RF power meter calibration

[Radio Wrangler]

For RF amplitude measurement, you'll find the probe compensation adjustment is rather twitchy. What makes a small effect in the waveshape of a 1kHz calibrator, and put you 10% off in voltage (=21% off in power) quite easily.

Tek made a passive probe for 50 ohm instruments. Effectively a BNC, a length of 50 Ohm coax and a 1k resistor with a stabby point and a ground shield. These are very good at RF. I've used them at 300W 1GHz (pulsed 1% duty factor) so their voltage capability is good. They get warm when someone touches a point on a transmitter board with 50v bias on it

David

[G0HZU_JMR]

I looked up in my notes and realised that it is about 3 years since I last checked (at work) the 50MHz 0dBm reference on my old Anritsu power meter so I took it into work today. At work we have dozens of the Keysight E4418/9 power meters and I selected two that were calibrated at Keysight within the last month. Both 50MHz 0dBm references on the Keysight meters agreed perfectly at 0.00dBm and they stayed like that for the whole afternoon.

I also checked the Anritusu 0dBm reference against the Keysight meters and it was within 0.01dB although I ended up tweaking it to be spot on. Normally it doesn't need adjusting. I then checked it again at home against the HP432A and HP478A sensor using the DC substitution procedure and a decent bench DMM. The last time I did this DC substitution check was 2 years ago and the result was spot on back then at 0.00dBm. I checked it again today and it was 0.01dBm. So virtually no change in 2 years.

The DC substitution procedure is a bit fiddly but I make it easier by using a spreadsheet and some BNC to 4mm adaptors at the back of the 432A. This makes it easy to make/change the connections at the back of the 432A as only one 4mm connector has to change as per the image below.

The results below are almost identical to the results I got two years ago.

[Bazz4CQJ]

I think I'm at the stage now where I'm just eagerly waiting for the flange resistors to arrive from China. I'll start by getting a good DC characterisation of them, but doing it on a heat sink which will give me options to easily install an attenuated RF output in one or more forms. My interest in this topic is not just curiosity; I'm planning to get my IC-718 out of the cupboard in the next few days and hook it on to the 5MHz dipole I've just put up. The 718 hasn't been switched on for ~3 years; I guess that it will be fine, but some test data would be good. Further back in the same cupboard is my TS-530 - don't ask how long it is since that was last switched on, but it would be good to check it out and then find a new home for it.

Thanks for everyone's inputs; I'll report back on how my cheapy resistors respond to a dash of power in to them .

B

[Radio1950]

Quote by Wrangler,

For RF amplitude measurement, you'll find the probe compensation adjustment is rather twitchy. What makes a small effect in the waveshape of a 1kHz calibrator, and put you 10% off in voltage (=21% off in power) quite easily.

That is a very good valid point, with, yes, a large impact on measurements.
I was assuming that most would take care when compensating their x10 probes.
One can always check the probe response with a sig gen if in doubt, after compensation is adjusted, at least up to the freq specs of the probe and CRO.


Baz
My smaller 5 watt HF Load with Florida RF 32-1157 termination has two extra BNC monitoring outputs.
One with a 910R off the RF input, and in series with a 91R to earth, giving a HF RF voltage (1/10 approx.) to the BNC "RF mon out".
It is rough but very useful, 5 watts giving approv 1.5 V RMS, and with a Zo of about 80 ohms.
This mismatches the Load Zin to about 47 ohms, SWR 50 is 1.06.
The other BNC is a crude diode detector using a schottky, and fed from the RF Mon Out.

This 5W load is a little crude but practical.


Quote by G0HZU

I don't see them as precision devices and care is required to compensate them correctly and old/tired probes can often have unstable compensation as the probe is handled.


Yes - appropriate.
When I originally saw the OP with a reference to "crossed needles", and meaning no disrespect to the OP, I knew immediately that this whole thread would not be about precision measurement.
All of my comments so far are in this context.

[G0HZU_JMR]

I interpreted the OP as an open invitation to discuss the good and bad points of using a scope or a diode detector to help calibrate a power meter.

Not sure where to put this but I know there’s a large collective of expertise on here. (Moderators, I wonder whether there would be any mileage in having a measurement/calibration/useful standards section ?)

I have one of those cross meter swr/power meter things that I need to check the calibration of.

Is it as simple as running the transceiver into a dummy load through the wattmeter, and tee’ing into the coax to measure the voltage waveform on the scope, or do I make a diode detector, and use a dvm to measure the dc output, or are there any other tips I can use please ?

I see that as requesting more info than the green light to 'just do it with a scope and a x10 probe, everything will be fine (because your power meter is probably a bit of a toy anyway)'

I consider my analysis of the loading effect of a typical scope and the various contributors to overall measurement uncertainty as the bare minimum required for anyone interested in taking this to a level beyond just hooking the gear together and hoping for the best

[Bazz4CQJ]

I thought that the most likely explanation of "crossed needles" is that he owns an MFJ tuner (their standard single meter, 2 needle, combined Power and SWR), as do I and may others. Whatever the case, I'd be surprised if many owners of budget ham radio equipment expect it to be very accurate re RF power. I'll live with a 5% error, 10% if we have to, but as soon as someone pipes in with 20% error (was that on Al's old thread?), then I'd like to understand the problem and try do better.

If I manage to knock together some 'half decent power meter', it will be interesting to see how good the MFJ's calibration proves to be; may not be too bad at 5MHz.

B

[G0HZU_JMR]

A ballpark estimate for the quality of any calibration gear is that it would ideally be ten times better (in ppm) than the device it is calibrating. In this case, it's probably OK to suggest it should be maybe two or three times better.

So it's not unreasonable to try and understand (and manage) the various contributors to overall uncertainty. Otherwise you could end up with a 10% power error caused by the load, another 6% power error due to the scope vertical accuracy, another 3% at 29MHz due to the loading effect of the scope, a few percent if the wrong diode detector type is used, quite a large error if a x10 scope probe is used and not compensated etc etc.

Significant uncertainty from harmonics when using the scope or diode method is unlikely because a commercial transceiver ought to have harmonics at least 40dB down (hopefully -60dBc?).

[G0HZU_JMR]

I measured a few scopes this evening with the VNA. It's tricky to set the reference plane accurately because the scopes have BNC inputs but the results are good enough.

See below for the Tek 465 result as this scope is probably quite similar to the 463. In the plots below I compare the (VNA measured) Rp and Cp of the Tek 465 against a crude model based on the input circuit of the Tek 465.

The input capacitance is about 20pF on the Tek 465 datasheet and there is a 51R resistor in series at the input of the scope. So the crudest model would be 51R in series with 20pF. For this RF analysis the impact of the 1Meg resistance can be ignored at RF. But for Rs I've cheated slightly and used 45R for Rs in series with 20pF because that's what I see with my best attempt at setting the VNA reference plane correctly at the input BNC connector.

But the results show a very close agreement and you can see that the Rp of the scope falls very quickly once the frequency gets towards the upper end of the HF band. So even if anyone reading this isn't interested in power meter calibration they might be interested to see the input loading of the Tek 465 over frequency. It isn't as simple as 1Meg in parallel with 20pF. This is because of the series 51R resistor at the input.

In reality, you don't really need the VNA for this because the crude series model based on 45-50R Rs and 20pF Cs (from the Tek 465 schematic and datasheet) nails it pretty well. At 30MHz the Rp is around 1700 ohms. You also don't need the exotic RF simulator. My earlier excel spreadsheet agrees with the simulator for Rp and Cp vs frequency of the 45R, 20pF series model extremely well. Certainly to 4 significant figures anyway.

My little Tek TDS2012 100MHz digital scope has a very similar response and the plot is not worth posting up because it is so similar.

[Bazz4CQJ]

Far away from the world of Tek and HP, for anyone not familiar with them, here is a picture of MFJ's cross-needle combined meter and the schematic of their 949 ATU (which is widely distributed on the internet and inside some of their user manuals) which shows two meters which switch between 30 and 300W ranges.

B

[G0HZU_JMR]

I think my Yaesu FC902 ATU has a similar coupler design (Bruene?) but it uses twin meters for power/SWR. My only other ham power meter is a little Welz SP-220. When I first bought the Welz meter I did spend some time testing it and it was remarkably accurate across 2-30MHz. I think it's spec'd to 145MHz but when I tested it up at 145MHz I think it was out by something like 20%. But across the HF band it was very good. I didn't write down any results but across the HF band it seemed to be very good. Sadly, I can't remember how good in terms of percentage error but it was much better than I expected. I was impressed despite the small and cheap meter movement. Obviously, the PEP feature on it was a waste of time and the SWR function is always dodgy on meters like this.

I have no idea how accurate the power meter is in the Yaesu FC902 ATU and I'm not sure it matters much.

[G0HZU_JMR]

I did some tests on a range of x10 scope probes today and I did write it all up. But the forum had a hiccup and it's all gone. But to summarise, some models of x10 scope probe are quite poor and only one was 'good'. A lot depends on the construction and the compensation technology. You can tell a good probe from a bad one just by looking at the way the compensation affects the leading edge of the square wave. A basic probe will affect more of the top edge of the square wave but a decent probe will only affect the very tip of the squarewave edge. The HP 10073C 500MHz x10 probe seems to be compensated very well (already) and the compensation only affects higher frequencies. So it is easy to get a flat response in the swept test. The other (branded) probes were poor, especially on a swept test where the level varied across LF-30MHz after compensation. They all used cheap and crude compensation that affected much of the waveshape of the scope calibration squarewave as the compensation was tweaked either side of a squarewave. So I found that no matter what I did with the compensation I couldn't get any of them to be as flat (in the 30MHz swept test) as the exotic HP scope probe. Note that all of the probes were grounded directly at the tip ground and I didn't use the long ground croc lead. These cheaper x10 probes therefore aren't suitable for this type of task. Danger, Will Robinson!

[G0HZU_JMR]

Note that I just checked my HP probe and it is actually an HP 10073A and not a C version. The worst performers were (unsurprisingly) the ones that had the cheap/basic/classic compensation cap adjuster at the tip of the probe across the internal 9 Meg resistor. The better ones (and the clear winner in the HP 10073A) have more elaborate compensation methods with a fine compensation adjustment at the scope end of the probe.

One other thing... I think the 10073 probes are very expensive when new and there are bound to be Chinese fakes of these for sale. I've had mine for many years and it must be >20 years old now because it has the HP logo on it. I think they are worth the money if you can get a genuine used one in good condition. Mine has lasted probably 20 years and is still going strong.

[G0HZU_JMR]

Today is a good day... I'm still playing with my HP 54825A scope and I had a rummage through the top pouch and found a bag containing a set of 4 x10 probes hiding in a zipped compartment meant to hold the manual and software.
I was looking for the original SW disks and assumed the bulge was the disks. But instead of disks it was a set of 4 brand new HP/Agilent 10073C probes in yellow, green, blue and pink to match the scope's colour coded CH1-4 inputs They must have been hiding there for years.

I rescued this old unloved Infinium scope from the company skip a couple of weeks ago. I'm now doubly chuffed to find the pack of hidden scope probes in the pouch!

A couple of days ago I did test a few of my regular x10 probes here at home to look at the input impedance across frequency.

They all seem to conform to the basic model of about 12pF in series with about 50R-75R. But some probes could be 12pF in series with 100R. This is the series equivalent model for how much the x10 scope probe loads the circuit.

Some were worse than others and my test results suggests that it probably isn't a good idea to use a x10 scope probe at 71V rms across the upper end of the HF band. This is the voltage across a 50R dummy load for 100W.
A lot of branded x10 probes come supplied with voltage charts showing the voltage limits vs frequency although none of my cheapo probes came with this data.

See below for a prediction that shows the expected limits for Rp and Cp for typical x10 scope probes. i.e. I'd expect most x10 probes to load a circuit in a way that fall inside the curves in the plot below. So by 29MHz a typical scope probe might have an Rp somewhere between 2200 ohms and 4000 ohms

I'm wondering if this subject deserves its own thread. i.e voltage handling limits vs frequency for various models of scope and various scope probe types.

[Bazz4CQJ]

Wow, Jeremy, your company skip is better than any of the ones I had access to.

Re probes, I was just looking at the Picoscope website re something completely different, but saw their range of probes https://www.picotech.com/accessories...loscope-probes.

Don't know where they sit on the quality spectrum?

No resistors yet .

B

[Radio Wrangler]

THe sheer variability across different makes and models of scope probes and the criticality of the compensation adjustment is why it's a better bet to use a 50 Ohm power attenuator to bring the transmitter down to a level of a few hundred mW, use a kength of decent quality 50 Ohm coax and have a small 50 Ohm load right close to the scope to act as a 50 Ohm termination if the scope hasn't got an internal one you can switch in,

It's a lot flatter and more predictable. Which is what you want if calibrating a power meter is the aim.

David

[Bazz4CQJ]

That's fine, but just remind me how we handle the ~35pF shunt capacitance of the scope. Does that need to be cancelled out within the attenuator?

I don't recall all these problems when F.G Rayer was powering up his 807 rigs

B

[Radio Wrangler]

You just have to be aware of the roll-off

50 Ohm source // 50 Ohm termination gives 25 Ohms thevenin impedance at the node.

1/(2*Pi*R*C) says the 3dB point is at 182MHz. Good enough for the HF bands without any further effort, if the scope itself is good enough. A small correction can be calculated for 2m, again if the scope BW isn't a bigger issue.

When G3OGR was playing with 807s, all he was required to do was measure the HT and the DC anode current. Life was simpler back then!

David

[G0HZU_JMR]

For calibration purposes above 30MHz I don't think it's a good idea to use a scope that doesn't have a decent internal 50R termination option. This is because of the increasing effect of the Rp of the scope. But this depends on the scope model obviously.

If you look back in the thread there's a link to a TDS2022 scope and even a TS2022 will have loading problems at VHF if you use an external 50R termination. By 30MHz this could cause a hidden 3% calibration error with this scope but the loading error will get much worse by 50MHz or 70MHz.

You could find that the raw scope input has a VSWR of less than 4:1 by 145MHz and this is before the 50R termination is applied. This shows how much that scope would degrade the effective Rp of the 50R load. My excel spreadsheet predicts a 200MHz TDS2022 scope (45R Rs and 20pF Cs?) will make the Rp of the 50R termination look like 35R by 145MHz. That's a big error?

[G0HZU_JMR]

Wow, Jeremy, your company skip is better than any of the ones I had access to.

Yes it is pretty good. I have spies that walk past the skip area every day and stuff doesn't stay there for long I can't just take it though. I have to fill in paperwork such that it is formally gifted to me by the company with a lengthy disclaimer.

It's a really nice scope and it has been given the official Agilent upgrade with a newer mobo, HDD, processor and memory and voice control HW. It's also got a really strange 3.5" floppy drive that has 120Mb capacity. I didn't even know these 'super' FDD existed but the 120Mb floppy disks look pretty much the same as 3.5" 1.44Mb disks. The scope drive can accept standard 3.5" 1.44Mb disks or the sexy 120Mb disks. However, you can't use the 120Mb disks in a standard 3.5" FDD, it has to be a special drive as in the scope.

[G0HZU_JMR]

Note that some BNC 50R 'through' terminations can help to offset this effect slightly as they aren't as simple as a 50R resistor inside. I think they can offset the scope capacitance a small amount. I don't know how they do it though. I've never taken one apart.

See below for an image of one of the 120Mb floppy disks next to a regular 1.44Mb disk. I had no idea these super FDD (floptical?) disks existed.