A Dilemma when evaluating a RF Power Sensor

[NY2KW]

I bought a Agilent 8481D diode power sensor at a great price and wanted to verify it’s linearity and sensitivity down to its rated -70dBm or better.  I have a R&S SM300 signal generator and a Siglent SVA1032X and I know these correlate well if I set the SM300 for any signal greater from -100 dBm signal and up.  I also have a recent NIST traceable calibrated HP438A (I am the original owner) that has worked fine for many years.  After calibrating the 8481D sensor at 50MHz to the HP438A, I found the lowest power I could read was with the SM300 set down to -55 dBM.  Any lower on the SM300 and the power meter/sensor just stayed at -55dBM.   OK maybe the sensor had seen better days,  but then I noticed that when I turned off the SM300 RF, the power level on my HP438A was still at -55dBm.    I disconnected the sensor from the SM300 output and the HP438A+sensor reading dropped to -90 dBm.     Now I wondered if my SM300 was leaking RF even with RF switched off,  so I hooked the SMA300 (with RF turned off) it up to my SA with same coax and looked at many frequency segments from 1 Mhz up to 3GHz and could not find any leaking signal and the floor remained below -90 dBm.   All instruments are grounded together and I did an quick check around the workbench using the SVA1032x and some EMI probes and saw nothing.   I even wrapped the sensor with copper foil but no difference. As a final check that maybe the sensor was somehow screwed up – I used a Rigol signal generator and some attenuators and found the 8481D sensor and HP438A read RF power accurately down to -80 dBM.   

Any suggestions as to what’s going on?

[Tony_G]

Did you check the whole range of frequencies? These sensors will measure average power across the entire range so if you have harmonics you'll measure those.

You might also want to set it to trace max hold in case you're getting spikes rather than constant values.

Finally, are you getting any DC out on the SM300? The 8481D will have a DC blocking cap but maybe that cap leaks.

TonyG

[M0HZH]

Is it possible that the SM300 leaks 50/100Hz from the PSU ? It wouldn't be visible on the SA as it starts from 9kHz.

I know it doesn't apply as they're not diode sensors, but even if they're specified from 100kHz onwards my 8482A and 8483A seem to be able to pick up signals that low.

Edit: that would be 60Hz/120Hz on your side of the pond .

[NY2KW]

Thanks for the suggestions.  DC block didn't help.  I did check different freq ranges and no spikes.  Since the sensor is a diode, I think it would be relatively impervious to spikes/transients, these would need to be pretty big to give an average power reading that is 20dB over the background noise power floor (-80 to -90dbM)

Weird huh?

[George Edmonds]

Hi

When calibrating the 8481D sensor to the HP438A power meter 0dBm output did you have the 11708A 30dB attenuator in place?

George G6HIG

[jjoonathan]

Old RF signal generators sometimes have considerable broadband noise in their heterodyne band: https://photos.app.goo.gl/f1iYq253oGGrXJ4a6

[NY2KW]

When calibrating the 8481D sensor to the HP438A power meter 0dBm output did you have the 11708A 30dB attenuator in place?

Yes, i should have mentioned that you always need the 30dB attenuator for the 8481D calibration

Old RF signal generators sometimes have considerable broadband noise in their heterodyne band

I appreciate that but wouldnt that be an issue with the SM300 RF 'On" and sweeping with switching?  My problem is some unknown source of RF at the output when RF is turned off.  Also, when RF is turned on, it doesn't add to the signal power, i.e if I set the RF outpit to -45dBm or -50 dBm then the HP438A/8481D reads correctly.  The problem is that as I lower the RF output below -55 dBm, the power reading does not change.  AND it stays at -55 dBM with RF powered off.    Yet if I measure the SM300 signal output on my spectrum analyzer the power levels set read all correctly all the way down to -100 dBm.

[jjoonathan]

The "old RF generator" noise floor comes from the combination of a mixer and amplifier with bad noise figure -- it does not require the RF signal to be on, just for the path to be active. The designers could have deactivated the mixer and amplifier when RF was turned off, but they often didn't, presumably to achieve the fastest/smoothest possible pulse modulation edges.

If you look at it on a spectrum analyzer, the -55dBm of noise power will be distributed across GHz of bandwidth, so a -55dBm sinusoid would rise far above the floor, even though the floor and the sinusoid each contain the same power. Also, as you reduce RBW, the sinusoid will remain at the same level on the analyzer while the broadband noise will fall, just like the noise floor of the spectrum analyzer falls, because that is the nature of broadband vs narrow band signals.

EDIT: I just confirmed that this is exactly the case on my HP8340B: https://photos.app.goo.gl/Get72X1k5dz8ojkA6

[NY2KW]

I will take a video of the SM300 sweeping both freq and level... i don't see the kind of background noise as you do with your HP8340B.     I will try to take a video.  I did the poor mans experiment and loosely coupled the SM300 output to one of my ham radio UHF transceivers, as a background signal of -55 dBm of broadband noise should be very noticeable but when I AC powered the SM300 on and off there is no change in noise level.    This is weird.

[jjoonathan]

What happens if you turn off the SM300? Not disabling the RF, I mean physically turning off the instrument.

In any case, I am still suspicious that you are looking for broadband noise with narrowband receivers and therefore not finding it, even though it's there. Both a spectrum analyzer and a UHF transceiver are extremely narrowband compared to the power sensor.

For instance, the 8340B (see attached -- a still from earlier) has -29dBm of broadband noise, yet even with RBW wide open my SA never sees more than -50dBm at a time, and reducing RBW would make it even worse.

[NY2KW]

My Siglent SVA1032X on full sweep has a max RBW/VBW of 1 MHz and the floor is just below -60 dBm.  Powering the SM300 on or off does not change the floor noise with it connected to the SA.    The correlation between the SM300 output, whats measured on the Siglent and the observed power on the HP438A/8481D has been incredibly consistent and less than a 1.0 dB difference between the 3 as long as the SM300 is set for -55 dBm or higher.  So there is a 5 db space between what i see on the SA and what the power meter is measuring.  I understand what you are proposing and I have tried to find the R&S SM300 specs on this.   I know the power meter/sensor works fine which was my original objective :-)

Thanks again for your generous help

Jerry NY2KW

[xmo]

Try this experiment:  First put a 50 ohm termination on the spectrum analyzer input.

Set the analyzer to some convenient center frequency with a span of your choice, 100 MHz would be fine.  It doesn't have to be full span.  Manually set attenuation to 0 dB.

Now go to the marker function menu and activate the noise marker function.  The instrument will correct for display factors and read out an actual noise floor value, normalized to a 1 Hz bandwidth.

If the instrument was "perfect", that is if it had a 0 dB noise figure, the reading would be the theoretical N=kTB value of -174 dBm/Hz.  The difference between -174 and your marker readout is essentially the noise figure of the spectrum analyzer.

Now, if you connect the the signal generator and turn on the AC power with the RF off - see if you get an increase in the noise marker readout.

A broadband noise source 18 GHz wide with a level of - 155 dBm/Hz would cause your power sensor to see -55 dBm or so.  If your instrument settings aren't showing you a noise floor better than -155 dBm/Hz, then you wouldn't see difference between the signal generator and a 50 ohm termination.

[NY2KW]

A broadband noise source 18 GHz wide with a level of - 155 dBm/Hz would cause your power sensor to see -55 dBm or so.  If your instrument settings aren't showing you a noise floor better than -155 dBm/Hz, then you wouldn't see difference between the signal generator and a 50 ohm termination.

I set the SVA1032x set to 50Mhz, span 100Mhz, Pre-amp on, Zero attenuation and RBW/VBW = 1MHz.  I did trace averaging.  With a 50 ohm termination to the SA - the noise floor was about -164 dBm/Hz.   Attaching the SM300 signal gen with a 3' coax, powered up but RF off, the SA noise floor rose to about -154dBm/Hz.   

[xmo]

The pre-amp definitely helps by increasing the instrument's sensitivity with a lower noise figure.  To get an accurate reading of noise power it is best if your instrument noise floor is 10 or more dB lower than the noise.  If it is closer, there are formulas and/or nomographs to help get a more accurate value, but because your reading with the pre-amp on meets that objective, I think that your -154 dBm/Hz reading is spot-on.

That matches what you would expect to see based on what your power sensor reading is giving you - so - you have your answer.  The issue you were seeing is based on broad-band noise - not spurs or harmonics above the spectrum analyzer's highest frequency.

I don't have an SM300 but I do have an SML (and an SMV) so I tried the same experiment but they don't put out a broad-band noise floor visible on either a spectrum analyzer or a power meter.

That still leaves the question - do all SM300's do this or just yours?  Perhaps someone else on the forum has an SM300 available to test.

[NY2KW]

Thanks to all for this informative discussion.   So with the SM300 powered up, I get ~10 dBm/Hz of added broadband noise over the SVA1032x floor.    I thought the noise power (dBm) was approximated by (−174 + 10 log10(BW)).  So to equate to a power measurement around -55dBm, the SM300 IF BW would have to be more than 25 GHz.   Is that a valid assumption?

[xmo]

That assumption would require the noise amplitude to be totally consistent across that entire frequency range which I suspect that it is not.   However, it certainly does appear to be very wide.

As an experiment, you could look at the noise at various frequencies within the range of your analyzer - say every 100 MHz or so - to get an idea of how flat the noise is for at least the first 3 GHz.

You might want to look at the Agilent / Keysight application note 1303 "Spectrum Analyzer Measurements and Noise" - 5966-4008E.  There are several versions out there (e.g. 2002, 2009, 2012) as it has been periodically updated.   

[TurboTom]

I know I'm late here to reply but I checked the noise issue with my SM300 on a Rigol DSA815TG.

Without the SG connected, I got similar noise readings as @NY2KW with someting like -163dBm/Hz. Yet, with the SM300 connected, there wasn't any noticeable increase of the NF observable. I tried different power settings while the output was off (in case the output amplifier gets reconfigured) and also output some small signals which all didn't affect the noise level.

But if I operate the setup at slightly higher power settings (had to change input attenuation of the SA) and I change the output level of the SG with the output active across the PA enabling threshold (i.e. from -26dBm to -25dBm), I get an increase of noise level by almost 20dB (-157dBm/Hz -> -138dBm/Hz) which can be expected due to the additional gain block routed into the signal path.

If your SM300 always has some raised noise present at the output, it may well be possible that the output RF switch is defective (SkyWorks AW002R2-12) and it doesn't disconnect the PA properly.

[RadioNerd]

Thats indeed a strange behavior.
Older synthesizers can have substantial  wideband noise, that's true.
However, I would not expect this noise to remain constant with output power setting but to decrease as the output attenuator steps are engaged.
Your unit may therefore have a weird step attenuator issue. As it seems to be a solid state attenuator, the switch ICs or their power supply might be degraded, damaged and/or noisy.

[Joel_Dunsmore]

Did you zero the sensor before you calibrated the sensor, and again after? (probably you did, but you don't mention it)  with no signal (e.g. put a load on the power sensor) and hit zero, it should read lower than the min power value on the power sensor label.  After it sits for a bit you will see it climb up and eventually it will reach some larger value but usually I see this on the order of no more than 10-15 dB above the minimum power label. BUT, the first time you connect a sensor there are lots of different offsets that have to be removed.  Some of the residual offsets drift so the specs only apply immediately after zeroing (and as far as i know, there is no specified time frame for "immediately").

[fenugrec]

I have a very similar question, and similar weirdness, with a Boonton 4210 and its 4210-4A sensor (full-wave diode based; -60 to 10 dBm range).

- sensor terminated, after zeroing the meter : -70dBm. Ok, everything normal.
- hooked to an hp 4195a generator : readings make sense until about -40dBm;
- I tried hooking up 40dB of attenuators between the 4195 and the power sensor :  I never saw any number below about -50. Even with the 4195 off I read -51 !!

Maybe some residual DC from the 4195 ? But then, how am I still reading -51dbm with the 4195 off , with 40dB attenuation before hitting the sensor ?

Could it be some kind of noise coupled into the shield / chassis on everything in the lab ? As soon as I disconnect the cable from the 4195, readings start to drop towards -65  as expected..

So if anything below about -50dBm is so troublesome, how does one even verify / use one of these ?

[xmo]

If you put the termination back on the sensor, does it go back to reading -70?

Try putting the attenuator on the sensor and then put the termination on the attenuator.  After zeroing, connect the sensor + attenuator to the signal generator.

[fenugrec]

If you put the termination back on the sensor, does it go back to reading -70?

Yes. (well it displays "-LO-" which means the same thing; switching the units to PWR shows 0 nW +/- 0.1nW. Readings jump around a bit at this level, obviously.

Try putting the attenuator on the sensor and then put the termination on the attenuator.  After zeroing, connect the sensor + attenuator to the signal generator.

No change : as soon as the shields make contact (sig gen on or off, doesn't matter), reading goes up to -57 . Maybe there's something wrong with the grounding / cabling in my power sensor+meter then ?

In the attached photo, I am holding the cable connected to the siggen, and as soon as I touch the attenuators, the reading changes.

[Joel_Dunsmore]

if touching the cable shield to the power sensor shield causes a reading, then ground loop is definitely a concern.  You can try moving both instruments to a single power strip that has nothing else connected to them and see what you see.   That will ensure they both have the same neutral and ground and nothing else is pulling current in either leg.  Then again, it could be low voltage: I see your marked location as Canada.  I don't remember the exchange rate, but isn't 120V Canadian only something like 94V U.S.?

[fenugrec]

if touching the cable shield to the power sensor shield causes a reading, then ground loop is definitely a concern.

Gross... I naively thought such a power sensor would be immune to that kind of problem, but it's exactly what it is. siggen is indeed on a different power strip.

I don't remember the exchange rate, but isn't 120V Canadian only something like 94V U.S.?

No no, we have proper Metric volts here, thank you : )

[xmo]

Possibly a fault in the sensor cable shielding is allowing noise ingress?