Standards Lab Measurement Dilemma #1

Despite all of the details you've given, it is a little difficult to say for certain what and why this ground braid made a difference without at least a schematic or block diagram of the circuit. I suspect that you've rediscovered what I call the myth of ground. For most circuits the slight impedance of the wire connecting different components together produces errors in measurements that are so small that they easily get ignored or remain undetected. But when one measures any effect over many orders of magnitude of frequency changes and to single parts per million resolution, ground at one frequency may not be ground at another. Now, here's a possibility why strategically replacing a part of your circuit wire with a braid gives you now acceptable responses. Likely this has to do with the difference in the impedance of the wire versus the braid. First the braid likely has a larger cross-sectional area than our original wire. So the DC resistance of the braid will be lower than the wire. With AC signals this improvement can be more dramatic. For the skin effect reduces the effective cross-sectional area of a circular conductor the most. A flat braid of the same cross-section of a circular wire may not have enough thickness for any skin effect to happen at 1MHz.

But without knowing precisely your circuit configuration, where the braid gets placed to improve the performance or being very familiar with any of the test equipment; I can only guess.

Obviously (?) there's a 500-kHz signal around somewhere* (I always said that the laboratory was the worst possible place to make measurements). Ground coupling can of course depend on the routing as well as the existence of a connection - or if this is cancelling a capacitively coupled source the impedance can be critical; for my own work I generally use balanced screened pairs - and the two screens of the four-wire connections closely coupled along their lengths, but even this is not foolproof at the sub-microVolt level if the IOs are not themselves screened.

(I had a colleague who routinely disconnected ground leads so he could have a single ground path along the same routing as the signals. This works very well once properly connected, but he "lost" quite a few static-sensitive mixers... and in the end his technician invested in a pair of insulating gloves

*Could it be a DC-DC converter in the Fluke?)

Hi Physicist?,

No, there is no 500kHz signal anywhere around, unless it is inside the Fluke. The problem was not just 500kHz, but was worse there than 1MHz or 300kHz. There is only 1 DC ground in the system (the tee case which grounds the TVC inputs). The nano-volt meter inputs are floating. All equipment cases are grounded through power cords and GPIB buss cables.

"I always said that the laboratory was the worst possible place to make measurements"

This lab was designed to make measurements. It is a metal room inside a metal building. Both metal walls are grounded through a ground rod. I had the room tested up to 3GHz for radiated power. Nothing of significance was found.

-S

a) The 50-400 Hz power-input range almost implies that the Fluke uses a DC-DC converter. But this should be quiescent when you are running from the internal batteries as recommended for precision measurements. There could be other reasons for internal RF generation, however (chopper amplifiers?).

b) "the inputs are floating" The signals coming to those inputs should be ground-referred, as should the case of the sensor. Ideally, there will be effectively a single path for this reference, and this should ideally be via a low-leakage screen surrounding the signals. The norm would be to route the screen from the TEE to the other equipment. Any other (e.g. mains) ground routes will of course degrade the situation. Short-and-low_impedance runs for the wanted routes help but can't completely remove issues.

c) Screened rooms are intrinsically good for radiated interference. Conducted interference is another matter, and has become substantially more problematical since computer-control was introduced. Maybe fibre-optic interfaces will eventually rescue this aspect?

d) The first task I give a measurement trainee* is to perform measurements in a noisy environment and to achieve a predefined level of interference. Once the techniques for this have been assimilated they are ready to make the most of the screened room.
*Typically, these have previously been power-supply designers, partly availability, and partly because they are supposed to have detailed knowledge of the issues - so they are ready for what can be quite a steep learning curve.

"a) The 50-400 Hz power-input range almost implies that the Fluke uses a DC-DC converter. But this should be quiescent when you are running from the internal batteries as recommended for precision measurements. There could be other reasons for internal RF generation, however (chopper amplifiers?)."

You must be very confused. The Fluke 5720A is a multimeter calibrator that I am using as a voltage source. It's power input frequency is 47 to 63 Hz. It has no internal batteries. Neither do the Agilent 34420A nano volt meters or the Datron 4700. So what if there was a DC-DC converter? Choppers haven't been used in any new designs since about 1970. So, do you suspect, as I do, that RF spurious signals are the culprit?

"c) Screened rooms are intrinsically good for radiated interference. Conducted interference is another matter, and has become substantially more problematical since computer-control was introduced. Maybe fibre-optic interfaces will eventually rescue this aspect?"

The room I am in is not exactly a screened (shielded?) room, but is next to one. Moving this station in there would mean moving something else out. It's not likely to happen, and may not be a solution anyway. Until I can prove what the problem is there is little hope. I can only wish for fibre-optic adapters. That is not likely either.

Please go to dilemma #2 in electrical engineering. I may not respond here anymore.

-S

I was assuming that you were running the A55 from a 540B, which is where the frequency range and the battery came from.

But yes, I too believe that spurious RF signals are the most likely culprit. Without going into your facility I have no way of telling whether these originate within your equipments or from a friendly(?) neighbour

Are all of the standards plugged into the same circuit? You may have to consider floating the voltmeters, hence using the signal (output from 5720) ground. A couple of ground busters could help. I've seen ground loop conditions caused by (supposedly) high input impedance devices (voltmeters) before. Are any of the cables in your setup shielded?

Hopefully this helps.

MRH

Hi MRH620, The only standard is one of the TVCs on the tee. The TVCs are passive. The nano-voltmeters and the calibrators are on the same power strip along with other equipment that is not used for this test. The computer, monitor, and printer are on another power strip, but use the same breaker. "A couple of ground busters could help" If you mean "nema plugs" with the safety ground open, they will not give isolation because of the GPIB cables which also ground each units case. However the voltmeters are floating already, but maybe not to an RF "spur" because of capacitance. The ground braid has solved the problem. I am just trying to understand why. But dilemma # 2 may be related, so I may try to find some unshielded GPIB cables to use for that problem. -S

I am thinking that the difference between the wire lead and the braid used to ground the TEE is due to their inductance. Flat braids have lower inductance than round conductors and of course shorter lengths also have less inductance. I would run a few experiments varying the lengths of the ground leads, their size, routing, etc. to characterize the differences. You may find that the frequency of where the measurement error occurs shifts as the ground wire placing is moved or otherwise varied.

The first time I tested coaxial cables on a network analyzer, I was absolutely amazed at how 50 ohm cables from different manufactures behaved so differently. And when simple bends were introduced into the cables, the differences were even more pronounced. I have never thought of wires as just simple conductors since that day. Every wire is an L-C-R network.

Hi Snave,

Inductance shouldn't matter in this case. It should affect both TVCs equally and cancel out any effect, as their inputs are in parallel. My theory was that a high frequency signal was present that could result in standing waves which would affect each unit differently because of their different lengths to the TC element. Since the problem was "solved", I saw no need to try different lengths but I moved on to other voltages which led to dilemma # 2 which I will post in a few minutes.

-S

You were having a ground loop which is picking up a 5oo kHz beackon from near by airport or airline guidance, common problem for high gain gain RF systems.

Hi StandardsGuy,

Just off the top of my head, have you 'accidentally' produced a radio signal after adding 'extras' to your proven setup, which hits a 500kHz node?

Sounds like the Fluke thing, (technical word that!) was inducing a current that was feeding back and causing problems at the 500kHz point, perhaps because of the length of the wire inside the 'T' or the length of the 'T' itself?

Good luck

Hi babybear,

The length of the A55 connected to one side of the tee differing from the length of the Holt 11 on the other side of the tee may have contributed to the problem. It looked like a resonance problem or standing waves from an RF "spur". I tried building a low pass filter for the Fluke 5720A, but it only made the problem worse (20,000 ppm error for one version). My best quess was that a signal was being generated inside the 5720A until dilemma # 2 came about.

-S

Hi StandardsGuy,

I appreciate your reply post, thanks.

I do not have that facility now, just can't do it, but I attempted to make a variable attenuator for one of my few HAM Radios. I was not able to measure it so precisely as you have, and no training in electronics to do so, but I made the attenuator and somehow must have induced a standing wave where it was joined to the Radio. I knew what was happening but did not know how to solve it!

Instead of attenuation I ended with severe 'interference' with a standing wave (and or extra waves) at almost the start of the HAM radios range and it would cause other problems at various frequencies. I guess I should have gone and joined a Local Radio Society where they would have known right away any possible problems which may occurred? I even started to think it was an 'unknown' Russian 'killer' signal? They tend to sound a little like a drum beat.

Anyway, you had the knowledge through experience not to mention skill, to know what to look for and consequently to remedy any troubles. Congrats!

Take care and good luck for the rest of the holidays.

Thanks for your inputs everyone. Some mysteries like this one may never be solved. Some day I hope to be able to explain it.