PDA

View Full Version : What is better than a ULN2803?



lardom
01-07-2012, 01:55 AM
There is a big improvement in motor speed and torque whenever I have replaced a ULN2803 with transistors. I would much prefer a single IC over 8 transistors, 8 diodes and 8 resistors for two steppers. Is there an IC that can match the performance of discrete components?

Phil Pilgrim (PhiPi)
01-07-2012, 02:41 AM
I take it that your steppers are either 6-wire or 8-wire. If they're of the 8-wire variety, you can also drive them in bipolar fashion with an H-bridge. (Actually 6-wire steppers can be driven this way, too, but with reduced performance.)

-Phil

lardom
01-07-2012, 04:17 AM
They are 6 wire. I will use bipolars for my next project especially there is a gain in power with the same current. I would place L293s in the same category as ULN2803s. I would rather have performance.

Phil Pilgrim (PhiPi)
01-07-2012, 05:57 AM
Here's an IC that might work better for you:


http://search.digikey.com/us/en/products/DRV8803DW/296-29743-5-ND/2766232

One of the problems with Darlingtons is their slow switching time, which may have contributed to the poor performance you received from the ULN2803.

-Phil

Tubular
01-07-2012, 06:15 AM
I've looked at this IC more than once but never used it. 6 Hi Side and 6 Lo side switches.
http://www.onsemi.com/pub_link/Collateral/NCV7708B-D.PDF
(http://www.onsemi.com/pub_link/Collateral/NCV7708B-D.PDF)
Doesn't have very low switch resistance, but a it has a flexible design with good monitoring and SPI interface.

lardom
01-07-2012, 03:11 PM
Thanks Phil Pilgrim and Tubular. Both chips use mosfets. Scratch the darlingtons for high speed inductive switching loads. The only downside is they are both SOIC. I have several QFN propellers which I can only admire. I have to find something to practice on. Until then I will track down DIP versions of the recommended ICs.

Tubular
01-07-2012, 08:12 PM
I don't think many will come in DIP version unfortunately. But don't despair, grab yourself an assortment of these little adapters.
http://www.futurlec.com/SMD_Adapters.shtml
(http://www.futurlec.com/SMD_Adapters.shtml)
Taking the plunge into SOIC (1.27mm) sized chips is well worth it and not as difficult as you might think. There are all sorts of really useful chips in soic.

The worst you can do is join a couple of pins together (solder bridge), and this is easily fixed with a bit of solder wick.

The adapters don't have good heat sinking, but you can always solder a 'wing' of copper to the heatsink pads of the driver, certainly for proof of concept / prototype testing

cheers
tubular

erco
01-07-2012, 10:26 PM
What is better than a ULN2803?

Petunias on your piano?

Sorry, it sounded like a joke.

lardom
01-08-2012, 04:02 AM
Erco, not bad...I program my Propeller with a long stem rose between my teeth. :cool:

Kalo
07-05-2012, 07:24 AM
Bump for the thread and the continuing search for a DIP packaged stepper driver.
I have a couple of Vexta PX243M-01AA steppers I want to run in a camera motion control system. I think this might be the driver IC of choice for this project:
http://www.digikey.com/product-detail/en/A4975SB-T/620-1435-5-ND/2783714

Sure, it doubles the chip count, but still worth a look.

Mark_T
07-05-2012, 10:38 AM
There is a big improvement in motor speed and torque whenever I have replaced a ULN2803 with transistors. I would much prefer a single IC over 8 transistors, 8 diodes and 8 resistors for two steppers. Is there an IC that can match the performance of discrete components?

TPIC2701 is pin-compatible DMOS chip, complete with diodes - hard to get hold of though. Looks like a brilliant replacement too.

Loopy Byteloose
07-05-2012, 10:48 AM
Having 8 MOSfets or 8 transistors in one package is not appealing to the manufacturers and engineers as the heat buildup for power is a problem.

Even the ULN2803 is limited in this way. The chip can only handle a fraction of what the individual Darlingtons are capable of. Seems very odd to provide so much gain when so little actual output is available.

To make matters worse, the DIP package is disappearing as surface mount devices eliminate the need to drill holes in PCBs. Those holes are a significant added cost.

If all 8 channels of the UNL2803 are on at the same time, you might be 60ma from each channel, but 8 2N3904s will provide 200ma per channel, and 8 2n2222s will provide something like 800ma or more.

If you were to locate a driver chip that would do the job, it may not have the straight through configuration that the UNL2803 has, so wiring gets more complex as well.

BTW, not every chip in a catalog with a PDF is available for retail. Many will only be manufacturer on order and that means one must buy a large quantity.

Mark_T
07-05-2012, 12:12 PM
Having 8 MOSfets or 8 transistors in one package is not appealing to the manufacturers and engineers as the heat buildup for power is a problem.

Even the ULN2803 is limited in this way. The chip can only handle a fraction of what the individual Darlingtons are capable of. Seems very odd to provide so much gain when so little actual output is available.

To make matters worse, the DIP package is disappearing as surface mount devices eliminate the need to drill holes in PCBs. Those holes are a significant added cost.

If all 8 channels of the UNL2803 are on at the same time, you might be 60ma from each channel, but 8 2N3904s will provide 200ma per channel, and 8 2n2222s will provide something like 800ma or more.

If you were to locate a driver chip that would do the job, it may not have the straight through configuration that the UNL2803 has, so wiring gets more complex as well.

BTW, not every chip in a catalog with a PDF is available for retail. Many will only be manufacturer on order and that means one must buy a large quantity.

Darlingtons are an anachronism these days, both superbeta transistors and MOSFETs are arguably superior in performance (especially saturation voltage). Put 8 0.15ohm MOSFETs in one package and you'd be able to supply 0.5A per output at < 0.1Vsat) with a total power dissipation of 300mW. Take 0.5A from every ULN2803 output and the dissipation would be about 5W until it failed with 1.3V Vsat. In practice you can't take more than 200mA per channel without risking overloading the chip.

I've used an array of e-line high-performance transistors (ZTX450) before now to get 8 drivers into the same space as one 16-pin DIL package, which is one approach...

For MOSFET there are a variety of dual n-channel devices in SOIC8 or smaller packages that can provide decent performance, but all surface mount of course... A high performance MOSFET+diodes driver chip that's pin-compatible with ULN2803 ought to be a winner though - even if in SOIC20 format. There are various non-pin-compatible options (mostly without diodes) which some enterprising soul ought to sell on a breakout board (SparkFun?)

lardom
07-05-2012, 01:15 PM
Loopy Byteloose and Mark_T, you both make a strong case for me to learn surface mount soldering.
For the next version of my machine I'll eliminate the belt and pulleys and double the speed. I suspect I won't be able to simplify construction with driver ICs.

Loopy Byteloose
07-05-2012, 04:49 PM
Surface mount? Not really, having the little transistors sticking up from the board provides better cooling. In many cases, a surface mount equivalent is the worst performer in a similar group as it cannot properly cool. For example, take a look at the ratings for 2N7000 in their different packages. The TO-92 out performs the SMDs.

I suspect that UNL2803 exists to get us all to really learn what the PDF says. When you really can comprehend that the high gain of the Darlingtons is rather pointless when you epoxy 8 together with shared silicon. They may have once been useful in dot matrix printers with big fat parallel cables or still useful to drive a matrix of 8x8 LEDs; but motors are not really an appropriate use - at least not anything that has much power. They are just handy for beginners as the wiring is so simple.

I still use some Darlingtons as the gain does come in handy. Instead of using two transistors, one Darlington works very well. They are great for H-Bridges. But most often I have no need of 1000x gain and simply 40x or 60x is just fine.

lardom
07-05-2012, 05:38 PM
Loopy Byteloose, I agree with your assessment of TO-92s. A transistor/diode combination is my preference. I don't mind the 0.6V drop. I get great power and the bias resistor calculations are easy.