TB6564AHQ (THB6064AH) PCB design

[lucas]

No, you haven't missed anything. A PMinMo+ interface doesn't exist yet. But hopefully we can encourage its creation...

It would be very nice to have the possibility to create an estop or tell Mach there's a problem.
But this needs new BOB designs, I could include it and adapt my BOB design to support it if there's a "market".
Why don't you start a poll in an appropriate thread?

Absolutely multiturn, I figured something like this one.

Yes similar to this one, but there are many versions, some will increase the current if turned CCW........

Doesn't that also open the PCB up for use with the TB6560 (I haven't checked every pinout) and using a sequencer within the PSU?

The TB6560 has a different pinout and other functions like current setting are completely different. Why would you use the TB6560, do you have some stock?

My understanding, with some drive chips, it's recommended to minimise noise in the ground plane of one drive having any effect on others. Your comments on wiring "star" fashion back to the main filter cap also significantly reduce this effect.

This might be needed if a design is using 2 chips for one drive like the LMD18245 or L6203 to synchronise the currents in the 2 coils, is this what you mean?

[boldford]

The TB6560 has a different pinout and other functions like current setting are completely different. Why would you use the TB6560, do you have some stock?

I was thinking of those with stock to use up on a well designed board. I.e decent optos as you indicated earlier.

This might be needed if a design is using 2 chips for one drive like the LMD18245 or L6203 to synchronise the currents in the 2 coils, is this what you mean?

Perhaps I mis-read things but my interpretation of articles elsewhere on such matters suggest it is desirable to synchronise the currents for all drives in a system. Obviously this becomes academic if it isn't possible to slave the oscillator of one TB6564AHQ/THB6064AH to another. I was thinking of the synchronisation to be in a broadly similar manner to that shown for the multiple L297s in Fig 3 of http://www.st.com/internet/com/TECHN...CD00000063.pdf accepting that a L297/L298 combo has far more crude current control.

Always happy to be corrected and learn.

[lucas]

I know that it is better to let people think you are stupid than to open your mouth and prove them right. but I am going to open it anyway and ask this question.
Could you not have inputted the signal to the anode of the opt diode ( cathode to gnd) and have a non inverted signal and not have the inverting transistors? I am thinking I am missing something.

Sorry, forgot to answer.
The transistors are there to clean up the output signal of the optocouplers, this isn't nice and square especially if the input is driven with a low current.
It was needed in a previous design with 3V3 logic. The TB chip uses 5V and has schmitt trigger input's so here they are probably not needed, I will scope their outputs and see if there's a nice transition around 2V.

[lucas]

I was thinking of those with stock to use up on a well designed board. I.e decent optos as you indicated earlier.

I do have a TB6560 design (single drive) with everything on board wich performs very well here, maybe I will post in the appropriate thread.

.
Perhaps I mis-read things but my interpretation of articles elsewhere on such matters suggest it is desirable to synchronise the currents for all drives in a system. Obviously this becomes academic if it isn't possible to slave the oscillator of one TB6564AHQ/THB6064AH to another. I was thinking of the synchronisation to be in a broadly similar manner to that shown for the multiple L297s in Fig 3 of http://www.st.com/internet/com/TECHN...CD00000063.pdf accepting that a L297/L298 combo has far more crude current control.

That ST synchronisation is only needed for boards with multiple drives and where good grounding can't be achieved. In a single drive setup you need to eliminate ground noise by a proper wiring scheme and it's almost impossible to sync seperate drives.

[lucas]

I have been working on another design whilst waiting for the boards to arrive.
The rather expensive heatsink issue (if you have to buy them) in the first one was the reason for it.

It's a totally other approach and a stripped down version:
- only one current sense resistor iso 2 in parallel.
- no diodes for heat reduction.
- the transistors after the opto's are gone, in normal circumstances they are not needed, checked that today.
- board traces are a bit thinner but still well within spec's.
- More resistor array's were used to reduce PCB size.
- Most resistors are standing up also to reduce the size..

I have made a DIY proto PCB and it was a pain in the ass to assemble and troubleshoot: shorts, open tracks and impossible to reach pins for soldering the component side traces. Won't do it again. I also somehow managed to cut the PCB a bit to small and solder the chip completely out of alignment, this doesn't affect it's functionality at all.

Attached is a pic of it, it looks messy but everything fit's nice and thight, on a profi PCB it will look just fine.

Some quick tests were done and it performs like the first one but the current sense resistors are a bit hotter.

[SteveWill]

Some great thinking on the driver chip mounting. Seem to solve a lot of problems. I don't think it is possible but is there a way to mount the 5V reg to use the same heat sink as the driver?

[lucas]

The LM317 heats up above 24V supply, a small heatsink is enough but on the same heatsink is almost impossible.

[Rambling mode on]
BIG problem with this type of layout is that all PCB traces go in one direction iso 2 if the chip is standing up.

Datasheet of the chip and other components say:
- Caps as close as possible
- Traces as short as possible
- Large ground plane to reduce noise
- Big copper area's to reduce heat
- Seperate signal and power traces with max. spacing between them.
- etc...

All the above = contraditcio in terminus.

So 25 pins = 25 traces to comply with the above.
This can only be done with a small PCB, use tracks wich are wide enough and route the traces to obtain as much groundcopper for noise and heat reduction. It isn't easy and this is the best I could produce, it will always be a compromise.
[Rambling mode off]

[SteveWill]

Datasheet of the chip and other components say:
-1 Caps as close as possible
-2 Traces as short as possible
-3 Large ground plane to reduce noise
-4 Big copper area's to reduce heat
-5 Seperate signal and power traces with max. spacing between them.
- etc...

All the above = contraditcio in terminus.

So 25 pins = 25 traces to comply with the above.
This can only be done with a small PCB, use tracks wich are wide enough and route the traces to obtain as much groundcopper for noise and heat reduction. It isn't easy and this is the best I could produce, it will always be a compromise.
[Rambling mode off]

I added number to you thoughts and thinking out loud myself.

1 and 2 without seeing the board layout, I am guessing they are as good as can be.
3. I think this is more an issue with sharp voltage edges. not sure it will help with the high current switching that being done on this board. On a board this size I would go with traces and spacing as big as possible and forget the ground planes.
4. Without 2 or higher oz copper boards you are are not going to get any any real heating dissipation. It may help a little on the sense resistors if they are close on heat. you could do feed thru's and pads on back side of board to add a little more copper but I don't think you have room. I would not worry about the rest of the board.
5. I think lihaijiang from the chip house said the signal and power traces had to be tied together at the chip so this is a none issues also.

the only other thought I had when seeing your last layout was it would be nice to see the PC board mounted to the heat sink somehow. I can see someone mounting the heat sink to a frame and not using the mounting hole on the PCB and with all the cabling breaking something. I think the new layout is the way to go. Looks good.

[lucas]

There is a power plane on both sides and a row of via"s to transfer the heat, you can see the pieces of wire in the via's and a solderblob on the right of the top CS resistor.

I tested it first with the alu block only, this was almost good enough @ 3A, then I added the piece of heatsinck.
This is indeed NOT the way to mount it, the pins of the chip almost broke due to me handling it.

[lucas]

Quick update to keep the thread going while I'm waiting on the profi PCB's to arrive, hope it will be soon.

There's now a nice pdf on my website for the flat version also:
http://users.skynet.be/ldt/CNC%20ele...4_flat_PCB.pdf
Remember that the PCB size is 50x50mm.

I am also working on the website to post all files and desribe the features and differences in detail, still a lot to do.

[lucas]

Today the first batch of PCB's arrived.
These are for the first version wich still has the fuse, they are already fully assembled apart from the TB chip, one has the diodes and the other not.
As mentioned before: this setup will be used to test the effect of the diodes on heat, with 2 identical heatsinks, motor and supply.

Pic's attached, it looks a lot better than a DIY one.

More tomorrow, hopefully..

[boldford]

Today the first batch of PCB's arrived.
These are for the first version wich still has the fuse, they are already fully assembled apart from the TB chip, one has the diodes and the other not.
As mentioned before: this setup will be used to test the effect of the diodes on heat, with 2 identical heatsinks, motor and supply.

Pic's attached, it looks a lot better than a DIY one.

More tomorrow, hopefully..

Nice work Lucas. Methinks you'll be in demand for kits, or at least PCBs, globally.

[lucas]

Nice work Lucas.

Thanks, I try to do the best I can.

Methinks you'll be in demand for kits, or at least PCBs, globally.

"Methinks" I will need the comments of 1 or 2 betatesters with DIY experience and the required tools before releasing anything to the public, I have my way of doing things and some are easy for me, difficult for others,
Feedback would be extremely helpfull.
I expect the PCB's for the other versions next week, same issue here.

Candidates? FIFO....

[lucas]

Found some time to do the comparitive tests.
Both drives have the same heatsink, see pic. Supply is 40VDC, 3Amp current and fast decay.
There's a difference in chip temperature but small; 70°C without and 66°C with diodes, not enough to justify the costs.
Now I wonder why de diodes are recommended in the Tips document. Ok it says above 3Amp but there should be a major difference at 3Amp also.

My conclusion: it can be used up to 3Amps with a decent heatsink, above 3Amp active cooling will be needed.
The CS resistors also get hot but there is a big difference between them: 50° if they are on a copper plane, 70° if it's only a wide track.

[H500]

The chip is supposedly good for 85c. Perhaps their recommendation is based on high ambient temperatures, such as inside an enclosure.

[lucas]

Inside an enclosure the temperature will rise significantly, some air flow will be needed and a larger heasink with some airflow will also lower the temp.
But why is the effect of the diodes marginal why they recommend it?

At 3Amp the chip has to dissipate 9Watt, quite a bit, no peanuts.
I did the comparison with a decent heatsink, not too big or small. I had to find a size where the difference would show.
Tests were done at 28° ambient. That gives a delta T of 42° without diodes and 38° with diodes.
Not spectacular at all.

Maybe lihaijiang could comment on these results?

[lihaijiang]

Dear Luc.

Could you please tell me Max reverse recovery time for your diodes? It has en effect in heat cooling.

[boldford]

Lucas may I suggest you retain the option on the final PCB to retain the diodes for those of a nervous disposition. Elimination of them doesn't make such a great saving. Assuming someone is building 3 channels for a straight-forward XYZ gantry mill there would be an overall saving of approximately £12.00 since BYV28-100 are available from Farnell for 50p.

[lucas]

@ lihaijiang: the diodes are BYV28 trr max. = 30nsec.
Did you experiment on the effect of the diodes? The results maybe different depending on the drive settings, supply voltage, motor induction... I just don't have time enough to test this all out.

@boldford: agree, they don't cost that much but are in the way for universal heatsink mounting. There is another version where I removed the fuse wich was really in the way and rearranged the diodes.
The PCB's of these are in, just need some time to charge batteries for a pic.

Luc.

[lucas]

Ok, batteries are charged.
Below is a pic of the pcb without fuse and diodes arranged to the center, this will allow a U-style heatsink, a stand-up one or a raiser style to clear the diodes.
See the 3 examples.

The heatsink can be mounted to the chip with full contact to it's metal tab and still leave space to clear the low height components but not the diodes.
A processor heatsink can then be used eventually with a fan on it.

The pic's just show some possibilities with some heatsinks I had, there are others available wich would fit better.
There's one requirement: the heatsink must be mechanically fixed somehow, you can't just leave a heavy one hanging on the chip as I did with a lightweight one.