Tristate Buffer Output if Input is High-Z?

Jeff Walther trag at io.com
Fri Sep 30 16:15:21 CDT 2005


>Date: Thu, 29 Sep 2005 17:51:13 -0400
>From: Paul Koning <pkoning at equallogic.com>
>Subject: Re: Tristate Buffer Output if Input is High-Z?

>  Tony> An analogue switch IC is a similar device, and those do
>  Tony> exist. The problem is finding one that will switch quickly
>  Tony> enough (I would guess in a few nanoseconds) for this
>  Tony> application.
>
>No problem.
>
>Quickswitches are rated to several hundred Megahertz.  If that isn't
>good enough, you can use microwave switch transistors.  I don't know
>the max switch rate of either, but presumably any transistor capable
>of carrying microwave frequencies can switch on/off in a nanosecond or
>better.

>Date: Thu, 29 Sep 2005 15:16:46 -0700 (PDT)
>From: "Dwight K. Elvey" <dwight.elvey at amd.com>
>Subject: Re: Tristate Buffer Output if Input is High-Z?

>>No problem.

>  The problem is as Tony mentioned, you get no buffering.
>This means you can't expect to put the mux too far from
>the inputs on either side. You are really better off
>using something like '245's.

Thank you, gentlemen, and that includes the respondents I did not 
quote above (previous digest).  I really appreciate all the great 
information and leads.

I started with Pericom (as one person suggested) as I am familiar 
with some of their products and worked backwards from there.  It 
appears that they are calling the product which I want "Bus Switch".

Once I knew that, there's a nice selection available on Digi-Key, 
although I'll check Mouser later for better pricing.

The selection is not quite as nice as I'd like, because there's 
nothing analogous to the SN74ABT241A.  The nice thing about the 241 
is that it divides its eight buffers into two groups of four with 
independent enables for the two groups of four.   Okay, that's like a 
244.  However, the 241 has one active high enable, and one active low 
enable, while both enables on the 244 are active low.

So I'll need an inverter in line with one of the OE pins on the bus 
switches, which isn't the end of the world, but its not as elegant as 
the SN74ABT241A would have been.  The candidates are the Pericom 
PI5C3244 or the TI SN74CBT3244.

The speeds on these components look fast enough, the time to switch 
from high-Z to conducting has a maximum value of about 6 ns on the 
Pericom and 8 ns on the TI.   I'm not certain they used the same test 
conditions.  Minimum times are 1 ns.  Both have a propagation time of 
.25 ns from A to B or B to A once the switch is enabled.

Dwight, can you elaborate on why I'd be better off with something 
like '245s?  I either don't understand your point, or I have not 
explained my intent clearly enough.  I am trying to emulate the SIMM 
on a IIfx.  The only difficulty is getting X 4 DRAM chips to look as 
if they have separate D and Q pins.

On stock IIfx SIMMs, the data signals are not buffered on the SIMM, 
they come directly from the DRAM chip to the SIMM connector.  On my 
SIMM, they will go from the DRAM chip, through a bus switch (or 
tristate buffer) and then to the SIMM connector.  So, unless too much 
signal strength is lost going through the bus switch, I do not 
understand why I'd need a buffer there to strengthen or clean up the 
signal.  Or are you saying that I should have a latch there to hold 
the signal for some time after my switch control signal goes inactive 
to make sure that hold times are supported?

The pinout on the SN74ABT241A (buffer) and the SN74CBT3244 (switch) 
are the same (the PI5C3244 is also the same but comes in a smaller 
package) and the same packages are available.   I jumped the gun and 
already have some SN74ABT241A on hand.

So I think what I'll do is design my first set of SIMMs with the 
posibility of taking the OE signal for the Data Out side of the 
buffer/switch either directly from WE, from WE run through an 
inverter (WE'), or from WE AND CAS'.   The inverters and AND gates 
are available in tiny SC70 packages (similar to SOT23).

Then I can test the Buffer with WE (OE is active high) as control for 
Din and Dout.  The Buffer with WE AND CAS' (OE is active high) as 
control for Dout.  And the Bus Switch with WE' (OE is active low) as 
control for Dout.

Jeff Walther




More information about the cctech mailing list