MOSFET equivalent to the ULN2803

[phillybustr]

The ULN2803 and other low side drivers all seem to be using darlingtons. Wouldn't it be better if these chips used MOSFETs / DMOS? I'm thinking a ULN type chip that used MOSFETs would be able to source or sink much more current, since there's no internal voltage drop? If so, what would be a good alternative to the ULN chips, using MOSFET drivers? I'm looking for both high side and low side drivers.

[mwilson]

Questions like this, I usually peek in at Digikey. I hope they don't mind. Looking for MOSFET arrays shows some kind of pricey H-bridges in DIP-16 that seem to be the closest thing. The package holds two pairs of N-channel FETs with a common source for each pair. About half as compact as ULN2803. The larger industry seems to be just sprinkling individual SMT MOSFETs around as needed. Big integrated packages aren't very popular.

[phillybustr]

Thank you for looking that up. I had tried that and found those same MOSFET, limited, arrays.

As you're saying, the industry tends to use individual SMT MOSFETs, which makes me wonder, who is buying these chips in enough quantity that it's worth manufacturing them? The ULN2803 has been around for at least 20 years. Maybe it's been around for 30 years or more. At some point, they were probably used in commercial products, but it seems unlikely that that's still the case? In particular if MOSFET sinks would make more sense than darlingtons, which I think it probably would in many cases?

Anyway, evidently, it's worth it to manufacture octal darlington arrays, so one would think it would be worth it to manufacture octal mosfet arrays as well?

[baldengineer]

Anyway, evidently, it's worth it to manufacture octal darlington arrays, so one would think it would be worth it to manufacture octal mosfet arrays as well?

Return on investment of developing a new design, new masks, documentation, and production tests for a new chip is probably much lower than just pumping out more chips designed 30 years ago.

There is enough people buying the old chip to keep producing it, but, apparently, not enough to develop something to replace it.

[phillybustr]

Sounds depressingly plausible

[easternstargeek]

Howdy!

Ah, this brings me back... Let's see- in the early 1980's when switching power supplies were in their infancy, I worked cheek-by-jowl with a wizened Senior engineer. One day, I asked him about this new-fangled thing called a power MOSFET, and why we didn't try using those to build a 25kHz SMPS using smaller ferrite transformers instead of the bigger tape-wound cores operating at 8khz? He immediately asked me to justify my question by asking what the benefit of the new technology was. As a young engineer, I balked at his question- somehow believing that MOSFETS were better. I never did convince him- I guess he was more comfortable with the familiar, Besides, this was a military avionics application, and the Air Force* was reluctant to adopt new technologies too quickly... but clearly I digress...

The ULN2803 is a great chip for what it was designed to do. Yes, you are correct that a MOSFET doesn't have an intrinsic gate-to-source voltage drop like bipolar transistors do, but they do have a non-zero channel resistance when they are on. (Rds-on). So, unfortunately, there's just no getting around the voltage-drop (and on-state power dissipation) problem. Both kinds of transistors will get warm when they are conducting current, so there is no free lunch. MOSFETS do have the advantage of high speed and high gain, but in your application, will it matter? Only your hairdresser knows for sure...

Anyway, to address one other of your comments, This device is the PNP compliment to the ULN2803, and makes a great high-side driver:
http://www.jameco.com/webapp/wcs/stores ... 1762681_-1
The catalog description, "Trans Darlington NPN/PNP..." is a bit peculiar, but it is definitely eight PNP darlington drivers.

If you are making blinky things, the Darlington-based devices should perform flawlessly. They also do very well switching relays and solenoids, thanks to their built-in freewheeling diodes.

* As an interesting (?) side note- during this time, I was required to have a Secret security clearance, because we were working on a high-tech, tippy-top secret method of securing communication transmissions between military aircraft, called... are you ready?
Frequency Hopping Spread-spectrum! http://en.wikipedia.org/wiki/Frequency- ... d_spectrum
Yes, that's right- the very same technology is now used in thousands of different wireless consumer gadgets (albeit without the fancy encryptioon methods used by the military), all the way down to cheapo $19.99 wireless telephones sold at Wal-mart. How times change!

By the way- if you really want to see an amazing example of very early secure-communications technology, check this out:
http://en.wikipedia.org/wiki/SIGSALY
Wild, bizarre, and brilliant in it's implementation!

[phillybustr]

I have found a series of MOSFET equivalents to the ULN2803. It seems to be a versatile family of chips, all having Power MOSFET (DMOS) outputs. Some are available in DIP packages. The part numbers start with "TPIC", from Texas Instruments.

Examples:
- TPIC6259, 8 latches, each sinking 250mA continuous when all are on, for a total of 2A.
- TPIC2701, pin compatible with the ULN2001.
- TPIC2810, drive 8 100mA loads directly from an I2C bus.

However, to my surprise, it doesn't seem like having MOSFET ouputs is that big of an advantage over darlingtons when looking only at how much current the outputs can sink when all outputs are on at the same time. The datasheets for the TPIC6259 AND THE ULN2803 don't have graphs that allow them to be compared directly, but these give an indication, if one uses the TA=100C line for the TPIC and 100% duty cycle for the ULN.

[baldengineer]

However, to my surprise, it doesn't seem like having MOSFET ouputs is that big of an advantage over darlingtons when looking only at how much current the outputs can sink when all outputs are on at the same time.

The advantage of a MOSFET isn't necessarily the total current capability, but wasted power. The difference comes in with the saturation voltage of a DP (BJTs) and Rds of a MOSFET.

The Saturation of a ULN2803 at 350mA can be as high as 1.6V. That's 560mW of heat dissipated by the pair.

Rds of the TPIC is less than 2ohms which comes out to be 245mW at 350mA. That's over 2x savings in heat/wasted power...

[phillybustr]

I read somewhere that it would be prudent to keep the dissipation in the entire package to 1W or less. Which would then mean that you could only run 2 outputs at slightly less than 350mA. Extrapolating, that would be only ~60mA per output.

I have found a couple more DMOS sink drivers: TLC59210 and TLC59211.

I have yet to find a DMOS high side driver.

[nevillep98]

Check out TI's TPL7407L. It uses NMOS transistors to get a lower VOL than ULN2003's Darlington NPN. Therefore it will run cooler than the ULN2003 in this application. The only reduced specification is maximum voltage, 40V instead of 50V. More data at http://www.ti.com/product/TPL7407L.

[rogerdahl]

Thanks for the pointer, @nevillep98! Do you know if these are new chips or if I missed them when I was doing this project a few years ago?

The ULN2803 would have worked out fine for my project. I was doing a LED display and only needed to "sweep" the low side driver with one line on at a time, so wouldn't have been affected by ULN2803's total heat dissipating limit. In the end, I changed things around a bit and went with a TLC5941 instead.