ULN alternatives for I/O handling

Hi all,

I have a number of boards that expose digital I/O to the outside world, with logic states signalled by open/closed contacts. For this my typical approach is to use ULN2804 darlington transistor arrays. For outputs I use the ULN to drive transistor output opto-couplers. For inputs, the ULN provides level translation and a measure of protection for my circuit.

The problem now is that the ULN2804 is obsolete and there doesn’t seem to be much in the way of alternatives. My question is what other better parts and techniques would people use for this?

Finding alternatives to the obsolete ULN2804 for driving digital I/O signals while providing level translation and protection requires careful consideration of various factors. Here are some alternative parts and techniques you can explore:

Alternative Transistor Arrays:TPIC6B595: This is a shift register with high-current, high-voltage outputs, commonly used in LED driving applications but can also be used for general-purpose digital I/O.

TPIC6595: Similar to TPIC6B595 but with latch functionality, allowing simultaneous output updates.SN75468: Quadruple half-H driver with built-in clamping diodes, suitable for driving inductive loads such as relays and solenoids.

Techniques for I/O Protection and Level Translation:Integrated Protection Diodes: Use discrete diodes or integrated protection devices to clamp voltage spikes and protect the circuitry from transient events.

Level Shifting ICs: Utilize dedicated level-shifting ICs like the SN74LV1T34 or SN74LVC8T245 to translate logic levels between different voltage domains.

Optocouplers with integrated Protection: Consider using optocouplers with built-in protection features such as overvoltage protection and reverse polarity protection for input isolation and signal conditioning.

Custom Interface Circuitry: Design custom interface circuits using discrete components like transistors, diodes, and resistors to achieve the desired level translation and protection.

Considerations:Voltage and Current Ratings: Ensure that the chosen alternative parts can handle the voltage and current requirements of your application.

Input and Output Compatibility: Verify that the alternative parts are compatible with the logic levels and signal characteristics of your system.Board
Space and Cost: Consider the footprint and cost implications of alternative parts compared to the ULN2804.

Availability and Longevity: Choose parts with good availability and long-term support to avoid future obsolescence issues.

Conclusion:Exploring alternative transistor arrays and techniques for I/O protection and level translation can help address the obsolescence of the ULN2804 while meeting the requirements of your digital I/O application. Evaluate the alternatives based on factors such as compatibility, performance, and cost to select the most suitable solution for your needs.

1 Like

I’m confused, digikey has 11k ULN2804A is stock and lists it as “active”: https://www.digikey.com/en/products/detail/stmicroelectronics/ULN2804A/599594
The part is made by ST and they also list is as active: https://www.st.com/content/st_com/en/search.html#q=ULN2804-t=products-page=1

That’s a DIP part, unfortunately. I should have been more specific. There is the 2803A in SOIC from TI, but that appears to be the only part.

Is there a reason why you can’t use the 2803? The SOICs are in plentiful supply from numerous companies - LCSC shows lots. Pretty much “jelly bean” status. You might have to use a semi house you’ve never heard of before, though. It’s so low tech, it’s hard to screw up. I use them in several of my products.

There is a mosfet pin “compatible” version though I don’t recall the part number off the top of my head.

Finally, why are you using the ULN to drive an opto coupler? Seems like massive 1980s overkill. Modern optos work down to 1 mA with decent CTR.

1 Like

There’s no reason I can’t use the 2803, only that I was put off by the seemingly limited options. I didn’t think to look at LCSC, only the more usual suspects (Farnell, Digi Key, Mouser, etc.).

The honest answer is that this is an old design, based off an even older design which works and I didn’t see a good reason to change it. The obsolescence of the BOM part prompted me to stick my head up and see what other people are doing.

The opto I’m using is an ACPL-247-500E, which only has a (maximum) CTR of 600%. The microcontroller I’m using can’t drive all eight photodiodes hard enough on its own. If you have a better suggestion, I’m all ears.

You know your application far better than I. My uses of optos (EL3H7) don’t need a huge CTR as I’m not driving anything very hard with the transistor (inputs to a microcontroller).

Though, even with the darlington (in the ULM) driving the opto you won’t get a huge amount of current from the phototransistor. If needing lots of current from the phototransistor, I’d be tempted to replace the ULN with NPNs and resistors - no need for a darlington. Probably cheaper and maybe less PCB space. Otherwise, I’d drive the opto directly from the microcontroller.

If you can live with 7 not 8 channels, there is TPL7407LA data sheet, product information and support | TI.com which is a “MOSFET ULN 2003”.

Some parts are “obsolete” yet continue to be in the market due to the sheer number of people using it I’m looking at you, 89C52…

(and, you, 555)

This reads very strongly of direct copy-pasted GPT output. What gives?

1 Like

For sure. There are 100s of Chinese companies that make a lot of low tech chips. Jelly bean parts. There are 12 suppliers of NE555 in the JLCPCB library. Only 3 I’ve ever heard of. $US 0.035 and up. 30 suppliers of LM358. $US 0.023 in 1K quantities. Transistors are all over the place too. mmbt2222 - I stopped counting at 50 suppliers, none of them you’ve heard of. 1K quantity - typically $US 0.007. And, the ULN2804 has 3 suppliers while the 2803 has 15. Prices for the 2804 are a bit high but the 2803 runs $US 0.15 and up.

Also, a lot of the microcontroller architectures are now available in VHDL with no IP royalties so they will live forever.

The big advantage to using this kind of stuff is availability. Even at the depths of chip-a-geddon, they were available in large quantities.

The EEVBlog guy has a bunch of jelly bean videos on YT. Worth watching if this is new to you. I like him a lot but he is a bit of an acquired taste for some.

Looks like it. Arrived by e-mail. Some mail clients are providing AI assistant writing tools. Browsers may provide them too.

I’m sensitive to it at the moment; at Sugar Labs we are an open source community engaging with Google Summer of Code, and some students with low English experience copy pasted a whole lot of responses to GitHub issues. The AI got it wrong every time, as they were bug reports with deep complexity.

On the subject at hand, ULN2804, we used them at the Anglo-Australian Telescope for LED drivers on alarm panels on every floor, with the common pin connected to a lamp test button. Lightning took out every one of them, and all the LEDs are lit. We are awaiting spares and time to fix them.

I went back and reread that post. First time I kind of skipped over when it started talking about shift regs. With your comment in mind, it really does sound like chat gpt BS. Goes way afield pretty quickly.

About 6 months ago I played with chat gpt a bit to see if it could help me with HW designs but it failed miserably - even getting things like the pinout for a microSD card wrong. HTF do you do that?

On the other hand, I find that it is helpful for improving my code but you have to ask the questions very carefully and give it follow on guidance to get to something that works well. I even fed it code with deliberate simple bugs and it found them.