AD8307 probe-How is it used?

[djsbriscoe]

I bought a couple of AD8307 based RF probes off that auction site a while ago.
They were supplied by www.60dbm.com
Now I am wondering how they actually work and how I can actually use them?
Does anyone have any experience with these probes or any advice they can pass on? Thanks.

David.

[Skywave]

Here, when I try to access www.60dbm.com, I get 'Unable to connect'.

Al.

[Julesomega]

Wonderful little logarithmic amplifier/detector device which had a period of being sold in many eBay modules, search this site for AD8307

[Radio Wrangler]

Start by reading the analog devices data sheet. Pay particular attention to the lower frequency limit. It can be a bit of a disappointment for general applications on some of this family.

I don't know the circuit of the probes you have, but I have used an awful lot of ADI's loggers over the last 30 years. They do what they say on the tin and the data sheets are honest, but you do need to read and digest the specs particularly regarding the scaling of input and output voltage. Notice that the input impedances ar higher than 50 Ohms, and the in/out graphs are usually in dBm now is this w.r.t. one milliwatt into a matched source to the device? or into a shunt bringing the device down to 50 Ohms? You'll find both variants in different places on some data sheets.

David

[Terry_VK5TM]

Originally Posted by Skywave

Here, when I try to access www.60dbm.com, I get 'Unable to connect'.

Al.

Same here - a check through isitdown shows it as down for everyone.

[mickm3for]

The 8307 gives 25mv out for every 1dB input data says from .1mhz to 500mhz
i used one in a rf power meter. so i think the output of the probe to a dvm then for every 1dB in gives 25mv out, the 8307 requires power for the ic

[Radio Wrangler]

The ambiguities come in the issue of the input termination and the intercept point of that 25mV/dB slope. Without sorting these out, it's fine for comparing levels but there is no reference for absolute level.

David

[orbanp1]

Here is a web-page with the original QST article of the W7ZOI power meter:
https://www.robkalmeijer.nl/techniek...e38/index.html
This also has applications.

This looks like an updated version of the above meter:
https://www.qsl.net/n5ib/AD8307%20RF...%20rev2r03.pdf

Regards, Peter

[Radio1950]

Some good replies there.
Reading required.


Addressing your specific questions, about your specific probe.
The active device is a log RF amp and detector, and your probe has a 50R input with what appears to be some freq compensation.
Your probe has SMA input socket?
You can just use adaptors, but it makes it physically fragile.
You can probably expect level accuracy to +/- 1.5 dB from 2 to 350 MHz at least.

You will have to arrange something to interpret the output voltage, ie DVM or voltmeter.
You can scale down the output to say 10 mV per dB rise if you like with just 1% resistors.
Easier on the brain.

It is best "calibrated" with an accurate sig gen.

These chips can also be found on PCBs as "experimenter boards", and sourced from the "Orient".
I have used these as input sensors on my Three Channel Milliwattmeter, a COVID project, with very good results.
Third channel is a probe.
So I have some direct experience with the actual chip.
It has a big brother by the way, the AD8318, dual channel, with phase output.

I have accidentally "blown" one sensor already, and have bought twenty AD8307 chips for replacements.
So they can easily be damaged in use.

I use a "Power Limiter" when working with transmitters etc.
It's just two 1N5711 diodes back to back in a male BNC plug, and used with a BNC "T" adaptor.
I remove it for low risk accurate measurements.
Purists will smile, but it works well.

Enjoy your probes.

Applause please for Analog Devices!
(No connections)
.

[G0HZU_JMR]

I've still got my old homebrew power head here that uses the AD8307. This must be nearly 25 years old now. I recall I optimised the input network such that the input VSWR was very low and the response was quite flat up to about 150MHz.

One thing to be wary of when calibrating the meter is that the AD8307 is very sensitive to (odd order) harmonic energy. It isn't a true average power sensor so I would recommend calibrating it with a sig gen that has low harmonic distortion. Ideally the odd order harmonics should be suppressed by 50dB for the purposes of calibration but 30dB suppression will be enough for most users.

If a decent attenuator is always fitted at the input to the AD8307 it will improve the VSWR and minimise measurement uncertainty. I often used mine to measure the insertion loss of low loss RF filters or cables. However, an RF filter is one example where the harmonic distortion of the test source needs to be really low or the result could be corrupted by the harmonics. If the source harmonics are as bad as -30dBc then this can cause significant measurement errors in this case. A lot depends on what you want or expect from the AD8307. If used carefully with regard for some of its limitations it is an excellent logamp.

[G0HZU_JMR]

To give an extreme example I've set up two decent Agilent RF sig gens to share the same 10MHz reference. These sig gens have a variable phase control which will be needed shortly. The RF filter under test is an 11MHz LPF (50 ohm) and I want to know the insertion loss at 5MHz.

If I measure it with a VNA it shows 0.19dB loss at 5MHz. If I measure it with a sig gen and a lab RF power meter it measures 0.20dB loss at 5MHz. The source impedance of the sig gen has bee improved with a resistive attenuator to have a VSWR better than 1.05:1.

If I measure it with the AD8307 power meter (input VSWR of the AD8307 power meter is also better than 1.05:1) I see a detector difference of 5.4mV.
If I assume 25mV/dB then this is a loss of 0.216dB. Not a bad result but I used a clean lab sig gen with harmonics better than -60dBc for this test. Next I'll repeat the test with a -20dBc third harmonic present.

I can combine the other sig gen using a high isolation combiner (still has 1.05:1 source VSWR) and then set the frequency of the second sig gen to 15MHz to mimic a third harmonic.

I can then control the amplitude and phase of this third harmonic. If I set it to -20dBc and connect the combiner to the AD8307 power meter I can then rotate the phase of the third harmonic such that the AD8307 detector gives the lowest reading for the combined 5MHz fundamental and the 15MHz third harmonic at -20dBc. By doing this I'm setting up the worst case extreme of the uncertainty window caused by the third harmonic.

I then note the AD8307 detector voltage and then insert the lowpass filter again and attempt to measure the insertion loss just like I did before. However, something crazy happens when I now look at the AD8307 readout. The detector voltage RISES by 23mV instead of falling by the expected 5.4mV.

Because there is a -20dB harmonic present at the worst possible phase angle, the AD8307 shows an increase in power level when I put the filter inline. In other words, If I trusted the AD8307 to be a 25mV/dB power meter I would have to assume the lowpass filter had about 0.92dB gain instead of 0.2dB loss.

If I set the phase angle of the third harmonic to the other extreme the AD8307 reports a -27mV change. This would indicate the filter had an insertion loss of 1.08dB. The correct result should be about 0.20dB but because there is a -20dBc third harmonic present the meter has an uncertainty window of about +/- 1dB for this simple measurement. The error is affected by the phase angle of the third harmonic.

There are equations to predict the uncertainty but I would just recommend that the AD8307 is always used with a sine wave source. Keep the harmonic energy of the source as low as possible.

[Radio1950]

David

I cant help but think that unless you really want to use these probe PCBs as actual probes like an old VTVM probe, that you may be better to convert one to be in a more rugged sensor box.

This is what HP, Anritsu, et alia, do with their sensors.

The originals, as per photo, if that indeed is yours, are way too fragile to use in practice.
And you may be better to change the SMA to another type eg BNC.

You can remove the SMA very easily with Junior Hacksaw, solder iron, and care.
You could saw that PCB in half, to alter its size, and just wire it together electrically.

And if you wished to convert one into a higher input impedance, you could perhaps remove the PCB terminating resistor, and convert to approx 1000 ohms Zin??

I probably wouldn't do this as most RF sources are 50 ohm by deliberate design, but you did ask for suggestions.
There is a place for a separate low capacity hi Z RF Probe.


The basic device in concept is very useful to measure RF Power, and coupled with an external power attenuator, or "T" Tap, can be useful for measuring transmitter power.

In its present form, it could be useful for work on RF Test Equipment internals, but you may have to get some adaptors eg SMA F to SMB M and F, etc.

Photo of my RF Sensor, using a similar PCB, attached to possibly give you ideas.

good luck

[Radio1950]

Jeremy,

following on from your conversation, may I please ask a question?

If we combine two unrelated sine CW signals, one L1 say 0 dBm, the other L2 -10 dBm, what will be the sum power level L3 as observed firstly on a voltage measuring sensor eg AD8307 etc, and then with a bolometer (thermal) sensor (eg HP43x and Sensor).

Next, at what level of the second signal L2, is the original sum L3 affected by 1 dB, ie if you vary L2 up and down.

The application of this scenario is course the practical effects of harmonics and spurious.

Thanks R50
.

[G0HZU_JMR]

Hi R50

For the regular average reading power meter if you sum two unrelated signals of 0dBm (1mW) and -10dBm (0.1mW) you should see 1.1mW.

The AD8307 will behave differently. The slope and intercept given in the AD8307 datasheet is for a sinewave. If you feed it anything else it will give a different result. In your example however, the error will be negligible.

It is much more pronounced when the second signal is an odd order harmonic (eg third harmonic)

I think it over-reads for square waves and under reads for triangle waves.
These last two examples are relevant to my earlier example because if you rotate the phase of a -20dBc third harmonic you can make the fundamental and harmonic look a bit like a triangle wave on a scope. This is why the AD8307 under reads in this case. If the phase is rotated some more the waveform will start to look a bit like a squashed square wave on a scope and the AD8307 will over-read this waveform. It will only read correctly at a certain phase in-between these two cases

Another extreme (but classic) example would be if you feed the average power meter 1mW of noise power spread over a 50MHz bandwidth (say from 200kHz to 50MHz) it should still read 1mW as long as the sensor is flat across that frequency range. If you do the same with the AD8307 power meter it will under read the average power by (in theory) 2.51dB because it uses a log detector.

Have a look at the pdf below from Analog Devices. There's some correction factors for waveforms like square, triangle and noise given in a table near the end.
Be careful with the correction factors given in the table. For noise it reads +2.51dB. This means it under-reads noise and you have to add a fudge factor of 2.51dB to get the true average power of the noise. For a square wave it says -3.01dB and this means it over-reads a square wave and you have to subtract 3.01dB from the readout for a square wave.

[G0HZU_JMR]

Don't feel too bad about these issues with the AD8307 logamp. The designers of spectrum analysers have had to deal with (as in correct) this issue with noise for decades.

If you read some of the old classic HP app notes for their spectrum analysers this issue is described in detail because a logamp will be scaled for a sinewave. When the logamp stage in the spectrum analyser has to deal with noise it under-responds to noise by 2.51dB just the same as in the table in the Analog Devices app note. Sometimes this can be a nice kind of error for the user because weak cw signals appear clearer on the analyser display due to the under-response to noise. However, when the analyser has to faithfully report the noise power with a noise marker function it has to add fudge factors to account for the detector type and the logamp effect.

[G0HZU_JMR]

I put together a quick demo using the two combined sig gens and the AD8307 power meter and an old HP digital scope (set to 50R input) that can read Vrms of complex waveforms. I set the first sig gen to give 5MHz at about 0dBm out of the combiner and this should be 224mVrms and this is the top left scope waveform and you can see the scope measures 221mV rms. The AD8307 sensor measured 1826.3mV.

Then I turn this off and turn on the other sig gen to 15MHz at -10dBm and this is about 70mVrms on the scope. The AD8307 sensor measured 1561.8mV.

I then turned both sig gens on together and rotated the phase of the harmonic to get the triangle looking waveform in the top right corner. This shows 234mVrms which is very close to the expected 1.1mW. The AD8307 sensor measured 1686mV. The level from the AD8307 has dropped (a lot) when it should have gone up very slightly.

I then rotated the phase some more and this gave the waveform in the bottom right corner that looks a bit like a ripply square wave. The scope still reads 234mVrms. However, the AD8307 sensor now reads 1876.6mV.

The swing shown on the AD8307 for the phase change is 1876.6 - 1686 = 191mV. This is an uncertainty window of 191/25 = 7.6dB in this case.

Clearly the AD8307 sensor is not the right tool to use for this waveform because it has a high level of third harmonic present. You can see the AD8307 under reads when the waveform looks like a triangle wave and it over-reads when it starts to look like a square wave even though the Vrms of the waveform is unchanged.

The scope waveforms can be seen in more detail in the link below.

https://www.qsl.net/g/g0hzu//RF%20Tr...07_scopeWF.gif

It should then be easier to read the text on the scope.

[Radio Wrangler]

Originally Posted by G0HZU_JMR

If I assume 25mV/dB then this is a loss of 0.216dB. Not a bad result but I used a clean lab sig gen with harmonics better than -60dBc for this test. Next I'll repeat the test with a -20dBc third harmonic present.

These devices are successive detection type loggers. Their error function within their good operating range is roughly a sinusoid with an amplitude comfortably smaller than 1dB and a period of several dB. So the usual assumption that error magnitudes reduce over smaller spans doesn't hold for certain ranges of spans and the rate of change of error versus shift in level can therefore be surprising. They are very useful, but you have to be aware of their quirks.

Essentially, each detector in the offset array of them is a rectifier, performing peak voltage detection of the signal. They aren't true RMS detectors detecting power. This comes to the fore when you start detecting complex signals with multiple components, whether they are independent, harmonics, modulated signals or intermod products. It's rather like a multimeter on AC volts. It reads peak, but the scale is made to give you RMS assuming the waveform is a sine, but it'll have an error for all other waveforms.

Successive detection loggers are the usual sort in spectrum analysers, and therefore they too fall victim to the peak/RMS difference. If you set the analyser up with narrow enough resolution bandwidth to separate all the components of a signal, then there is only one component, a sine, getting to the logger at any one time and all is well. If you open the bandwidth up and let multiple components in, for example, trying to find the total power in a modulated signal, then you have to remember it's peak sensing.

Fancier analysers allow you to choose peak detectors or sample detectors.

Especially on noise, another quirk of different loggers arises. The only tool which can get a grip on noise is statistics. Averaging is essential. The loggers in many spectrum analysers do the logging before detection, and then average afterwards. The way people think of detecting noise level is to have a wide dynamic range detector, then an averager, then a log converter.

These two schemes aren't the same. The having the logger ahead of the averager creates an error of a dB or so on gaussian, unclipped noise. If the crest factor is limited, then the error factor changes. Fancy analysers have a 'noise marker' function which applies correction factors for the noise bandwidth of the IF filters not being quite the same as their 3dB bandwidth, and for the effect of logging before averaging.

These little loggers are amazing beasties, but they have a few foibles.

Log detectors are also useful in receiver AGC as they complement the roughly linear dB of gain/volt of IF amplifiers and give you more uniform AGC loop gain across the range of signal levels.... so your AGC is more stable and uniform in action.

David

[G0HZU_JMR]

Yes, it's worth investigating the limits of these logamps when used as a power meter.

Last night I tried a triangle wave from a function generator and the AD8307 under-read by 0.89dB. The table in the AD app note shows the error is 0.9dB. This was very close but there will be some uncertainty in the test gear so there's some luck involved in seeing a result so close.

I also dug out a pulse generator and set the duty to 50% and measured a square wave. The AD8307 should over-read by 3.01dB but the closest I could get was +2.87dB. I think this is partly because the square wave isn't perfect and the bandwidth of the AD8307 isn't infinite. It was still a pleasing result though.

A couple of years ago I wrote a VB app to test my AD8307 detector across a fair chunk of the log range. I'm lucky to own some very accurate sig gens in terms of the attenuator step sizes so this means I can try and replicate what is on the AD8307 datasheet. The results were very good. Looking at the program code it looks like I've used a slope of 26.2mV/dB for my AD8307. I must have used this value to get the best fit for the slope. I'll check it again later today.

It's also worth understanding how the AD8307 power meter can get things wrong with certain types of modulation. I'll post up a few examples later.

[G0HZU_JMR]

I tried AM modulation but the time constant after my AD8307 is too fast so it only gives a steady reading with high modulation frequencies. I tried 60% modulation and the error was quite small.

What isn't so good is a two tone test. With spacing of (say) 100kHz the power after a second tone is added should go up by 3dB but the power level hardly changes with the AD8307. It only goes up by a small fraction of 1dB. It really is meant to be used with a sine wave.

I also tried the noise test using a noise source covering 0.2-50MHz. I've done this test a few times in the past so I already know the result. I measured the noise with a thermocouple power meter with a LF head fitted and adjusted the power to -6.00dBm or 0.25mW. I then measured it with the scope and it reported 110mV rms which is a very good result at 0.24mW.

The AD8307 measured the power of the noise waveform with an error of -2.44dB and this is very close to the theoretical -2.51dB.

[G0HZU_JMR]

Here's some old trivia from work about the logamp accuracy of a spectrum analyser. About 30 years ago at work the best spectrum analyser we had was a nearly new HP8568B and this has a noise marker function.

However, because this analyser uses a traditional logamp and analogue resolution bandwidth filters it wasn't very consistent at measuring absolute noise power. The overall measurement uncertainty was quite significant and it often showed slightly different readings with the noise marker depending on span and RBW setting.

To overcome this the company bought a couple of very expensive Telonic bandpass filters centred on 70MHz (where the noise power accuracy was needed).

These filters were regularly sent to the National Physics Laboratory (NPL) to have the noise bandwidth offset factor measured in dBHz. The idea was to put a precision/flat noise source through the 70MHz BPF and feed this into a thermocouple power meter. This can measure the noise power (after the filter) very accurately. NPL were able to provide us with an accurate dBHz fudge factor where we subtracted (say) 66.15dB from the reading on the power meter and this then provided a very accurate noise power level at 70MHz in dBm/Hz. The analyser could then be calibrated against this. The idea was that the accuracy of the noise measurement was transferred from the power meter to the analyser as long as the filters were consistent. Luckily they were effectively the same every year.

These days this isn't needed any more because all the log display and noise detection is done in the digital domain and the overall level calibration of a modern spectrum analyser is a lot better today.