On 6/29/20 10:41 AM, ed breya wrote:

Couldn't find the AC390 - did find HC390, but it only goes to 35 MHz (at
room temp, less over temp)

However, you got me started hunting for decade counters, and there's
things like the 74HC4017 which claims fmax=60 MHz on the data sheet
front page, but only claims 35 MHz down in the AC specs

You might try the 74AC161, which works to 73MHz at 3.3V or 103 MHz at 5V,
-40 to 85C.

Set the data inputs to DCBA = 1011 and connect an inverter from the carry
output (pin 15) to the Load input (pin 9) to divide by 5. See
http://www.techlib.com/electronics/74161Divider.htm

On Mon, Jun 29, 2020 at 12:49 PM jimlux jimlux@earthlink.net wrote:

--

--Jim Harman

Am 29.06.20 um 18:43 schrieb jimlux:

In the picture is probably what you need, and maybe more.
The left third is a comparator that generates valid CMOS levels from a
vaguely defined sine signal.
The right third determines if there is a valid reference and tells the
PLL (not in the picture) to use
or ignore it.
If you have a valid CMOS signal, the middle is all that is needed.
LVC163 + LVC04. Good enough for
150 MHz+.  On terminal count, the inverter forces the counter to load
the P0..3 pins on the next
clock.  The value on P0..3 determines the division ratio, from 2 to 16.
Large numbers  = few clocks
until the next terminal count. The example is divide by 5, which happens
to fit your problem.

IIRC, the 74AC191 could have done that without the external inverter,
but it did not make it
into the 74LVC series.

That's what I used to lock my DG8SAQ vector network analyzer V2+  to an
external 10 MHz reference
before  that was available in V3. The PLL chip was a 74lvc4046.

https://www.digikey.de/product-detail/de/nexperia-usa-inc/74LVC163PW-118/1727-3097-2-ND/946683

Cheers, Gerhard

To divide by 5 with a '161/'163 counter, connect
the 8's bit output to the /preset enable input.
Then set the input bits to 12.  The counter will
count:  12, 13, 14, 15, 0, 12, 13, 14, 15, 0 ...
This is the fastest configuration.  It avoids
external gate delay and the slower carry output.
You can only divide by up to 9 this way.

I have been doing this for at least 40 years with
various logic families du jour.

Rick N6RK

On 6/29/2020 3:32 PM, ed breya wrote:

Ed,

For division, there's less need for a dedicated divide-by-10 counter
since the '161 and '163 are presettable synchronous binary counters.
As such you can wire them to divide by anything from 2 to 16, which
includes 10. In addition they are cascadable, which means that you can
create synchronous 8- or 12- or 16- or 20-bit or even larger dividers.
The datasheets for these counters sometimes show examples of multi-chip
dividers.

Attached is a photo of a 'LS00 plus 3 'LS161 counters configured as
divide-by-1173 counter. The 2nd photo is a divide-by-3295 count. You can
see it's just a matter of changing the jumpers on the preset pins.

/tvb

On 6/29/2020 3:32 PM, ed breya wrote:

On 7/1/20 1:41 PM, ed breya wrote:

That's sort of the design goal for the 22V10 and earlier PAL devices -
keep them in familiar DIP packages, power on the corners like the IC
gods intended, and you can program it to replicate a whole variety of
MSI functionality, often with the same pinout.

Am 02.07.20 um 00:35 schrieb jimlux:

16 bore holes just to deploy 4 flip flops is not what I'd call
simplicity. And Fairchild
recognized that as well. The 390 did not make it into the 1987 Fairchild
Advanced CMOS
(FACT) data book.
They expected other chips to sell better, like the 74ACT488 GPIB / HPIB
/ IEEE488 bus
interface. Ever used one? The market for Nixie clocks with one counter +
one decoder
per digit must have been smaller, even back then.

Corner pinning for Vcc and GND is not what any gods intended. In FACT,
(pun intended)
it's evil. Remember ground bounce? The corners are the worst locations
on a DIL chip
you can find for that job. And fig leaf capacitors across the chip are
just that.
At 100 MHz, they simply are not there. The optical illusion helps only
to hide that.
The 3  ACT chips on the experiment board posted yesterday reminded me of
the
last pages of this:

<
http://www.hoffmann-hochfrequenz.de/downloads/experiments_with_decoupling_capacitors.pdf

The golden times of logic design are now, not then!

Gerhard

On 7/1/20 11:21 PM, Gerhard Hoffmann wrote:

Certainly, but when wire wrapping that big panel, and using a different
color of wire for the V+ and V-, it sure makes it easier. <grin>

I'm surprised nobody has suggested using the 12AX7 or 6J6 dual triodes.

Jim,

Funny, just yesterday I was looking at the design of a laboratory cesium
beam standard from 1963. Sorry, there's no divide-by-5 example in there.
But the attached images show the 108x multiplier (8.5 MHz to 9180 MHz).
Sure enough, spot the 12AX7 and 6J6 tubes in use...

/tvb

Hi

For vacuum tube divide by 5 / 10 circuits, take a look at the schematics of the
Beckman EPUT meters…..

Bob

That's a very early use of "MHz" rather than "Mc"!

On Jul 2, 2020, 9:50 AM, at 9:50 AM, Tom Van Baak tvb@leapsecond.com wrote:

Makes me think one could use the input signal edge to synchronize a 2N6027 based programmable unijunction oscillator, thus effecting a divide by 5. Unlikely the 6027 would be fast enough for 50 MHz, but maybe a 2 transistor equivalent using RF transistors? Output would be nowhere near the 40%-60% range, though. Might just have to build one for fun.

...

On Thu, 2 Jul 2020 19:23:03 +0200, you wrote:

And that makes me think of an early Navy airborne Loran set that
used step counters to make the time base.  Adjustments were through
holes in the side of the case, because some poor soul would have to
tweak them now and again.

Gary Woods O- K2AHC  Public keys at home.earthlink.net/~garygarlic, or get 0x1D64A93D via keyserver
fingerprint =  E2 6F 50 93 7B C7 F3 CA  1F 8B 3C C0 B0 28 68

--
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

On 7/2/20 6:48 AM, Tom Van Baak wrote:

Yeah, but I think those are multipliers, not dividers.

So then, what shall we call it. PICDIV draws from the PIC.  Next thing
from TAPR,  catering to the retropunk style, will be a TUBEDIV or EJDIV
(Eccles-Jordan homage) - what better to drive your nixie (or stacked
neon bulb) display than a 19" rack full of modules.

I'm sure if one looks at pictures from WW 2 era, there's plenty of this
around.

If I didn't have many other projects going, that might be a cool one.
Lots of folks have binary clocks on their desks at work, but they're LED
and solid state.  A few have Nixie clocks, but solid state driven.

Another way to achieve divide-by-N is with a non-retriggerable one-shot,
adjusted to the appropriate time value.
Back in the 40's/50's, the common tube circuit was called a Phantastron
(really, look it up!).
Phantastron dividers were used in several of the early HP counters (i.e.
HP524B), because you could achieve (for example) a decade divider with a
single tube, whereas a 'binary' solution requires 4 tubes (like in the HP
AC-4A decades).

If a 50% duty cycle is needed, then do a divide by N/2 with a Phantastron,
then a single flip-flop gives you 'N' with a 50% duty.

Note that the Phantastron type divider only works if the input freq is
known and fixed, because the one-shot delay has to be adjusted to a
specific value.

Pete

On Thu, Jul 2, 2020 at 2:29 PM Gary Woods garygarlic@earthlink.net wrote:

On 7/2/20 11:37 AM, ed breya wrote:

You forgot to add rad-hard.. I was the OP - This has been a fascinating
thing - we have a breadboard that uses a fancy clock distribution chip
that consumes close to a watt (and has too much jitter, as well)..

I was thinking on the long commute to (tele)work from downstairs to
upstairs (or maybe on the commute home going the other way).. You know,
we don't need all that functionality in the clock chip, there must be a
way to do a divide by 5, and that has good noise properties - 2 and 4
are trivially easy (I thought), there's probably some easy way to hook
up 3 (or 4) flipflops and get a nice divide by 5, and maybe even 50/50
duty cycle out.

So I posted the question - because I've seen discussion of good divider
designs here, and I was sure that someone would come with a novel
suggestion.

What I have learned is that

1. All those clever handbook designs and data sheets that I grew up with
   in the 70s,80s, and 90s are just the ticket, but you can't actually get
   the SSI MSI parts any more.

2. One can brute force design simple functions by just trying all
   possible connections and see if it works. What a clever idea!  All that
   work with Karnaugh maps, etc. trying to come up with minimalist designs,
   and you can let your idiot savant assistant (the computer) just grind
   through all possible designs.  Of course, now that you have that clever
   efficient design, because of #1, you can't build it.

3. It's hard to even find programmable logic that is simple and small.
   All the mfrs tout their latest tiny parts with only half a million
   gates (I exaggerate, but you get the picture)

4. What I'd be happy to do for a room temperature breadboard probably
   won't work over temperature - and an "existence proof" that it can be
   done at room temp doesn't mean you can find parts to build it to work
   over temp (See #1, again)

We are building the eventual system to fly, and yeah, we've got tubes
and tubes of old ICs at work (JPL) or, as Rick mentioned, there's always
Rochester Electronics, who have a warehouse full of old wafers and dice.
But I'd rather not.

For those who come after me, and are perusing the archives (thank you
google) - here's some reasons why old parts are a pain (and a curse).

1. Reliability people freak out about packages that have not been kept
   in absolutely pristine conditions with a full paperwork trail of
   certifications. The humidity might have gone up. Oxygen or Helium might
   have leaked in. There might have been latent ESD damage.  So you'd have
   a tube of parts with date codes from the 80s or 90s that work (over
   temp, etc.), but the mission assurance folks will want a bunch of them
   to do destructive analysis. Making sure there's no latent degradation,
   etc.  That can cost a lot.

2. You CAN get parts from Rochester, and they're freshly packaged, from
   known good dice, etc. That's not cheap either.  But, is probably cheaper
   than #1.

3. the biggest reason - There are innumerable cases where someone used
   "end of life" or "flight spare" parts, just this once.  And then, the
   next mission comes along and wants to do a "build to print" to claim
   heritage - and you spend a lot of time trying to track down NOS parts,
   or in design reviews trying to say "well even though all the parts are
   different, except the resistors, it really isn't a new design".

We used the Xilinx Virtex II in several of our radios that have flown to
Mars (in the Electra UHF radio on all the rovers since 2003, for
instance).  I think we have the largest stock of flight qualified
Virtex IIs in the world, because people still want to use Electra
radios, as a "build to print".  This is not healthy.  One has to have
development systems to test software, one has to have spare units, etc.
All consuming those few remaining XQR2V3000's in the world. And one has
to have the tool chain as well (Long since obsoleted by Xilinx, and
doesn't run on any "new" versions of Windows).

As a project manager I get the desire to use something that is known -
you know what it will cost, you have the procedures, there's very little
uncertainty in the cost and schedule, which is what I care about as PM.

As a design engineer, I despair of having to support a design that is 20
years old. I never know whether the next one is the one that fails in
test or otherwise not work, and there's no "fix-it" parts, and it's
going to be my job to tell the management - uhh, you're gonna need a
different radio.

Back to divide by 5 - If I can offer a suggestion to the design team,
they can run with it, look for parts that are reasonably available, and
feel comfortable that when somebody says "let's do a build to print" in
10 years, they'll still be able to get the parts.

Now, for myself, making something with vacuum tubes is something I've
not done since the early 70s. I am intrigued.. I grew up in the 60s
reading Millman and Taub (the big yellow book) as a child (EE professor
as a parent), I read about all the vacuum tube circuits, but what I
built were transistors. My parents were no fools and not about to let me
work with a B+ supply of 100V.  Those 2N404s and 2N1613s were what I
used. And some early UJT, but I can't remember the number.  And the RTL
digital logic parts like the uL914 dual 2 input nand.

But I never went back to the vacuum tube designs.. Ah, one of those
things for retirement, perhaps.

Hi

If they want to sell you a fully self contained "million gate" device for a couple bucks, is that really
a bad thing?  Sure, if it’s in a thousand pin BGA, it’s a bad thing. If it’s in a < 40 pin package that
you can get on a small board …. maybe not so much.

Bob

On 7/2/20 2:13 PM, Peter McCollum wrote:

And you can be clever if you can trigger on both rising and falling
edges, so you get basically a x2 and a divide by 5 in the same circuit,
followed by the divide by 2 to make the duty cycle 50/50.

On 7/2/20 2:50 PM, Bob kb8tq wrote:

yes, if it's in a small pinout package. One other peculiarity that I've
been burned by is that a lot of the modern devices with large logic and
few I/O pins have power dissipations that are clock rate independent
for the internal logic - it's the leakage current that dominates and
that's VERY dependent on die temp.

I think though, that the marketplace is driving towards increased
functionality on one chip, with bigger die size. Those of us who want 50
gates at medium or low speed are distinctly in the minority.

Hi

There are very few applications that require that sort of device. Volume
matters ( a lot !!)

Bob

For fixed frequency operation there's always Wenzel's divider using a D FF with LC feedback:
http://www.wenzel.com/wp-content/uploads/dividers.pdf
At least the power consumption is low.

Bruce