Spartan3 driving mosfets

You can also switch with NPN transistor + resistor, and a reverse connected one has lower saturation voltage ( and lower beta too ).

What resistance do you expect of the on devices ?

-jg

Is there a failure mechanism?

John

Once I start adding parts, I may as well go with an octal level shifter and get all the way to 5 volts. The pcb would route nicer if I could use just the pullups.

The 2N7002's go to about 2 ohms with 4 volts or so on the gates, and that 2 ohms will have a roughly +6000 PPM/K tc, net 12 mohm/K, which is good enough here... that will cause just a few PPM/K dac gain error.

John

You could always use something like a 74AC05 open drain buffer - the device has about 4 ohms of resistance to ground (at 12mA sink). If a couple more ohms isn't an issue, you can dispense with the FETs entirely, but it would add more error than the typical 2 ohms of a 7002 in full drive.

Of course, the output FET in a logic gate isn't usually characterised 'thoroughly' :)

Cheers

PeteS

Use an other FET as FDV301 ... in sot23 too ! The Vgs of this one is much better than your one.

Laurent

Have a look at out TechiTip

John Adair Enterpoint Ltd. - Home of Tarfessock1. The Spartan-3E Cardbus Development Board.

Does a FET work better in this app than a bipolar would? I guess it is better to have a resistance (FET) in the path than a low constant voltage (saturated NPN).

Personally, I would find a FET with a lower Vgsth. They are not hard to find. I seem to recall the only thing you give up is a higher Vdss rating.

You could also use a voltage divider circuit. A series resistor between the FPGA and the FET's gate, and a pull-up resistor to a higher voltage. This would get you 4 or 5 V for the high, and 2.5 V or less for the low, depending on the values selected. As long as you don't need fast switching the resistors can be fairly high values.

Jon

But amongst all this John faces the standard hardware design issue: cost. A 2N7002 is 3 cents and thus easy to drop in by the 10s of units (layout costs aside). As to a voltage divider, I don't trust a FET to be off unless it's pulled to ground (it's actually in a sub-threshold region otherwise with a very undefined [generally] drain-source resistance) so I would not personally countenance that.

Much depends on the cost sensitivity of John's product. In some cases, one can justify a full solution, in others one can not (and some vendors would do well to learn that lesson).

Amusingly, on the subject of vendors and cost, I was comparing a MAX device against it's second source at TI today: MAX wants $3.11, TI wants $0.74. Hmm.

Cheers

PeteS

Grumble, snarl, I guess I'll use an FDV301 like Larry suggests... somebody else down the hall has just selected that same part for another board. Damned things cost 9 cents!

Roger the Maxim parts; they're expensive, but the good news is that you can't get them, so you save a lot of money.

John

The FDV301's has esd protection, and is a tigher spec than the 7002. I see digikey have them for 5.4c/10K - and you are making >10K, to be worried about single digit cents, right ? :)

You gan get the Transistor+Resistor I mentioned in a single package, look for PDTC11xx, these are ~1.7c each.

-jg

I don't see how 9 cents vs. 3 cents can make a difference when you are interfacing to a $20+ chip! If you really need to save the $0.06 per on a few pins, just beat on Xilinx a little harder and I am sure you can get them to drop their price another $0.50. Dammed things costs

Or you could just spend another couple of hours to locate a lower threshold FET that is just as cheap as the 7002!

John Larkin wrote: ...

Check for availabilty of the FDV301 too. Neither Digikey nor Farnell has stock.

So perhaps you save on the FDV3012 too, like on the Maxim parts ;-)