Author Topic: How to Find the Noise Floor with a Spectrum Analyzer (Read 15352 times)

Electro Fan · « **on:** February 13, 2016, 11:28:31 pm »

With a spectrum analyzer, what is a good method for determining the local noise floor (in the lab area around the spectrum analyzer)?

If for example we wanted to determine the noise floor from 0 to 1 GHz would it be best to run a sweep using a full 1 GHz span, or would it be better to pick various frequency ranges with smaller spans? The reason for this question is in part to find a practical way to determine the noise floor across the full range, but also it is driven by the consideration of antennas. One antenna might be good for low frequencies (say a standard AM antenna that is designed for roughly 500kHz to 1.6MHz), another antenna might be designed for FM radio from about 88-108MHz, another might be better further up in VHF and/or UHF, etc.). I'm primarily interested in HF, VHF, and the lower end of UHF. It is as simple as using an appropriate antenna for each frequency range, adjust the span and center frequency accordingly, and take a look-see? (Or is there a preferred method that doesn't use an antenna?) Is there a better approach?

Another question, how would you recommend setting RBW and VBW?

Also, any wild guestimate on how much noise should be found in a standard home lab? Does -85 to -90 dB for HF and -95 dB for VHF sound plausible?

Thanks

uncle_bob · « **Reply #1 on:** February 14, 2016, 01:31:28 am »

Hi

OK, backing up a bit. For a spectrum analyzer you do it fairly simply. You put a 50 ohm termination on the input and see what you get. The floor will be some number in dbm per square root Hz. This is also called "normalized to 1 Hz" on some analyzers. As you change bandwidth settings on the analyzer, the displayed number will change if you do not have a "normalize to bandwidth" button on the scope. If you don't have that button you will also need to calculate the actual noise bandwidth of the filters in the analyzer. That can be more or less exciting depending on the vintage of the analyzer.

Next up is video bandwidth. This is just an averaging process. Depending on how it is done in your specific analyzer, there may or may not be a correction for it. On older analyzers there is also a video peak to average correction (don't ask if you have a modern unit).

So let's assume you now have an analyzer that is calibrated and properly measuring noise in dbm / sqrt(Hz). You take a look and you might be getting -171 dbm. If so your analyzer has a 3 db noise figure front end. You might also get -131 dbm. In that case, your analyzer may still be in spec. It all depends on which one you bought and which options you got with it. Before we ever do anything with an antenna, there could be a 40 db delta....

Now we grab an antenna. At least it says it is an antenna. Consider that a half wave dipole at 1 MHz is 150 meters long. That thing in your hand is *not* quite the same thing, regardless of what it says it's lower frequency limit is. If you are reading low frequencies, you are using a probe rather than an antenna. The probe is likely either designed for E field or H field pickup. It has some magic calibration factor of "this field equals that dbm" buried in the spec sheet. It may have a 20 db amp with it, it could have a 60 db amp. There's another 40 db delta

We move up to a frequency (say 500 MHz) where we *can* build an antenna. It's either a monopole or a dipole. We read up on all the design details and have the directivity and gain all worked out. We hook it up to our -171 dbm analyzer and in the quite spots we get .... -171 dbm. Everyplace else you get discrete signals that are related to this or that. You can see the local TV stations. You can see all sorts of stuff. How well can you see them? It depends on your building construction and where you are in the building. I have seen people claim (and the FCC actually accept !!!) that they have 120 db buildings.

So what will you read in your experiment (after calibration) ... -171 to -91 with the 80 db fudge factors above. Does this matter? ... err ... do your people carry around cell phones? If so, the phones are likely checking in with this and that on a regular basis. When they do, you have at least +20 dbm on your analyzer. Do you have an in-plant pager system (RF not audio). Any bets what happens when it goes off? Same issue with in-plant cell repeaters. Then there's WiFi (outside your band).

Assuming you have tossed out all of everybody's RF gear on a permanent basis, what is reasonable? No matter how hard you try, the broadcast stations are going to be there. If you can hunt down every source in your lab and get them below what the broadcasts peak at. That's about as good as it gets. Good luck tracking down that final UPS that's spraying crud from DC to light ....it can be done .. just keep at it.

Bob

Electro Fan · « **Reply #2 on:** February 14, 2016, 06:52:13 am »

Quote from: uncle_bob on February 14, 2016, 01:31:28 am

Hi

OK, backing up a bit. For a spectrum analyzer you do it fairly simply. You put a 50 ohm termination on the input and see what you get. The floor will be some number in dbm per square root Hz. This is also called "normalized to 1 Hz" on some analyzers. As you change bandwidth settings on the analyzer, the displayed number will change if you do not have a "normalize to bandwidth" button on the scope. If you don't have that button you will also need to calculate the actual noise bandwidth of the filters in the analyzer. That can be more or less exciting depending on the vintage of the analyzer.

Next up is video bandwidth. This is just an averaging process. Depending on how it is done in your specific analyzer, there may or may not be a correction for it. On older analyzers there is also a video peak to average correction (don't ask if you have a modern unit).

So let's assume you now have an analyzer that is calibrated and properly measuring noise in dbm / sqrt(Hz). You take a look and you might be getting -171 dbm. If so your analyzer has a 3 db noise figure front end. You might also get -131 dbm. In that case, your analyzer may still be in spec. It all depends on which one you bought and which options you got with it. Before we ever do anything with an antenna, there could be a 40 db delta....

Now we grab an antenna. At least it says it is an antenna. Consider that a half wave dipole at 1 MHz is 150 meters long. That thing in your hand is *not* quite the same thing, regardless of what it says it's lower frequency limit is. If you are reading low frequencies, you are using a probe rather than an antenna. The probe is likely either designed for E field or H field pickup. It has some magic calibration factor of "this field equals that dbm" buried in the spec sheet. It may have a 20 db amp with it, it could have a 60 db amp. There's another 40 db delta

We move up to a frequency (say 500 MHz) where we *can* build an antenna. It's either a monopole or a dipole. We read up on all the design details and have the directivity and gain all worked out. We hook it up to our -171 dbm analyzer and in the quite spots we get .... -171 dbm. Everyplace else you get discrete signals that are related to this or that. You can see the local TV stations. You can see all sorts of stuff. How well can you see them? It depends on your building construction and where you are in the building. I have seen people claim (and the FCC actually accept !!!) that they have 120 db buildings.

So what will you read in your experiment (after calibration) ... -171 to -91 with the 80 db fudge factors above. Does this matter? ... err ... do your people carry around cell phones? If so, the phones are likely checking in with this and that on a regular basis. When they do, you have at least +20 dbm on your analyzer. Do you have an in-plant pager system (RF not audio). Any bets what happens when it goes off? Same issue with in-plant cell repeaters. Then there's WiFi (outside your band).

Assuming you have tossed out all of everybody's RF gear on a permanent basis, what is reasonable? No matter how hard you try, the broadcast stations are going to be there. If you can hunt down every source in your lab and get them below what the broadcasts peak at. That's about as good as it gets. Good luck tracking down that final UPS that's spraying crud from DC to light ....it can be done .. just keep at it.

Bob

Ok then, no problem, I'll just get right on that process... sounds easy...

Just kidding, thanks for the detailed info.

I can't get to -171 dBm; best I can do according the specs on a HP8561E is ?145 dBm

I'm not expecting to find no noise or eliminate all noise. Maybe I worded the heading wrong. I'm just trying to see what my actual noise floor is given whatever noise is present - I don't need to see if HP was right about the -145 dBm

As for how accurate is the SA, fwiw when I put reasonably small signals into it from a couple signal generators the SA seems to line up with the sig gens to within about a dB (and I think most of the difference comes from the cable).

So all I'm trying to figure out is how much noise is in the vicinity. The reason for this is that I'm starting to get interested in amateur radio and before I get too carried away with antennas and receiving and sending signals I'm curious to see what the baseline environment is all by itself.

A question back for you: when you say that at "low" frequencies the "probe" (antenna) is likely designed for E field or H field, how "low" is the frequency threshold you are speaking of what are the typical reasons an antenna would be designed for one (E) or the other (H)? (I was under the impression that an antenna creates both E fields and H fields.)

Thx again for the very thoughtful and informative reply.

T3sl4co1l · « **Reply #3 on:** February 14, 2016, 02:06:09 pm »

FWIW, specs are usually designed for amplitude accuracy more than noise floor, so as long as you can attach a LNA ahead of the spec, and measure the noise floor of the amp and/or whatever it's attached to, you can perform accurate difference measurements. Of course, dynamic range suffers by the same token (the LNA might have excess distortion or clipping for > ~0dBm signals, say), so keep that in mind too.

If you're doing an EMC sort of survey, you normally want RBW = 9kHz (if you don't have it, 10 is close enough) with quasi-peak detector, for the low frequency range (150kHz to 30MHz; which is normally measured as conduction along wires, not with an antenna). Peak detector is okay; it will overestimate some, depending on the modulation waveform. 100kHz RBW and peak detection is used for the radiated range (30-1000MHz). This would normally be done only to calibrate a shielded test room, with a generator outside and a detector inside. But open air tests do need to subtract ambient sources from the measurement, so this would be an important step.

As far as what to expect, for HAM purposes -- you will be able to tell, qualitatively, with a very basic antenna (a hunk of wire poked into the BNC, perhaps?), what the major broadcast stations are. You probably won't get a good idea of nearby switching noise and whatnot (in the < 100MHz range, say), for which you'll need a suitable antenna (conical dipole?) and enough sensitivity to detect it (really, what you're ultimately concerned about is, how low is the ambient noise floor, at levels where your receivers can detect it?).

Tim

uncle_bob · « **Reply #4 on:** February 14, 2016, 02:31:50 pm »

Hi

Ok, -145 is not surprising. Often there is an optional external pre-amp that will get you another 20 db or so. As mentioned above that preamp does not help the dynamic range at all. You don't see a lot of them out there.

The 2 meter ham band is at 144 MHz. A half wave antenna there will be about a meter long. If you are playing around informally, that's about the biggest thing you want to wave around the lab. There are indeed setups that go lower and stay somewhat compact. To me, not worth the effort.

The E field and H field stuff is actually fairly simple. You either couple to things with a capacitor (E field) or with a coil (H field). In the good old days, we just wound little coils on the end of pieces of coax to probe things. It was called a often called a sniffer (coil or capacitor). For an E probe, put a clip (from a broken clip lead) on your scope probe. Wave it around the circuit. Instant capacitive coupling. Why use one over the other? Voltages create E fields, currents create H fields. Close to a circuit (near field) you can indeed have one without much of the other.

Oddly enough, the guy with all the neat toys ... Dave .. has a video on using E field and H field probes. He's also got a really nice set of probes and an amp. It's certainly on my list of things to steal if I ever drop in for a visit

(let's see .. hop in the car ... start in Pennsylvania ... head for Australia ... hmmm ..).

In terms of chasing things down, I was not kidding about that stupid UPS. You will find things around your lab or house that are nasty RFI emitters. Some can be dealt with. Some can not be dealt with (for various reasons). Before you put up antennas, it's nice to know what is where. Normally, the RFI in your lab (or shack) is secondary to the area around the antennas. Modern shielded cable works pretty well. You will always have things like computers and the like making crud locally.

E fields do not go through houses very well at all. Most LF and VLF you can pick up is E field. A 1/4 wave vertical at 138 KHz is not practical even if you are a national government. As you go up in frequency, you start to get true E/M reception. An outdoor antenna will still massively out perform an indoor one. At least you can pick up something without a giant outdoor structure. By the time you get to VHF, antennas are getting pretty small and are a lot less difficult to deal with. Line of sight counts so elevation is a good thing.

One other hint ... check the title to your house / dwelling / cave / hut. Does it have a magic clause in it that reads "no antennas". If so be a bit careful. Sometimes those clauses get enforced. Yes this is empirical knowledge ...

Have fun !!!

Bob (the guy with the *very* small antennas)

G0HZU · « **Reply #5 on:** February 15, 2016, 09:23:11 pm »

Quote

With a spectrum analyzer, what is a good method for determining the local noise floor (in the lab area around the spectrum analyzer)?

Assuming you would be mostly working on RF PCBs and RF modules then my advice would be to just carry on working and only investigate noise and spurious pickup if/when it becomes an issue.

If your work involves waving antennas about then maybe the same advice applies. Only worry about the signals that actually bother or interfere with your measurements. If this becomes unmanageable then track down any local signals in the workroom and turn as much stuff off as you can.

The worst offenders are usually things like laptop PSUs, chinese wall wart battery chargers and anything with an active USB cable connection on your bench. Wifi and PCs and monitors and plasma TVs are also well known offenders.

cdev · « **Reply #6 on:** February 15, 2016, 11:56:33 pm »

A wideband SDR or capture program that can walk up and down the spectrum saving numeric values that can visualize as much as possible of the entire HF spectrum is best for getting a picture of what's floating around. Try running your device for a while capturing very slowly or very fast. Then you might get clues as to whatever things occur in patterns.

YouTube user SM5BSZ has four videos entitled "Sensors" that show how to make simple, good performing E and H probes, and look for noise using an SDR.

Electro Fan · « **Reply #7 on:** February 17, 2016, 08:08:55 am »

I don't want to derail the discussion so this is just a brief interruption for a light hearted thought....

after following some of the posts here I came across this:
http://m.ebay.com/itm/AGILENT-KEYSIGHT-HP-8510C-SYSTEM-SCALAR-AND-VECTOR-ANALYZER-/131727315933?nav=SEARCH

http://www.keysight.com/en/pd-1000002059%3Aepsg%3Apro-pn-8510C/vector-network-analyzer?cc=US&lc=eng

I'm wondering if one of these might be able to get the equivalent of 2 scopes credit toward the 20 scope merit badge?

ok, returning now to the original thread in progress ...

uncle_bob · « **Reply #8 on:** February 17, 2016, 12:43:46 pm »

Quote from: Electro Fan on February 17, 2016, 08:08:55 am

I don't want to derail the discussion so this is just a brief interruption for a light hearted thought....

after following some of the posts here I came across this:
http://m.ebay.com/itm/AGILENT-KEYSIGHT-HP-8510C-SYSTEM-SCALAR-AND-VECTOR-ANALYZER-/131727315933?nav=SEARCH

http://www.keysight.com/en/pd-1000002059%3Aepsg%3Apro-pn-8510C/vector-network-analyzer?cc=US&lc=eng

I'm wondering if one of these might be able to get the equivalent of 2 scopes credit toward the 20 scope merit badge?

ok, returning now to the original thread in progress ...

Hi

If you get one of those (or a similar vintage unit) the display is it's main weakness. They go dim and are a bit expensive to replace (If you can find them). They also have a tendency to have been used by some guy named Bob in an auto test system. Said knuckle head ran tests on them constantly for years. The poor relays in the attenuators got a real work out ... they also are a pain to replace.

The same issue applies to some generations of spectrum analyzers. The ones with relay attenuators (not knobs, not solid state switches) do have a wear out mechanism. It shows up often as one range bring wrong.

Bob

kazam · « **Reply #9 on:** February 28, 2016, 12:05:32 pm »

Get a unit with handles, not rack mounts. For exactly the reason uncle_bob mentions.

Quote from: uncle_bob on February 17, 2016, 12:43:46 pm

Quote from: Electro Fan on February 17, 2016, 08:08:55 am
I don't want to derail the discussion so this is just a brief interruption for a light hearted thought....

after following some of the posts here I came across this:
http://m.ebay.com/itm/AGILENT-KEYSIGHT-HP-8510C-SYSTEM-SCALAR-AND-VECTOR-ANALYZER-/131727315933?nav=SEARCH

http://www.keysight.com/en/pd-1000002059%3Aepsg%3Apro-pn-8510C/vector-network-analyzer?cc=US&lc=eng

I'm wondering if one of these might be able to get the equivalent of 2 scopes credit toward the 20 scope merit badge?

ok, returning now to the original thread in progress ...

Hi

If you get one of those (or a similar vintage unit) the display is it's main weakness. They go dim and are a bit expensive to replace (If you can find them). They also have a tendency to have been used by some guy named Bob in an auto test system. Said knuckle head ran tests on them constantly for years. The poor relays in the attenuators got a real work out ... they also are a pain to replace.

The same issue applies to some generations of spectrum analyzers. The ones with relay attenuators (not knobs, not solid state switches) do have a wear out mechanism. It shows up often as one range bring wrong.

Bob

KJDS · « **Reply #10 on:** February 28, 2016, 01:14:50 pm »

Quote from: Electro Fan on February 17, 2016, 08:08:55 am

I don't want to derail the discussion so this is just a brief interruption for a light hearted thought....

after following some of the posts here I came across this:
http://m.ebay.com/itm/AGILENT-KEYSIGHT-HP-8510C-SYSTEM-SCALAR-AND-VECTOR-ANALYZER-/131727315933?nav=SEARCH

http://www.keysight.com/en/pd-1000002059%3Aepsg%3Apro-pn-8510C/vector-network-analyzer?cc=US&lc=eng

I'm wondering if one of these might be able to get the equivalent of 2 scopes credit toward the 20 scope merit badge?

ok, returning now to the original thread in progress ...

That's missing a source and a test set

Mosaic · « **Reply #11 on:** March 13, 2016, 03:34:53 am »

I did some calibs on my Rigol DSA 815 recently...using a Boonton 4210 -4b Power meter and an HP8753D as a sweep (CW) gen.

I did a full range cal to 'smooth' the response down to around 0.1 db precision via the correction constants entries in the Rigol. Took < 30 cal points.

Of course, for tighter ranges of interest I'd have to do another set of cals to optimize the results, but it can be done, in about 30 mins!
Ok, onto the subject at hand.

I can get the Rigol down to -150dBm/Hz or so with a small VBW (300Hz), internal preamp on, and a 1Mhz span which takes abt 25 secs to sweep.

I hope to use this approach to doing NF work:
http://www.keysight.com/main/editorial.jspx?cc=US&lc=eng&ckey=1000001900:epsg:faq&nid=-536902970.536880671&id=1000001900:epsg:faq

Perhaps I'll score a calib. noise source one day.

uncle_bob · « **Reply #12 on:** March 13, 2016, 03:43:12 pm »

Quote from: Mosaic on March 13, 2016, 03:34:53 am

I did some calibs on my Rigol DSA 815 recently...using a Boonton 4210 -4b Power meter and an HP8753D as a sweep (CW) gen.

I did a full range cal to 'smooth' the response down to around 0.1 db precision via the correction constants entries in the Rigol. Took < 30 cal points.

Of course, for tighter ranges of interest I'd have to do another set of cals to optimize the results, but it can be done, in about 30 mins!
Ok, onto the subject at hand.

I can get the Rigol down to -150dBm/Hz or so with a small VBW (300Hz), internal preamp on, and a 1Mhz span which takes abt 25 secs to sweep.

I hope to use this approach to doing NF work:
http://www.keysight.com/main/editorial.jspx?cc=US&lc=eng&ckey=1000001900:epsg:faq&nid=-536902970.536880671&id=1000001900:epsg:faq

Perhaps I'll score a calib. noise source one day.

Hi

Noise diodes are not that hard to come by. Building up a source that covers the range of an 815 with one is a very doable thing. For the adventurous, there are a lot of structures (diodes etc) that generate fairly flat noise spectra. With some time spent sorting, you may not even need a fancy noise diode.

Will your home-brew widget be flat to 123 THz? Most certainly not. You have the gear to work out just how flat it is over the range of your other stuff. The result likely will be a chart of noise output versus frequency. For a simple / cheap lash up ... that's probably just fine.

Bob

Mosaic · « **Reply #13 on:** March 13, 2016, 04:40:43 pm »

Hmm, can you suggest a type of diode that's not too hard to source? I could get 1 and select from them for a decent unit.

uncle_bob · « **Reply #14 on:** March 14, 2016, 01:08:15 am »

Quote from: Mosaic on March 13, 2016, 04:40:43 pm

Hmm, can you suggest a type of diode that's not too hard to source? I could get 1 and select from them for a decent unit.

Hi

First hit from Mr Google when searching for "noise diode":

http://www.noisecom.com/products/components/nc100-200-300-400-series-chips-and-diodes

They are the guys I bought from in the past. There are a number of hits further down on the first page. I believe the diodes were in the $30 each range.

Looking at their app notes .. the diodes are just zeners. They have a low capacitance, but none the less they are a diode that is reverse biased and breaking down. The key point is that they are low capacitance junctions that are breaking down. Diodes are not super complex things. Lots of junctions go into breakdown.

Hmmm .... where to find a junction we can break down at a low voltage *and* that has a low capacitance? I wonder about these three terminal gizmos. Odd that the "reverse Vbe rating is so low. I wonder if a bipolar transistor's base to emitter junction breaks down like the data sheet says?

So: Best bet, grab some RF transistors. The ones you want are the older parts that have less emitter resistance (ballast R) than the newer ones. Anything that talks about "broadband optimized" likely is not a good bet. Cheap / high Ft and a 2N number with 3 digits or in the low 4 digits is a good sign. Those parts are now made by who knows who to what sort of specs nobody knows. It's going to be a "buy ten for a dollar" and see what you get. You can do a lot of experiments that way before you pay for one $30 diode from NoiseCom.

If you dig fairly deep into the past, there is a BBC (of all people... the BBC) app note on building noise sources with very conventional zener diodes. Their findings were that the noise is a fairly binary thing as you change current. At one current it's a good noise source, move up or down from there ... it's a lousy noise source. I would not be at all surprised to find that a current sweep is a really good idea when testing junctions as broadband noise generators.

Bob

Mosaic · « **Reply #15 on:** March 14, 2016, 03:58:26 am »

I have some 2n2369 on hand which are fast avalanche, (as per Jim Williams pulse gen for scope cals). Also some high Ft (up to 25Ghz) transistors.
Anyhow, this document was very interesting as it describes the build of noise gens as well as their cals.
http://www.w1ghz.org/noise/noise99.pdf

uncle_bob · « **Reply #16 on:** March 14, 2016, 12:52:11 pm »

Quote from: Mosaic on March 14, 2016, 03:58:26 am

I have some 2n2369 on hand which are fast avalanche, (as per Jim Williams pulse gen for scope cals). Also some high Ft (up to 25Ghz) transistors.
Anyhow, this document was very interesting as it describes the build of noise gens as well as their cals.
http://www.w1ghz.org/noise/noise99.pdf

Hi

The 2369 is a switching transistor with a relatively low Ft. I probably would try a 2N918 before I played with a 2369. They both are from the same era and neither one is expensive. In both cases, you want the surface mount versions (sot-23 or better still sc-70). The lead inductance on leaded part will give you all sorts of crazy problems at microwave. Mouser has MMBT-918's at a bit under 7 cents each for 100 pieces. Shipping at $4.99 will cost you almost as much as the transistors. A sort on Ft brings up the PBR-941 with an Ft of 8 GHz in a SOT-23. They are twice as expensive. A hundred pieces of both devices including shipping still costs you less than what my (possibly defective) memory of the price for a real noise diode was.

One thing not mentioned in the paper you linked: You want thin film chip resistors rather than tick film for the attenuators. They are also a good thing to toss in on a Mouser order.

Bob

Mosaic · « **Reply #17 on:** March 14, 2016, 04:10:32 pm »

Ok, and i understand soldering the SMT , top dow, provide s a shorter& straighter inductance path.

uncle_bob · « **Reply #18 on:** March 14, 2016, 07:40:34 pm »

Quote from: Mosaic on March 14, 2016, 04:10:32 pm

Ok, and i understand soldering the SMT , top dow, provide s a shorter& straighter inductance path.

Hi

Another trick that the paper uses, but does not explain:

He runs the DC current "return" for the diode through the output attenuator. This lets him avoid finding an inductor that works broadband to take the DC to ground. The key point is that you need a DC block later on in the attenuator or you have DC on the output. For the microwave frequencies he is playing with, that will be a pretty good capacitor. I suppose you always could test things that are DC blocked already. That seems like a somewhat dangerous assumption...

Bob

Mosaic · « **Reply #19 on:** March 14, 2016, 09:48:12 pm »

Ok, went fishing for RF transistors in stock. Have several of each.

BFR93A - 6Ghz SOT23
BFG540W - 9Ghz SOT-343
2.5GHZ GaAs FET AF002-32
BFP 650 - 42GHz NPN
http://www.infineon.com/dgdl/Infineon-BFP650-DS-v01_01-en.pdf?fileId=db3a30431400ef6801142743b0330715

BFG21W - 18Ghz, SOT-343R
BFG 425W - 25Ghz, SOT343R

uncle_bob · « **Reply #20 on:** March 14, 2016, 10:07:44 pm »

Quote from: Mosaic on March 14, 2016, 09:48:12 pm

Ok, went fishing for RF transistors in stock. Have several of each.

BFR93A - 6Ghz SOT23
BFG540W - 9Ghz SOT-343
2.5GHZ GaAs FET AF002-32
BFP 650 - 42GHz NPN
http://www.infineon.com/dgdl/Infineon-BFP650-DS-v01_01-en.pdf?fileId=db3a30431400ef6801142743b0330715

BFG21W - 18Ghz, SOT-343R
BFG 425W - 25Ghz, SOT343R

Hi

The FET's are not going to help in this case. You need a bipolar device in order to get a reasonable breakdown. I would start with something cheap and work up from there. For testing, sweeping the current is just a knob twist on a lab supply. When you build a real source, you should do a constant current driver to go with it.

======

A fairly common way to do the basic testing is to set up a broadband amp in front of your spectrum analyzer. Make sure it's stable and you know what the gain curve looks like (flat is best, but not absolutely needed). Also consider how much power will saturate the amp. If you come up with some sort of super diode, you want to know when things are not reading correctly due to overload. Put at lest a 6 db pad in front of the amp to get the impedance somewhere near 50 ohms.

The amp *should* have enough gain to show a rise on the spectrum analyzer when it is turned on. If you have an analyzer with a -150 dbm / sqrt(Hz) noise floor, that means you will need at least a 24 db amp, 30 would be better still. There are standard formulas that will let you work out what the net result on the screen means.

Early on all you really are looking for are two things:

Noise that is as flat as possible.

Within reason, as much noise as possible.

If your amp has a 6 db pad in front of it, it would be nice to know the noise figure of the amp. The net noise figure will be 6db plus the amp's noise figure. If it's 3 db you have a 9 db setup. If your diode puts out 9 db of noise, the output of the amp will go up by 3 db. Your analyzer reads a combo of it's floor and the amp's output so it goes up by a little less.

Net result, If you have 15 to 20 db of excess noise from the diode, you should easily see it on the analyzer. You can sweep current and sort diodes nice and fast.

Bob

Mosaic · « **Reply #21 on:** March 14, 2016, 10:37:47 pm »

Ok. I have this wideband amp:
http://www.minicircuits.com/pdfs/ZX60-6013E.pdf

Some of these as well:
http://www.minicircuits.com/pdfs/ERA-2SM+.pdf

I can try to copy this assy with FR4 0.8mm.
http://www.minicircuits.com/pcb/WTB-408-2+_P02.pdf
But this Ebay item seems better for my app...given the gain:
252075077721

uncle_bob · « **Reply #22 on:** March 15, 2016, 12:04:05 am »

Quote from: Mosaic on March 14, 2016, 10:37:47 pm

Ok. I have this wideband amp:
http://www.minicircuits.com/pdfs/ZX60-6013E.pdf

Some of these as well:
http://www.minicircuits.com/pdfs/ERA-2SM+.pdf

I can try to copy this assy with FR4 0.8mm.
http://www.minicircuits.com/pcb/WTB-408-2+_P02.pdf
But this Ebay item seems better for my app...given the gain:
252075077721

Hi

If you take a look at the eval board, it's done on a teflon laminate board (I have one in front of me). I would pick the packaged amp. It's got good enough input return loss that the 6 db pad probably can be backed off a bit. That might let you get away with one amp. If not, then cascade them.

Bob

Mosaic · « **Reply #23 on:** March 15, 2016, 02:22:21 am »

I ordered one of these, Ebay 181861155271, so I'll have options when I start the project. Don't like to get stuck in the middle needing something that can take 3 weeks to arrive.

Mosaic · « **Reply #24 on:** March 15, 2016, 05:44:49 pm »

This example uses a transistor i do have:
http://ham-radio.com/k6sti/nsrc.htm


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: How to Find the Noise Floor with a Spectrum Analyzer (Read 15352 times)

Share me