In this blog post, I will show you how to build a 0 to 5.8 GHz tracking generator for the HP 8566B 100 Hz to 22 GHz spectrum analyzer using off-the-shelf components for under $100. Although this tracking generator is specifically designed for my HP 8566B spectrum analyzer, the method discussed below is applicable to pretty much any spectrum analyzer that has an LO output (typically the 1st LO).

A tracking generator, as its name implies, tracks the frequency of the spectrum analyzer’s sweeping oscillator (typically 1st LO) so that the tracking generator’s frequency output matches the center frequency of the bandpass filter in spectrum analyzer’s IF stage. Thus at any given moment, the spectrum analyzer sees the same frequency input as what it is currently sweeping at. The combination of a spectrum analyzer and a tracking generator is often referred to as a scalar network analyzer (SNA).

Tracking generators are very useful for measuring the frequency response of filters and amplifiers. In my previous blog post, I discussed how to perform this kind of measurements using the manual sweeping method, without using a tracking generator. Although there are some benefit sweeping through the frequency range manually, the obvious downside is the slow speed. Because of this, the manual sweeping method is only useful for characterizing devices with time-invariant frequency responses as a single sweep can take minutes to complete.

The basic principle behind a tracking generator is rather simple: the tracking signal can be obtained by subtracting the IF frequency from the LO output via a mixer. Take the base band (0 – 2.5 GHz) of HP 8566A/B for instance, the 1st LO output goes from 3.6214 GHz up to just above 6.1 GHz during each sweep. So if we subtract the 1st IF frequency (3.6214 GHz) from the 1st LO, we get our precise 0 – 2.5 GHz tracking signal.

John Miles at KE5FX documented his 2.5 GHz tracking generator build for the HP 8566A/B on his website. In his build, John used the mixer and automatic level control (ALC) circuitry from an HP 86222A RF plugin. And you can find a similar build by CuriousMarc using the mixer from an HP 83522A plugin for the HP 8350B signal generator.

If automatic leveling is not a primary concern (most of the time, leveling is not critical since almost all spectrum analyzers have builtin functionalities to calibrate and thus correct the leveling of the sweeping source), a tracking generator can be built using just a suitable microwave frequency mixer and a reference frequency that is determined by the frequency band in use. This approach is what I used for building my tracking generator.

Before going into more details, let’s first take a look at the different base and harmonic bands HP 8566B can operate under:

Band f1 (GHz) f2 (GHz)
0 0 2.5
1 2.0 5.8
2 5.8 12.5
3 12.5 18.6
4 18.6 22

The frequency ranges listed above are specified in HP8566B’s manual and are the typical frequency intervals for the harmonic bands during the default automatic sweeping (i.e. band unlocked, SHIFT+Q). Since the 1st LO can sweep between roughly 2 GHz and 6.2 GHz, the actual achievable range for each harmonic band is actually much wider. The following table lists the frequency ranges when each harmonic band is locked (SHIFT+t):

Band f1 (GHz) f2 (GHz) f1 (LO GHz) f2 (LO GHz)
0 0 2.5785999 3.6214 6.1999999
1 1.67860005 5.8785999 2.000000050 6.1999999
2 4.32140005 12.7213999 2.000000025 6.199999950
3 6.32140005 18.9213999 2.000000016 6.199999966
4 8.32140005 25.1213999 2.000000012 6.199999975

By examining the data above, you will notice that the IF for the base band is at 3.6214 GHz and for all the harmonic bands the IF sits at 321.4 MHz.

The base band (band 0) and the first harmonic band (band 1) are of particular interests, as the LO output can be used as the tracking generator output signal directly after subtracting the IF frequency without the need of harmonic mixing.


Choosing a Mixer

So a suitable mixer for our tracking generator should be able to mix signals in the entire LO range (2 GHz to 6.2 GHz) with either an IF frequency of 3.6214 GHz for the base band or an IF frequency of 321.4 MHz for the first harmonic band. Because the LO output from HP 8566B is greater than 5 dBm, a level 7 or level 10 mixer is suitable.

I ended up choosing Mini-Circuits’ ZX05-83-S+ 2300 MHz to 8000 MHz level 7 mixer as my work will mainly be dealing with frequencies under 2 GHz. This mixer has excellent performance in the 0-2.5 GHz base band for mixing the 3.6214 GHz to 6.2 GHz LO signal with the 3.6214 GHz IF. This mixer is relatively inexpensive, retails for less than $50. As you will see later, even though this mixer is not rated for RF signals as low as the IF frequency for the first harmonic band (321.4 MHz), it is still usable in the 5.8 GHz frequency band, but the performance does suffer suffer a bit.

If you need to work in the first harmonic band more often, you probably want to consider a wide band mixer such as Mini-circuits’ SYM-63LH+ level 10 mixer, which has a reasonable performance curve between 1 MHz and 6 GHz. It only costs slightly more than $10, but only comes in surface mount package.

For any other frequency bands (other than the base band and the first harmonic band discussed above) you will need to find a suitable subharmonic mixer for the desired frequency ranges, but it will likely be very pricey unless you can find a used one on eBay.

Choosing an RF source

An ideal RF source for the tracking generator should be phase-locked to the spectrum analyzer, this is especially important when the resolution bandwidth (RBW) used is narrow. It also should have sufficient output power (e.g. at least 0 dBm) so that the tracking signal is strong enough to provide adequate dynamic range.

There are a couple of solutions available. One is to get a PLL signal source that is capable of generating the 3.6214 GHz signal for the 0 – 2.5 GHz range and/or the 321.4 MHz signal for the 1.68 GHz to 5.68 GHz range (band locked to the first harmonic band). Such phase locked oscillators are widely available on the used market and can be had for less than $30. If you only need to use a relatively high RBW, then the oscillator does not even need to be phase-locked to the spectrum analyzer clock, as long as the oscillator is stable enough.

Alternatively, you can use a signal generator or synthesizer to generate the desired frequency. As a bonus, most signal generators can be phase locked to an external clock source. Since I already have a synthesized RF signal generator (HP 8642B), I chose to use that to generate the required frequency. Of course, since HP 8642B only goes up to 2.1 GHz whereas the signal required is at 3.6214 GHz. So I used a Mini-Circuit frequency doubler (ZX90-2-19-S+) to get the required signal by doubling a 1.8107 GHz 10 dBm input signal.


This setup is not as ideal as using a proper 3.6214 GHz signal source however as there will be some fundamental frequency leak-through in addition to the higher harmonics generated by the multiplier. Also, due to the conversion loss of the frequency doubler and its maximum allowed input level, the strength of the output RF signal is only at around 0 dBm. Of course, this issue can be alleviated by adding in a cavity filter after the doubler and inserting an amplifier to further boost the signal strength before inputting into the mixer. But this would add considerable cost to the tracking generator. As is, this tracking generator is already quite usable.

LO Isolation

Although it is possible to connect the LO output directly to the mixer’s LO input, the reverse LO leakage will reduce the available dynamic range. So I decided to add an isolator between the LO output from the spectrum analyzer and the LO input to the mixer. I happen to have a Harris A12390 isolator on hand, it offers 20 dB isolation from 3.6 GHz to 6.4 GHz which covers almost the exact LO frequency range for the base band. Although from my experiment data shown later, the inclusion of the isolator only added marginal benefits. I suspect that adding a narrow bandpass filter centered at 3.6214 GHz would help as the output from the frequency doubler has a lot of unwanted signal components.

Alternatively, an attenuator in conjunction with a low gain amplifier could also be used to reduce the back feeding of the LO signal.



Putting It Together

The diagram below shows the components used in my tracking generator: the mixer (ZX05-83-S+) takes signal from the 1st LO and mixes it with the 3.6214 GHz signal from the frequency doubler (ZX90-2-19-S+) and generates the 0-2.5 GHz tracking signal for the base band. The input to the frequency doubler is taken from my HP 8642B which is phase-locked to the HP 8566B.


The maximum signal strength of the tracking signal is limited by the LO signal strength and the RF signal strength (the output from the frequency doubler). The frequency doubler used here limits the maximum input power to 10 dBm, and the output 3.6214 MHz frequency is just under 0 dBm. Adding the conversion loss from the mixer, the tracking signal is at around -7 dBm which still gives at least 25 to 50 dBm of dynamic range.

The quality of the tracking signal is actually quite good. It is reasonably flat considering that this is taken from the LO output directly, without using a leveling amplifier.


The picture to the left below shows the actual tracking generator components. The picture to the right shows the noise floor when the spectrum analyzer’s input is disconnected. The baseline noise floor is not as ideal as I had hoped. At between 1 GHz to 1.5 GHz there is a significant rise in the baseline noise and because of this we only get about 25 dBm of dynamic range within this frequency region.

As mentioned earlier, I believe this added noise can at least be partially attributed to the multiple frequency components in the multiplied signal. I will need to do a couple of comparison studies by either using a phase locked oscillator to provide the 3.6214 GHz signal directly or using a cavity bandpass filter centered around 3.6 GHz to see which method yields the best dynamic range.

trackinggencomponents basebandnoisefloor

The addition of the RF isolator (3.6 GHz to 6.4 GHz) does seem to have improved the performance a little bit, although not as much as I had thought. The picture to the left below shows when the isolator was taken out of the equation and the picture to the right below shows the measured noise floor without using the isolator. As you can see, the isolator seemed to have narrowed the noise in the “problematic” band quite a bit but only managed to reduce it by a few decibels in the 1 to 1.5 GHz region.

But this picture does suggest that you probably can do away with the RF isolator, which could reduce the BOM even further.

basebandwoisolator basebandnoisefloorwoisolation

For the 1st harmonic band, a 321.4 MHz signal from the HP 8642B is coupled directly into the RF port of the mixer without using the frequency doubler. While the tracking signal’s strength varied wildly between 1.67 GHz and 5.87 GHz (with the 1st harmonic band locked), the performance is actually quite reasonable between 2 GHz and 4.5 GHz.

Note that for the first harmonic band the 321.4 MHz signal was adjusted to 10 dBm which was inputted into the mixer directly (as opposed to the 0 dBm signal via the frequency doubler in the base band), this additional power helps improve the dynamic range.

Since the mixer (ZX05-83+) used here is technically not rated for frequencies under 2.3 GHz (the IF for the harmonic band is only 321.4 MHz), the performance we see is actually quite good. Of course, you can always swap for a better spec’d mixer if you find yourself needing to characterize devices in the first harmonic band more often. I think you should see additional performance improvement with a properly rated mixer.

firstharmonicband firstharmonicband1

Here is a picture showing the noise floor of the tracking generator in the 5.8 GHz harmonic band.




All the experiments I did in my previous blog post using the manual sweeping method can now be done with the tracking generator instead.

The first experiment below shows the characterization of an LC filter. Because the tracking generator’s output is very flat between 150 MHz and 450 MHz, no trace adjustment was made before the expeirment:


Here is the measured result. As you can see from the screenshot below, the resonant frequency of this particular LC filter is at 323.1 MHz.


The next experiment shows the gain-bandwidth characterization of an ADL5536. ADL5536 is a 20 MHz to 1 GHz IF gain block with a fixed gain of 20 dBm.


For the measurement, the baseline is lowered to just below -40 dBm (by lowering the RF input into the mixer) and trace arithmetic was used to normalize the baseline between 10 MHz and 1.5 GHz. This is condition is slightly different than what I used during my video demonstration.


And the picture below shows the output level from the RF amplifier between 10 MHz and 1.5 GHz. There are some losses due to the coax used, but the shape of the curve is largely in line with what is specified in the datasheet. And as you can see, the gain of this amplifier rolls off very gradually as frequency increases and thus this amplifier could be used all the way up to at least 1.5 GHz. Of course, at the higher frequency end, the gain is limited to just around 10 dBm.


Final Thoughts

While the work done here is for the HP 8566B spectrum analyzer, the same method can be used to build tracking generators for virtually any spectrum analyzers with an LO output. As I illustrated, a very basic tracking generator can be built from off-the-shelf components for less than $100. Additional components (e.g. RF isolators, bandpass filters, leveling amplifiers and subharmonic mixers etc.) can be added to improve the tracking generator performance further.

Be Sociable, Share!

29 Thoughts on “Building A Tracking Generator”

  • Hi Kerry,
    Excellent article.
    You explained your approach to cheap tracking generator very well.
    Will try your approach on an old TEK specy.
    I have a Wavetek 904 which i may be able to use as a stable Local Oscillator
    for the mixer as you descibed (Wavetek freq range 3.7Ghz to 7.6Ghz)…but
    have not gone into it yet)…too many projects to little time)
    Have been following your blogs for several years.
    I am currently working on your Arduino based Constant current dummy load.
    I liked your recent article on measuring ESR.
    I like the way you approach a subject using the basics.
    I built a micro based 100khz ESR meter years ago but never thought
    to try using other frequencies like you did.
    Good to see you on U Tube these days
    Wish You the very best for 2016
    George (Sydney, Australia)

  • i have an agilent 8594e spectrum analyzer but it does not have a tra,cking generator. would your tracking generator work for my unit.

    • Hi Dennis,

      I just checked the service manual for 8594e quickly. It has a 1st LO output and the 1st LO is 3.9214 GHz above the input signal frequency (e.g. roughly from 3.9 to 6.8 GHz). So it looks like you can use pretty much the same components. The only thing I would double check is your 1st LO’s output power, the mixer I used is a level-7 mixer so ideally the LO output power needs to stay under 7dBm. You may need a level-10 mixer (e.g. ZMX-7GLHR) if the power is closer to 10 dBm.

  • Hi kerry. February 8 I asked if your tracking generator would work for my spectrum analyzer. You researched it and got back t ok me. I never said thanks. That was rude of me. So I’m kinda late with it but I’m saying thanks now.


  • Hi Kerry,

    a defekt Anritsu MS710C SA is fallen into my hands.So I was lucky to repair it.Now I need a Tracking Generator (the MH680B is of no use,because it ends at 2GHZ).
    Because I have good experince
    with Mini-Circuits so I plan a Tracking-Generator to 6 GHZ.
    The exact frequency is 4.3 GHZ for a Radioaltimeter and
    testing of the needed Patchantennas !The Anritsu has two
    LO#2 outputs 2.5 GHZ (no output at high bands)and LO#1
    2.2 to 6 GHZ .What mixer and what VCO do I need to buy ?
    Many thanks

  • Hello Kerry.

    I just got myself old HP8558B analyzer. In start it didn’t work at all, but fortunately only 1.5A fuse in +15 volt line in the mainframe was burnt. Now it is working ok.
    It does not have a tracking generator built in, but it does have 1st LO output in the front panel. Manual says it should be around +10 dBm and 2.05 – 3.55 GHz range.

    So, because I’m bit new with these instruments, would Mini-Circuits model ZX05-C42LH-S+ be a good choice as a mixer for this purpose ? Also I have found that circulator model AMF5283 (HP-0960-0084) is commonly available on ebay. Is this okay?

    The 2.05 GHz oscillator, I already have ADF4351 PLL based signal generator, which maximum output is + 5 dBm.

    Many Thanks


    • So it looks like the IF for HP8558B is 2.05 GHz. So ZX05-C42LH+ would be fine. And the one I used (SYM-63LH+) would also work nicely although you probably do not need as wide of a range. I could not find the datasheet for AMF5283 so I don’t know whether it is rated for 10dB LO input. Also the range is on the 2.05G is on the lower end so which might not be as good as the minicircuits part.

  • Hi,

    Very interesting article!

    I currenty got a HP 8594E from a friend. I does not seem to have 1st LO output. Is it a factory installed option or can I retrofit it?



    • For HP 8594E I think it’s option 009 for LO output. Unfortunately, you will have to access the LO to ensure the sweep is in sync.

  • Hi,Kerry.
    Advantest R4131:
    1st LOCAL OUT 4 GHz to 7.5 GHz more than -5dBm
    2st LOCAL OUT 3.77 GHz (fixed) more than -5dBm

    Which mixer to use with such low levels? Can I use yours?
    I would use the second local oscillator as the mixer’s RF input.
    Confirm the possibility?
    Thank you.

  • Thanks for the answer, but sorry Kerry, i do not understand why to use mixers for large input signals when my outputs are at -5dBm and not at + 5dBm..

    • Balanced mixers typically require higher LO input level. At -5dBm it is a bit too low to obtain useful dynamic range but Schottky diodes typically can operate well below the forward voltage drop point specified so it would still work but just not as well. If you need better performance, you will need to ensure the LO is amplified to the correct level.

  • Or Advantest means to say +5dBm in output and then i understand your tip for the standard level 7 mixer. Just can I measure it. Thank you!

  • Kerry,

    You made several reference as to spectrum analyzer needed to be phase locked with the signal generator. Please confirm you are talking about 10Mhz reference oscillator needing to be common, so that both equipment operates on the same clock. You are not talking about phase of signal needing to be locked to each other. Correct?? I can’t see how clock (or any other signal) being a problem if it is 180 degrees out of phase. Thank you.

  • Hi,Kerry,
    I am going follow your schematic to build a tracking generator for my spectrum analyzer. Both the first and third oscillators are swept. Could I ignore the third oscillator?
    Thank you.

  • Hi Kerry,

    I recently obtained a second hand Agilent E4407b which unfortunately came without a TG. I’ve been investigating options to build a TG myself for it and found your blog very resourceful. The good thing is that this particular E4407 has an auxiliary 1st LO output(External Mixing (Option AYZ)) which would allow me to connect it to an external mixer. I’ve been investigating different options and I’m a bit puzzled with the service manual, according to my understanding 1st LO has a range of 2.4214GHz – 3.9214GHz.
    Which mixer and PLL range I should go for? I’d like to purchase a AD PLL evaluation board also for experimenting and would like to buy one which can synthesize frequency suitable for this range. Any suggestions more than welcome.

    Thank you!
    Kind regards,

      • Hi Berkin, sorry didn’t get a chance to reply to your original question. I receive a lot of emails every day :-). Anyway, if the range is 2.4214 – 3.9214 GHz you can use the same ZX05-83 mixer as it can handle between 2.3 to 8 GHz. you would need a reference of 2.4214GHz as the RF input into the mixer.

        • Hi Kerry,
          Thank you for your response. To doublecheck I connected another SA to the 1st LO output (external one in the front panel) of E4407b. However this SA has a range of 30Hz-2.9Ghz. I was expecting to be able see at least lower range of 1st LO frequencies. To my surprise, I saw nothing :(
          I went through the service manual and calibration manual again, but there’re controversial statements in there. I know this is very basic and it’s very essential information in order to have a TG at the end. May I kindly ask you if you could check the service manual for E4407B and find out what is the 1st LO frequency output from the External LO output from this SA? I googled and there’s no information I can find which gives this information reliably.
          But I heavily suspect, the fact that I’m not able to see any LO frequency on the other SA is that the range of E4407B is outside the measurement capability of this old SA.
          Kind regards,

          • Hi Berkin,

            If you take a look at
            Page 95, yo uwill see that the First LO is 2.4214 to 3.9214 GHz as you staged. There’s also some instruction on Page 40 on how to check the LO:
            6. Check that the TG LO signal is present.
            a. Set the analyzer to a center frequency of 50 MHz and zero span.
            b. Use an active probe, such as the HP/Agilent 85024A, and a microwave
            spectrum analyzer to probe A2A1J4. J4 is not loaded, but the holes for the
            jack are present; probe the center hole. One of the other four holes may be
            used for a ground connection.
            c. Set the microwave spectrum analyzer to a center frequency of 2.5 GHz and
            a span of 1 GHz.
            d. An LO signal of 2.45 GHz at approximately 20 dBm should be present.
            e. If the LO signal is not present, or is significantly off frequency, suspect the
            A2A1 tracking generator control board.
            f. If the LO signal is OK, suspect the A2A2 tracking generator RF board.

  • Kerry,
    I build something similar. I used a cheap variable attenuator in the fixed LO path to adjust the trace. The 1st LO output of the spectrum analyzer had a difference of ~10db between the start of the sweep and the end. This resulted of course in a trace variation of 10db over the 1.7-8Ghz range I build it. I got around that by using 2 mini circuits amplifiers back to back using the second amplifier in compression mode. This reduced the overall trace difference to about 1.5db over the full range. More than enough to look perfectly straight at lower spans. Care must be taken to terminate the input and outputs of the mixer, I use 3db-10db pads and adjust the levels accordingly. I can adjust the trace from -5 to -35db, ample enough adjustment for any measurement I want to make.

Leave a Reply

Your email address will not be published.