby for me and designing stepper drives also.
I have a 30 years career in electronics repair and know where the weak points are, where they save $$'s using cheap components resulting in premature failures.
Good electrolytic capacitors and trimmers are expensive, my employees and me replaced several thousands of these, no kidding.
We had to give warranty after repair and for some models of computer terminals we simply changed approx 20 components to be sure they wouldn't come back.
Big problem is that a simple electrolytic capacitor fails and sometimes causes 10+ others components to blow at the same time, avalanche effect and thus an expensive repair due to a 2$ cent saving.

Back to the topic:
I have designed several THB6064 drives for me and MassMind.org, a TB6600 drive is coming soon.
For questions like the one below it's best to know exactly what the possible problems are with the cheap Ebay drives.
This is the reason I bought one to analyse and compare.




PS: I'm now working on the schematic of this drive and it looks like I bought a heatsink, some connectors and 2 optocouplers in a socket.

Lucas,
Is it possible in the future you'll be analyzing and/or constructing some DSP-based drivers?
Jack

[bobgrant]

For comparison here's the waveform capture across the sense resistor (0.226R) of the Haoyu driver.

Settings are 1/8 ustep, 30RPM or a step rate of 800Hz, 2A per phase, Vsupply = 24.5V.

In case you are wondering about the square yellow tiles - that's the current ramping up in the motor.

[tholling55]

Do not assume that the schematic posted by Lucas matches your HY-DIV268-5A driver. There are multiple versions out there. Mine for instance has a different capacitor in the TOQ timing circuit. It is a 100 nF instead of the 1 nF in the schematic from Lucas. There are also other differences. The LED current limit resistors for the 6N137 are both labeled 560 ohms on mine, but they are actually 56 ohms. Please just note that the manufacturer has changed some of the design, and in other cases the board does not actually have what the component is labeled to be. The only way to be 100% sure is to measure it. I am a bit concerned that mine has 56 ohms, I would appreciate hearing back from others if they can measure those resistors.

Shorting out pins 3 to 4 can be detrimental for some machines. if you have the correct capacitor timing circuit that matches your machine it should be left alone. Shorting it out will lead to the drive being set at full torque at all times. For users with Linux CNC this can lead to excessive heat build up. The problem is that Linux CNC does not have a default setup for axis by axis enables. That would allow a motor to cool while not moving. In the default setting for Linux CNC all three motors will be pulling current, loading the power supply, and getting hotter until the machine is shut off killing the Amplifier Enable pin. If you have a scope you need to do exactly as Lucas did and verify the waveform across the current limit resistor. Watch for the spike indicating activation of the torque reduction. If this occurs during a step, the capacitor in that timing circuit needs to be increased. It is surface mount, but not that hard to replace. I
Good Luck.
Tommy

[lucas]

Do not assume that the schematic posted by Lucas matches your HY-DIV268-5A driver. There are multiple versions out there. Mine for instance has a different capacitor in the TOQ timing circuit. It is a 100 nF instead of the 1 nF in the schematic from Lucas. There are also other differences. The LED current limit resistors for the 6N137 are both labeled 560 ohms on mine, but they are actually 56 ohms. Please just note that the manufacturer has changed some of the design, and in other cases the board does not actually have what the component is labeled to be. The only way to be 100% sure is to measure it. I am a bit concerned that mine has 56 ohms, I would appreciate hearing back from others if they can measure those resistors.

I rechecked the schematic for those capacitor values and there's one 100 nF missing on the 5V line. There are only 2 1nF capacitors on the unit, both are in the torque reduction circuit, could you indicate wich one has another value on a picture or so?
Regarding the resistors for the 6N137: my mistake, they are 56 iso 560 ohm here also. This is to low, the led current would be 62mA and the datasheet mentions 50mA maximum. But I think that the optocouplers will survive this, a value of 270 to 680 would be best.

@ludovanginderen: I agree with most of what you are saying, some components will be hot so use a fan to cool the top of the PCB to extend their lifetime.
I only have one wich I bought to check it out, I won't use it :-)

Guys,
I worked out the problem with my TB6600 only driving at 30% current. The Clock Pulse is too short from Mach3 (1-5ms).

Solution is change the value of one of the caps to 1uF.

I'm afraid that this won't work for long. The transistor just below that capacitor has to discharge the capacitor wich is charged to 5V. There are no current limiting resistors or other features to reduce the discharge current. When the transistor is active it will short that 1µF capacitor and create a very high peak current, value could be something like 10A or more depending on the type of capacitor used. The transistor has a max current specification of 200mA, it will also limit the current but there are no specifications in the datasheet for this situation.
How long did you test this change?
Datasheet:
http://www.fairchildsemi.com/ds/2N/2N3904.pdf

[Arquibaldo]

Just a quick observation about post #63: as we all know, a resistor labelled 560 IS a 56 ohm resistor:
The final digit is the number of zeros after the first two digits; 0 means no zeros, so the value is 56 ohms...
A 560 ohm resistor would be labelled 561...

By the way, I also ordered one of these accursed drives on dec/2, before finding this thread... my bad...
I hope the info provided here by all the brilliant and dedicated people posting on this thread will
enable me to fix some of the design errors in this drive and recover a part of my (modest) investment...

Thanks a lot to you all, particularly the OP: Lucas.

Nelson

[tt-boy]

I cut a track and inserted a 100 ohm current limiting resistor before the 1uf cap and it seems to be OK. On that note I haven't got around to adding this resistor to my other machine and the modules are still working (haven't blown up yet)

[lucas]

Didn't have much time lately but didn't forget this thread.

I cut a track and inserted a 100 ohm current limiting resistor before the 1uf cap and it seems to be OK. On that note I haven't got around to adding this resistor to my other machine and the modules are still working (haven't blown up yet)

How long that it will work depends on your machine settings. The transistor has some work to do only at low speed.
If your start speed and acceleration is high enough then it has to work for a few pulses at the beginning of a move, after these initial pulses the transistor has an easy job doing almost nothing. If you want a "will always work" solution: test it with the machine running at 1-5 step pulses per second.

I did some measurements and it takes approx. 70msec for a stepper motor to stabilise after a single step.
The step pulse width is only 1µsec for some 3D printer firmware.
The entire torque circuit must respond within 1µsec and remain active for 70msec (preferable longer for safety reasons).
This is quite difficult to get right with simple circuits, even a retriggerable one shot has difficulties to operate within these margins.

I assume that if I want a kit (the one you designed) I go to MassMind and order it there. One question maybe you can help with, Can a kit be purchased (controllers and BOB) for 4 Axis, as Im thinking of driving the y axis with 2 motors (it runs the gantry, will be using linear rail system and ACME lead screws direct drive via motor). The machine will be about 24" x 24" max and used for PCB milling only, so I think smaller motors may work for me (once I figure that part out).

24" x 24" for PCB's only? That are large pcb's :-)
I would suggest to visit the "PCB milling" forum here, you will probably find very usefull info there.
Your setup is no problem, you can slave the 2 Y axis in software (Mach3 etc..) or connect the 2 drives to the same BOB output.
There are some differences between my BOB-drive setup compared with the Massmind one. Take a look and decide wich one suits you best.

Is there any information around when the Dir-Signal should be stable prior the positive edge of the clock (Step-Pulse) and how long it should stay stable (in relation to the positive or negative edge of the clock (Step-Pulse)?

I had a quick look at the latest datasheet and can't find it neither.
There are more spec's missing, like the max. output current that can be drawn from the internal 5V regulator.
On top of this they have changed some other parameters wich required some component changes in my design.

I have a design ready and it's in use on a Mach3 mill and 2 3D printers, all 3 are working fine. It will be released when I find some time to compile the documentation and update the website.

[lucas]

There's a CD included with the drive, it has Mach3, some files for other products and a users manual for the drive.
Great: a Chinese Ebay product with documentation, maybe something positive?
But just noticed something today, see attached pictures and find the difference. :drowning:

They are different and on top of that: neither of them is correct.

[lucas]

No, not at all.

The problem is in the design: wrong component values wich cause the chip to remain in idle current all the time wich is that 30%.
I think they never used a scope to verify their design, the concept isn't wrong but some component values are waaaaay off what they should be.
Also the designer didn't read the datasheet, there are at least 5 totally unnecessary components.

I did complain about the wrong documentation regarding the microstep settings, the manual and the drive's cover are different.
They were happy that I pointed that out and issued a partial refund but they are still on offer on Ebay with the same error.
It was already really hard to convince them on this obvious error and thus I didn't pursue the other problems like lack of torque.

They just don't care about the quality and performance: selling is important and if somebody complains, it takes 10 or more mails back and forth and if the customer doesn't give up then they just propose a partial refund: Issue closed and they continue selling the same crap.

The very same thing happened with the TB6560 drive's: Complaints, forum posts on the reliability issue, lack of and even wrong documentation in a lot of cases etc.. have no effect at all. ( see the posts on the forum here for TB6560 ).
Did they do something about this? No, not at all. The same stuff is still for sale on Ebay and other sources.
$$'s is the only thing that's matters.

[ludovanginderen]

Beste Lucas,

ik heb mij eveneens laten vangen door de goedkope driver,er wordt gezegd 50V 5A ,maar dit blijkt wel een illusie te zijn gezien de datasheet maar 42V en max 4,5A geeft.
De driver blijft op 30% vermogen staan doordat die TQ op 0V blijft ,ze hebben gedacht onderdelen uit te sparen en dit met enkele transistors op te lossen.
Verder zijn de senseweerstanden veel te licht en de diodes over de low side van de H-brug worden ook gloeiend heet.
Ik heb de transistor die de TQ input laag houd losgesoldeerd en de TQ ingang met 4K7 aan de +5V gelegd,nu heb ik wel de volle 100% ,de diodes heb ik ook losgesoldeerd deze hebben een negatieve invloed en zijn in de datasheet ook niet gebruikt.
Zoals je al hebt aangegeven de staptabel blijkt ook niet kloppen.
Verder blijkt het wel te werken maar het is toch niet zoals een duurdere driver.
Ik ga hem toch maar gebruiken met max ca 35V en max 3A ,ook nog wat warmte pasta aanbrengen en dan maar hopen dat het blijft werken.
Heb jij ze nog in dienst of liggen ze al in de vuilbak of heb je nog verbeteringen aangebracht.
groeten,

Ludo Van Ginderen

[lucas]

I finalised the schematic and there are several things wich could be improved but these are not important.
Just one wich confirms the error in the microstepping table: look how the switches 5,4 and 6 are out of sequence with the M1,M2 and M3 inputs. They simply copied the microstepping table from the datasheet wich doesn't match the PCB layout and schematic.

The real error is in the circuitry for idle current reduction, the components wich define the delay are a 2K resistor and 1nF capacitor.
Their time constant is 2 µsec, this means that the drive will go into standby mode approx. 2µsec after receiving a step pulse.

Time to measure the effect of that.
Attached are 2 waveforms measured on the current sense resistors, one is from the Chinese drive the other from my TB6600 drive.
Drives are set for 1/8 usteps, 4 Amp and are driven at 30RPM.
Vertical scale is 0.2V/div. , 1 Amp on a 0.2 ohm resistor = 0.2V so 1 division =1 Amp.
Only the postive part is relevant, the negative part of the waves are a result of the decay mode.

The difference is clear, The wave form should reach 4 divisions for 4 Amp but the chinese waveform goes to 1.2 - 1.3 divisions with some hairy spikes.
These spikes are the 2µsec ones where the drive should reach maximum current after each step pulse.
What we are seeing here is a drive that basically runs in standby mode or 30% of the torque.

I'm thinking of an easy way to compare the torque, a weight attached to a lever seems the easiest.

[doorknob]

What is the horizontal scale timing set to?

A question about the waveform that I'm looking at - I think that you're differentially measuring instantaneous voltage across a series current-sense resistor to get the current through a single winding - on the screen shot for your driver it appears to me that we may be looking at multiple sweeps (perhaps four?) with a bit of triggering jitter, and not one sweep (whereas the shot of the Chinese driver looks like a double sweep). Otherwise I'm not sure that I understand the vertical overlap with offset of adjacent 'square' portions of the trace on the shot from your driver.

>>Only the positive part is relevant, the negative part of the waves are a result of the decay mode.

I'm a bit confused at that, because I would have thought that the negative part of the waveforms would have represented the bipolar nature of the driving. But then I have not previously looked at waveforms from a chopping bipolar microstepping drive. I'd appreciate any clarification of that point.

My assumption is that the width of each 'square' part of the trace is the duration between one step pulse arriving and the next as it holds a certain position and current level until it is advanced to the next microstep, and there is chopping (and decay) going on at each current level as it awaits the next step pulse which is represented by the height of eqch 'square' of the trace, and that if we were also simultaneously looking at the other winding's current we would see it out of phase with the first trace in order to handle the microstepping.

[lucas]

Horizontal is at 5msec/div.

I'll try to explain, English is not my maternal language and choice of words can cause misinterpretation.

This is the waveform at the current sense resistor for a single phase, not a series resistor.

on the screen shot for your driver it appears to me that we may be looking at multiple sweeps (perhaps four?) with a bit of triggering jitter, and not one sweep (whereas the shot of the Chinese driver looks like a double sweep). Otherwise I'm not sure that I understand the vertical overlap with offset of adjacent 'square' portions of the trace on the shot from your driver.

There are indeed multiple sweeps overlapping each other, that's due to the automatic shutter timing of the camera in combination with the poor triggering of the scope, I also adjusted the intensity on the scope and that must have changed the shutter timing. They look better in real maybe a movie iso of picture would have been better.

>>Only the positive part is relevant, the negative part of the waves are a result of the decay mode.

I'm a bit confused at that, because I would have thought that the negative part of the waveforms would have represented the bipolar nature of the driving. But then I have not previously looked at waveforms from a chopping bipolar microstepping drive. I'd appreciate any clarification of that point.

The waveform is taken from the current sense resistors, if you have a detailed look at the current paths in the datasheet then the forward and reverse current through the motor both produce a positve half sinewave over these resistors and the decay currents generate the negative one.

Edit: you can also see some spots at zero current.
The TB6600 uses mixed decay, fast creates a negative current through the resistor and there's no current for slow decay.

My assumption is that the width of each 'square' part of the trace is the duration between one step pulse arriving and the next as it holds a certain position and current level until it is advanced to the next microstep, and there is chopping (and decay) going on at each current level as it awaits the next step pulse which is represented by the height of eqch 'square' of the trace, and that if we were also simultaneously looking at the other winding's current we would see it out of phase with the first trace in order to handle the microstepping.

That's entirely correct, if both windings would be visible there would be a sine/cosine waveform with a 90° phase shift.

The important thing here is that the chinese drive never reaches full current, you can see the "hairy" pulses after each step trying to get to full current.
But they don't due to the inductance of the coils.

[doorknob]

Thanks for the clarification, that was most helpful.

[lucas]

Almost the same, differences are in the control signals and some features:
- different microstep possibilities, THB6064 has more.
- TB6600 lacks settings for decay modes.
- TB6600 has an internal 5V regulator but it's maximum current capability is not clear.
- TB6600 lacks automatic idle current and this is where the Chinese designs fail.

[lucas]

That's a nice shot, what equipment are you using?
800Hz is just on the limit when the TQ input could become low on that drive.(if my calculations are correct)
Could you measure it at 400Hz or lower?

[silcnc]

I was thinking of buying some of these drivers (hy-div268n). Do you think they will work well or should I steer clear?