Re: TG501 enigma :)


Sergey Kubushyn
 

On Thu, 24 Dec 2015, David davidwhess@... [TekScopes] wrote:

How can you align 20ns period pulses with 50ns ones? They are different
frequencies not even evenly divisible by each other.

You have a "SET" (or reset that makes no difference) signal every 50ns. This
is applied to a pulse train with 20ns period. That means that reset signal
comes after 2 20ns pulses, 10 ns after the second pulse. What you have in
result is a constant stream of 20-20-10ns pulse sequences that is hardly
something that could've been called 20ns time mark signal... You can clearly
see it with a scope. And this is _ONLY_ present on 20ns range; none of the
others have this weirdness.

TG501 generates a _CONSTANT_ pulse train, not some sequences so it makes no
difference when each particular pulse train started.

Then, there is a switch that selects which of many available frequencies is
connected to the output. Only one can be selected at a time, all others are
just not connected to anything.

All pulse trains are derived from the same 100MHz VCO so they are all
synchronized to that base frequency but that is just a consequence of
deriving everything from the same base signal that has no practical purpose.
Those signals can synchronized or not -- that doesn't matter. Only one of
those signals is used at a time so it doesn't matter what other signals are
and if they exist at all.

100MHz VCO can be either locked to a _SEPARATE_ reference clock (coming from
a separate reference crystal oscillator, either a simple or ovenized one
depending on Option 1 presence) when in "calibrated" state or free-running
with a frequency changed by a voltage from a potentiometer when in
"variable" state. It allows for something like +/-8% tuning from the center
100MHz frequency. All subdivided pulse trains are changed the same
percentage because they are just a VCO signal divided by a constant.

Faster signals (5-2-1ns) are also derived from the VCO signal using a snap
diode for frequency multiplication followed by tuned bandpass filters. As
those are tuned variable frequency can not be implemented so the VCO is
forced locked to the reference when in 5-2-1 range i.e. variable timing is
disabled.

I cut that trace on all 3 TG501s and now they all make perfect regular time
marks on all ranges. I'm still puzzled what was the original intent on
crippling that particular 20ns range pulse train.

I looked at it quickly and I think it was done to keep 20ns divided
output phase aligned with the 50ns divided output. The other outputs
are also phase aligned to the 50ns output through Q330.

Otherwise depending on the state of the divider for the 50ns divider,
the 20ns output could be misaligned with the trigger.

On Thu, 24 Dec 2015 16:13:39 -0800 (PST), you wrote:

Hi everyone!

I'm cleaning up my collection of PG506 and multiple TG501 and SG503s
keeping one each for myself and readying the remainder for selling on EPay
(don't want to sell those in "as-is, don't know if it works" state so
checking/fixing/calibrating everything to offer them as "guaranteed working
and calibrated".)

While working on TG501 I found something strange that I have no explanation
for. _ALL_ of my 3 TG501s were giving crappy output at 20ns range (multiple
overlapped pulses on a scope) so I decided to investigate and fix it. And
here is where something weird showed up.

10/20/50ns range pulses come from 100MHz VCO. It is fed straight to output
for 10ns and through /5 and /2 dividers for 50ns and 20ns ranges. 10ns and
50ns ranges are clean and nice so VCO works fine. It is only 20ns that is
crappy so the trick is definitely in that /2 divider. And it is indeed :)

If you look at schematic sheet 2 you will see that /5 divider at the left
top is nothing special, made of 3 ECL flip-flops (U310 and half of U315) and
it works just fine.

If you look at the /2 divider below the /5 one you would notice something
very weird. A /2 divider is just a flip-flop so it should've been
straightforward. And indeed it is a flip-flop made of the remaining half of
U315. However there is very strange fragment made of U290A and U290B parts
attached to it. It generates a short pulse using a propagation delay in
U290A every 50ns and that pulse is used to forcefully set the U315B
flip-flop every 50ns. 50ns is not an even multiple of 20ns so that is why
U315B output is not a series of pulses with 20ns period.

And that is not something that just happened or had been overlooked; it is
done _DELIBERATELY_ by design. Furthermore, it had been modified in newer
instruments so there are 2 versions of U290A connection on the schematic,
one for S/N below B036900 another for S/N B036900 and above. For later units
it is even crappier because it adds /10 output to the mix...

Cutting a trace at U315 pin 12 predictably makes 20ns pulses nice and clean
like all other ranges. That trace is on the LF board solder side so it is
easily accesible without taking the entire thing apart thus making it an
easy fix.

The question still remains -- what was an intended purpose for crippling
that particular 20ns range pulse train? It is 100% sure that had been done
deliberately and even "enhanced" in later versions of TG501 plugins.

Does anyone has some idea why?
---
******************************************************************
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
******************************************************************

Join TekScopes@groups.io to automatically receive all group messages.