Probes for checking voltage references

[rhb]

What are recommended probe techniques for checking a voltage reference?  I tried probing the LM399 reference for noise  this morning with a DSO and was promptly handed my head on a platter.  With a regular probe the best I could do was ~11 mV of noise picked up  mostly from the LED overhead light.  The LM399 is holding nicely at 6.917 V on my 4 1/2 digit Tenma meter + 2 mV if I disconnect the meter or power down the LM399.  But it settles back after a few minutes.  Probably about a minute if I disconnect the DMM and a bit longer if I power off the LM399.

I finally did a search and reread Bob Pease's comments on the subject.  I followed Bob's advice and made up a socket to fit on the power header of the board Jason designed that slipped over the probe tip with the grabber removed.  Even getting that to work was not a slam dunk.  I think I'm going to put hot melt glue on it to hold things in place.

But I'm still a factor 0f 100 from the 3478A headed my way. I realize that the DMM has a much longer time constant, but clearly EMI becomes a very serious issue playing with voltage references.

[David Hess]

The low input capacitance and low bias current which are necessary for the 1 megohm input buffer of an oscilloscope also yield high noise.  References are not particularly quiet but a low noise amplifier will still be necessary to make a meaningful measurement with an oscilloscope.

[cellularmitosis]

Hey Reg,

You want a "spring tip adapter".  Looks like Mouser has a kit with four sizes for $3.60.

https://www.mouser.com/ProductDetail/Cal-Test/CT3668?qs=sGAEpiMZZMtbPwtnQMMkA0r1RXhinUnLxXKNNibaEQMIqT%252bnMgkUXQ%3d%3d

Insert the probe tip and ground spring into the holes labelled "test".

Also, to reduce EMI, you'll want to bodge in a 0.1uF cap on the output of that board.

Edit: alternatively, if you are trying to measure "1/f" noise (0.1Hz to 10Hz), you'll need a low-noise amplifier (LNA).  I just did a write up about measuring some LTZ1000 references for 1/f noise.  https://www.eevblog.com/forum/metrology/diy-low-frenquency-noise-meter/msg1461828/#msg1461828

[BravoV]

You want a "spring tip adapter".  Looks like Mouser has a kit with four sizes for $3.60.

https://www.mouser.com/ProductDetail/Cal-Test/CT3668?qs=sGAEpiMZZMtbPwtnQMMkA0r1RXhinUnLxXKNNibaEQMIqT%252bnMgkUXQ%3d%3d

Or much-much cheaper alternative, use steel wire for guitar string, wind it on nail with similar or smaller diameter, and/or reheat if needed to fix the coil from unwinding by it self.

[cellularmitosis]

this video from TI does a good job of covering spring tip probing (starts about 1 minute in)



oooooooh fancy!

[montemcguire]

When measuring LF 'signals' like a DC reference, the type of cabling from the DUT to the meter is more important than the actual contact and grounding to the DUT. At frequencies above a MHz, e.g. an oscilloscope probe, getting a proper wideband ground connection and HF signal 'launch' is probably more important, but below a couple of kHz, none of this matters.

Shielded twisted pair cables can essentially place an enclosure around the probe leads, all the way to the meter input jacks, and hopefully also remove a bunch of interference. By using a dedicated conductor (one conductor of the twisted pair) for 'ground' from the DUT, as well as an overall shield, more interference can be rejected. The standard procedure is to ground the shield at only one end of the cable, usually the measuring instrument side, so that it behaves as a pure electrostatic shield. At the DUT end, the twisted pair is used for signal + and 'ground' connections, just as in a normal probe.

In the audio world, star quad cable is used as an improvement over standard shielded twisted pair cable. Star quad uses two conductors for each of the conductors in a twisted pair, and because of the mechanical symmetry of such an arrangement, star quad is able to reject about 20dB or more of random EMI compared to a similarly shielded twisted pair. An overall shield will reduce electrostatic interference, as in the case of a coaxial cable, but unless a twisted pair (or quad) is used, there will be no rejection of current induced interference. The advantage of a star quad cable is that the twisted pair geometry is more precise, has more effective 'turns per inch', and thus rejects EMI better.

A good plan would thus be to use a twisted pair (or quad) connected with one conductor to the V+, one conductor to the V-, and the shield cut at the DUT end and connected only to the guard, chassis, or V- connection as appropriate to your meter or test instrument. Mogami 2799 is a nice miniature star quad cable that is very small but still not a huge pain to terminate. If your connectors can handle larger diameter cable, Mogami 2534 is a better cable - it has less capacitance, and because of its larger overall size, you get more copper and better dimensional stability. I have some 2534 cables that are 30 years old and they still work beautifully despite being flexed and used without great care. Conventional shielded twisted pair cables can work well, probably better than coax, but if their construction uses a drain wire, the mechanical displacement of the drain wire against the real twisted pair will mess up the symmetry of the twisted pair, and thus it will not reject as much EMI as possible.

[hwj-d]

@rhb

first of all avoid ground-loop: pwr-earth-psu to grnd-dut to pwr-earth-dso. Don't connect shielding, dut-casing, at both sides to ground. Maybe take batteries to prevent that. Scope ground is pwr-earth in principle, dmm is floating measure.

Keep in mind, that dmm measuring in 1ppm/µV area is done normaly under 10 or 100 plc condition, what you don't have with a scope. Take 20mhz bandwith limitation.

Good luck 

[rhb]

Thanks to all.

I'd already made a spring tip from copper wire and put a pin header socket on the end.  I play guitar and *really* should change the strings as they have been on there for a year or two.  It's an archtop, so muted treble is part of the sound.  So I'll wind a bunch of the ground springs for all my probes.

I have a lot of noise pickup on the power from the PS to the DUT.  I twisted the zip cord, but it didn't seem to help much.  The PS and DMM are floating.   I'll look for some shielded twisted pair to use for power with the shield grounded  at the PS only.

I'm not familiar with "10 or 100 plc condition".  Could you explain the acronym?

I hope I can get some long historical records of voltage refences.  When I read the comment about "seasonal" effects in the Fluke paper that @cellularmitosis linked I got very interested in getting records for multiple devices at one location.  That told me the author at least did not know what was causing the changes.

Even without temperature, pressure and humidty, I should be able to make some useful inferences just based on the location. While humidity shouldn't effect a reference in a hermtically sealed enclosure,  the op amps and other parts usually  are susceptible.  At ppm, everything matters.  A good analogy is the water elevation in a well.  At first glance that would seem to be quite independent of the phase of the moon.  But you'd be wrong.  It is easily measured and observed.  As is the effect of a train pulling into the station and then pulling out.

Geophysics is full of stuff like that.  If you have a large mass underwater, there is a *very* small bump on the surface of the water.  That's how they make maps of the seafloor using satellite altimetry data.  In my primary playground, reflection seismic data,  one is measuring motions in the subnanometer range and the noise is much larger than the signal.  It's only possible by having multiple channels and a very precise model of the physical relationships among the channels.  Many millions of dollars are spent every year creating models of the elastic wave propagation velocity in the earth so that each data sample can be summed with the correct samples.  The direction dependence of velocity is a major cause of problems and topic of research.

To create a 3D image of the subsurface,  10e7 to 10e9 samples get summed together to produce each of 10e12 to 10e14 image pixels.   The summation gives a 1000 to 10,000x improvement in SNR  There are warehouses scattered around Houston filled with computers doing this.  A major limitation is how much power you can get to the building from the grid.

[hwj-d]

I'm not familiar with "10 or 100 plc condition".  Could you explain the acronym?

"Power Line Cycles (PLC) or Number of Power Line Cycles (NPLC) indicates how long an input signal is integrated to obtain a single measurement.  Noise introduced from the power line tends to be periodic.  If the A/D converter integrates for an amount of time equal to one cycle of the power line noise, then the signal components from the periodic noise can be canceled.  Generally speaking, the longer a signal is integrated by the A/D converter, the more accurate the reading result. The (N)PLC setting of an instrument allows adjustment of the tradeoff between speed and accuracy."

https://de.tek.com/support/faqs/speed-what-nplc

Same measurement of one of my ltz's with 1 plc, 10 plc, 100 plc



For 60Hz power, an instrument operating at 1 NPLC can report a new value no faster than at 16.67 msec intervals.  For 50Hz power and 1 NPLC, a new value can be reported no faster than at 20 msec intervals.

[Conrad Hoffman]

Maybe too basic, but remember that noise levels are entirely meaningless unless the bandwidth is specified.

[hwj-d]

Maybe too basic

Yes that's right. Sometimes it's difficult to clarify these basic things without immediately compromising someone. The last sentence in the op prompted me to do so.

[rhb]

I haven't got far enough to worry about quantifying the noise.  I'm still trying to suppress it.  At the moment I'm debating wire sizes to get  for shielded twisted pair power supply cables.  My bench supply is 6 A max if i parallel the outputs, so 20 AWG ought to do.

I never worked on the acquisition side of things, so I'm not used to specifying or measuring SNR.  Just dealing with whatever we actually got.  But at several million dollars for a survey, the specs are quite specific and comprehensive ranging from allowable sea state to tow noise of the cables and noise from other vessels.

I have enough education and experience to realize that there are always gaps in what one knows and remembers.  So I'm not inclined to take umbrage at having something very basic pointed out unless I get beat over the head with something which is not relevant to the instance in question.

[cellularmitosis]

A trick I picked up from Conrad is to use the twisted pairs from either cat5 ethernet cable or telephone "bell" wire.  Usually 26AWG, solid core.  Great for hooking up voltage references, and more than adequate for powering them too.

When you get into situations where leakage in the cable causes measurable error (I think this comes into play when measuring something with a high output impedance), the PVC insulation on those twisted pairs is no longer good enough, and you'll want something with Teflon (PTFE) insulation.  EEVBlog forum member "ap" sells some cable like that.  https://www.ab-precision.de/products/accessories/

[rhb]

Thanks.  I was looking at 20 AWG silver plated shielded PTFE on eBay at $16 for 10 ft with shipping.

[montemcguire]

I have a lot of noise pickup on the power from the PS to the DUT.  I twisted the zip cord, but it didn't seem to help much.  The PS and DMM are floating.   I'll look for some shielded twisted pair to use for power with the shield grounded  at the PS only.

Maybe try grounding one of the devices? If everything is floating then there's a lot of trash on the shields / enclosures that isn't going anywhere but into the common mode of the devices. I'd ground the PSU if there's a choice, just to see if anything changes.

[hwj-d]

Thanks.  I was looking at 20 AWG silver plated shielded PTFE on eBay at $16 for 10 ft with shipping.

I'm looking for somewhat of this type too. But see the prices in germany ...

https://www.ebay.com/sch/i.html?_from=R40&_sacat=0&LH_Complete=1&LH_Sold=1&_nkw=10m+HF+TEFLON+LITZE+%C3%98+5mm%2F2x1%2C8mm+GESCHIRMT%2FVERSILBERT&LH_PrefLoc=2

[rhb]

Ouch!

[Conrad Hoffman]

I don't remember which ones, other than AN-47, but read anything you can find on probes and noise from the late Jim Williams of Linear Technology (now part of Analog Devices).

[thermistor-guy]

...
In the audio world, star quad cable is used as an improvement over standard shielded twisted pair cable. Star quad uses two conductors for each of the conductors in a twisted pair, and because of the mechanical symmetry of such an arrangement, star quad is able to reject about 20dB or more of random EMI compared to a similarly shielded twisted pair....

Cat5 cable outperforms star quad here:
https://www.prosoundtraining.com/2010/03/15/the-emi-project-part-2-point-source-field/

[cellularmitosis]

pasting the graph in case that article suffers from bit-rot in the future

edit: interesting that they could measure the difference between the various twist-rates of the CAT5 pairs!  Looks like the blue pair is the clear winner.

edit 2: part 1 of that series: https://www.prosoundtraining.com/2010/03/15/the-emi-project-part-1-introduction/

[rhb]

Thanks.  I'm a big Jim Williams fan.  I grabbed AN-47 and took a quick look.  Very nicely done.  I wish EDN & AD  would put out all of Jim's work in book form.  I've pointed a lot of times to "Max Wien, Mr. Hewlett and a Rainy Sunday Afternoon".

[Svgeesus]

Thanks.  I was looking at 20 AWG silver plated shielded PTFE on eBay at $16 for 10 ft with shipping.

I'm looking for somewhat of this type too. But see the prices in germany ...

For good cable at good prices in Europe, try Thomann Music who sell high quality Sommer cables for microphones etc.
https://www.thomann.de/de/kabel_meterware.html

[CalMachine]

Here is what I use    Not sure why the datasheet claims it is tinned copper conductors, though.  It is most definitely bare copper conductors.

https://www.ebay.com/itm/162965561759

https://catalog.belden.com/techdata/EN/5500FE_techdata.pdf?ip=false

[rhb]

I have read online that oscilloscope cables are typically 200 ohms to reduce reflections at the 1 Meg scope input.  But I have been unable to find a source of 200 ohm coax.  The highest impedance I could find was 95 ohms.   Moreover, the magnitude of the reflection coefficient is almost unity with 200 ohm cable.  So it seems rather silly even if scope makers have secret sources of cable.

The 2 & 4 conductor PTFE shielded twisted pair takes care of power and DMM probes, but I'd like to make up a set of test cables with various connectors at the DUT end.  Should i just use 50 ohm cable with a thru terminator?  That's a lot simpler.  And seems to me to make more sense so long as the DUT side is close to 50 ohms.

What do people who do this stuff for a living do?

[splin]

Thanks.  I'm a big Jim Williams fan.  I grabbed AN-47 and took a quick look.  Very nicely done.  I wish EDN & AD  would put out all of Jim's work in book form.  I've pointed a lot of times to "Max Wien, Mr. Hewlett and a Rainy Sunday Afternoon".

Well they have put a number of them in book form:

https://www.amazon.com/Analog-Circuit-Design-Applications-Solutions/dp/0123851858

I don't have volumes 2 and 3 but there are 41 Llnear Technology application notes in the first including AN47 and  AN86. Vol 1 includes 21 power management, 5 data conversion, 15 signal conditioning and 2 HF/RF  design AN's

[RandallMcRee]

Silver-plated teflon wire is great when you need it, you really need it:

http://www.apexjr.com/wire.html

(Satisfied customer, no affiliation, etc)

I use Mogami cable, its great stuff. Markertek is a good source.

Randall

[rhb]

Thanks for the Amazon link.  It's pricey, but not too bad.  I like books which is why I have 5000+.

[macboy]

I have read online that oscilloscope cables are typically 200 ohms to reduce reflections at the 1 Meg scope input.  But I have been unable to find a source of 200 ohm coax.  The highest impedance I could find was 95 ohms.   Moreover, the magnitude of the reflection coefficient is almost unity with 200 ohm cable.  So it seems rather silly even if scope makers have secret sources of cable.

You got it all wrong. The characteristic impedance of the coax isn't 200 ohm, but the end-to-end resistance of the center conductor is around 200 ohm (roughly 100 ohm per metre length). This is due to a very fine, relatively high resistivity center conductor. The high resistance makes the cable a lossy transmission line, which is what reduces the reflections. The extremely fine center conductor also has a vanishingly small surface area, which reduces the capacitance to ground/shield to improve high frequency characteristics (reducing cable capacitance will reduce the need for additional probe tip capacitance to compensate the frequency response).

[rhb]

Thanks for the explanation.  That makes a *lot* more sense than what I read on the internet.  I don't normally use Wikipedia, but they have a reference to the 1959 patent by Kobbe and Polits on the idea.

What's a suitable cable and who sells it?  I can find nothing.  In fact I have not found any specification of the resistance per foot  anywhere.

Edit: I measured a Rigol and an Instek probe center conductor at around 230-240 ohms  Surely I don't have to buy probes and behead them to get cable.

[Conrad Hoffman]

I've never seen that probe cable available anywhere. You also can't solder it, as the center conductor is Evenohm, nichrome or some similar alloy. That's the reason you always find crimps when you take a scope probe apart. The properly terminated 50 ohm coax is fine, so long as the circuit can handle a 50 ohm load. No good for much DC or any high impedance circuits.

[rhb]

It looks as if this is a machining problem making widgets that fit my probe tips.  Or buying probes and cutting off the tips.  Likely a combination of the two.

I made  wire springs per Bob Pease and put a 0.10" header socket on them, but I still have ~0.75 mV of noise picked up from the LED lamp overhead.  Some of this is probably pickup from the board traces, so I need to put the DUT in a shielded enclosure.  In cany case a huge improvement over 12 mV.

[hwj-d]

Hm, again,

if you want to measure references for noise, think about, what a buffered or unbuffered reference can drive. And that's definitely not 50 , 200 or something like that. A buffer like ltc2057, that's used in some forum references like kx, px, etc drives about 20mA, unbuffered about max 2mA. The 50 impendance from RGB-58 cables should closed with a 50 termination resistor to prevent reflections. So, don't do that to measure references directly.

I don't see anyone in the forum, who want to measure a reference with a scope that way. Correct me if I see something wrong.

As David Hess and others mentioned above, you need a LNA for decoupling and measure noise correct with a scope. Pipelie developed a LNA for that. kj73 and others show it in action, for example:
https://www.eevblog.com/forum/metrology/diy-low-frenquency-noise-meter/msg1462484/#msg1462484

And even here, the caps of this LNA should preloaded, to prevent damage an unbuffered reference:
https://www.eevblog.com/forum/metrology/diy-low-frenquency-noise-meter/msg1461828/#msg1461828

So, i don't know the characteristics of your reference, but note his driving possibilities.

[TiN]

hwj-d have good points. Just lil remark, as my FX REF can drive over 50 mA and don't have problems with charging pipelie's cap in few seconds.

[rhb]

Thanks. 

I was connecting to the output op amp, not the reference itself.   I would not consider a direct connection to the reference with a probe.  Even 1 Meg is too much load.

I've tracked down the noise to the power cable coming from my PS.  I've ordered 20 AWG shielded twisted pair PTFE wire to make PS leads and 22 AWG 4 wire PTFE for resistance measurement cables.

The source of the noise is the LED lamps I just installed when a  fluorescent tube went out.  I really needed more light, so I installed a pair of 4' LED lamps.   I'm going to get some metal screen to cover the lights to see if that helps.

Having a multitude of nearby RF noise sources is unavoidable for me right now.  So I'm going to have to put a bunch of work into suppressing EMI at the source and at the DUT.

[Conrad Hoffman]

FWIW, I installed a bunch of LED lights in my lab. They save lots of money compared to the 6 flood lamps that were there before. Still, when I do FM alignments or any very sensitive measurements, I have to turn them off and resort to an incandescent in a desk lamp.

[Echo88]

I got rid of the led-lamps in my measurement room and now use halogen-incandescent lights instead. So now i dont have to fear led-lamp-noise but can see spikes on my resistor measurement when i switch the lights on...so, not a perfect solution. 
Might need an input current impulse limiter for the lights.

[hwj-d]

That's my little workbench corner with selfmade ledlamp. The lamp is made quick and dirty with some aluprofiles, very cheap 4x 10W led modules, a cheap 3A buck converter inside the aluprofile, silicon wires, an old laptop power supply and a sonoff module. In the right corner are some servers, synology, odroid xu4, cubitruck, pizerow's, lan switches, under the table fritzbox router, wlan, dect, laptop, monitor on the left, bluethoth, cell phone ...

Believe me or not, i didn't see excessive spikes  from that all if i measure my resistors, references on that table with my 34461a (nlpc 10-100).
Ok, that's not a profi like lab, were I don't implement that all in a spot like that.

[eurofox]

That's my little workbench corner with selfmade ledlamp. The lamp is made quick and dirty with some aluprofiles, very cheap 4x 10W led modules, a cheap 3A buck converter inside the aluprofile, silicon wires, an old laptop power supply and a sonoff module. In the right corner are some servers, synology, odroid xu4, cubitruck, pizerow's, lan switches, under the table fritzbox router, wlan, dect, laptop, monitor on the left, bluethoth, cell phone ...

Believe me or not, i didn't see excessive spikes  from that all if i measure my resistors, references on that table with my 34461a (nlpc 10-100).
Ok, that's not a profi like lab, were I don't implement that all in a spot like that.

Did you put 0.01% resistors in your lamp? 

[hwj-d]

Did you put 0.01% resistors in your lamp? 

psst, that's the trick ... 

[tautech]

Reg
Check out this great old book about Tek probe design:
https://www.davmar.org/TE/TekConcepts/TekProbeCircuits.pdf

[rhb]

Thanks to all.

My Dad left me a bunch of large aluminum chassis' which are large enough to have left me wondering what to do with them.  Based on input from a number of people, I'm going to use  a couple to make a Faraday  cage test box.  Given the prices I've seen for mu metal I'll make  a  smaller Faraday cage with magnetic shielding.

[ArthurDent]

Regarding the discussion of what type of cable/wiring to use, years ago I had an electrometer which had an obscenely high input impedance. I made an input cable for the electrometer using what I thought would be a great piece of cable which looked like RG-58 size shielded coaxial cable. Under the black outer jacket was a silver plated braided shield, under that was a polyethylene tube, and fitting loosely inside the tube was a small teflon insulated stranded silver plated wire. Maybe it looks like a little overkill but I had a piece of the cable on hand and this seemed like a perfect application for it. 

I noticed I was having a lot of trouble getting stable readings and found that it was the triboelectric effect caused by me moving the cable. I found that if I just had the tip of my probe disconnected and wiggled the wire slightly the analog meter needle would wildly zing from one end of the scale to the other. The slight flexing and/or movement of the different insulating materials used in the cable was producing the triboelectric effect, like rubbing a balloon on your hair to make it stand up. 

If you’re trying to make precision measurements everything from LED lighting, SMPS, to wiring may have an effect.

[splin]

Instead of using costly to run incandescents, why not modify suitable LED lamps/striplights etc to run from a mains transformer + linear regulator(s)? Some lamps use a relatively high voltage so might not be ideal from a safety point of view but certainly many cheap Chinese linear light fittings should be suitable. Shouldn't be expensive either given the low power levels required.

Obviously you couldn't use the existing mains light fittings and wiring but in a lab environment it shouldn't be too unsightly to install replacement or suplementary noise free lighting.

[zhtoor]

That's my little workbench corner with selfmade ledlamp. The lamp is made quick and dirty with some aluprofiles, very cheap 4x 10W led modules, a cheap 3A buck converter inside the aluprofile, silicon wires, an old laptop power supply and a sonoff module. In the right corner are some servers, synology, odroid xu4, cubitruck, pizerow's, lan switches, under the table fritzbox router, wlan, dect, laptop, monitor on the left, bluethoth, cell phone ...

Believe me or not, i didn't see excessive spikes  from that all if i measure my resistors, references on that table with my 34461a (nlpc 10-100).
Ok, that's not a profi like lab, were I don't implement that all in a spot like that.

nice and cosy setup 

regards.

-zia

[rhb]

The LED lamps run straight off the mains.  I'm going to install a 1/8" hardware cloth screen today and see how that does.  At least if the cloth arrived as promised.

I'm acutely aware I've started playing in the land where you have to hold your mouth just right to get things to work.

[MisterDiodes]

Just some general food for thought, since RHB seems to be just starting out:

If you run LED lighting, even if you run RAW (directly from mains) be aware there is a small current transient when the diodes switch "on", so pay attention that the power source is quiet DC (and no switch mode power supply allowed near low-ppm test area - batteries or regulated linear + isolation XFMR only).  Also pay attention to the LED CRI - If it's not at least 90+ it becomes more difficult to see the difference between some shades of wire or component code colors - say orange and red.

For those reasons we generally only use halogen incandescent from 12V, 24V or 48V battery banks during critical tests.

As far as cabling goes for chasing low ppm - you're generally not ever going to be measuring a Vref with a 'scope probe (not without an LNA).  Cat5 or even shielded Cat6-7 with faster twist rates is fine in most cases, and a decent DMM is required. CLEAN Copper to Copper connections only, watch out for oxide Copper or tarnished silver plate on that silver-plated wire (that creates thermal EMFs and noise).  Gold plated connectors are easier to keep clean, but add a small amount of noise.  Keep the leads dressed SHORT and use DIRECT wiring, keep power cords away from sensitive circuits, make use of shields / guard circuits as required, and understand ground loops.

Remember:  Every circuit loop that encloses a non-zero area is a potential magnetic pickup point, and every PN junction is a potential demodulator.

Copper / Aluminum shielding is not the best at low freq - for that application look at steel and or MuMetal.  If you're really serious about cable shielding at low freq you start to look at MuMetal braided cable sheathing - depending on your work environment.  In our work environment we have to use that since the 3458a's are being used near some servo drives, and we're trying to grab a low ppm DC measure.  In that case the servo wiring is on one shield system, and the measuring cables are on a separate shield / guard system.

More Suggested reading - Keithley's Low Noise Measurement handbook, and Art of Electronics, 3rd edition.

[cellularmitosis]

even shielded Cat6-7 with faster twist rates

Ah, the fast twist rates should be even better than the "blue" pair in CAT5 https://www.eevblog.com/forum/metrology/probes-for-checking-voltage-references/msg1462766/#msg1462766

[rhb]

Well, 1/8" galvanized screen over the LEDs accomplished very little. :-(   Not really surprised, but I was hopeful.

Unfortunately, I don't know what's in the lamps.  They apparently have some sort of SMPS internally because the noise is definitely not 60 Hz switching transients.  I just ordered an HP 8560A with TG, so with some probe construction I should be able to track down and squelch most of the noise.

I would not consider measuring Vref with a scope probe.  I just wanted to see how much noise there was.  It's high enough to not be measurable without considerably more finesse.

Mu metal is shockingly expensive.  I plan to make a large Faraday cage which is just aluminum and a smaller one which is aluminum clad with mu metal sheet and mount assorted feedthrus.

I have bothof the references, but more time spent reading AoE 2nd than 3rd.  Only glanced briefly at the Keithly monograph.  I have another monograph on low noise electronics circuits.  Clearly time to give it a read.

Thanks to all.
Reg

[MisterDiodes]

Mu metal is shockingly expensive.

Welcome to the world of low ppm and high precision measures...get your wallet out! 

MuMetal and to a certain extent steel are your very true friends at low freq - aluminum / copper shields won't really do it for mag shielding, but they can help for electrostatics.  It depends on what problems you're facing. We use screened isolation transformers in the work areas for bench mains power (keeps out noise coming in on utility mains and the utility ground - usually the worst offender), and sometimes you'll need a dedicated lab ground and/ or a ZSRG system (Zero Signal Reference Ground - usually a copper grounded grid in your lab floor), depending on what you're measuring. 

Low ppm measuring also means you're shielding against air drafts and typically you've got a place with good temperature control.  Pay attention that all your test lead connections / binding posts are at the same temperature for lowest thermal EMF effect.

Here's another one for your homework reading list describing at least some of the grounding techniques - this is only a tiny tip of the noise-control iceberg. 

https://www.embedded.com/print/4014229

Have fun!

[eurofox]

I'm building a new LTZ1000 10V board and my intention is to mount it in a steel massive block completely hermetic, no humidity allowed 

It will be temperature compensated and all critical resistors at 2ppm.

The power supply will be a very low noise type.

I will post the results after finishing and aging for minimum 1 month.

eurofox

[rhb]

Thanks.  Is it worth getting that book or is there something better?   It looks as if it would fill a gap in my library. I have "Low Noise Electronics Design" by Motchenbacher and Fitchen, but I don't see anything devoted just to grounding and shielding. The usual intro stuff in a lot of books, but nothing in depth.   Even expensive books are cheaper than trial and error.

I found a couple of 2" x 11" x 7" 29 gauge steel chassis'.   How effective would those be for a test chamber?  I have in mind putting piano hinge between the two.  Should they be annealed?  How much mu metal shielding  would be needed to make a chamber using those reasonably quiet?    Would copper foil be justified?  I don't *think* I have any large fields to contend with other than computers and lab equipment.

This is all the classic "tools to make tools" exercise. but I don't mind spending $150-200 for a good test chamber for precision measurements.  From the pricing I've seen I could cover both chassis' with copper foil and 1-2 mm of mu metal for that price if I look hard enough.

[MisterDiodes]

...I heard that sometimes you can Google the title and you might find a lower cost PDF version, and then you can decide if it's worth the purchase.  I doubt that's what the publishers had in mind, and I wouldn't post those links here.

Otherwise it's a relatively good introduction to various grounding / noise control techniques.  Not at all what I'd call 100% step-by-step comprehensive but at least an overview to start thinking about the concepts involved.

[rhb]

I like books as evidenced by my 5000+ volume technical library.  I'll order a copy.  If you know of other references you'd recommend having please let me know.

"A Mathematical Introduction to Compressive Sensing" by Foucart and Rauhut is quite comprehensive, but even that doesn't cover all the applications of sparse L1 pursuits.  Indeed, it would take 4-5 volumes to do that. And it is hardly "step by step",  My general experience is that I need 2-3 references for a casual interest and 4-5 or more for anything serious.

[Echo88]

If were talking about shielding and low noise stuff, this book is also very noteworthy: Henry W. Ott: Electromagnetic Compatibility Engineering
Especially the shielding comparisons for steel enclosures and wiring comparisons are interesting.
Im still searching for good and thick-walled steel cases (little bit bigger and more massive than the RF-steel-cases fom TEKO), which can house smaller aluminium cases like the Hammond-case-series.

[MisterDiodes]

Sometimes deep drawn or weldied steel standard electrical mains general purpose junction / equipment boxes can work very well, and look nice when they come powder coated.  The better ones to use are the type that have a rated dust-tight close-fitting lid and -don't- have pre-punched knockouts - it's usually better to do your own machining.

[hwj-d]

My opinion about it,

don't see any prof. massive steel cases to prevent RF shielding today. Look at sharias yt-channel, what the industrie is doing.

I mean, this is about voltage references, right? So keep it simple and cost effective. That's for the pcb's too. (€: I don't think, we really want to beat a fluke reference in the first steps... ).

It's more important to pay attention to the basic things and how to measure them correctly. Know, from where the noises really comes from. Read the books, then build a reference, calibrator, or what ever. But not a heavy expensive tank, that doesn't really help much, but tries to solve the problem like under a large pile of sand.

Keep it simple,
have fun,
 

[rhb]

I've got a mid-size U-haul pile of old '40's and '50's electronics.  Quite a few NOS E.F. Johnson, Bud and similar chassis'.  So in part I'm trying to use up some of Dad's treasures.  The general idea is to fabricate a good basic enclosure and make a series of tests before and after adding refinements like copper foil and mu metal.

I plan to order Ott's book along with a 2000+  page tome entitled "Electromagnetic Compatibility Handbook".   

There's another book I'm considering,"Grounding and Shielding: Circuits and Interference" by Ralph Morrison.  It looks to be very basic physics oriented. which is what I would like in addition to the others.  The more detailed the math the better.  Any comments on that book or suggestions of alternative treatments?

[eurofox]

My opinion about it,

don't see any prof. massive steel cases to prevent RF shielding today. Look at sharias yt-channel, what the industrie is doing.

I mean, this is about voltage references, right? So keep it simple and cost effective. That's for the pcb's too. (€: I don't think, we really want to beat a fluke reference in the first steps... ).

It's more important to pay attention to the basic things and how to measure them correctly. Know, from where the noises really comes from. Read the books, then build a reference, calibrator, or what ever. But not a heavy expensive tank, that doesn't really help much, but tries to solve the problem like under a large pile of sand.

Keep it simple,
have fun,
 

Using massive block of steel is not only for shielding but to get it hermetic to humidity.
It is simple for me, I have a milling machine, lathe, tig and stick welding and many other mechanical tools, it is easy for me 


I just want to test something that I think was never done from volt nuts:

1/ heavy shielding

2/ avoid humidity influence, te massive housing with wall thickness should be about 10 mm, build with 2 parts with gasket (base and cover), silicone isolated wires hermetic sealed, area with silicate gell to remove the humidity after closing the "container", 1 taped hole just above the 10V trimmer but closed with a screw and hermetic gasket. Power supply isolated in a different area shielded from the LTZ1000 pcb.

3/ have a "stable" thermal mass and if needed could easy be heated very accurately.

It will take some months before I can really make some test

And this will cost me just time and maybe 10-20 Euros of metal.

eurofox

[rhb]

Touche!  Never underestimate the skill set of an EEVblog forum member.

I'm looking at $200 to $350 for books.  So $100-200 for a chamber seems pretty reasonable and in line with other costs.

[hwj-d]

Touche!  Never underestimate the skill set of an EEVblog forum member.

Keep calm, boy!
I'm definitly not interested in some "fight club" here ...

[Echo88]

Noise-related throw-in: A few weeks ago i tested a mains-based AM-stereo-radio to see if it could detect RF noise and make it audible, which didnt work for any device i brought close to the radio.
Today i tested a portable battery-driven AM-radio and am amazed how nice it indicates all devices which emit rf-noise. You can hear the noise-spikes from switching the lighting, usb-wall warts (you can hear how the dcdc tries to hold the 5V after being disconnected from the 230V) and even tv remotes make sounds like the phasers from star trek. 

Since it such a nice indicator for RF-noise and -peaks im gonna get the radio a bigger battery and will log the rectified speaker-output together with my longterm resistor measurements on which i still see sharp peaks despite my attempts at shielding and use of CAT-5-cable.

Used handheld radio is the Sony ICF-S22. But i guess every handheld radio should do the trick. Would be also interesting to test the radio in a shielded box and look for rf-peaks which get trough the shield.

[MisterDiodes]

You haven't lived until you played Trek (a basic Star Trek game, mostly text) on an IMSAI 8080.  The sound system was an AM radio placed next to the computer, and it worked better with the lid off the motherboard box. The programmers had intentionally built various loops into the code to toggle a few unused address bus lines at the correct rate, and that created enough noise on the AM band to be clearly picked up by the radio...So when you fired phasors, you could hear that,  photon torpedos has another sound,  the Klingons yammering at you could even be made out.   The "Red Alert" noise worked surprisingly well too.

Pretty darned clever way  to get a multimedia game experience at a time long before computers had any sort of audio output jack.

[ArthurDent]

Your Star Trek story reminds me of when my brother worked as a programmer on an IBM360 at a local university years ago. I was visiting him once when another program was stringing together a bunch of variable length do loops and listening to the RFI produced by a cheap 6 transistor AM radio. By varying the length of the loops he could generate different ‘notes’ and the final product was ‘Daisy’. Phasor noises were still in the future in the mid 1960s.

There were Teletype 33 terminals scattered around campus and one of the games available for students to play (with the limited sign-on time allotted) was a text only adventure game. You had to wait for an interrupt to get a reply so the game was very slow and you had to use your imagination because there were no images.