Electronics > Electronics

Motor driver circuit - suggestions for improvements

<< < (2/3) > >>

sys 49152:
Thanks for your input.

I'll do some more work on the circuit before I build the next revision.

Tim.

jwatte:

--- Quote from: Soeren on December 04, 2012, 11:24:46 PM ---
--- Quote from: jwatte on December 04, 2012, 09:22:07 PM ---If you want fast diodes, Schottkys are kind-of hard to find in high voltage grades.

--- End quote ---
You're the one mentioning Schottky diodes here. Schottkys are usually not the answer and a lot of them are a bit on the slow side as well.

--- End quote ---

--- Quote ---the 1N4148 is blindingly fast compared to Schottkys.

--- End quote ---

That's counter to all the design advice I've read until now. Schottkys are generally recommended as fast barriers, in data sheets for anything from switching power supplies, to gate drivers, charge pumps, etc.

Looking at the data sheets:
1N4148 has a switching speed of 4 nanoseconds, and a max current of 450 mA. It has an 8 pF capacitance.
Meanwhile, a Schottky like CSD01060 is rated at "essentially no switching losses" and 1.7A. It has an 11 pF capacitance, which is slightly higher. Maybe that's the "slowness"? The capacitance per rated ampere is lower, though.

Also: I take "essentially no switching losses" to mean "zero recovery" but maybe that's not what it means. If you could qualify your suggestion to look away from Schottkys with a more detailed reference, that would be great!


--- Quote ---But what do you consider high voltage in that respect?
--- End quote ---

I really should proofread my posts, as I meant "current!" The above diode is rated for 600V -- I meant high current. Schottkys rated for > 10 A seem to be hard to find and expensive to purchase when you do.
A small-signal switcher like 1N4148 is cheap, but it's also small :-)



--- Quote ---
--- Quote from: jwatte on December 04, 2012, 09:22:07 PM --- Although if we're talking 250mA/12V unipolar stepper motors, this probably isn't really a problem. I've driven such things with BS-170s on breadboards without mishaps...

--- End quote ---
BS170 is good for driving LEDs (providing a high enough voltage), as they need current limiting anyway and they're good for fast driving due to their switching times, but I wouldn't use it for a motor, as it has a higher voltage drop than a bipolar.

--- End quote ---

That depends on driving current, doesn't it? Although the reason I mentioned BS170 was that it's not a high-performance part, yet worked fine for a small, 12V, unipolar stepper motor. Most unipolars tend to be small, because if you want real power, you will want bipolar. I think we agree on that?


--- Quote ---It's a common misunderstanding that MOSFETs can be properly driven by a microcontroller pin.
If you want the fast switching it's capable of (which is what gives it a high efficiency), you need to poor buckets of current at it when it hits the Miller Plateau. If you don't, a bipolar solution will be better.

--- End quote ---

You need to build up the gate charge quickly enough that you don't develop heat. If you can switch "slowly" (we're still talking less than a few microseconds) -- either PWM at perhaps a few hundred Hz, or even just on/off switching -- then low currents like 20-40 mA may be perfectly fine, and that can be delivered by many microcontrollers.
After all, the microcontrollers are built entirely out of MOSFETs themselves, and seem to switch their own internal MOSFETs just fine :-) Yeah, those are smaller-current integrated devices, with smaller gate charges, but I point this out to show that it's a numbers game, not a black-or-white.
Calculate time in transition zone times power lost while in transition zone, multiply by switching rate, subtract cooling, and you end up with "it works" or "it doesn't."


--- Quote ---
--- Quote from: jwatte on December 04, 2012, 09:22:07 PM ---Speaking of which: Is there a good, through-hole, 5V gate driver that works with 100% duty cycles? I previously used some decent 5V drivers from International Rectifier, but they use a bootstrap that needs a lower duty cycle -- so, only for PWM and continual stepping, not for standing still.

--- End quote ---
MOSFET drivers are usually made to be able to provifde a high peak current for switching where the high current is needed. For holding the device in conduction (after the switch) it only takes a very modest current compared.

--- End quote ---

Yes, that's the whole benefit of MOSFETs! But a driver that uses a bootstrap capacitor will eventually see the gate voltage drive if you keep holding it high, and eventually the switched MOSFET will dip into the transition zone and then quickly overheat. This is why drivers have a rated duty cycle -- the time needed to re-charge the bootstrap capacitor. 100% duty cycle drivers can re-charge the capacitor without driving the gate low, which is clearly preferrable, but I only found those in surface mount packaging back when I was looking.

[quoite]TI makes the UCC2742x (where x is 3, 4 or 5) dual 4A drivers in 8-PDIP, the 3, 4 and 5 covers Dual Inverting, Dual Non-inverting and one of each respectively.
Microchip makes TC4421 (Inverting) and TC4422 (Non-inverting) 9A single drivers, both comes in 8-PDIP TO220-5.
[/quote]

I looked at those, and they are not bootstrapping (high-side) drivers, only low-side drivers. Which is fine for a unipolar driver, but not for bipolars that need H-bridges. I should probably have been more clear that I was looking specifically for high-side drivers with 100% duty cycle.

Soeren:
Hi,


--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---That's counter to all the design advice I've read until now. Schottkys are generally recommended as fast barriers, in data sheets for anything from switching power supplies, to gate drivers, charge pumps, etc.

--- End quote ---

Yes, well it's a relative world... Compared to regular PN diodes they're fast (with a few exceptions), but compared to fast/super fast/ultra fast diodes, they're slow.



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---Looking at the data sheets:
1N4148 has a switching speed of 4 nanoseconds, and a max current of 450 mA. It has an 8 pF capacitance.
Meanwhile, a Schottky like CSD01060 is rated at "essentially no switching losses" and 1.7A. It has an 11 pF capacitance, which is slightly higher. Maybe that's the "slowness"? The capacitance per rated ampere is lower, though.

--- End quote ---

When you read datasheets, you need to keep your eye on what you need the component for - in this case an inductive recoil - and hold that up against the (right) parameters the datasheet reveals. In this case, look at the pulse handling.
The 1N4148 has a max rating of around 1A+ for short pulses and up to 2A for a 1Ás (non repetitive pulse.

The CSD01050, one of the "new breed" silicium carbide Schottkys, id not what I consider a normal Schottky, in the sense where amateurs can lay their paws on them, visiting a high street parts pusher, so in this context, you shouldrefer to regular (Old Skool) Schottkys, which is way slower than the SiC's.
If you have access to SiC's, I don't understand your problem of finding the appropriate current handling clamping diodes.
I'm not sure where you got the 1.7A rating, as the datasheet from Cree gives different numbers depending on temperature, but no 1.7A?



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---Also: I take "essentially no switching losses" to mean "zero recovery" but maybe that's not what it means.

--- End quote ---

Terms like "essentially" and "virtual" means (in this context) mens "little" compared to the older generations of ("Heavy Metal") Schottkys.
Saying a diode have zero recovery time (they talk about ZR current), would mean you could pour daylight through it ;)



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---If you could qualify your suggestion to look away from Schottkys with a more detailed reference, that would be great!

--- End quote ---

I'm not saying that you shouldn't use Schottkys! I was just referring you to fast (and faster) recovery diodes, which is a much older device than SiC Schottkys, as you wanted high voltage/current/whatever, but perhaps I should rather have asked you how you arrive at a need for >10A diodes for clamping a motor driver, as there's a widespread misconception around diode selection among amateurs (and  way too many pros I hate to say), which is usually realized in people installing 10A diodes if the motor is rated at 10A (whether unloaded or stalled).

A clamping diode needs to be able to safely handle the flyback pulse, which is a short duration, it doesn't need to handle the peak as if it was DC.

If you want the lowdown on the fast and faster diodes, I'll suggest reading a few of the papers about them (not datasheets and app notes, but what some of them call "linear briefs" - they're usually a lot of info in a light package.



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ----- I meant high current. Schottkys rated for > 10 A seem to be hard to find and expensive to purchase when you do.

--- End quote ---

The beefiest Schottky in my drawers are 60A (drawn from some long forgotten piece of electronics), but Fairchild (among others) have several 40A devices - one example here which is under  a buck at 1k units.
So, easy to find and fairly cheap I think ;D



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---That depends on driving current, doesn't it?

--- End quote ---

Yes, you referred to a "250mA/12V unipolar stepper motor".



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---Most unipolars tend to be small, because if you want real power, you will want bipolar. I think we agree on that?

--- End quote ---

I don't think that's why they're small :P
If I want "real power" I don't use steppers at all.
(It would probably be good to write eg. "bipolar motor", as diodes are unipolar (Schottkys) and bipolar (PNs) as well



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---You need to build up the gate charge quickly enough that you don't develop heat. If you can switch "slowly" (we're still talking less than a few microseconds) -- either PWM at perhaps a few hundred Hz, or even just on/off switching -- then low currents like 20-40 mA may be perfectly fine, and that can be delivered by many microcontrollers.

--- End quote ---

For on/off with a period of say a second or more between shifts it's OK.
For driving a stepper at anything beyond super slow, it's poor design, plain and simple.



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---After all, the microcontrollers are built entirely out of MOSFETs themselves, and seem to switch their own internal MOSFETs just fine :-) Yeah, those are smaller-current integrated devices, with smaller gate charges, but I point this out to show that it's a numbers game, not a black-or-white.

--- End quote ---

The internals of I/O is completely unrelated here and in a sense, it is black and white, as it boils down to knowing the components you work with and the environment they'll live their life - or not... Leave bad design to the Chinese, they're the experts in that discipline.



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---Calculate time in transition zone times power lost while in transition zone, multiply by switching rate, subtract cooling, and you end up with "it works" or "it doesn't."

--- End quote ---

No, with any design, you end up with something you'll stake your rep on, or add to the already too large pile of crap design that isn't totally in the "doesn't work" category, but is still a far cry from anything I'd put my name on.



--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---I should probably have been more clear that I was looking specifically for high-side drivers with 100% duty cycle.

--- End quote ---

What? You're using N-ch. devices for high side?  ::)
That would explain it.

Oh well time for a short nap.

jwatte:

--- Quote from: Soeren on December 06, 2012, 11:47:48 PM ---If you have access to SiC's, I don't understand your problem of finding the appropriate current handling clamping diodes.
I'm not sure where you got the 1.7A rating, as the datasheet from Cree gives different numbers depending on temperature, but no 1.7A?

--- End quote ---

"access" in the sense that they're within my hobby budget in quantity "a handful." The 1.7A came from me searching Digi-Key, then finding a data sheet, and copying the part number for the cheapest I found.


--- Quote ---perhaps I should rather have asked you how you arrive at a need for >10A diodes for clamping a motor driver, as there's a widespread misconception around diode selection among amateurs (and  way too many pros I hate to say), which is usually realized in people installing 10A diodes if the motor is rated at 10A (whether unloaded or stalled).

--- End quote ---

Right! And that explains where I'm coming from: I've used these guys for switching power supplies and high-side N-channel drivers. Motors are pretty new to me, and the advise I've heard is "put a fast diode across it if you use bipolar transistors; MOSFET body diodes are good enough otherwise."


--- Quote ---
--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---That depends on driving current, doesn't it?

--- End quote ---
Yes, you referred to a "250mA/12V unipolar stepper motor".

--- End quote ---

But here we were talking about the current driving the gate of the MOSFET that in turn drives the motor.



--- Quote ---
--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---Most unipolars tend to be small, because if you want real power, you will want bipolar. I think we agree on that?

--- End quote ---
I don't think that's why they're small :P
If I want "real power" I don't use steppers at all.
(It would probably be good to write eg. "bipolar motor", as diodes are unipolar (Schottkys) and bipolar (PNs) as well

--- End quote ---

Clearly, assuming context leads to misunderstandings. Transistors, too, can be bipolar :-)

Yes -- if you need a stepper motor, and want something more powerful, I would suggest a bipolar stepper motor over a unipolar stepper motor, because unipolar stepper motors are pretty slow even when driven as hard as you can. Chopping drivers and high voltages for bipolar steppers is what the CNC and servo guys seem to use these days when specifying steppers.

This is the background to the question to the original poster: I think he's driving a unipolar stepper, and I'm wondering whether it performs well enough.


--- Quote ---
--- Quote ---BS170 as an example of a small use case
--- End quote ---

For driving a stepper at anything beyond super slow, it's poor design, plain and simple.

--- End quote ---

It worked fine for my egg plotting 'bot, driving small 12V unipolars. Engineering, in the end, is the art of solving the actual problem as effectively and cheaply as possible. The problem statement does not always include "must last 100 years and stand up to abuse outside normal operating parameters." (Even though the world would be a less frustrating, and much more expensive, place if it did!)



--- Quote ---
--- Quote from: jwatte on December 05, 2012, 02:41:31 PM ---I should probably have been more clear that I was looking specifically for high-side drivers with 100% duty cycle.

--- End quote ---
What? You're using N-ch. devices for high side?  ::)
That would explain it.

--- End quote ---

Yes, I do, if I drive something like a bipolar (so I need an H-bridge or similar.)
I've found that P-channel devices have a similar problem, because you have to somehow switch their gate up to the source voltage to turn them off. Pulling down to turn on is easy with a small signal N-channel switcher, but pulling up to 24V or whatever when your control signal is 5V is harder. For example: A pull-up resistor that is small enough to switch the P-channel fast, means that you waste a lot of current through the signal N-channel when pulling the P-channel gate down.
Hence, if you're going with drivers anyway, you might as well go with high-side N-channel drivers, which lets you use the faster, cheaper, lower on-resistance N-channel power MOSFETS.

But, I've only found high-side drivers that require duty cycles less than 100%. When stepping motors for locomotion, as the OP does, this is probably fine, until you want to stand still...
So, to be able to recommend bipolar stepper motors as an alternative to the current unipolar design, the question of how to actually drive those bipolars probably should be solved. Integrated motor controllers can do it, of course, but that's significantly more complex than just a few MOSFETs that he's using now.


--- Quote ---Oh well time for a short nap.

--- End quote ---

Naps are the best!

MrWizard:
Now it's your turn Soeren  ;D

I am learning lots from this....   ;D

Navigation

[0] Message Index

[#] Next page

[*] Previous page

Go to full version