SN74LVC1G123: Fast pulse data for CEXT = 100 pF

Hi Steve,

Figure 4 will help you get started finding the right resistor (or by using the equation K*R*C = pulse width with K being 1). The best thing to do is to test it if you want to get more precise pulse widths. As you said there will be some skew between devices, but that is heavily outweighed by the change in RC components due to their tolerance.

Dylan-

Thanks for your response.

Figure 4 is of little help as I mentioned in my original post:

First, the vertical axis is logarithmic so it is very difficult to resolve the curve.
Second, also as I mentioned in my original post it is clear from the R = 1K curve that below C = 100 pF there is non-linearity.

Figure 6 doesn't go below 1000 pF, so I have no value of K to start with, and it is clear from Figure 6 that K is NO WHERE NEAR a value of 1 for small values of CEXT.

Assuming K = 1 will not work as a starting point.

Do you have a suggested value for K when C = 100 pF and VCC = 3.3V?

Thanks, Best, Steve

Dylan-
K = 5.7? Doesn't this seem a odd to you?

-Steve

Steve,

Odd in what way? Here is the application report the data was gathered for: www.ti.com/.../slva720.pdf

Referring to SLVA720, Figure 4, I digitized this graph using Engauge Digitizer and precisely measured where the R = 1K curve intersects Cext = 100 pF.

Since Tpw = R*C*K(Vdd, C),

K(Vdd, C) = Tpw / (R * C) for a given Vdd graph.

At this point in Figure 4, K(5V, 100 pF) = 247 ns / (1K * 100 pF) = 2.47

Of course, this is at Vdd = 5V, and I'm looking for Vdd = 3.3V.

Referring back to SLVA720 Figure 5, we see that as VDD drops from 5V down to 3.3V, K increases by about 5% in the cases shown (notice the 5% per division on the vertical scale).

So, I would expect K(3.3V, 100 pF) to be on the order of 2.47 * 105% = 2.6. Which, by the way, is far off from the ideal K = 1 value, and far far off the the chart of Figure 5.

Your suggestion that K = 5.7 is more than double the above, so that doesn't seem right to me.

Perhaps what this is revealing is that C = 100 pF might not be a very good choice. I can increase C to get a better K, but will need to use a smaller resistor to achieve the same KRC product.

If I use K = 1.12 for Vdd = 3.3V, and C = 1000 pF (per Figure 5 of SLVA720) to achieve a 125 ns pulse requires R = 111 Ohms.

What happens when R < 1K, like when it gets down to 111 Ohms? Anything else to go haywire?

-Steve
(PS, I'm a former TI FAE :-) )

Per the data sheet, the recommended max current source/sink at 3V is +/- 24 mA. 3.3V/24 mA = 137 Ohms. That is substantially less than 1K, but of course its probably not a good idea to operate the part at it's recommended max.

I suspect what happens is that there is finite time required to trigger the one-shot (i.e. reset the capacitor), and then there is some fixed propagation time associated with the internal comparator. These fixed time delays are not well characterized by the simple t = K R C formula. For longer delays, these non-ideal artifacts wouldn't matter. For short pulses, they may become a large portion of the overall time delay.

If you're not going to recommend using a resistor less than 1K for the reasons you state, than I need some better data for C = 100 pF, or at least a recommended combination of R and C to achieve a 125 ns pulse.

Unfortunately, I need something a little more scientific than some simple bench testing. Propagation delays will vary over process, so unless you know you have a variety of process spread in your bench sample, it may be of limited use.