How to find a power level reference?

[rjmusto]

So in my workshop I have a couple of scopes, a sig gen, spectrum analyser and now a TinySA. None would be classed as high end……

How then can I find a way to calibrate the amplitudes produced and read by this lot? I need some kind of reference obviously, but how to do it without expensive kit?

Any clever ideas?

[Cruisin Marine]

Start with the scope, I used to use straight TTL square wave 5v, calibrate from there and from there you can work out AC levels and .707 and 1.414 and all that jazz. From that with known good attenuators you can cal the other stuff.

Now I have a DC reference source, can't remember where I got it from tbh but it does 1.25v +- 2mV.

[Radio Wrangler]

HP RF power meters model 435A and later have a thermal power detector (or diode for low lwvels down to -70dBm)

THe detectors are heads mounted on a cable to the meter unit.

The detector heads are easily damaged, and many have been destroyed. They are getting in short supply and second hand prices of working ones are rather high.

BUT

The meters parts of these systems have a pretty accurate power reference source in them, and the meter units, sans RF heads are quite cheap. THey give an accurate 0dBm into 50 Ohms,at a not-so accurate 50MHz.

I have a few of these systems and cross-comparing measurements of their reference sources gives a lot of confidence in their stability. In the plant where they were made there were several references running continuously as a long-term (multi-year) check on their stability.

David

[rjmusto]

The voltage reference into the oscilloscope route has been pretty much what I’ve been doing and perhaps is good enough for non-professional situations.

But funnily enough I have an HP 435 sitting in my eBay watchlist at the moment but didn’t know about the accurate reference. Was just aware that the probes are crunchingly expensive….. so haven’t taken it further. But will look at it in a new light.

So thanks for the input.

[rjmusto]

I guess I should clear the decks one afternoon and pull together a spreadsheet of readings across all the bits of kit and see if I can come up with a sensible calibration table. That might clear things a bit in my own head at least!

[G0HZU_JMR]

It's also possible to use a sig gen as a power reference across a large frequency range if you feed it into a homebrew levelling amplifier.

The levelling amplifier will level the signal to the performance of its power sensor and this sensor could be something as simple and cheap as a low barrier Schottky diode. If a diode detector like this is used it helps if the sig gen has low harmonics as the diode detector will be very sensitive to harmonics. The source impedance of the sig gen also gets cleaned up to a very accurate 50 ohms as long as a fairly decent power splitter is used as part of the levelling system. This minimises mismatch uncertainty. The absolute accuracy of the diode detector can be checked against a cheap power reference. This could be done at a very low frequency.

A basic levelling amplifier could be made from salvaged or surplus parts from pretty much any electronics workroom. Nothing exotic is needed. It's just a low frequency error amplifier that sits inside a feedback loop. The total cost including a basic metal box and connectors would be about £15. If built from salvaged parts the price could be close to zero.

I made my own levelling amplifier a while back to check the flatness of my RF power meter heads at low frequency. I've also used it to level a sweep generator across 1GHz to about 18GHz using a power meter as the reference inside the ALC loop. The only requirement for the sig gen is that it should support external AM modulation down to DC. Otherwise, a basic voltage controlled RF attenuator would have to be fitted to the levelling amplifier. The attenuator doesn't have to be accurate, it just has to have a reasonable amount of attenuation range. 15dB would be more than enough and there are lots of ways to make this cheaply.

It's possible to achieve remarkably good results even with a basic, low cost system. A lot depends on what performance you want to achieve and over what frequency range. +/- 0.2dB would be achievable up to 1GHz or so with a basic system and this could be used to check the frequency response of a spectrum analyser or another RF power sensor.

[G0HZU_JMR]

I also have a crude lashup levelling amplifier somewhere that uses a PIN diode for the attenuator. This system only works up to about 100MHz but it can level an RF sig gen extremely well. It was built on some bare copper board as a prototype levelling amplifier to help me make a better one that I have in a nice die cast box. Every RF workbench should have a levelling amplifier as a verification tool especially if none of the test gear gets formally sent out for annual calibration.

[G0HZU_JMR]

I can't find the original lashup board but see below for my levelling amplifier. I've used this at work and here at home to provide RF levelling.

I'm afraid my construction skills are fairly ropey (they haven't improved much in decades) but the levelling box below only contains a resistive trimpot, a basic voltage reg and a few resistors and caps and a FET opamp.

External to this the system does also need a resistive power splitter (can be homebrew) and also a wideband power meter/sensor and this can also be homebrew.

This is based on the same (very simple and elegant) RF levelling system used in metrology/calibration labs across the globe. See below for a recent attempt to check the level accuracy of my Agilent PSA spectrum analyser to 3GHz. This spectrum analyser is incredibly flat over this frequency range and to test it requires a very accurate levelling system. See the typical response graph provide by Agilent and also my attempt to check it using my levelling amplifier. It wouldn't be possible to do this without a levelling amplifier and a decent RF power splitter. I used a precision homebrew splitter here that cost very little to make using SMD resistors costing 2p each.

Note that my spectrum analyser plot covers LF through to 3GHz and the scale is 1dB per division. This shows how well the levelling amplifier works.

[rjmusto]

Hi Jeremy,
Thanks for your interesting input.
As you say the circuit could be fairly simple and seems able to achieve remarkable accuracy.

Couple of questions though;
- how can you be sure the detector for the error amp has a flat response across the frequencies of interest?
- and what have you used as the 'cheap power reference'?

Thanks,
Ralph

[orbanp1]

Hi Jeremy,

Care to share the details of your levelling amp?
Like circuit diagram, and internal pictures of the build?
I certainly could use myself something like that.

Thanks, Peter

[G0HZU_JMR]

Originally Posted by rjmusto

Hi Jeremy,
Thanks for your interesting input.
As you say the circuit could be fairly simple and seems able to achieve remarkable accuracy.

Couple of questions though;
- how can you be sure the detector for the error amp has a flat response across the frequencies of interest?
- and what have you used as the 'cheap power reference'?

Thanks,
Ralph

Hi Peter and Ralph, it's probably best I start a dedicated thread on levelling amps and I can show how I made mine and I can draw out the circuit for you. I also managed to find my old prototype but it is very ugly and was just intended as a low frequency concept demonstrator. This has everything on it including the variable attenuator and the RF splitter.

This levelling technology dates back many decades and it is typically used by HP to level the source in network analysers and also for calibration/checking of RF power sensors up to many GHz.

A good place to start is to look at application notes based on the classic (but very expensive!) HP 11667 (A) (B) (D) RF splitter. I made my own homebrew version of this using SMD resistors and it works quite well.

The classic setup for this stuff is the HP 83752A 20GHz sweeper and the HP 11667 RF splitter as in the image below. The HP 83752A sweeper supports two types of power sensor to use in the ALC feedback path, the low barrier Schottky diode sensors like the HP8473C (these output a negative voltage) and it also supports the 'recorder out' from the back of a classic lab power meter. This outputs a positive voltage.

My levelling amp shown in the previous post only supports the positive recorder output from a lab power meter but it is easy to convert it to a 'negative' sensor.

The key goals for a successful system are accurate levelling and a very low source VSWR. The levelling system is very elegant because it uses a feedback path and this creates a voltage source at the input to the 11667 splitter when the ALC loop is closed. This means the impedance at the common port of the 11667B is close to zero so the source VSWR is defined by the series 50R resistor inside the 11667 and this VSWR is very, very low at a few GHz. Check out the 11667B VSWR on the datasheet. Sadly I don't own a 11667 but it has been on my watch list for many years. I made my own homebrew version and it works quite well up past 6GHz.

To preserve the source VSWR and the levelling accuracy the DUT has to be connected directly at the 11667. No cables are allowed here.

See the image below. The sweeper RF level doesn't have to be accurate and the loss in the RF cable from it is automatically calibrated out. This is a very elegant system and has been around in various guises since the 1960s at least.

I'm lucky to own a lot of the classic HP lab gear and this includes the 20GHz 83752A sweeper. This sweeper doesn't need my levelling amp as this is already built into the sweeper.

I designed my levelling amp to allow the same functionality from any lab sig gen that supports external AM down to DC. I've used my homebrew levelling amp with a Marconi 2022, 2019 (2024 doesn't work?) Agilent ESGD series and the Keysight PSG sig gen at work.

I'll set up a thread covering all this stuff as it is really quite easy and cheap to set up a system for just a few quid that will level almost as well as the exotic stuff up to a couple of GHz or so.

I can show how to make a cheapo version of the HP8473C LB Schottky detector but it will only work up to a couple of GHz or so rather than the 26GHz capability of the exotic HP 8473C. I've got a HP8473C here to compare against.

This stuff can be built really easily at home and cheaply and it can still work remarkably well up into the UHF band.

[G0HZU_JMR]

It's also possible to make a fairly useful power reference (for a 50R spectrum analyser) using logic gates and a few resistors although this will probably only work well up to 10MHz or so.

The key goal here would be to produce something that has a very low source VSWR (defined by the resistors) and a fairly accurate output and this would be defined by the resistor tolerances and the ESR of the logic gates.

This would never win any awards for accuracy but I'd expect it to be within +/- 0.2dB across typical workroom temperatures and the source VSWR would be very low.

[factory]

Originally Posted by Radio Wrangler

HP RF power meters model 435A and later have a thermal power detector (or diode for low lwvels down to -70dBm)

THe detectors are heads mounted on a cable to the meter unit.

The detector heads are easily damaged, and many have been destroyed. They are getting in short supply and second hand prices of working ones are rather high.

BUT

The meters parts of these systems have a pretty accurate power reference source in them, and the meter units, sans RF heads are quite cheap. THey give an accurate 0dBm into 50 Ohms,at a not-so accurate 50MHz.

I have a few of these systems and cross-comparing measurements of their reference sources gives a lot of confidence in their stability. In the plant where they were made there were several references running continuously as a long-term (multi-year) check on their stability.

David

Are these the references that they used in the plant? pictures are from the web. I've got one & a half 435 power meters here & was thinking of making something similar with the bits.


David

[Radio Wrangler]

THose are boxed up power references. The one on the left has, I think, also got s crystal oscillator for 10MHz. But the 50MHz power references are from free-running LC oscillators and will be a bit off in frequency. The levelling is the accurate bit. Note spec'd power is into an accurate 50 Ohm load.

The power ref circuit worked so well HP never dared change it. All those 2N3866 I scattered around were from the dumping of life bought components for the reference oscillator. Someone up high decided they weren't going to support them any further.

At one point, the glass capacitor used in the tank went unavailable, and Tony changed it over to an ATC porcelain job, hence the multi-year comparison run.

David

[Cruisin Marine]

If you want to spend a bit more money than on the bum basic stuff, the Marconi 6960 units are extremely accurate as far as I can tell.
I got one with a good sensor for less than £100, and judging by measurements on vg sig gen's it seems extremely accurate indeed.
They can be found for reasonable prices if you are patient.

You could also get a Marconi TF2603, if lucky you will find a good one, but they do need calibrating and need a source to do so. They are very useful for level measurements, but nowhere near the accuracy of the 6960 though.

[rjmusto]

Not come across these Marconi beasts - had a quick look on eBay and prices for ones with sensors were pretty high. But, as you, patience is needed here.

For the HP boxes, I was thinking it odd that no-one has come up with a homebrew sensor, since they are so popular. For hobby use and in my case just ham radio HF work, top end accuracy isn’t really needed, so why not make something.

And have found this thread;
https://groups.io/g/HP-Agilent-Keysi...2C0%2C0&jump=1

So might be nice to have a go at making my own sensor 😁

[rjmusto]

I would also like to try the levelling amp - Jeremy; could you maybe share the schematic for what you did?

Thanks.

[Radio Wrangler]

The diode sensors are OK, but it's the thermal ones which are best for accuracy. They are micromachined silicon structures having a bridge with a precision load resistor on one side and a chain of temperature sensing either thermocouples or sensing diodes on the other. Far beyond a home lab, I'm afraid.

The more approachable sensor is the type 478 which goes with the 432A power meter.

These are thermistor bridge DC substitution machines and qualify as transfer standards between RF and DC power. Can be switched between 50 Ohm and 75 Ohm modes, thouth the RF transmission line is really 50. I remember David xxxxx at the NPL asking (almost in desperation) for HP to never delete the 50/75 switch.... they used it in a couple of smart tricks for compensating for RF source impedance. - He was up visiting about standards traceability for out new noise sources and was getting the two and sixpenny tour when he spotted the power meter production line and his eyes lit up.

David

[G0HZU_JMR]

Originally Posted by rjmusto

I would also like to try the levelling amp - Jeremy; could you maybe share the schematic for what you did?

Thanks.

I've started a thread about it in the link below and this includes a schematic. Note that this version only supports the classic lab power meters with the 0-1V DVM/recorder output BNC connector on the rear panel.

https://www.vintage-radio.net/forum/...d.php?t=187865

It really is a crude circuit hastily fudged into a diecast box as I already have a 83752A sweep generator with a levelling amplifier built in. However, it does work with several sig gen types.