RF equipment calibration

[G0HZU]

But this thread isn't about shipping electrical products or ISO. The OP hasn't specified any requirements. I interpret this as a question as to whether it's possible to maintain reasonable calibration of test gear over longer intervals than one year by doing basic checks in a lab.
There's no mention of wanting to ship products or conform to any regulations. Just a means to boost confidence of a given measurement beyond a 1 year calibration.

FWIW, when I see a sig gen at work with maybe 6 months left on its annual calibration I see an instrument that passed a health check six months ago and it also means I can use it for certain applications without violating any company rules/procedures for another 6 months. That's all it means to me. It doesn't tell me how flat the sig gen response is over frequency or 'prove' how accurate it is when set to 0dBm or -120dBm at various frequencies.

Same if you work in a universtiy and do research you wish to publish, you have to know your traceable error budget.

Going off topic a bit here (and maybe I'm being pedantic but...) you don't 'have' to use traceable test gear to do valid research work in RF.  Accuracy and innovation aren't tied together.

[LeoUCDavis]

But this thread isn't about shipping electrical products or ISO. The OP hasn't specified any requirements. I interpret this as a question as to whether it's possible to maintain reasonable calibration of test gear over longer intervals than one year by doing basic checks in a lab.

That's exactly what I meant.

Going off topic a bit here (and maybe I'm being pedantic but...) you don't 'have' to use traceable test gear to do valid research work in RF.  Accuracy and innovation aren't tied together.

Yes indeed, but at some point, you always need an accurate measurement to show that your work is valid.

[G0HZU]

Yes indeed, but at some point, you always need an accurate measurement to show that your work is valid.

I don't agree

Some things don't even need test gear or physical tests. Go back to the 1960s and look at a few classic patents. Some things just needed basic/classic theory and nothing else. eg the classic Sontheimer range of directional couplers. Even a student can explain how these work using a pen and paper based on theory dating back to the days of Faraday with no need to build anything let alone the need for traceable calibrations of any test gear.

[G0HZU]

That's exactly what I meant.

Maybe it would be best to describe this without using the word 'calibration' because this word is open to interpretation.

For checking RF power here at home I have two RF power meters and three power heads covering up to many GHz. I can also use a packaged Agilent detector diode to check for flatness up to several GHz although these diodes are of limited use for making absolute measurements as they are temperature sensitive. However the detector flatness and input VSWR is extremely good (as in pretty awesome) from 100MHz up to 5GHz. It does prove useful when cross checking against my various power heads for flatness across a few GHz. I use one of the diodes in the range below (HP 8473C)

http://cp.literature.agilent.com/litweb/pdf/5952-8299.pdf

One of my power meters is the classic HP431C with HP 478A thermistor sensor. This runs in closed loop and when used where the input VSWR is very low you can use DC substitution to improve accuracy although I have never bothered to go this far with it. I use this meter as a reference because it stays consistent year after year (for maybe 20 years so far). Note that I use it with a DMM connected to the rear DMM port rather than look at the analogue dial at the front.

Unlike modern (open loop) thermocouple meters it doesn't need a 50MHz 0dBm onboard reference to self calibrate. So it doesn't have one fitted. The 478A sensor dates back 50 years or so and is often used to calibrate modern power meters because it is so simple and accurate. You can still buy one new from Keysight for about $4000 and the most accurate power meter made by Keysight uses a 478A based thermistor sensor

I also have precision power references built into some of my spectrum analysers and these are very consistent year after year when cross checked. So I have references at -20dBm, -10dBm and 0dBm. I also have a way of homebrewing a reference at HF and VHF that is based on simple tests and theory. But I'm not really that interested in chasing down beyond a few percent accuracy. So the cross checking results I get year on year are much better than I think I need

In my experience, the ability to make fairly accurate relative measurements is also very important and I find that cross checking several instruments with precision attenuators is one way to boost confidence.

But I'd probably better repeat that none of the stuff I do above is 'traceable' to any calibration house or standard. It just meets my needs for achieving high confidence in the measurements I make

[G0HZU]

If anyone is interested in reading more about the old 478A sensor it's worth googling

dc substitution 478A NIST

http://cp.literature.agilent.com/litweb/pdf/5990-9078EN.pdf

The N432A operates on the basis of a DC self-balancing Wheatstone bridge
which shall be elaborated on later. Hence, the N432A is the only RF power
meter with a DC substitution measurement, enabling it to convert RF power to a
DC measurement such as voltage (V) and resistance (?). This ability is why the
meter is widely used in metrology and calibration labs worldwide to calibrate
power meters and sensors, and for any application which requires accurate RF
power measurement.

Provided you use a test frequency where the sensor VSWR is very good (eg 100MHz to 1GHz?) you can measure RF power with fairly high confidence using DC substitution.  Of course the sensor could go faulty or be damaged but you could always buy several of them Mine has been amazingly consistent for over 20 years.

The standard 478A sensor is rated from 10MHz to 10GHz but the input VSWR is quite poor at the extremes of this range. They make a special version for very accurate use at 50MHz that has low VSWR at 50MHz.

The important (elegant) thing is that the RF power fed into into the sensor is the same as the DC power you input to balance the sensor. So you can effectively use a DMM on the DC range to measure RF power very accurately instead of the HP431 or HP432 meter's internal metering. So there is no need to have any 50MHz or 100MHz 0dBm reference built into the meter

[G0HZU]

Here's an image of my little HP8473C detector diode.

This device has a very flat frequency response although it isn't very accurate in absolute terms over temperature. So I use it for cross checking the flatness of various items of test gear. You also have to be wary of uncertainty caused by harmonic distortion terms in the signal under test if you use the detector up towards the top end (linear) part of its range.

This one has an input return loss of better than 30dB right up to 5GHz and it gives best performance across 100MHz to 5GHz.

[LeoUCDavis]

Yes indeed, but at some point, you always need an accurate measurement to show that your work is valid.

I don't agree

Some things don't even need test gear or physical tests. Go back to the 1960s and look at a few classic patents. Some things just needed basic/classic theory and nothing else. eg the classic Sontheimer range of directional couplers. Even a student can explain how these work using a pen and paper based on theory dating back to the days of Faraday with no need to build anything let alone the need for traceable calibrations of any test gear.

I agree, but such elegant innovations don't come by everyday and you'd still have to keep the more mundane work going. 

[LeoUCDavis]

The important (elegant) thing is that the RF power fed into into the sensor is the same as the DC power you input to balance the sensor. So you can effectively use a DMM on the DC range to measure RF power very accurately instead of the HP431 or HP432 meter's internal metering. So there is no need to have any 50MHz or 100MHz 0dBm reference built into the meter

Thanks G0HZU for the writing on power measurement. I can see that the 478A is fairly inexpensive on eBay and I'm interested in getting one to play with. How would you ensure that the reading from a second hand sensor is reasonably accurate? Do you have to compare/adjust it with a "traceable" standard?

[G0HZU]

Because these sensors are so old and so easy to damage you may find it difficult to get a working one. From what I have heard they (usually) either work or they are obviously toasted. I would expect to see lots of dead ones on ebay. You also have to get a compatible power meter to go with it and I would recommend the classic HP432A from the 1970s. Don't worry about the old school analogue dial it has. You can use a DMM on the recorder output at the back and this is going to be good enough for most users. I don't like the 432B with the dated and basic digital display and would always go for the 432A.

I did a quick return loss measurement of my 478A sensor using my VNA up to 1GHz. The plot below does say C? for cal but this is because I turned down the sweep speed after I calibrated and I experimented with different power drive levels after the calibration to see if it affected things. But the RL plot didn't alter. I'm too lazy to recal the VNA for the purpose of the plot

Note that this VNA actually has a year left of a 2 year Keysight calibration so this is one of only a few items of test gear I have with current calibration stickers. I didn't pay for the cal, it was an existing cal on the VNA and the Ecal kit.

But check out the return loss of this fabulous old power sensor...

Even at 50MHz this one is at 29dB return loss which is easily OK for my needs in terms of uncertainty when used with a decent 50MHz source with good VSWR. The HP478A and the old HP432A power meter from the 1970s are one of the most impressive pieces of RF engineering I've seen because of the simplicity and the accuracy they offer

I think I have a little nugget of gold here with mine and it just keeps on working every year returning consistent performance and it is my gold standard for power because of the simple/elegant way it works

But I don't use it for general measurements because it is quite clunky and slow to zero and set up for every measurement and I'm nervous I might damage it. The overload/damage level is marked as being only 30mW. These things are fragile!

There are some tests you can do to test a sensor on its own (without a power meter) and I'll see if I can find the info.

[G0HZU]

How would you ensure that the reading from a second hand sensor is reasonably accurate? Do you have to compare/adjust it with a "traceable" standard?

I guess only you know what you would accept as accurate. I just want consistency and for the first few years I owned this meter I used to bring it into work to check against the 0dBm reference used in the modern calibrated power meters at work. But it was boringly repeatable and I don't bother any more.

I suppose the correct way to check it would be via dc substitution using an HP432A and a fairly decent DMM because this would exploit the true qualities of this sensor without worrying about how well adjusted the HP432A is in terms of the dial accuracy or its recorder output accuracy. But I'm much too lazy to bother to do this with my meter even though it has connections at the back marked up for operation using DC calibration.

[LeoUCDavis]

Because these sensors are so old and so easy to damage you may find it difficult to get a working one. From what I have heard they (usually) either work or they are obviously toasted. I would expect to see lots of dead ones on ebay. You also have to get a compatible power meter to go with it and I would recommend the classic HP432A from the 1970s. Don't worry about the old school analogue dial it has. You can use a DMM on the recorder output at the back and this is going to be good enough for most users. I don't like the 432B with the dated and basic digital display and would always go for the 432A.

I did a quick return loss measurement of my 478A sensor using my VNA up to 1GHz. The plot below does say C? for cal but this is because I turned down the sweep speed after I calibrated and I experimented with different power drive levels after the calibration to see if it affected things. But the RL plot didn't alter. I'm too lazy to recal the VNA for the purpose of the plot

Note that this VNA actually has a year left of a 2 year Keysight calibration so this is one of only a few items of test gear I have with current calibration stickers. I didn't pay for the cal, it was an existing cal on the VNA and the Ecal kit.

But check out the return loss of this fabulous old power sensor...

Even at 50MHz this one is at 29dB return loss which is easily OK for my needs in terms of uncertainty when used with a decent 50MHz source with good VSWR. The HP478A and the old HP432A power meter from the 1970s are one of the most impressive pieces of RF engineering I've seen because of the simplicity and the accuracy they offer

Looks very nice. I see several listings on eBay that claim a working condition. Maybe I'll get one of them to find out. Thanks for all the helpful information.

[G0HZU]

Note that if you already have access to a few modern thermocouple power meters you probably aren't adding any useful day to day value by getting a thermistor based power meter.

The thermistor sensors do require care to avoid damage and they also require regular and careful zeroing. Also, on the sensitive power ranges the zero will drift if you handle the sensor (thermal effects) or even if you move the whole meter/cable/sensor assembly. So you may need to keep zeroing the meter several times a minute on the lower ranges! The 432A has an auto zero feature which helps a bit here but you still need to allow time to zero itself regularly. It really is a labour of love to use them as a day to day power meter and these are some of the reasons they aren't used any more in modern RF labs.

Also, the interconnecting cable between the sensor and meter can be quite hard to find and a good one can be expensive because of this. The cable can also be intermittent. Some of this stuff is 40-50 years old.

I just use mine as a cross checking meter. If I was seriously into calibration like some of the calibration petrolheads are on here then I suppose I would have tried cross checking via DC substitution. But I have many references and sources here that I can cross check to adequate confidence without needing to do this.

However, for some people on here, calibration is like a religion (make your cheques payable to 'Calibration Heaven') and they chase down every $$$ avenue that gets more and more accuracy in terms of voltage and frequency etc etc. It's almost as if calibration 'is' the hobby

[acbern]

However, for some people on here, calibration is like a religion (make your cheques payable to 'Calibration Heaven') and they chase down every $$$ avenue that gets more and more accuracy in terms of voltage and frequency etc etc. It's almost as if calibration 'is' the hobby

Well, I guess you miss the point. The initiator of this posting was asking how to avoid costs for calibration of equipment. I would assume that they do this (getting gear calibrated) to get a meter that they know is within (certain) specs. Otherwise it would be useless. He was not asking: is calibration necessary?

Now of course not everybody needs calibrated equipment for all their measurement jobs. You can write a physics document about e.g. a new theory or basic research without it.  You cannot do that if you have e.g. re-measured a certain particle parameter to achieve a better accuracy. It all depends on what you do. And it certainly has nothing to do with 'calibration is the hobby' (except for those who have it as a hobby, if these exist, and thats fine too).

But if you negate the need for calibration for those who need it (for whatever reason, and as the initiator of the post), then in deed all calibration is worthless and a waste. Certainly not an engineering/scientific approach.

[G0HZU]

I wasn't missing the point. I drew a stark contrast between my needs and the extreme needs of some others. It's up to the individual to decide what they 'need'. If I want to make a humerous comment about it then that's up to me

I've offered up the methods I use (and the limitations). It's up to the OP to decide if any of the info is useful.

But if you negate the need for calibration for those who need it (for whatever reason, and as the initiator of the post), then in deed all calibration is worthless and a waste. Certainly not an engineering/scientific approach.

That was a pretty lame strawman IMO