TB6564AHQ (THB6064AH) PCB design

Printable View

Show 40 post(s) from this thread on one page

07-31-2012, 10:14 PM
lucas

1 Attachment(s)

TB6564AHQ (THB6064AH) PCB design

We continue the discussion here on my PCB design and device evaluation from several posts in this thread:
http://www.cnczone.com/forums/open_s...ing_motor.html

The 2nd design shown in the picture includes freewheel diodes and has the current sense resisitors on the top side for better cooling.

The chip with a heatsink like in the picture below doesn't get hot at 2Amps, I measured 45°C approx. with freewheel diodes installed.
But the power dissipation rises exponentially with the current: a rough calculation (see datasheet) gives 3.2W for 2A but at 4Amps it becomes 12.8W = times 4!!

A very quick comparison between the first (no diodes) and second design shows that the freewheel diodes reduce the chip's temperature @ 4Amp by 7°C after 5 minutes running, 63 iso 70°C. I used a tempgun to obtain comparable measurements. This needs to be tested in more detail with exactly the same design and environment. The diodes (8 pcs) are big and expensive. I will order "real" PCB's today and compare this when they arrive.

Quote:

Originally Posted by lihaijiang

I think you'd better to use 2X0.47ohm 2512 in parallel instead of 1206.

If necessary, you should use schottky diodes or Fast recovery diodes(below 100ns)to reduce the chips power dissipation .

If you want to reduce the temperature, you can add the fast decay proportion or add the decay time .That means you can increase the resistor value which is connected with PIN23.

The ones wich are on order are 2 of these in parrallel: 0.40 ohm, 2W, should be a lot better. I've been thinking to reduce the value lower than 0.2 ohm. This would reduce their heat significantly but in the tips it's not recommended, the datasheet doesn't mention it only in the typical application diagram.

The diodes are already installed: BYV28-100 trr= 30nsec. You can see them in the picture behind the heatsink.

I don't fully understand your last sentence, pin23 OSC2 changes the fchop. The settings wich I use are fast decay, I think this is better at higher speeds.
Do you mean that the heat will be reduced by lowering the chop frequency?
Thinking about it: There are less switching cycles and thus less switching losses, makes sense. What are the disadvantages?

Thanks,
Luc.
07-31-2012, 11:07 PM
H500

I would expect that the freewheel diodes would not lower the chip temperature if you were using fast decay.

Do this driver produce a hissing sound, or is it quite quiet?
08-01-2012, 12:19 AM
lucas

Hello H500,

as far as I understand: there are at least 2 things heating the chip:
- Lost power due to the RDSon of the fet's, standard issue.
- Switching losses due to parasitic diodes beeing too slow.
The diodes should improve the second one, it might be minimal and not economically worthwile but I will try it out.
Please correct me if I'm wrong.

Didn't pay much attention on the sound while testing, was more focussed on smoke, blowing fuses, etc ...:)
Can't remember hissing sounds, I will look (hear) for it tomorrow.

Regards,
Luc.
08-01-2012, 03:01 AM
lihaijiang

Quote:

Originally Posted by lucas

I don't fully understand your last sentence, pin23 OSC2 changes the fchop. The settings wich I use are fast decay, I think this is better at higher speeds.
Do you mean that the heat will be reduced by lowering the chop frequency?
Thinking about it: There are less switching cycles and thus less switching losses, makes sense. What are the disadvantages?

Yes,you got it.The heat will be reduced by lowering the chop frequency.But maybe it will cause hiss noise.So you need to find a balance point.

I think your heatsink is not big enough.You 'd beter add a bigger surface heatsink,eg.dentate heat sink.
08-01-2012, 01:01 PM
lucas

2 Attachment(s)

Should have mentioned this: The heatsink is just a sturdy base, fixed to the PCB. It's done like this for further testing: just screw another one on top, size as needed, evt. with active cooling.
The one in the pic below seems just enough for 4Amp, chip temp is approx 65 °C with ambient at 25.
This style of heatsinks are available in different sizes and with fins, I've adapted the PCB to match their fixation location. Attached is a sample drawing.

Please bear in mind that the thing is running at 40V and 4Amp per phase wich is very close to the limits, there are almost no heat issues at 2 - 3A per phase.

@H500: can't hear any hissing when the motor is idle at 3A, there might be if cheap wirewound sense resistors are used (as you probably know).

I really like it, it's working fine. A price guestimation for all parts, including PCB, chip and heatsink is approx 35 Euro. Wich isn't too bad for a full drive with these spec's.
08-01-2012, 07:17 PM
vegipete

Thank you Luc for posting your results so far! Could you post a circuit diagram so we can see the changes you've made from the application example in the datasheet? (I'd like to know how the diodes are connected.) What are the part numbers for the diodes and current sense resistors you've used?

Thanks for the info and great work!

P.
08-01-2012, 09:46 PM
lucas

The schematic is not yet finished and is still a mess, some values need to be changed: this will be done when the PCB's arrive and final testing can be done.

The 8 diodes currently installed are: BYV28-100 trr= 30nsec. Smaller ones like BYV27 might be enough. Also 4 pcs iso of 8 could be sufficient, there might even be none: tests, measurements and economic evaluation will tell. Calculations for this is close to impossible.

Definitive current sense resistors are ordered, not yet installed. So can't tell much about their result.

I could clean up the schematic and post a preliminary version tomorrow if time permits, this design has taken too much of my time recently.
08-01-2012, 11:14 PM
H500

Lucas, I think you are correct about the switching losses, but I'm not sure how much power is involved. If you have a good scope and probes, the spike should be visible across the current sensing resistor. If the diodes help, the spike should be substantially lower.
One way to reduce heat in the resistors is to use the slow decay mode. That causes the current to circulate through the lower fets without going through the resistor. On my drives, I'm not sure if the fast decay results in better performance or not.
08-02-2012, 01:11 AM
lucas

The effect of the diodes and the effective current sense resistor power depends on a lot of factors:
decay mode, speed, motor inductance, etc...
One could "accidently" try a drive in optimal configuration and be very pleased whilst somebody else has all the bad luck because everything is just on the marginal side and is "pissed off".

EDIT: this is exactly what occured with the TB6560, some are really satisfied, for others it's a piece of junk. The TB6564 looks much sturdier build and won't fail as much like the TB6560. ( I hope coz I've never had an issue with the TB6560 even at 35V 2.5A...)

It's real difficult to compare, some changes give minimal result but several together might be worthwhile.

Wich are the worst, most difficult settings for a drive? Hard to say, some settings and configuration will make it easy for the drive but reduce performance, others are the inverse.
It's a give and take situation, like reducing the chop frequency to reduce the heat just a bit but then possibly end up with an unstable drive...

I learned the theory on switching losses from a few posts by Mariss on the CPLD thread.

My aim is to try and develop a reliable drive as good as possible but that takes a (huge) lot of time.

This chip is definitly a good one if used a bit below it's max. ratings. Like everything else, eg a formula one car, it's very hard to drive it to it's limits reliably in all possible circumstances.
Do we have to do this? Or limit the spec's to something like 35VDC and 3Amp? Price/performance it's still a good deal.
The chip can do more, but everything has to be spot on or we slip out off the corner into the barrier.
Right now I'm pushing it to the limit and it still works, even with different motors and settings, tomorrow I might blow it with a marginal change.

My design is finished, some minor details to be sorted. I will use it "as is" and see how far it goes in the real world.

I will publish the schematic, might offer Pcb's and maybe complete component kit's. Need a few nights of sleep on this.

Thoughts, comments, recommendations ... anyone?
08-02-2012, 03:45 AM
SteveWill

Quote:

Originally Posted by lucas

This chip is definitly a good one if used a bit below it's max. ratings.
Do we have to do this? Or limit the spec's to something like 35VDC and 3Amp? Price/performance it's still a good deal.

I will publish the schematic, might offer Pcb's and maybe complete component kit's. Need a few nights of sleep on this.

Thoughts, comments, recommendations ... anyone?

One: I am a newbie so I don't know much about drive circuits. With 162 views of this thread I would say there is some interest.
I am looking forward to seeing your schematic.

If this board is produced in a kit is there some protection from reverse power connection? I have seen a diode tied to ground so it blows a fuse when the power is reverse and the diode is forward biased.
Spec are always a problem, First off I do not believe them. If I have a motor that needs 3 Amps I will be looking for a driver that say 3.5 to 4 Amps.
So if I wanted to hold the spec to 3 Amp and 35 volts for a board I was supplying. I would say something like the driver is tuned for 3 Amp motors but the chip is rated for 4.5 amp and 42 volts but would need a bigger heat sink and a fan.

What process did you use to make your raw PC board?

Good job.
Steve
08-03-2012, 08:01 PM
lucas

A reverse polarity protection is a good idea: I will look if there's still some empty space on the pcb.
It would be optional: install it just to be sure.

All my proto's pcb's are made using photosensitive PCB's, toner transfer doen't work well enough for me.

@vegipete:
Didn't have time yet to clean up the schematic, for me it's even not needed: will do it when the PCB's arrive and the design details are finalised.

For the diode configuration: have a look here:
http://www.sparkfun.com/datasheets/R...8_H_Bridge.pdf Fig 8. It's exactly the same.

Luc.
08-03-2012, 09:26 PM
vegipete

Quote:

Originally Posted by lucas

Didn't have time yet to clean up the schematic, for me it's even not needed: will do it when the PCB's arrive and the design details are finalised.

For the diode configuration: have a look here:
http://www.sparkfun.com/datasheets/R...8_H_Bridge.pdf Fig 8. It's exactly the same.

Luc.

Don't worry about the schematic too much - Fig. 8 of the L298 data sheet answers my question. That's a lot of diodes! At least they're not overly expensive. How fast do they need to be?
08-05-2012, 07:23 PM
lucas

The boards are shipped and I expect them next week.

All other parts are ordered also.
The heatsink is a bit of a problem, the fuse is in the way for an easy mounting of readily available ones.
I will have to drop the fuse and rearrange the components behind the chip, I searched for another type of fuse holder but can't find any, they either are partly off the board and a vertical one will interfere with some types of heatsink.
With the fuse gone there are a lot of heatsink possibilities, from a simple 2" or 50mm wide U profile to more sofhisticated ones.

It can be done as I did it using a "heatsink raiser block". The one you see in the pictures is a piece of one coming from a 5 phase, Berger Lahr drive.It had 10 mosfets on each side (20 total !) and a big 100 x 160mm heatsink mounted flat on top of it, I've seen similar arrangements in other equipment too.
But it's a bit of a hassle, needs accurate drilling and tapping.

I will need some more input before finalising: Is this the design you guys want or would there be more interest in a cheaper design with less options:
- No optocouplers?
- No on board voltage regulation?
- Other?

Please let me know.
Luc.
08-07-2012, 10:17 PM
lucas

Just a quick update:

The layout is modified: fuse is gone, diodes moved to the center behind the heatsink and added diode for reverse polarity.

The preliminary schematic and a PCB layout print are on my website:

http://users.skynet.be/ldt/CNC%20ele...%20-%20PCB.pdf
http://users.skynet.be/ldt/CNC%20ele...0Schematic.pdf

Regards,
Luc.
08-08-2012, 09:49 AM
boldford

Quote:

Originally Posted by lucas

. . . .

I will need some more input before finalising: Is this the design you guys want or would there be more interest in a cheaper design with less options:
- No optocouplers?
- No on board voltage regulation?
- Other?

Please let me know.
Luc.

A couple of comments.

1. I would have thought complete opto-isolation (i.e. separate grounds) is a preferred option, to protect the PC's LPT port, so long as the selected optos have a high enough slew rate.

2. In using the fairly robust THB6064 the worry over power sequencing is removed. I would say keep the 5v regulator to simplify connections to the board.

3. One thing I have missed is whether or not the oscillators of multiple drivers using this chip may be/should be synchronised. If it is a requirement provision should be included.
08-08-2012, 08:15 PM
vegipete

A couple of thoughts:

I don't know about the optocouplers. Since the interface is the PMinMo header, pc parallel port systems will need a break-out-board which is where the opto's can be placed. This way, the board would be a simple drop in replacement for numerous other boards using that interface, such as Linisteppers.
(As an aside, and request, I'm love to see a 'PMinMo-plus' interface which changes pin 7 to be a fault output - low when all is well, high when the drive faults. So connect it to the Alert output of the THB6064AH. Maybe with a jumper for older systems that force pin 7 to ground.)

Any chance of positioning the DC power in connectors to make busing a set of driver boards easier? The easiest might be V+ and GND terminals duplicated on opposite sides so that short pieces of solid copper wire could link boards together.

Should the current adjust pot screw be side mount or top mount? I guess the pad pattern allows either.

p.
08-08-2012, 09:09 PM
lucas

The interface is indeed the PMinMo header, this one has been adopted by others like usbcnc and planet cnc, but these "expensive" boards copied the signals and don't have opto's, they need drives with opto's on board, that's why I want them also. Another reason is to be "compatible" with leadshine and many others wich also have them.
They are optional, a few wire runs will do if you don't want them.

I must have missed something: never heard about the 'PMinMo-plus' interface,
this feature is very usefull, will look into it tomorrow.

Bussing drivers is a bad idea, every drive should have its power wires connected individually to the power supply where they should join as close to the filter cap as possible.

The current adjust pot is a critical one, imagine what would happen with the motor current if it fluctuates value.... so I used a multiple turn quality brand. I had some with side screw, there are top screw ones wich use the same layout.

@Boldford:
1: The opto circuit as on this drive performs excellent, not like the ones on the chinese TB6560 PCB's.
2: I fully agree but others might have another opinion. It's always optional, don't install the regulator and related components, then it's just a matter of injecting +5V from a seperate supply into the drive without creating ground loops.
3: Don't understand what the benefit would be, I have a vague memory of a 3 axis drive where they used only one oscillator for cost saving. I think it's impossible with this chip .

Thanks and regards,
Luc.
08-08-2012, 11:31 PM
SteveWill

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.
08-09-2012, 12:01 AM
vegipete

Quote:

Originally Posted by lucas

I must have missed something: never heard about the 'PMinMo-plus' interface, this feature is very usefull, will look into it tomorrow.

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

Quote:

Originally Posted by lucas

Bussing drivers is a bad idea, every drive should have its power wires connected individually to the power supply where they should join as close to the filter cap as possible.

Yah, I figured as much. I'm too use to jumpering terminal strips together in control cabinets.

Quote:

Originally Posted by lucas

The current adjust pot is a critical one, imagine what would happen with the motor current if it fluctuates value.... so I used a multiple turn quality brand. I had some with side screw, there are top screw ones wich use the same layout.

Absolutely multiturn, I figured something like this one.
08-09-2012, 08:25 AM
boldford

Quote:

Originally Posted by lucas

2: I fully agree but others might have another opinion. It's always optional, don't install the regulator and related components, then it's just a matter of injecting +5V from a seperate supply into the drive without creating ground loops.

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?

Quote:

Originally Posted by lucas

3: Don't understand what the benefit would be, I have a vague memory of a 3 axis drive where they used only one oscillator for cost saving. I think it's impossible with this chip .

Not for cost saving, but if not possible of no consequence. 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.

Great work!

Show 40 post(s) from this thread on one page