[time-nuts] PDIP package 100 MHz decade dividers

[Magnus Danielson]

Hi Rick,

On 2020-07-19 04:30, Richard (Rick) Karlquist wrote:
>
> Yes, the simple pair of resistors doesn't work well on TTL.
> That was probably done by a production engineer, who was
> working above his pay grade.  We can't all be Len Cutler.
Sounds very reasonable. It's a crude hack and works OKish, but to fix it
not much would be needed to be done.
>
> OTOH, that circuit works well with 74ACXX circuits.  We did that
> at 80 MHz in the 5071A and it worked well.

OK, that can be useful to know. I should probably do some tests then.
Maybe replacing the 7474 with a 74AC74 will make it perform better. It's
simple and straight-forward enough, but I would like to measure and
compare to know. Maybe someone already done that. The problem is to
lower the added noise enough, and slew-rate limited signals into a
trigger point is a know source, then the gate itself can contribute
naturally.

Anyway, thanks for that little insight.

>
> Back in the 70's when I worked for Zeta Labs, my boss taught
> me to use a 1 transistor buffer that had a pull up resistor
> on the collector and a resistor from the collector to the
> base and a resistor from the base to ground.  We used it
> in all our designs and it worked well, despite being
> ridiculously simple.  My boss was really smart.
>
> That is what they should have used in the 5065.

Sure. It's not very hard to do. The 5065A synthesizer input have the
same challenge, so it has higher noise because of that, but if one has a
00105 oscillator that will cover that up anyway. The 5065A synthesizer
isn't very quite for sure, and just using a 3325B is replacement I
dropped the ADEV floor by over a factor of 2 on a 00105 based 5065A,
because the 3325B lock-up filtered much of the 00105 noise out and did
not have as terrible sidebands and input treatment. That said, the 3325B
isn't particularly "clean" but sufficiently clean to achieve that
improvement. Sometimes "sufficently" is the key word, and applying it
one can get cheap fixes that moves things out of the critical path for
performance. Sometimes all it takes is a 2N3904 or two with a few
resistors and caps. Figuring out where the key performance limiters are
and address those sufficiently well may achieve most of the gain at times.

Cheers,
Magnus