Why do you need a low pass filter before the first mixer of a spectrum analyzer?
Life is full of learning opportunities and I'd like to share one I had today. I'm trying to measure the signal quality of a SI5351 based signal generator set to 12MHz and using a 30dB attenuator at the input of the SA. Resolution is set to 15kHz What I see is a real mess, the 12MHz signal is there at -23dBm but also a vast amount of strong spurs at about all possible frequencies.  The SI5351 is not a very bad signal generator so I'm puzzled why it is that bad. With 12MHz input and the first IF of the SA at 110MHz I should have plenty of bandwidth headroom I think. So I connected a rather clean 12MHz signal from an old analog signal generator That looks much better!! The spur at 8.3MHz is coming from some leakage because of the rather long cable between the signal generator and the SA but for the rest it is about clean.?  Then it occurred to me that I am not using a low pass filter before the first mixer of the SA so I add one that should do -40dB at least above 100MHz and reconnect the SI5351 The picture changes completely. There are many spurs, but at 70dB below to the 12MHz signal so this measurement is consistent with this artikel:?  So I learned today that harmonics go a LONG way..... For the technicaly inclined, In the SA I'm using two ADF5351 LO's to steer two AD831 mixers with first IF at 110MHz and resolution filter of 15kHz at 10.7Mhz followed by the log detector These AD831 mixers (on 12$ modules from Bangood.com) are VERY good, unfortunately thy don't go above 1GHz. I'm considering shifting the first IF to 500Mhz to make maximum use of them The two ADF5351 seem to create much cleaner LO signals in the SA then one SI5351, that is why I started to investigate the output of the SI5351
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
That thing I threw out there:
?> with just 18ma of current draw on Rx,?it can't do a good job of
> looking at weak signals in the presence of strong signals.
doesn't quite hit the mark.
The si4432 just can't handle very large signals, but that's what attenuators are for.
Dynamic range could be quite large, determined in part by the ADC's they use
and the noise figure.? So could be pretty good.
And as you say, with a mixer and local oscillator and filter to upconvert,
it could easily be extended for use below 240mhz.
An appropriate local oscillator for it might be a second si4432 set up as a transmitter?
Qualcomm RF360 series SAW filters (perhaps for the 315mhz ISM band) are around $1 up on Mouser.
A promising direction to go for a cheap SA.
Jerry
toggle quoted message
Show quoted text
On Mon, Jan 14, 2019 at 07:37 AM, <erik@...> wrote:
An FCC document with actual measurements.
Seems that for an adjacent signal 1MHz away the rejection is at least 50dB
?
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
An FCC document with actual measurements.
Seems that for an adjacent signal 1MHz away the rejection is at least 50dB
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
I'd looked at the SiLabs ISM band data transceivers pretty hard before,
hadn't dawned on me that they could be useful as a UHF spectrum analyzer.
Very cool.
Datasheet says works from 240-930 mhz, has an "RX Channel Bandwidth" down to a min of 2.6khz.
And "RSSI Resolution" of +/- 0.5dB, accuracy not stated but that could be calibrated.
Indeed, they are cheap.
And lots of breakout boards for them too.
On the downside, with just 18ma of current draw on Rx,
it can't do a good job of looking at weak signals in the presence of strong signals.
For example, looking at -50dBc spurs in the presence of the carrier.
Also, "Image Rejection" of 30dB with "IF=937khz".
But still, quite useful.
Way better than a diode RF detector.
Or the 8 bit DVB-T dongles as well (except for that 240mhz minimum spec).
Jerry
toggle quoted message
Show quoted text
On Fri, Jan 11, 2019 at 12:28 PM, <erik@...> wrote:
I read somewhere about using a si4432 as log detector with 120dB range directly at 400mHz or higher with buildin sw defined resolution filter 6khz to 600khz.
This way? you only need one mixer, once LO, one roofing filter and the SI4432 directly after the roofing filter
I ordered some si4432 to test, they are dirt cheap
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
I read somewhere about using a si4432 as log detector with 120dB range directly at 400mHz or higher with buildin sw defined resolution filter 6khz to 600khz.
This way? you only need one mixer, once LO, one roofing filter and the SI4432 directly after the roofing filter
I ordered some si4432 to test, they are dirt cheap
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
> Now I'm using 110MHz as second IF and 10.7MHz as 3rd IF (for the resolution filters)
Should work fine, cheaper too.? But the PX1002 is drop in with no tuning required.
> I did this for my own build VNA that can run in SA mode with switchable DSP resolution filters
Have you put up a description of that thing?
> The limit actually does not exist. I'm using the ADE-11X in another project and it goes down to audio frequencies
Sounds right.
They fully characterize their parts, and something undoubtedly doesn't meet spec at audio.
But likely good enough, especially if you calibrate the variance
Worth exploring the various parts.
But since I am mostly interested in HF at this time and not 1ghz+, I'd probably tend toward the ADE-5+
Asking it to deal with a 2ghz LO is cheating by 33%, whereas asking a 10mhz minimum part to do audio might be much worse.
Incidently, G0UPL found that the si5351 doesn't crap out till 292mhz, well above the 225mhz spec.
I'd guess the loop filter is not optimized for that, so jitter could be getting bad.
If the ADF4351 needs replacing, I'd probably go for the si5340 (plus doubler, if need over 1024mhz).
Much cheaper than the si549, multiple independent outputs, has a fast spi interface so you can sweep?
as fast as your filters allow.? Jitter spec on the si5340 is an order of magnitude better than the si5351.
Though likely a few days of fun involved to write suitable firmware for either the s549 or the si5340.
?
> How would you do the resolution filters?
Well, the PX1002 bandwidth seems about right as is for looking at spurs and harmonics.
To see IMD and other close in stuff, I was thinking a second mixer from 86.85mhz down to baseband audio.
So a simple DC receiver, no attempt at image rejection.
Inspect the result on a soundcard based audio spectrum analyzer as you scan through the signal,
images will be moving the opposite direction and thus easily ignored.
An idea stolen from Arv, of HBTE fame.
But mostly I'm interested in harmonics and spurs, and probing RF signal levels through each stage of a receiver.
None of that requires the final DC convert to audio, so scans can be fast and fully automated.
And once bored with that simple spectrum analyzer, maybe move on to a full featured direct to baseband IQ
using something like the AD8348 quadrature demodulator, 50 to 1000mhz in.
Below 50mhz, could upconvert first to 86.85mhz using the front end of that PX1002 design.
Don't hold your breath, will likely get distracted with a dozen other unfinished projects
before any of that happens.
> Suggest to look at the AD8307 data sheet and consider adding a AD603 ....
Yup, that caught my eye as well.
Good to know it works well!
To bad they haven't integrated the two of them into one device.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Thu, Jan 10, 2019 at 11:53 PM, <erik@...> wrote:
On Thu, Jan 10, 2019 at 12:07 PM, Jerry Gaffke wrote:
Something like PX1002 SAW filters at 86.85mhz might be a better choice than 10.7mhz for a second IF, avoid image trouble.
Now I'm using 110MHz as second IF and 10.7MHz as 3rd IF (for the resolution filters)
Or just mix down from 915mhz to I/Q baseband and use SDR techniques?
I did this for my own build VNA that can run in SA mode with switchable DSP resolution filters
The W7ZOI spectrum analyzer front end uses the mixer's IF port for the incoming signal,
RF port is outgoing, because the IF port can get down to lower frequencies.
Doing the same
Here's the entire line of MiniCircuits mixers:??
On the ADE-1+,? LO/RF ports are 0.5-500mhz, the IF port is good from DC-500mhz.
The ADE-11X is suitable for higher frequencies, but your minimum signal into the SA will be the 10mhz limitation imposed by the IF port.
The limit actually does not exist. I'm using the ADE-11X in another project and it goes down to audio frequencies
The ADE-5+ might be a better choice,? LO/RF ports are 5-1500mhz, the IF is good from DC-1000mhz, so our SA could see from 0mhz to 1500-915=585mhz
And could perhaps press further than 1500mhz on the LO port.
Cheap as I might be, spending the extra buck on an ADE-5+ to be able to see the bottom 10mhz seems like a good deal.
.
All diode double balanced mixers will go down to DC on their IF port
I'm curious how well behaved that ADF4351 local oscillator of yours is.
Ah, indeed, not so easy to get rid of the spurs, low noise AND have a fast sweep. Lots of register tweaking. SI5351 are much simpler to use!!!!.
The new si549 (like an si570, but solid coverage up to 1500mhz) would also work, with much finer resolution in selecting the frequency.
Or an si5340 (4 channels, up to 1024mhz) followed by a clock doubler, cheaper and get multiple LO's for 2'ed IF, tracking generator, ...
?
Mostly armchair engineering here, but I am trying to build a simple single conversion SA?
using the ADE-1+, two PX1002's, and an AD8307.? (Also a couple MMIC's, front end LPF, si5351, Arduino Nano clone, etc)
How would you do the resolution filters?
Those PX1002's are $12 quantity one, but otherwise this is throw-away cheap to play with.
AD8307's can be had on ebay for down to $0.30 each, reportedly work well enough? Factory seconds, slightly out of calibration?
Suggest to look at the AD8307 data sheet and consider adding a AD603 as shown in one of the examples in the datasheet. I did and it gives you 130dB measurement range, even when build in dead-bug style!!!!
Success with your build
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
On Thu, Jan 10, 2019 at 12:07 PM, Jerry Gaffke wrote:
Something like PX1002 SAW filters at 86.85mhz might be a better choice than 10.7mhz for a second IF, avoid image trouble.
Now I'm using 110MHz as second IF and 10.7MHz as 3rd IF (for the resolution filters)
Or just mix down from 915mhz to I/Q baseband and use SDR techniques?
I did this for my own build VNA that can run in SA mode with switchable DSP resolution filters
The W7ZOI spectrum analyzer front end uses the mixer's IF port for the incoming signal,
RF port is outgoing, because the IF port can get down to lower frequencies.
Doing the same
Here's the entire line of MiniCircuits mixers:??
On the ADE-1+,? LO/RF ports are 0.5-500mhz, the IF port is good from DC-500mhz.
The ADE-11X is suitable for higher frequencies, but your minimum signal into the SA will be the 10mhz limitation imposed by the IF port.
The limit actually does not exist. I'm using the ADE-11X in another project and it goes down to audio frequencies
The ADE-5+ might be a better choice,? LO/RF ports are 5-1500mhz, the IF is good from DC-1000mhz, so our SA could see from 0mhz to 1500-915=585mhz
And could perhaps press further than 1500mhz on the LO port.
Cheap as I might be, spending the extra buck on an ADE-5+ to be able to see the bottom 10mhz seems like a good deal.
.
All diode double balanced mixers will go down to DC on their IF port
I'm curious how well behaved that ADF4351 local oscillator of yours is.
Ah, indeed, not so easy to get rid of the spurs, low noise AND have a fast sweep. Lots of register tweaking. SI5351 are much simpler to use!!!!.
The new si549 (like an si570, but solid coverage up to 1500mhz) would also work, with much finer resolution in selecting the frequency.
Or an si5340 (4 channels, up to 1024mhz) followed by a clock doubler, cheaper and get multiple LO's for 2'ed IF, tracking generator, ...
?
Mostly armchair engineering here, but I am trying to build a simple single conversion SA?
using the ADE-1+, two PX1002's, and an AD8307.? (Also a couple MMIC's, front end LPF, si5351, Arduino Nano clone, etc)
How would you do the resolution filters?
Those PX1002's are $12 quantity one, but otherwise this is throw-away cheap to play with.
AD8307's can be had on ebay for down to $0.30 each, reportedly work well enough? Factory seconds, slightly out of calibration?
Suggest to look at the AD8307 data sheet and consider adding a AD603 as shown in one of the examples in the datasheet. I did and it gives you 130dB measurement range, even when build in dead-bug style!!!! Success with your build
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
These SAW filters for the 915mhz ISM band are quite cheap at $2, under $1 in quantity, 40dB down at 15mhz out:
? ??
A clean 50 ohms so easy to stack.
Something like PX1002 SAW filters at 86.85mhz might be a better choice than 10.7mhz for a second IF, avoid image trouble.
Or just mix down from 915mhz to I/Q baseband and use SDR techniques?
The W7ZOI spectrum analyzer front end uses the mixer's IF port for the incoming signal,
RF port is outgoing, because the IF port can get down to lower frequencies.
Here's the entire line of MiniCircuits mixers:??
On the ADE-1+,? LO/RF ports are 0.5-500mhz, the IF port is good from DC-500mhz.
The ADE-11X is suitable for higher frequencies, but your minimum signal into the SA will be the 10mhz limitation imposed by the IF port.
The ADE-5+ might be a better choice,? LO/RF ports are 5-1500mhz, the IF is good from DC-1000mhz, so our SA could see from 0mhz to 1500-915=585mhz
And could perhaps press further than 1500mhz on the LO port.
Cheap as I might be, spending the extra buck on an ADE-5+ to be able to see the bottom 10mhz seems like a good deal.
.
I'm curious how well behaved that ADF4351 local oscillator of yours is.
The new si549 (like an si570, but solid coverage up to 1500mhz) would also work, with much finer resolution in selecting the frequency.
Or an si5340 (4 channels, up to 1024mhz) followed by a clock doubler, cheaper and get multiple LO's for 2'ed IF, tracking generator, ...
?
Mostly armchair engineering here, but I am trying to build a simple single conversion SA?
using the ADE-1+, two PX1002's, and an AD8307.? (Also a couple MMIC's, front end LPF, si5351, Arduino Nano clone, etc)
Those PX1002's are $12 quantity one, but otherwise this is throw-away cheap to play with.
AD8307's can be had on ebay for down to $0.30 each, reportedly work well enough? Factory seconds, slightly out of calibration?
Jerry, KE7ER
toggle quoted message
Show quoted text
On Sat, Jan 5, 2019 at 07:57 AM, <erik@...> wrote:
1: Between IF1 and IF2 so: 2510.7MHz -> 110.7MHz -> 10.7MHz
2: Got rid of leakage of LO's (bumps in first part of spectrum of first post) and got rid of mirrors. The IF1 filter is actually 100MHz wide, too wide for 10.7Mhz as second IF but OK for 110MHz as second IF?
Specan is OK.?
Its actually simpler and cheaper to aim for 1GHz as max input frequency. This enables the use of cheaper ADE-11X mixers (better isolation and higher conversion) and cheap but narrow ADS? SAW filters (see eBay, 10$ each, use two in series), or even active mixers (AD831) allowing to get rid of the IF amplifier(s) and you can go directly from 1090MHz to 10.7MHz because of the narrow ADS filter
I am having problems with the SIM-83+ mixers I bought. Not from cheap Chinese supplier so I assume they are original but only one has 7dB conversion los, the others 15dB and 25dB. I complained and will receive at least a replacement for the worst mixer. This may solve the current -10dB offset and improve dynamic range
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
Thanks for taking the time to explain further Erik.
So sharp filters and mixers with good isolation is the key then, good to know.
Please keep us posted on your progress. And if I get around to building a SA myself I would certainly be interested in reusing your GUI SW even though I write SW for a living I find myself hard to motivate to when it comes GUIs :-)
Br
Marcus
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
1: Between IF1 and IF2 so: 2510.7MHz -> 110.7MHz -> 10.7MHz
2: Got rid of leakage of LO's (bumps in first part of spectrum of first post) and got rid of mirrors. The IF1 filter is actually 100MHz wide, too wide for 10.7Mhz as second IF but OK for 110MHz as second IF?
Specan is OK.?
Its actually simpler and cheaper to aim for 1GHz as max input frequency. This enables the use of cheaper ADE-11X mixers (better isolation and higher conversion) and cheap but narrow ADS? SAW filters (see eBay, 10$ each, use two in series), or even active mixers (AD831) allowing to get rid of the IF amplifier(s) and you can go directly from 1090MHz to 10.7MHz because of the narrow ADS filter
I am having problems with the SIM-83+ mixers I bought. Not from cheap Chinese supplier so I assume they are original but only one has 7dB conversion los, the others 15dB and 25dB. I complained and will receive at least a replacement for the worst mixer. This may solve the current -10dB offset and improve dynamic range
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
Erik,
?
can you please elaborate the following:
1. The 3rd IF@110MHz, was put after IF2@...? Or in between IF1 and IF2?
2. How and why did that affect performance? (Trying to learn here so links to relevant info is perfectly ok).
I have a stalled Specan build that might get some new inspiration from your build.
Br
Marcus, SA5PMG
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
I added a third IF at 110MHz.?Now performance is acceptable. Two examples: Wide scan:?200MHz output of a SI5351 through 50dB attenuator to see the harmonics. The actual maximum scan frequency is 1890MHz  The spurs at 25MHz and related harmonics are probably from the 25MHz reference clock on the ADF4351 module. Limited mixer spurs. Small scan: Measuring a 1 MHz sinus signal at calibrated level of -80dBm  The power scale is 10dB off, actual noise floor is thus at -105dBm. Max input signal is -30dBm (actual -20dBm) so dynamic range is not very good (85dB, I was hoping for more then 90dB) but low frequency performance is usable and the zero hertz response small enough. Now I have to put everything in a nice box, not my biggest competence
|
Re: junk box noise generator
This morning it occurred to me -- an MMIC on micro-stripline will basically amplify ANYTHING from DC to daylight, so I googled on "MMIC noise source."? It led me to a site which showed a HB noise gen that used an emitter - base junction in an NPN transistor (thanks to Jim Tibbetts for reminding us of this) and three chained MMICs to deliver amplified noise.? But the builder mentioned creating more than 100 thru-board vias by hand.?? So ... look down the search results and -- voila!? A group discussion not unlike what we have here that referenced "you can buy this stuff on ebay."
One click later, and I saw that pre-built, MW-capable noise sources can be had from China for twelve bucks (and twelve weeks).
So I'll probably buy AND build, since I have MMICs and zeners in the junk box, and I am not in a hurry for this. A comparison might be interesting.
It's a nice day when I find that something I could imagine -- could be built, AND if I wanted one today I didn't have to invent or even re-invent it.
73 Jim N6OTQ
PS -- When you "google" something, be sure to use duckduckgo as well, because their search algorithms are different, and BOTH search engines come up with true gems that the other one missed.? Difference is -- you have to wade through all of google's paid placements and SEO-enhanced results.
|
Re: junk box noise generator
You can also use the base/emitter junction of a random transistor in place of the zener for the noise generator portion
Jim
I want to test filters and once again Santa stiffed me on all the fancy test gear I wanted.? It could be my fault as my chimney leads to the boiler.? Anyway, I have an rtl-sdr and an upconverter so all I needed was some clean noise.? Searching lead me to N5ESE's description of using a zener diode.? He references an earlier design by N0SS.? N0SS is now SK and his website is no longer but thanks to I found this I breadboarded N0SS's circuit  and lo and behold it works great.? Below shows without and with the noise circuit.? The upconverter adds an offset of +125MHZ  if you are a Spielberg fan and live on a Native American burial ground you should probably skip this. Doug AC9RZ
|
Re: junk box noise generator
Jim,
I do not have much more than what I posted.? As to a design for UHF, I am not the best to answer that.? You guess looks reasonable to me but that is like asking Ronald McDonald to check over your beef wellington recipe.? There are some pretty savvy UHF/VHF/microwave folks that pop in here from time to time, perhaps one of them will jump in and help.
Also, in case it wasn't apparent from my post, the receiver is actually tuned to 7.2MHz not 132.2MHz? due to the offset from the upconverter.
|
Re: junk box noise generator
Doug,
Do you have a sample circuit for a noise generator that goes an octave or two higher??
Would you be willing to speculate on stripline / SMD construction techniques to go up to or past 1 GHz?? My guess is -- duplicate the schematic with SMDs and use some common microwave transistors like MRF901.
73 Jim N6OTQ
I want to test filters and once again Santa stiffed me on all the fancy test gear I wanted.? It could be my fault as my chimney leads to the boiler.? Anyway, I have an rtl-sdr and an upconverter so all I needed was some clean noise.? Searching lead me to N5ESE's description of using a zener diode.? He references an earlier design by N0SS.? N0SS is now SK and his website is no longer but thanks to I found this
|
I want to test filters and once again Santa stiffed me on all the fancy test gear I wanted.? It could be my fault as my chimney leads to the boiler.? Anyway, I have an rtl-sdr and an upconverter so all I needed was some clean noise.? Searching lead me to N5ESE's description of using a zener diode.? He references an earlier design by N0SS.? N0SS is now SK and his website is no longer but thanks to archive.org I found this I breadboarded N0SS's circuit  and lo and behold it works great.? Below shows without and with the noise circuit.? The upconverter adds an offset of +125MHZ  if you are a Spielberg fan and live on a Native American burial ground you should probably skip this. Doug AC9RZ
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
Some pictures: The complete analyzer:  Signal left to right: Input low-pass filter (green), mixer (blue module), IF1 filter (big black brick) ,mixer (blue module), IF2 amplifier (green module), IF2 filter (tin box), log detector (tin box) ->arduino analog input Below two adf4351 modules (black PCB's) Above Arduino? The prototyping board is temporary till I get an Arduino Due (3.3Volt) to get rid of the 5v->3.3v resistors The IF2 filter. relays to switch, ceramic for 300kHz and chrystals for 15kHz. Not very beautiful but performance is good.  The log detector, hope you can see the 0602 components. Its loosely based on the AD603+AD8307 reference design but with single sided supply of 12 volt.? 
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
For the PC I started from scratch using .net libraries.
I like the Specan SW as its pure C but that also implies it is too much effort to add functionality?
The arduino SW can easily be made to work with the specan SW, it may even work now (did not test lately)
|
Re: 2 GHz spectrum analyzer mostly build out of eBay modules
The arduino SW is a rewrite/combination of various SW's
It can be used as a standalone tool controlled over the serial interface
The ADF4351 library is a rewrite of something I found but spur elimination and speed has been a big headache
The SI5351 library is partly rewritten to go to 300MHz (if you device supports it) and with substantial speed improvement.
Zip file with ino and libraries for SI5351 and ADF4351 attached. The libraries are still in continuous development.
|
