USA CAl Club: Round 3 suggestions

[Magnificent Bastard]

Perhaps you should read the USA Cal Club Round 2 thread.  That kit is pretty close to what you describe, though neither reference is 10 V.  It's quite a lot of work to design and build a device to achieve 30 ppm uncertainties from 10 microvolts to 100 volts which is the basic DC spec for the Fluke 515A.

What you describe is useless for anything but DC.   Using that to calibrate RF gear would take a lifetime of building instruments.  And without a time reference, not even possible.  The subject of "metrology" covers a lot more than the volt, ohm, ampere and temperature.

I'm not even remotely a volt or time nut.  I'm  concerned with practical calibration of a lab full of LCR meters, VNAs, spectrum analyzers, RF synthesizers, semiconductor testers, etc.  That requires accurate volt and time references, but those alone are quite inadequate. It requires quite a lot more and it is vastly more difficult to get 100 ppm RF measurements than 10 ppm DC measurements. And in many instances national laboratories struggle to get 10,000 ppm uncertainties.  Try to find a 0.1 dB step RF attenuator good from DC to 18 GHz.  After all, at DC that's just a 0.25% accurate  resistor ladder.  I want one, but I'm still looking.  They may not exist above 6 GHz at any price.

I started this thread to discuss aspects of speeding up circulation of the references  among people and encompassing at least basic AC measurements up to RF.  Primarily the latter.

Have Fun!
Reg

Understood.  Your needs are probably different that most of the people in the Cal-Club I think.  When I said "everything can be derived from that" I assumed everyone already knew "...with the proper equipment."  A super-accurate GPSDO frequency standard can be purchased off of evilBay very cheap (and everyone should probably have one of these)-- so I did not include that.  Thermal based AC and RF standards can be MADE, (again, if you have or are willing to build the right equipment, which includes an FRDC calibrator, etc.-- just like NIST does it).  At the end of the day, you still need that DC standard to calibrate the AC thermal converter to.  AC and DC current require an accurate shunt-- and you need the 10K standard to derive that (..."with the right equipment").  With a stable power supply and a 100:10:1 divider, you can derive all 5 cardinal points-- 1000V, 100V, 10V (from the standard), 1V, and 100mV.  With these and a stable AC power supply, a set of thermal converters (that you can make yourself if you also make a FRDC calibrator)-- you can derive the 5 cardinal points for AC voltage-- and then on to AC current.  I never said it would be easy, and I never said that you would not need to make or buy some equipment, but with those standards, you CAN derive everything needed.  If you are willing to accept higher uncertainties for the standard for temperature, you can use an distilled-water ice bath and boiling distilled water (pressure corrected) to calibrate an NTC thermistor-- so the calibrated temperature probe is not needed in that case, then you only need the 10V and 10K standard.

Here's what I am assuming (and please correct me if I'm wrong)--- most of the people in the "Cal Club" are participating because they want to have an accurate volt and ohm in their lab (for DC and LF-AC work), without having to spend the money to have an accredited Cal-Lab provide that.  So, it's kind of a "share the love" project-- one person provides the laboratory standards, get's them calibrated, and everyone shares from that.  (OR-- No initial calibration nor bona fide lab standards are required, and everyone just compares each other's readings just for fun and entertainment).  I don't think anyone in this club wants to buy or own a professional-grade calibration laboratory-- but building "hobby-grade" equipment appears to be (at least for some people) an interesting pastime.  The more equipment you make or buy, the more calibration points you can generate (or measure).  Different people will need different calibration equipment based on what they have and what they want to do.

For RF, building and calibrating the thermal converters is quite a bit more difficult, as a much more technical FRDC calibrator is needed-- but you can MAKE that if you are so inclined-- and you will STILL need the DC standard to calibrate THAT.

I have an idea-- why don't you design some home-brew RF calibration equipment, and post your designs in this forum?  Obviously, the uncertainties will not be the same as a good calibration from Keysight, but for hobby purposes you CAN design something that is quite good!

[TiN]

Can only add to Magnificent Bastard words that having standards is only half of the story. Maintaining history and doing uncertainty budget for all results is remaining half, which somehow very lacking on Cal-Club activities.
So it's more of spending pastime than any actual calibrations, which are defined as comparisons between STD and DUT with known uncertainty. And rhb is not correct, one of standards in current Cal-Club R2 kit is in fact 10V

[Magnificent Bastard]

Can only add to Magnificent Bastard words that having standards is only half of the story. Maintaining history and doing uncertainty budget for all results is remaining half, which somehow very lacking on Cal-Club activities.
So it's more of spending pastime than any actual calibrations, which are defined as measurements with known uncertainty. And rhb is not correct, one of standards in current Cal-Club R2 kit is in fact 10V

Well, I didn't mention the "paperwork", because most every engineer I know hates doing paperwork.  My wife is an accountant-- she would LOVE to do this stuff.  But, you are right.  As I said, I doubt people on this board are wanting to own and operate a "real" (accredited) Cal-Lab-- but just want to know (within reason) that they have an absolute value of the volt and ohm in their lab that is close to correct.  This, and a warm glass of milk will help you sleep at night...

One thing I can think of that is really GREAT, is if you have the 10V and 10K standards, AND you have a 3458A having a dubious calibration state, you *could* use those 2 standards to do an artifact calibration of the 3458A.  Maybe not as good as sending it in for a "real" calibration at Keysight, but at least you would know that it is closer to correct (and maybe less "dubious").  This alone would be a valuable service to volt-nuts.

That said, IF you are building something for sale that may be adversely affected by your in-house values of the volt and/or ohm being incorrect, then for heaven's sake don't be CHEAP-- get your equipment professionally calibrated at an accredited Cal-Lab!  {...sheesh!...}   

[TiN]

Cheap and high-end calibrations are not mixing well I'd add 1 Ohm standard to have a complete kit, so one can use that with 10V and 10kOhm to perform artifact calibration of 5700/5720A.
Then calibrator + 3458A data can be verified against each other and can deliver enough points to keep busy with paperwork and uncertainty calculations until milk glass goes bad  .
Oh wait....somebody at Fluke calibration facility had same idea....



Equivalent for RF world would be calibration kits and high-end VNAs.

[rhb]

I'd forgotten that the FX was 10 V.

I do plan to design low cost OSSW/OSHW RF test gear.  That is my primary interest.  However, to do that you need T&M kit which is reasonably in cal.  It can be done without it, but I would not live long enough to finish the task.  Been there and tried that 35 years ago.

At present I'm trying to quantify what can be done with existing low cost stuff from ebay.  The BSIDE ESR02 Pro is slated for extensive evaluation relative to a 4284A LCR meter and a 4145B semiconductor analyzer.  I'll do similar testing of the nanoVNA against my 8753B and 11801/SD-24 with an eye to seeing what I can improve. At least if someone else doesn't beat me to it.  I ordered a nanoVNA with the intent of implementing TDR on it, but before it arrived someone had already done it.  And quite well. However, I've not seen anyone invoke the shift theorem to solve for the travel time of an event to picosecond resolution.  So I may still have something to contribute.

Recently I've been playing with a new noise source from ebay with a couple of RTL-SDRs V2 and V3.  The V3 has more impedance mismatch than the V2.  This appears to be a significant issue with both designs.

I know that the noise source is flat from 250 kHz to 1 GHz to within +/- 1 dB which is the spec for my 8560A.  It looks flat to within +/- 0.5 dB, but no way to be sure yet.  That will have to wait on setting up and calibrating my 8970B with the 346B noise source and measuring the new ebay noise source with that.

There is *no* resemblance to the 8560A spectrum when using the spektrum program and the RTL-SDRs.  At the moment I don't understand why, though I suspect the front end of the RTL-SDR dongles is anything but constant impedance over the full frequency range.  The question now is how to measure that.  Ideally with cheap kit.  Then comes the question of whether I can correct it without redsigning the entire dongle.

I haven't posted it, but I did some TDR with a Tek 11801 and SD-24 of two  N-F OSL sets, an Anritsu and a Chinese ebay set.  The results were rather surprising.  The reference planes for the open and short are different and the Chinese load is pretty crappy, but it looks to me as if I can correct cal kit errors mathematically.  I spent a couple of hours on the math a few days ago and it looks tractable.

The price of a good quality 3.5 mm set for my 8510C is a strong incentive to see if I can pull off determining a correction for an arbitrary OSL set from first principles without reference to a better set.  Very likely I could not use it in the 8510C software and would have to export the data to a PC.  But it's quite similar to a standard problem in reflection seismic processing.  If it works, it could significantly improve the results from  a nanoVNA.  Even the SDR-Kits Rosenberg cal set is 2x the cost of a nanoVNA.  The idea may not work, but it's an interesting problem.

Have Fun!
Reg

[guenthert]

[..]
What you describe is useless for anything but DC.   Using that to calibrate RF gear would take a lifetime of building instruments.  And without a time reference, not even possible.  The subject of "metrology" covers a lot more than the volt, ohm, ampere and temperature.
[..]

  All good points, but the time reference is thankfully a solved problem, isn't it?  For a modest investment one can easily get a GPSDO which yields exact time +/- a few ns and a 10MHz frequency standard better than 0.1ppb.  I'd think only the most ambitious time nuts have more stringent requirements.

  Given, deriving e.g. a 6GHz signal from that 10MHz standard isn't trivial, but isn't that a very specific need and should that really be part of the travelling calibration kit?

[Kosmic]

There is *no* resemblance to the 8560A spectrum when using the spektrum program and the RTL-SDRs.  At the moment I don't understand why, though I suspect the front end of the RTL-SDR dongles is anything but constant impedance over the full frequency range.  The question now is how to measure that.  Ideally with cheap kit.  Then comes the question of whether I can correct it without redsigning the entire dongle.

The R820T inside the RTL-SDR was originally made to watch TV not being the front end of a spectrum analyzer. I suspect that a lot of the non-linearity you are seeing come from the IC itself. Also the open source driver everybody use was made before the specs of the R820T & RTL2832U was widely available. When you start crosschecking the datasheet to the implementation you realise it's a miracle that it's actually working. The implementation is full of errors. I started reimplementing the driver some times ago, but like a lot of my project it was shelved 

Personally, I think you would probably have more success implementing your own devices (like this guy: https://hforsten.com/improved-homemade-vna.html) than trying to fix those cheap devices.

[rhb]

The GPSDO solves time for practical lab cals.  I'd expect that anyone concerned about lab cals at RF has one.  I have two, one of Leo Bodnar's dual output devices and a multi-output 10 MHz unit.

Hard stuff at RF:

power/amplitude

complex impedance (i.e. capacitance and inductance in combination with resistance)

I'm sure there are other things, but those are the ones I'm most concerned about at present.  The new ebay noise source I've been testing and a step attenuator should cover 250 kHz to 1 GHz  quite well for power with the addition of a clean 1 or 2 frequency 0 dBm source.  I have a pair of HP DC-18 GHz attenuators which I plan to send to Keysight for cal data so I can cal a cheaper unit for the cal kit. That leaves a handful of stable capacitors and inductors.  That will all easily fit in a medium flat rate box. 

At  DC  I'd like to see two  1-10-100-1K-10k-100k-1m resistor sets and a 2nd PX reference added with some synchronization of distribution so the PX references and the resistor sets get swapped between the high precision and low precision sets each trip.  The voltage references and the resistor set would make it possible with modest effort to calibrate other DMM ranges.  Adding a 3 V output to the PX would be very helpful for people with 3478Asas their best meter.

I was rather alarmed at a post in the Round 2 thread which appeared to indicate that one person has  had the Round 2 kit for 3 months.  I hope that is erroneous, but unless there is strong justification for that, measures will need to be taken to prevent a recurrence.  TiN contributed a rather posh piece of kit. I'd be very upset if it were stolen.  We might need to require security deposits via PayPal to protect against theft and to penalize excessively long possession.  I felt huge pressure to turn the kit around ASAP as did Conrad and most others.  If someone has a 3458A and environmental chamber I'm fine with their keeping the kit for a longer period.  But I think there is an obligation to present data to justify it.

I have recently realized that when I get set up, I'll actually have most of a state of the art mid-90's cal lab if I simply get a suitable set of  references with NIST traceable cals and correctly implement a test program to compare all my instruments.  And suffer through the adjustment process for anything out of spec.

My primary interest is making 0.1% accuracy affordably available across the full spectrum of electrical measurements.  If that happens I'll push for 0.01%.   I'll leave 1 ppm to TiN et al.  While I find it entertaining, it's not something I care to spend money on.

Have Fun!
Reg

[rhb]

Kosmic,

I'm very aware of the history of the RTL-SDR.  I'm not trying to fix it, I'm trying to give people  a reasonably accurate idea of what it is capable of.  There are a lot of people for whom a noise source and RTL-SDR are all they can afford.  I want to help them.  There was a period of many years when that was my situation.

If someone wants to fix the RTL-SDR, I'll do what I can to help.  As a first step, I plan to measure the input impedance vs frequency of the units I have with my 8753B VNA and post those results. And see if a pair of RTL-SDRs on the reflected and incident ports of a 60dB.com RF bridge will allow better corrections.

Have Fun!
Reg

[Kosmic]

Kosmic,

I'm very aware of the history of the RTL-SDR.  I'm not trying to fix it, I'm trying to give people  a reasonably accurate idea of what it is capable of.  There are a lot of people for whom a noise source and RTL-SDR are all they can afford.  I want to help them.  There was a period of many years when that was my situation.

If someone wants to fix the RTL-SDR, I'll do what I can to help.  As a first step, I plan to measure the input impedance vs frequency of the units I have with my 8753B VNA and post those results. And see if a pair of RTL-SDRs on the reflected and incident ports of a 60dB.com RF bridge will allow better corrections.

Have Fun!
Reg

FYI, the R820T has been discontinued and is not in production anymore. I wonder for how much time they will be able to produce the RTL-SDR and similar products.

[rhb]

The devices in the 2 versions I got are marked R820T2.  I've not bothered to investigate the chip.  I remember the Elonics E4000 chip that got it all started.  The future status of the RTL-SDR hardly matters.  The SDRplay series is much better. though also more expensive.  I only bought the 2 RTL-SDR versions to test them.  You can be certain someone will find a replacement.

Actually relevant to this thread, I opened up an HP 435-K05 which is  5 and 50 MHz 0 dBm reference sources for a look inside.  With modest effort I can repackage it to fit in a small or medium flat rate box.  So that neatly fits the requirement for a 0 dBm reference.  The 50 MHz 2nd harmonic is 50 dB down and the 7th, 8th and 9the are over 60 dB down.  The power and frequency are off, so it may do better after a bit of TLC.

Reg

[rhb]

FYI

https://www.eevblog.com/forum/metrology/usa-cal-club-rf-division/new/#new