Minimilist Level Shifting

Does your 3.3V end of the problem even have a +5V supply?

There's 5V tolerant inputs on the particular parts, CD4049 and CD4050, when running on 3.3V, was the point. If, by 'it' one means a general 4000 series part, or 74HC14, on 3.3V those 'it' items are not 5V tolerant.

Yeah, it's a nuisance; some of the 'digital transistor' items can lower the resistor count, or if you have a 3.3V supply available that eliminates a resistor. The inverter isn't required to do the level translation, it just assures clean output (and inversion, of course).

A common-base digital transistor like Rohm DTC014Y with emitter to the 3.3V logic input, base to +3.3V, and a pullup resistor from collector to +5V is the level-translator part of the JT design (for 3.3V in, 5V out) that you'd want. One resistor only.

As I've said, I have to support MHz on the RS-422 signal. Slow devices are not suitable.

Two resistors and two capacitors would start to be significant real estate. Ii could use 0402 devices, with 1.6 x 0.7 mm pads. Just two such devices are the size of a 6-XSON and the eight I would need would be four times larger. I think using a LVC2G14 is a good way to go.

I'd like to get the RS-422 device in a QFN, but supply is not clear at this point. But heck, this is the one part in the BoM that continues to be a problem in any package. Most of the parts I'm using seem to have settled down and I can get what I will need for this year. I may be stuck with the TSSOP package for this part.

"In stock"? Perhaps, but nothing suitable for this project. I think you are trying to design without knowing the requirements. Never a good idea.

< >

There is also an 1G04.

Gerhard

What value? Do you guarantee this for the part I'm using? Does it matter how many lines are terminated this way?

That is the sort of short cut that I don't like to do. Injecting current into the ESD diodes always has risk, from unexpected events. Two resistors for two channels, is about the same board space as a single X2-DFN1410-6 which is very similar to the SOT886, 1 x 1.5 mm. I think the 74LVC2G14 is probably the best solution without playing games with ESD diodes. If the chip maker of the receiving part had an app note for interfacing to 5V this way, I might do it. But I have no interest in taking liability for replacing thousands of boards that start failing in the field after some months or years.

So other than one less resistor, how is the series resistor better than a voltage divider? The 200 ohm value I see recommended, would result in around 10 mA of high level current. I'd like to use much less than this.

Ah, I was not aware that any parts in the CD4000 series were 5V tolerant when run from 3.3V. So that's like a 74LVC2G14 then, which is the part I'm thinking of using, but in a MUCH smaller package.

The circuit with the transistor, inverter and four resistors, is a bit absurd. If the design had an unused inverter and lots of room, fine. I have neither.

But the pullup is either slow, or consumes excess current, or both. The 74LVC2G14 is a much better solution for the two channels I need (I can even buy them). Part of my frustration is the incredible plethora of logic families. I find it hard to track what is what and why one is better than another. Maybe I'm just getting old. I remember when 74LS was the hot thing and then they came out with 74ALS! WOW! I think we had one design that used a 74S part for speed!

Thanks for the comments.

you already said two lines ...

And...

I have 2 full unopened reels (2,500 pcs each) of SN74CBTD3384CDBQR parts that I have no use for if somebody needs those :) And yet another partial reel of 2,000 on top of those. That is 7K parts total.

They are dirt chip, probably even cheaper than single 74LVC1G parts... And they are fast enough for PCI bus and beyond...

(mirror host takes bow)

Cheers

Phil Hobbs

That's what I used for the previous revision on the interface to the main board. But that's 3.3V now, so no more need. But I was running these to signals through that, which is why I'm looking for a 2 bit version.

Great minds think alike! :)

piglet

And do things the easy way.

Too bad they come up with the wrong answer. What will be the current when the TTL output is high? Does anyone consider the requirements?

Parameter TEST CONDITIONS MIN TYP(1) MAX UNIT VOH High-level output voltage VID = 200 mV, IOH = –6 mA 3.8 4.2 V

We'll see if that makes it through the usenet filter. Likely not.

I don't get your point? If the TTL Voh is 4.2V then the current in 200R would be 1mA (assuming 0.7V drop for the input clamp diode to 3.3Vdd)

If even 1mA is too much then one could increase the resistor until RC delay becomes significant.

piglet

Real 74xx style TTL doesn't pull up hard. Modern CMOS will tolerate 50 mA into its ESD diodes and not latch up. A direct connection would work fine.

Yes, add 200 ohms for the faint of heart (or the math impaired.)

I use the M74VHC1GT50 an there's plenty in stock.

Cheers

The spec on most chips is Vdd+0.3V abs max. The 4.2V is at 6 mA.

1 mA would be fine, if I believed that's the number it would be. What point will the RC delay be significant? If this part actually is only passing 1 mA, it isn't doing much regulating of the current. A small change in either power supply voltage and the current will easily double or triple, or more.

Every approach I looked at has issues like this, which make it a bit awkward, or even potentially unreliable. I know people talk about injecting current into the IO pins and into the power supplies this way, but a bit too much current and it goes into latchup... fatally.

There's just not enough utility in the single resistor approach to justify using it. The alternative is a single IC (a very small one) for the two lines. I started out looking for a two channel version of the 10 bit chip I had been using. It's not an easy thing to look for on Digikey or Mouser. This voltage conversion switches don't have a category of their own. So I got frustrated and started looking for a short cut. Not finding any of those that I liked, I searched for a digital buffer type part and they are small, cheap and plentiful. That's the route I will be going. The only thing that bugs me is adding another line item to the BOM, but the 200 ohm resistor does that as well.

Thanks for your suggestions.