Author Topic: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial  (Read 44126 times)

0 Members and 2 Guests are viewing this topic.

Offline EEVblogTopic starter

  • Administrator
  • *****
  • Posts: 37626
  • Country: au
    • EEVblog
EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« on: August 30, 2012, 10:52:19 am »


Dave.
 

Offline JuiceKing

  • Regular Contributor
  • *
  • Posts: 233
  • Country: us
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #1 on: August 30, 2012, 11:39:02 am »
Very neat demo! Thumbs-Up!

Some time back, I did a similar experiment with a sweep generator and oscilloscope and was puzzled by the high-frequency response anomalies you explain so well here.

Thanks!!

For your blog, Dave, it might be interesting at some point to explain the practical limitations of this "low-end" Rigol vs. an old HP/Agilent that you might find at a flea market.

 

Offline EEVblogTopic starter

  • Administrator
  • *****
  • Posts: 37626
  • Country: au
    • EEVblog
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #2 on: August 30, 2012, 11:50:54 am »
Some time back, I did a similar experiment with a sweep generator and oscilloscope and was puzzled by the high-frequency response anomalies you explain so well here.

Damn, that reminds me, I forgot to include a shot of the inductor datasheet showing the self resonant point.

Dave.
 

HLA-27b

  • Guest
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #3 on: August 30, 2012, 12:46:05 pm »
Quote
Anything below 1GHz is basically DC to the RF guys.


...and yet we don't have a smooth tracking analyzer on a measurement instrument.
If they were that good we ought to have Ghz RF in our toasters by now. What am I missing?


I mean imagine if a civil engineer said "Anything below 100 floors is basically ground level" and at the same time all his buildings were crooked and wobbly?
 

Offline Short Circuit

  • Frequent Contributor
  • **
  • Posts: 439
  • Country: nl
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #4 on: August 30, 2012, 01:33:40 pm »
If they were that good we ought to have Ghz RF in our toasters by now. What am I missing?
Microwave at 2.4 GHz close enough  ;D
 

Offline olsenn

  • Frequent Contributor
  • **
  • Posts: 993
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #5 on: August 30, 2012, 01:46:48 pm »
Quote
If they were that good we ought to have Ghz RF in our toasters by now. What am I missing?

Correct me if I'm wrong, but I think DC power dissipation in a resistive load is good enough to heat toast.  As for microwave ovens, they don't rely on the same circuitry that you typically see in low power RF transmitters for instance. Look up "magnetron" on google to find out more about how this is done.
 

HLA-27b

  • Guest
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #6 on: August 30, 2012, 02:34:00 pm »
Correct me if I'm wrong, but I think DC power dissipation in a resistive load is good enough to heat toast.  As for microwave ovens, they don't rely on the same circuitry that you typically see in low power RF transmitters for instance. Look up "magnetron" on google to find out more about how this is done.

I was alluding to the fact that as technology matures and gets cheaper it finds itself on most absurd places, as absurd as wifi on a toaster for example. So if they can't put a linear tracking gen on a 1 GHz measurement instrument (not a toaster mind you) then their technology is not mature and their "1GHz = DC" simply isn't true. Thats why a 13GHz scope costs 140.000 bucks and not 400.

 

Online jahonen

  • Super Contributor
  • ***
  • Posts: 1052
  • Country: fi
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #7 on: August 30, 2012, 02:43:19 pm »
Dave,

I think it would have been easier to set the desired start and stop frequencies for this kind of stuff by using start freq and stop freq settings in the frequency menu, instead of the span menu (center/span method).

Regards,
Janne
 

Offline hpux735

  • Contributor
  • Posts: 39
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #8 on: August 30, 2012, 03:35:30 pm »
Correct me if I'm wrong, but I think DC power dissipation in a resistive load is good enough to heat toast.  As for microwave ovens, they don't rely on the same circuitry that you typically see in low power RF transmitters for instance. Look up "magnetron" on google to find out more about how this is done.

I was alluding to the fact that as technology matures and gets cheaper it finds itself on most absurd places, as absurd as wifi on a toaster for example. So if they can't put a linear tracking gen on a 1 GHz measurement instrument (not a toaster mind you) then their technology is not mature and their "1GHz = DC" simply isn't true. Thats why a 13GHz scope costs 140.000 bucks and not 400.

I fully agree.  I think that phrase is pretty annoying.

About the high frequency response:  I don't know how much you can reasonably expect to achieve with that topology.  As 1GHz is not DC, you have to think about isolation.  Having the input and output terminals of the filter un-shielded from one another means that the high frequency response is likely through the air.  In addition, the waviness of the response is likely caused by various propagation modes.  The metal can is acting as an RF cavity.
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 8515
  • Country: us
    • SiliconValleyGarage
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #9 on: August 30, 2012, 04:48:00 pm »
The rf output amplitude is ot flat. The ampmcan only deliver so much power into the system.
This kind of measurements needs to be done by hooking up the dut through a power splitter , then taking a baseline measurement the get the output behavior of the tracking generator and storing that in memory. Then doing a measurement. This gives you the true in/out ratio.

That'swhy the. Heapo spectrum analusers like those 500$ attens on ebay are crap. You can't store the reference curve. So anything you measure is wrong as it is impacted by losses in the machine and wiring.

I normaly use an active probe. Hok up tracking gen to system under test.
Probe on input, store cal data , then probe on Output. That way even your cabling and probe is compensated.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline nitro2k01

  • Frequent Contributor
  • **
  • Posts: 843
  • Country: 00
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #10 on: August 30, 2012, 07:37:57 pm »
Maybe I'm crazy and/or blind but I think it looks like the signal input and output are shorted through that big copper plane. Or is there a slit in it that I'm not seeing?
Regardless, a big "I know what you're thinking" would have been to put the unit under test without any coil at all and see how much of that RF goes through.
Whoa! How the hell did Dave know that Bob is my uncle? Amazing!
 

Online jahonen

  • Super Contributor
  • ***
  • Posts: 1052
  • Country: fi
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #11 on: August 30, 2012, 07:43:27 pm »
I think that the ripples come from some kind of mismatch. The box is way too small to have any internal resonances at such low frequencies. Anyway, I bodged it on a copper clad board, and got a slightly different results (with 910 nH inductor and 220 pF caps) using a R&S FSV-7 with TG option, using a log frequency scale:



It is clearly a 3rd order filter:



Regards,
Janne
« Last Edit: August 30, 2012, 07:46:43 pm by jahonen »
 

Offline JackOfVA

  • Supporter
  • ****
  • Posts: 350
  • Country: us
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #12 on: August 31, 2012, 12:13:51 am »
You can also improve the noise floor by reducing the resolution bandwidth (RBW). Since the tracking generator will presumably match the spectrum analyzer frequency closely, you can reduce the RBW from 1 MHz to say 1 KHz - the result should be a major improvement in noise floor without using the preamplifier.

Let me add some Australian content ... there's a great study available at http://www.dsto.defence.gov.au/publications/2597/DSTO-TN-0531.pdf titled "Construction Techniques for LC Highpass and Lowpass Filters used in the 1 MHz to 1 GHz Frequency Range." 

It's well worth reading for anyone interested in frequency filters and thanks to Dave for paying the taxes that funded the work.

I should also add that the "one hung low" coaxial adapters Dave used (N to BNC) are not to be trusted for anything close to precision work at UHF. And, you have to watch the coax cable and BNC connectors as well. At HF, you can get by with most anything for cables and connectors, but as you move up in frequency, cheap cable with poor braid coverage and nickle plated Chinese adapters must be avoided. (Amphenol silver plated brass connectors are readily available and adequate for most casual work, and if you are really into precision measurements, then some very pricey products are available.)  I'll also note that there's a lot of coax cable out of China that may be marked  RG-58, implying 50 ohms impedance, but it is far from that. I measured a length of "Digiwave" RG-58 cable for a friend and found it close to 100 ohms impedance. Details at http://www.cliftonlaboratories.com/the_curious_case_of.htm


« Last Edit: August 31, 2012, 12:20:22 am by JackOfVA »
 

Offline JackOfVA

  • Supporter
  • ****
  • Posts: 350
  • Country: us
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #13 on: August 31, 2012, 12:35:40 am »
Many subtleties have to be considered when making a swept measurement. Agilent has an excellent application note covering error compensation in VNAs, available at
http://literature.agilent.com/litweb/pdf/5965-7709E.pdf

A tracking generator and spectrum analyzer does not make these sort of error correction measurements, other than the simplest type. A true VNA will have a bridge arrangement with a sample of the generated signal being used as the reference for a comparative level measurement.

Some improvement with the tracking generator and spectrum analyzer might be found by adding say 6 or even 10 dB attenuators to the TG output and spectrum analyzer input. This will help the return loss seen by both ends of the filter. (Probably the spectrum analyzer input return loss isn't too bad, but I would suspect the TG output return loss.) Mini-Circuits has a useful line of "Adaptenuators" that combines an attenuator and a between-series adapter. For example, a pair of NM-BF-10 might be used in Dave's setup. These are 10 dB attenuators, rated to 2 GHz with a male N on one end and a female BNC on the other.
 

Offline Zad

  • Super Contributor
  • ***
  • Posts: 1013
  • Country: gb
    • Digital Wizardry, Analogue Alchemy, Software Sorcery
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #14 on: August 31, 2012, 04:41:29 am »
Each of the peaks on the HF end of the response seems to be 75MHz apart. Maybe that corresponds with the electrical length of the coaxial cable? I would definitely try another cable on it, if only to check that the peaks didn't (or did) move.


Offline sagdahl

  • Contributor
  • Posts: 26
  • Country: se
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #15 on: August 31, 2012, 03:52:15 pm »
Dave, brilliant tutorial.

I have never used a spectrum analyser but now I want one.
Just to play around with and measure filters.
Rigol should really give you a percentage of all their sales due to the blog.
I have already bought an oscilloscope. Guess the logistics can be difficult.

/Roger
That, that is, is. That, that is not, is not. Is that it? It is!
 

Offline SoftwareSamurai

  • Regular Contributor
  • *
  • Posts: 170
  • Country: us
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #16 on: August 31, 2012, 05:41:00 pm »
I would love to see how the DSA-1030 compares to the DSA-815, in terms of the tracking generator and response of the SA itself. (i.e. Is the 1030 worth the extra $$$?)
 

Offline Andrei

  • Contributor
  • Posts: 12
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #17 on: August 31, 2012, 05:55:39 pm »
Hi All,


I am a bit confused as to what that tracking gen does. Does it measure the input to the spectrum analyzer and output it? If that is the case, how does it address the positive feedback issue when you connect it to the input to the input of the spectrum analyzer and now blow up the waveform ?

Thanks

 

Online jahonen

  • Super Contributor
  • ***
  • Posts: 1052
  • Country: fi
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #18 on: August 31, 2012, 06:25:38 pm »
Hi All,


I am a bit confused as to what that tracking gen does. Does it measure the input to the spectrum analyzer and output it? If that is the case, how does it address the positive feedback issue when you connect it to the input to the input of the spectrum analyzer and now blow up the waveform ?

Thanks

A tracking generator basically just generates exactly the same frequency what the spectrum analyzer is currently measuring. So the generator is just a generator which is synchronized to the local oscillator of the spectrum analyzer. Some higher-end spectrum analyzers even let to use an external signal generator as a tracking gen (so spectrum analyzer controls the RF gen via GPIB bus or by some other means during the sweep). So there is no feedback of any kind.

Regards,
Janne
 

Offline JackOfVA

  • Supporter
  • ****
  • Posts: 350
  • Country: us
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #19 on: September 01, 2012, 04:55:43 pm »
I built a version of Dave's filter, 200 pF shunt C and each end and 1.06uH series inductance, with decent results well up into the UHF range.

The shunt C consists of two 100pF 1206 C0G capacitors in parallel, mounted 180 degrees apart. (Capacitors are shown in the red circles). Inductor is 17 turns of # 20 AWG wire on a 0.5" (12.7mm) diameter powdered iron core, Micrometals mix 10. Adjusted by squeezing turns to be 1.06 uH at 10 MHz, measured on an HP4192A impedance meter.

The two end pieces of scrap PCB have a small diameter hole drilled through them, and an annular ring cut into PCB centered on the hole, using a diamond dust core drill 1/4" diameter (6.35mm). (Single sided stock.) The shunt capacitors bridge the ring gap and the inductor lead passes through the hole and is soldered to the ring pad, along with the BNC connector center conductor.

Response is measured with an HP8752B vector network analyzer, a 75 ohm instrument, with MiniCircuit adapter-attenuator matching pads for 50 ohms. The pad loss is calibrated out during the instrument zero loss calibration process.

Frequency sweeps from 300 KHz to 1 GHz and 300 KHz to 3 GHz are shown. All are log scale frequency.

I should add that in addition to concern over the inductor becoming a capacitor at these frequencies well above its SRF, it's important to prevent the input and output connectors and leads from coupling to each other, either via loop inductance or stray capacitance. Hence the end post BNC connectors and the the PCB "wings" provide protection against unwanted coupling (sometimes called "filter blowby")

At 15 MHz, the measured loss is 2.8 dB, pretty close to the expected 3 dB. End capacitance is a bit below the design value and that bumps up the -3 dB corner frequency.

The first point of resonance is at 430 MHz (peak) and then further resonances are seen up to 3 GHz.

However, it's not that bad a filter, providing at least 50 dB attenuation up to 1.7 GHz. And up to 976 MHz, at least 70 dB attenuation is seen.

I'll put it in the mail to Dave next week for his amusement.

The end wings and choice of connector are intended to minimize inductive and capacitive coupling between the input and output thereby reducing filter leakage.
« Last Edit: September 01, 2012, 04:59:47 pm by JackOfVA »
 

Offline G7PSK

  • Super Contributor
  • ***
  • Posts: 3859
  • Country: gb
  • It is hot until proved not.
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #20 on: September 01, 2012, 05:35:54 pm »
For best results you should also use coax that is cut to the correct length for the frequency in use that is why that  bounce was the same frequency apart (circa 75 Mhz as some one else noted)
 

Offline ruku

  • Contributor
  • Posts: 25
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #21 on: September 01, 2012, 07:57:24 pm »
Somehow, I knew that when I came to this thread, people would be dying to show off their analog filter design-fu. I'm so excited!

I built a version of Dave's filter, 200 pF shunt C and each end and 1.06uH series inductance, with decent results well up into the UHF range.

...

However, it's not that bad a filter, providing at least 50 dB attenuation up to 1.7 GHz. And up to 976 MHz, at least 70 dB attenuation is seen.

I'll put it in the mail to Dave next week for his amusement.

Oh my goodness, please do this. Are you related to Jim Williams per-chance?

Could you write up a small tutorial on how you made those design decisions? I'd love to pick your brain.
 

Offline JackOfVA

  • Supporter
  • ****
  • Posts: 350
  • Country: us
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #22 on: September 01, 2012, 08:29:44 pm »
Somehow, I knew that when I came to this thread, people would be dying to show off their analog filter design-fu. I'm so excited!

I built a version of Dave's filter, 200 pF shunt C and each end and 1.06uH series inductance, with decent results well up into the UHF range.

...

However, it's not that bad a filter, providing at least 50 dB attenuation up to 1.7 GHz. And up to 976 MHz, at least 70 dB attenuation is seen.

I'll put it in the mail to Dave next week for his amusement.

Oh my goodness, please do this. Are you related to Jim Williams per-chance?

Could you write up a small tutorial on how you made those design decisions? I'd love to pick your brain.

My only connection with Jim Williams is that we  both attended Wayne State University  -- he dropped out after a year of non-technical study and then went later to MIT, whist I completed my EE degree at WSU. Never met him, though.

As far as the filter goes, if you ever take a lumped component VHF/UHF low pass filter apart, you will find that it is built with multiple bulkhead shields strung out through the filter length. (The filter is housed in a long cylindrical or rectangular RF tight box. Between each bulkhead is the inductor and on each bulkhead are the shunt capacitors.)

The shunt capacitors are often made from disks of brass or copper and Teflon or mica dielectric. The bulkhead forms the ground side and there's a hole through the bulkhead for the inductor wire to pass and to be soldered or mechanically attached to the shunt capacitor disk.

The series inductors are air wound from heavy gauge wire, preferably silver plated.

The idea is that the bulkheads shield each filter section (inductor) from interacting with each other and provide a low impedance path for the ground return path from the shunt current.

In adapting this configuration for a 15 MHz filter, some changes have to be made.

1. The shunt capacitors are much easier to make for a one-off filter by stock surface mount parts. 0805 might be better from an electrical performance prospective, but are a bit too small for a quick kludge. 1206, C0G types are about the right size mechanically.

In order to more uniformly distribute the ground return current through the bulkhead, it's better to use several capacitors in parallel, radially spaced around the pad. But there's not a lot of room for the BNC connector posts so 2 or maybe 3 at most capacitors can be used at each bulkhead. 

It might have been better to drill holes for the BNC posts and mount it on the non-copper side so that more flexibility in capacitor placement and number was possible. That didn't occur to me until after I had built the prototype.

2. An air wound solenoid style 1uH inductor is starting to become a bit large - so this suggests a toroid, either with a ur=1 phenolic core or a powdered iron core with a VHF rated material. MicroMetals Mix 10 is a reasonable choice. 

A toroid is not perfectly self-shielding, but it's noticeably better in that regard than a solenoid form.

Regardless, at some frequency the distributed capacitance of the inductor is going to be sufficient that the self-resonant frequency will be exceeded. At that point, the filter looks like a capacitive voltage divider. You may still get some filtering action but it won't be what is expected. And once you are up in the GHz range, the inductor may even start looking like an antenna and will radiate. Those factors suggest you should use the minimum wire length possible. But that is associated with a higher permeability core which will usually have other problems as the frequency increases. So, it's a matter of where you decide to take the hit.  And, the shunt capacitors start to look like tuned LC circuits when the inductance of the bulkheads is considered.

Regardless, at some frequency the lumped element filter will no longer perform like a filter. If you really have to have effective filtering from DC to light, it probably will be necessary to use more than one filter in series - perhaps in the same physical package - each optimized for a different frequency range.

Still, this one turned out better than I thought it would.

Jack K8ZOA

 

Offline T4P

  • Super Contributor
  • ***
  • Posts: 3697
  • Country: sg
    • T4P
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #23 on: September 02, 2012, 06:34:16 am »
Gen1 Cable Modem (back in 2006 i think?)

Gen2 Cable Modem (2010)


The first one runs at about 800MHz* and the 2nd one at 1GHz

The 1GHz was introduced as a new system and the old model phased out then they looked for a more reliable system although it can be easily done on through hole and single sided boards
« Last Edit: September 02, 2012, 04:05:57 pm by T4P »
 

Offline vk6zgo

  • Super Contributor
  • ***
  • Posts: 7547
  • Country: au
Re: EEVblog #343 - Spectrum Analyser Tracking Generator Tutorial
« Reply #24 on: September 02, 2012, 06:38:22 am »
 I started to say "By & large,a 15MHz  single section LPF is unlikely to be used in a situation where it has to have a lot of loss around 70MHz & above."-----but I realised as I wrote it,that is exactly the situation faced by an Amateur Radio operator on 14MHz or so,when a neighbour is trying to receive Australian CH2 analog TV,which has its vision carrier around 64.25MHz,& sound carriers around 69.75MHz.
Remember TV front ends are broad as a bull's backside! ;D

Of course,it's not quite so bad,as the 5th harmonic will not have a lot of energy,but it definitely seems like a good idea to have several sections,as is commonly done.

As to the bumps & hollows in the TG,if you can correct them,all is well,particularly as you usually only looking at a fairly restricted bandwidth,very seldom at full scan.
I had a chance to buy an old Polyskop sweeper at the last Hamfest,but at 64kg ,& $300,I thought better of it.
I only would have got all stirred up,messing around with filters,anyway! ;D
« Last Edit: September 02, 2012, 06:51:21 am by vk6zgo »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf