Tom
I have not done this. But you're on the right track with differential
inputs.
The approach I would take is to measure each meter position without the
meter in circuit to ground.
This will give you the maximum differential voltage to watch out for with
respect to ground that might damage the diff opamp input.
Pretty sure all of the test points are easily reachable on the bottom pc
board A7 or 17??
Then figure out the meter resistance and use that value across each meter
function connected to the differential input.
This all sounds messy but really isn't. Just leave the meter in the battery
position or maybe the 100 KHz position.
Good luck sounds like a fun project.
Regards
Paul
WB8TSL

On Wed, Jan 4, 2023 at 10:20 AM Tom Van Baak via time-nuts <
time-nuts@lists.febo.com> wrote:

Hi Tom,

While I have not been looking at 5061A/B in particular, this do touches
on an idea I've been having.

Several of the analog rubidiums and cesiums have the same thing, a
rotary switch and things you can monitor. However, I've found that even
a very simple monitoring gives usually a great insight. For
Oscilloquartz cesiums, all the various voltages is available on a DB25
connector, so a project would be to use some small suitable processor,
scan voltage and log them similar to what we have from say FTS-4040/4065
etc. Many other rubidiums and cesiums could probably do the same thing.
A certain amount of signal-conversion in analog is needed to scale it
in. The EFOS A/B masers do the same thing really.

Being able to dump things into InfluxDB/Grafana help a lot, including
storing data and post-process them at need.

So, I think that once a suitable core code for such a thing is done,
doing the necessary analog adaptations for the various boxes and
document the offset/scaling factors to "real units" should make the
"catch script" relatively easy to adapt and we could lower the
threash-hold and crowd-develop it as we go. I would love to modify all
my running analog clocks to have this. None of my HP506[0-2] has cesium
thought.

Cheers,
Magnus

On 2023-01-04 16:17, Tom Van Baak via time-nuts wrote:

Having dealt with this on the 5065A recently (
https://www.eevblog.com/forum/metrology/the-hp5065a-and-its-ted-some-info!/m
sg4605040/#msg4605040 ) I'm sure the 5061B can be instrumented the same way.

The meter is a 2000-ohm load, 50-0-50 uA full scale.  So a reading of +/- 50
corresponds to +/- 100 mV at the terminals.  The circuit treats the measured
quantities as voltage sources with series resistances of a few thousand ohms
at most, using the rotary switch to connect one at a time to the meter
through additional series resistances on the A7 card.  These are
factory-selectable parts on the order of 100k.

So my strategy was to make the ADC behave like the meter, acting as a
low-resistance load that could be driven through high resistances to avoid
overdriving the ADC or loading down the various sampling points.  Because
the ADCs measure voltage and not current, it made sense to use a higher load
resistance than 2k in order to develop more voltage across it so as not to
waste too many bits.  I used 8.2k to ground on the Arduino analog input
pins, bypassed by 4.7 uF tantalum caps.  For series resistors, I used values
similar to those on the metering card, 220k on A7-6 for the photocell
current and 100k on A7-13 for the 2nd harmonic.

The code used to emulate the meter readings looks like this:

float ADC_to_2nd_harmonic(void)

{

//

// OEM meter = 2 mV/uA @ 1940 ohms, FS=+/- 100 mV) driven via 100K, so FS
@ 50 uA is 5.1V @ A7-13

// ADC load = 8.2K driven via 100K, so 5.1V yields 0.387V at input or
359/1023 on 1.1V scale

//

S16        A    = analogRead(2);

const float scale = 50.0F / 359.0F;

return (float) A * scale;

}

float ADC_to_photo_I(void)

{

//

// OEM meter = 2 mV/uA @ 1940 ohms, FS=+/- 100 mV) driven via 200K, so FS
@ 50 uA is 10.1V @ A7-6

// ADC load = 8.2K driven via 220K, so 10.1V yields 0.363V at input or
338/1023 on 1.1V scale

//

S16        A    = analogRead(3);

const float scale = 50.0F / 338.0F;

return (float) A * scale;

}

100k thermistors are connected from 3.3V to two more ADC input pins, each
with 2.7K resistors to ground:

float ADC_to_temperature(S16 A)  // Tracks Fluke 51 to within +/- 2C from
0-100C

{

const float Vref  = 1.1F;    // Full-scale Arduino ADC voltage

const float Vccth = 3.3F;    // Typical 3.3V rail voltage at thermistor
divider

const float Rdiv  = 2.7E3;    // Ground leg of voltage divider

const float Rt1  = 100E3;    // Thermistor nominal resistance at T1

const float T1    = 298.15;  // 25C in K

const float B    = 3950.0;  // B-constant for Vishay NTCS0402E3104JHT,
25C to 85C

float VADC = (float) A * Vref / 1023.0F;

float Vth = Vccth - VADC;

float Ith = VADC / Rdiv;

float Rt2 = Vth / Ith;

return (1.0F / ((1.0F / T1) - (log(Rt1 / Rt2) / B))) - 273.15F;

}

These are used to read the temperatures at both ends of the physics package,
primarily to serve as overtemp protection but also good for (very)
approximate datalogging purposes.

This worked very well, although because the source resistances aren't zero,
there is some slight interaction between the rotary switch and the ADCs.  If
you switch the meter across the lines being monitored  you'll get a small
shift in the recorded levels (and of course the meter readings will be
slightly lower as well.)  Fixing this would require adding voltage
followers, so it's easier to just leave the switch parked at an unmonitored
setting while recording data.

-- john

Ugh, all the markup got stripped.  Here's a shorter version of the link for
those who haven't seen it (I know tvb and jra have):

https://tinyurl.com/4t3vkb34

-- john