On Mon, 14 Jul 2025 11:23:05 -0700
Tom Verbeure via time-nuts time-nuts@lists.febo.com wrote:

I don't have a 5370A. This could all be bad advice:

It sounds like that could be rifa caps, which are common smoke makers
"after plugging it in" on a lot of devices that were made in a year
that starts with 19. They're usually relatively easy to replace with
something that won't smoke out on regular mains voltage if thats the
case - once you get past the burnt particulate...

Tom

Dont give up on it

The smoke is probably just the mains input filter (Rifa Capacitors)

In some designs these can easily be removed or just change the mains
input socket if they are integrated into the socket.

 HP instruments often have a complex mains socket with configurable
outputs to the transformer taps for different input voltages. The casing
can be opened with care (desoldering the screening) and the caps can
then be removed

The instrument will probably be unaffected  by the fault as the
capacitors fail open cct

The lingering smell will go away fast once they are removed.

On 14/07/2025 19:23, Tom Verbeure via time-nuts wrote:

--
This email has been checked for viruses by AVG antivirus software.
www.avg.com

Hi

There are now a range of TDC chips out. They come from a lot of sources. It tends to be
a “you get what you pay for” sort of thing. If your budget can fit something in the $20 to $30
range, you are into the same sort of timing resolution as the 5370. You hit the “darn close”
range before you get to $10.

Cost tends to make these the “go to” answer for designs these days. There are a number of
“home made” counters out there that use these chips. The TAPR TICC is one famous example:

https://tapr.org/product/tapr-ticc/

There is a lot more to a full up bench instrument like the 5370 than just the TDC section. You
very much do not get “all of that” for your $30.

Bob

On Tue, Jul 15, 2025 at 10:24 AM Mark Hughes via time-nuts <
time-nuts@lists.febo.com> wrote:

Definitely not! :-)

Yes, the mains sockets are usually from Comcor. I haven't checked on the
5370A yet, but I replaced one on my HP 8656A with a new one that's form
factor compatible but the contacts are in the back instead of in a cut-out,
so there's much less room for the wire. On the 8656A, I solved that by
removing all the multi-main voltage options and just hard-wiring the
transformer for 120V.

Good! I was hoping that this would be the case.

2 days later, and it's lingering still in my throat. But it has happened
before...

Tom

For those who do not know that RIFA caps are: these are X, Y safety caps, that run at mains line voltage and are used to filter out EMI entering/exiting via the mains feed. Time makes them degrade, they develop cracks in their epoxy case. A some point they tend to blow, and/or catch fire. Given that they are sometimes on the mains inlet, so before the mains power switch, they can do that even when the equipment is switched off.

They look like this:

Get rid of them, all of them. Burning the house down is not worth it.

Wilko

Hi,

On 2025-07-14 20:23, Tom Verbeure via time-nuts wrote:

Service powersupply. Should be relatively trivial.

I developed a mod to reduce the noise on the reference 10 MHz, as the
signal detect circuit generates a spurious 5 MHz. It's in the archives,
but I just grounded a base on the diff-pair for the cursuit with a
simple solder joint.

I would say that the particular vernier method of the 5370 is not widely
used. Pulse-widening has been used in some other counters, but not to
the same resolution. The 5371, 5372 and 5373 uses a variant of the
technique where the difference of delay of clock and trigger feeds a
bunch of D-flip-flops (DFF). This technique was there used to achieve
200 ps resolution or a 10 fold increate of resolution from the coarse
clock of 500 MHz / 2 ns. This delay-difference technique have since been
further developed in the FPGA and ASIC field, and is similar to the
Vernier approach used for interpolation in the 5370, but you do not need
to wait for 256 cycles and the single detection, instead this is done in
parallel with a lot of DFFs which also have the benefit of significantly
higher timestamping rates. In the case of 5371/5372 you can operate at
13.33 MHz rate (75 ns).

The ADC technique took over, offering higher resolution and improved
linearity. SR620, CNT80/81 and CNT90/91 are example of these, but the
Wavecrest DTS 2070C/2079 brings it further. Wavecrest pushed it to 200
fs resolution, even if trigger jitter is worse.

Further down the line, the oversampling ADC technique of Sam Stein that
is the Symmetricom line of measurement instruments, later the TimePod
and PhaseStation pushes this even further, avoiding aliasing problems
and immense noise removal. These may not have the classical TIC feel,
but we get that type of data out of it for ADEV and friends processing
and the noise floor is way down.

You should consider that to push performance with the technology at hand
then, custom hybrids was the way to go. Not the same need later, and we
even have different approaches.

The Vernier approach used is very smart for the technology at hand, but
it has been outperformed by other technologies. Then again, it would be
fun to experiment with tuneable delay and see what can be done. :)

Cheers,
Magnus

On Tue, Jul 15, 2025 at 5:00 PM Magnus Danielson via time-nuts <
time-nuts@lists.febo.com> wrote:

I developed a mod to reduce the noise on the reference 10 MHz, as the

Do you happen to remember which thread that was? The closests post I could
find was this one:
https://www.febo.com/pipermail/time-nuts/2015-May/091973.html.

I'll put that method on the list of things to study. :-)

The special PLL is the real interesting part IMO.

Thanks!
Tom

I removed all the boards except the power supply controller and the 15V was now present. Unfortunately…

I then reinserted boards one by one to find the one that’s pulling the 15V down: it’s the 10MHz detector board that sits next to the OCXO power supply board. All other boards are plugged in.

I would expect that with only this board removed, the CPU should be up and running and maybe displaying an error message? But that’s not the case: except for a 4 status LEDs, the display stays dark.

Does anyone know if that’s expected?

The 10 MHz detection board is not connected to the CPU bus, so the lack of it shouldn’t result in a bus hang.

Tom

Assuming you are talking of the "A8  Reference Frequency Buffer
Assembly": The clocking of the CPU is generated from the 10MHz
Reference. So it is normal that without taht board the CPU is not working.