Options to measure frequencies beyond 6GHZ - LMX2594 15GHZ signal generator

[mehdi]

Hello.
I am building an RF signal generator based on TI's LMX2594 eval kit.
It generates up to 15GHZ, while my spectrum analyzers go up to 6GHZ.
I can only think of 2 options to be able to measure and check my signal generator's output above 6GHZ, with my tools:

1. Buy a spectrum analyzer: the problem is the price and size. I can't afford a new 10+GHZ device, and I don't have enough space for old devices like HP 8593e. The only realistic option in my budget (less than 3K) is a Signal Hound SA124B which covers up to 12.4GHZ which is fine for me, but I have doubts about the impact of its phase noise on my measurements. It has -80dBc/Hz on 10GHZ in 1KHZ offset.

2. Buy a down-converter to check the signals with my Signal Hound BB60D. I have found one https://www.lotussys.com/products/budc2g14g which solves my exact problem as a frequency extender for spectrum analyzers, but I have doubts about its conversion loss (~9dB at 10GHZ) and noise figure (~11dB at 10GHZ). I don't know whether these numbers are good or not, and what their impact on my measurements will be.

Now, there might be other options as well, that I have missed. And also, what are the pros and cons of the above options.
I'd appreciate some help here.

[coppercone2]

I think the problem is going to be that you need more power to compensate for those problems.

buy both, pay twice, learn twice as much

if you own a SA and you wanna get your moneys worth from it, education wise, you end up owning the other thing you are describing too. its just like having a transistor to experiment on

for a signal generator, I would be surprised if there was a problem. for picking up a signal from an antenna then you have way less power to work with.

I assume you can put out like 0dbm at least. a signal from the air is going to be mad weak (even for a cell phone its often almost nothing)

you can lower noise, reduce losses, or increase signal strenght. since you have a generator hooked up by a coaxial cable, #3 is like way more practical for a lab source. that is like getting a bigger dish, or just turning up the output potentiometer, for a lab source

for instance, gps basically stops working at -160dbm. works good at -140. I think a useful lab source should do 10dbm, and then have an attenuator. there is no reason to reduce the power until you have to because it just makes everything more difficult.

infact your generator does 7dbm. That is like a ship horn compared to ant walking.

[tszaboo]

You can rent one. I've rented 26GHz spectrum analyzers, and if you need it for a week, they are quite reasonable cost.
Though you might need to be a company to do that, discuss and negotiate with the rental company.

[fant]

Good morning;

I am very interested about the generator; are you planning to provide it with modulations ( AM, FM, Digital?)

About the converter, I think it is easy to do it, with a 6 GHz generator and a mixer, you have the same attenuation (about 8-9 dB) and it cost much less.

Remember that the  spectrum is upside down.

Mandi

[tggzzz]

Good morning;

I am very interested about the generator; are you planning to provide it with modulations ( AM, FM, Digital?)

About the converter, I think it is easy to do it, with a 6 GHz generator and a mixer, you have the same attenuation (about 8-9 dB) and it cost much less.

Remember that the  spectrum is upside down.

Mandi

Compensating for attenuation in the instrumentation is easy and cheap.

Compensating for phase noise in any of the instrumentation is not easy and not cheap.

I too have a 2594 eval kit. I remember being surprised at how low its phase noise was at 15GHz. But, I only did quick and dirty experiments using an HP8562B and Tek492AP, haven't saved any results, and wouldn't feel justified in using them for anything more than amateur curiosity.

There was much more noise visible on the HP, including spikes visible at +-30kHz. I presume, without evidence, SMPS noise. The 492 display was much cleaner, for reasons I haven't investigated.

[mehdi]

I am very interested about the generator; are you planning to provide it with modulations ( AM, FM, Digital?)

For now, I just want to set it up and running, locked to an ultra low phase noise OCXO.
I have no plans for modulation yet (I already have a vector signal generator that covers all my needs up to 6GHZ)

I too have a 2594 eval kit. I remember being surprised at how low its phase noise was at 15GHz.

They managed to get very low phase noise with the eval kit, but it's mainly due to the Wenzel OCXO they used (it's pointed out in the eval kit application note)

[Bicurico]

You can use a TinySA Ultra in the enhanced mode, which will allow you to see signals up to 12GHz. You won't be able to carry out measurements, though, as the signal shown will be attenuated and is uncalibrated.

This, however, is your cheapest option.

Then you could go the down-conversion route, using a suitable mixer and signal generator. However, the cheapest signal generator I know for such an application would be the BG7TBL series of signal generators, which cost a few hundered Euro on AliExpress.

You would then need a suitable mixer - with luck you get a second hand on on Ebay, but pay attention to the supported frequency range.

This will not allow you to carry out meaningful measurements, either, because of attenuation and lack of calibration.

I agree that the SignalHound would be probably your best option, apart from buying a second hand spectrum analyser with desired bandwidth (but these might still be more expensive than the SignalHound).

[mehdi]

We live in interesting times!
Last week, SignalHound started pre-ordering for the SP145 spectrum analyzer (14.5GHZ, with 40MHZ realtime bandwidth and 200GHZ/s sweep)
Today, Aaronia announced pre-order for their new 18GHZ real time spectrum analyzer with 60MHZ of realtime bandwidth and 3THZ/sec sweep!
The former for around 10K USD, and the latter for ~6K EUR in the base version (1 RX)

[Gerhard_dk4xp]

Hello.
I am building an RF signal generator based on TI's LMX2594 eval kit.
It generates up to 15GHZ, while my spectrum analyzers go up to 6GHZ.
I can only think of 2 options to be able to measure and check my signal generator's output above 6GHZ, with my tools:

1. Buy a spectrum analyzer: the problem is the price and size. I can't afford a new 10+GHZ device, and I don't have enough space for old devices like HP 8593e. The only realistic option in my budget (less than 3K) is a Signal Hound SA124B which covers up to 12.4GHZ which is fine for me, but I have doubts about the impact of its phase noise on my measurements. It has -80dBc/Hz on 10GHZ in 1KHZ offset.

2. Buy a down-converter to check the signals with my Signal Hound BB60D. I have found one https://www.lotussys.com/products/budc2g14g which solves my exact problem as a frequency extender for spectrum analyzers, but I have doubts about its conversion loss (~9dB at 10GHZ) and noise figure (~11dB at 10GHZ). I don't know whether these numbers are good or not, and what their impact on my measurements will be.

Now, there might be other options as well, that I have missed. And also, what are the pros and cons of the above options.
I'd appreciate some help here.

I'm facing pretty much the same problems.
I have also built a synthesizer based on the LMX2594. I did not yet write software for it but operated
it with the controller board of the eval kit. The controller board could not tell the difference between my board
and the original RF board. I published it in DUBUS and was flooded by mails that the LMX2594 was
pure unobtainium as a chip. It still is. That did not stimulate my will to write software for it. I was told that
some broker had bought up all of the tiny supply and now wanted 100s of $$$ for a chip.
But last time I looked, there was now the LMX2595 available at Digikey. Same thing, with integrated doubler
to 20 GHz, footprint and mostly register compatible. Price is around $100, but the '94 ws not that much
cheaper.

My SNA-33 spectrum analyzer is defunct. The saga is in the metrology section, with the stories
of other people. So I built a down converter for my Agilent 89441A, a FFT analyzer with tuner to 2.7GHz.

100 MHz xtal osc, *5 , 500 MHz bandpass from my student time, Mult to 2 GHz with overdriven MMIC,
pipe cap filter, multiplier to 8 GHz, pipecaps to clean up, 17 dBm with SKY MMIC, Hittite HMC-220 ring mixer and voilà.

To measure phase noise, I have a Timepod, but it only goes to 30 MHz. The Timepod can do cross correlation,
It uses 2, but uncorrelated channels and it measures the noise that is common to both channels nevertheless.
It can measure oscillators that are 20 dB better than it's own two uncorrelated references (Morion 5MHz double
ovens in my case).

I'm working on a pair of down converters from 0..20 GHz or so  to 5-25 MHz. They are dual conversion. First IF
is 900 MHz with SAW filters. There aare 2 100 MHZ ECS crystal ovens that feed the LMX2995 synthesizers, and
also 100 MHZ * 3 * 3 = 900 MHz 2nd LO.  The SAWs start at 900 Hz, so they can be used to clean up the
2nd LO, too.  $1.50 a pcs.  Mixers are MCL from short wave to 3 GHz and LT5553 above.
The LT5553 has a built-in LO driver and 1 dBm is enough for it. The LMX can deliver that easily.

Because of the cross correlation, this setup can measure the LMX phase noise with some averaging.

I see, you are in .de.   I'm in Saar county.

cheers, Gerhard, DK4XP

[mehdi]

>LMX2594 was pure unobtainium as a chip. It still is.
The chip itself is available on Mouser for around €90.

>I see, you are in .de.   I'm in Saar county.
I am in Berlin!

>Because of the cross correlation, this setup can measure the LMX phase noise with some averaging.
Did you manage to measure and compare the phase noise by using different input clocks? I will also write my results here when I set it up.


Cheers.
DF2HF

[Gerhard_dk4xp]

I'll be in Berlin for a DK0TU veterans meeting next weekend...
(ham radio club station of the tech. univ)

I intend to do a Dubus writeup when I'm done. The board will be 5*10 cm or so.
These SAW filters are awsome.

Dubus started in Berlin, 50 years ago or so.

 :-)
Gerhard

[Verticon]

Hello,

it is interesting to see that there are others around to build microwave generators based on the LMX2594/95, too. I've started to build mine approx. two years ago. For the design I was strongly inspired by the Erasynth++ generator. Because I have other generators up to 6GHz I restricted the frequency range  to 4 to 18GHz. The main synthesizer is a LMX2595 that is driven by either a 100MHz low-noise oscillator directly (which is part of an ADF4002 PLL, stabilised by a 10MHz low phase-noise oscillator) or in the low-spurs mode via a 2nd synthesizer with a LMX2594. For signal enhancement there is an output amp module (2x electronic attenuators RFSA 2113 and 2 integrated amplifiers). The RF part was built on 3 separate shielded 4-layer FR4 boards in order to be flexible in case changes or improvements are neccessary. By the way, soldering the components, amongst them the many QFN packages and 0402 parts to the boards was a bit of a pain. The generator is controlled by an Arduino Due and the software includes the controlling of all functions, a sweep routine, a level calibration routine and allows communication via serial interface.

Although I regularly use the generator since more than half a year there is still plenty of room for optimisation. Because the generator has no ALC circuit the level calibration process is especially important but also very time consuming. Therefore one important next step is to implement an automatic calibration process.

Despite the use of FR4 for an 18Ghz device is a bold venture I am quite happy with its performance. The maximum output level is beyond 13dBm until 16GHz and 10dBm until 18GHz. The two electronic attenuators allow a useable level range of below -10 to +13dBm ( this is limited because room for AM modulation and flatness correction is neccessary). After calibration the level is constant within 1dB between 4 to 18GHz. The closed loop phase noise at 6GHz is well below 100dBc at 10kHz offset. The signal between 7,5 and 15GHz where no divider or multiplier is used is sufficantly clean. Harmonics around 10 GHz are below -35dBc. Above 15GHz and below 4,5 GHz the amount of subharmonics and harmonics is high. Originally I prepared a 4th board with switachble filters between synthesizer and output board but because of the high signal attenuation it was not very successful and I skipped it, But a simple external high-pass or low-pass filter, respectively, is doing the job. AM modulation is possible with an external signal. For FM modulation the PLL tune signal of the main synthesizer principally can be modulated by an external signal, too. But the active loop filter which is especially used for the case of FM modulation is extremely noisy and has to be further optimised. In the first approach the generator had an SMPS for the main supply. Meanwhile, I replaced it by a linear supply because of annoying noise in the rf signal. Unfortunately, this demands to add a 2nd transformer which resulted in a bit weird mechanical arrangement of the components.

I am very interested to hear about your experiences with building your own  signal generators.

FYI I've added some pictures of the 3 RF boards (reference, main synthesizer and attenuator/amp board), a bottom view of the generator with the shielded modules, a top view and also a picture of the main display page.

[mehdi]

@Verticon
Thanks for sharing your setup. It looks very nice!

I am a beginner in microwave, so I'm going step by step, and using modules.

Yesterday I tested the LMX2594 eval board and checked the output signal on 5.8GHZ just to make sure the setup and everything works well.
Now, the next step is to get a good reference for it (an OCXO)
Then, I need to test its output on >6GHZ (using a down-converter)
I also got a power meter to check the amplitude (would be important in the later stages)

My plan for now is a simpler version of what you've built: OCXO, the synthesizer, and 2 step attenuators (0-11dB and 0-70dB) controlled by an MCU, inside a box.

The closed loop phase noise at 6GHz is well below 100dBc at 10kHz offset.

I am aiming to go as much below that as possible (that's why I also emphasize on using a low noise OCXO)

Is there anything you want to share? (any tips? mistakes to avoid?)
Do you know the acceptable amplitude range for the reference oscillator?

Regards,
DF2HF, Mehdi

[mehdi]

By the way, what have you done for the harmonics? (any filtering?)
THD is not good at all!

[Verticon]

@ mhedi

thanks for showing the nice SA plot. A similar plot from my signal generator looks a bit worse. There are still some small single spurs in the low MHz-regime around the carrier. Though there level is below -70dBc but they are there. After having excluded other sources my guess is it is caused by the arduino.

Now to your question concerning the reference oscillator. I used a 100MHz ABLNO-V-100 as reference oscillator. According to the data sheet it is called an "extremely low phase noise oscillator" with typically -162dBc/Hz @10kHz offset. From a comparison of the data sheet values of the LMX2594/5 with my measurements (simple measurements with an SA) an improvement of 5 may be 10 dB can be expected. But it is not clear if the difference is caused by the reference oscillator or by peripheral circuits in the generator. The ABLNO-V-100 in my generator is part of a PLL where I used a 10MHz/0,5ppm VCTCXO as reference. So nothing special and especially no OCXO because for precise measurements I use a Leo Bodnar GPS module for synchronisation.

I haven't tested the minimum reference voltage for the LMX2594 but the data sheet says 200mVpp which shouldn't be a problem for most of the available oscillators.

What to recommend for your final generator? I think a modular design makes things easier, especially in the rf area. In case you add additional self-designed rf circuits up to 15 GHz on FR4 material keep the signal path short. And be aware that there a quite a lot of potential sources of noise and distortion from power supplies, microcontrollers, bad grounding or even from fans which are capable to degrade your nice signal quality. So keep an eye on it all the time. By the way the LMX2594/95 need cooling especially when they are mounted in a closed housing.

By the way, what have you done for the harmonics? (any filtering?)
THD is not good at all!

As I already mentioned my generator has no integrated filters. A first approach with a 3-fold switchable filter board where a band-pass filter should especially cut harmonics between 4 and 7,5GHz (where the VCO signal is divided) and a high-pass filter with an additional amplifier should cut subharmonics (signal is doubled) for signals above 15HGz was not succesful. I think mainly because of too long signal pathes on an FR4 board. So I didn't use it. This is no problem for the 7,5 to 15GHz area because here the signal is acceptable clean. But for serious work below an external filter has to be used (although 5 - 7,5GHz is not too bad without filter). For a pcb version a minicircuits LFCW-7500+ in series with a HFCN-4600+ is a cheap and sufficient solution.

[Gerhard_dk4xp]

Did you manage to measure and compare the phase noise by using different input clocks? I will also write my results here when I set it up.

No, not yet. For phase noise measurements, my Timepod goes only to 30 MHz. I'm trying
to build a cross correlation dual channel down converter to at least X-band. That will involve
some LMX...

I have made a 100/300 MHz clock generator since 300 MHz is the fastest phase comp.
frequency of the LMX2594 in fractional mode. It remains to be seen if my tripler is a real
advantage or if I can just inject the 100 MHz CMOS signal into the synthesizer.

The clock generator can use either an ECS 2522 100 MHz crystal oven (Digikey) or a
CVHD950 or Abracon osc. Or an El Cheapo DIL-14.
I got the phase noise plots from ECS for the 100 MHz oven. Quite OK.

The 100 MHz osc can be locked to a 10 MHz REF from my Lucent GPS.
The 100 MHz -> 10 MHz divider (74LVC161) seems to make some unwanted birdies.
Maybe I join the flock of ADF4002 users.


BTW a customer of mine bought a Signal Hound and it was off by > 10 dB fresh
out of the box against URV-35 level meter & SMIQ04. Check that.

2nd. pic is the 8GHz +-2.7Ghz down converter mentioned above, still in
statu nascendi.  +-2,7 GHz because of my Agilent 89441A.

[Gerhard_dk4xp]

Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".

Ah, attribution gone wrong. Should be tggzzz 

Come on. Making a visit with the cows is more fun with a B-Falke than with a ASH-25.

:-)   Gerhard

[tggzzz]

Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".

Ah, attribution gone wrong. Should be tggzzz 

Come on. Making a visit with the cows is more fun with a B-Falke than with a ASH-25.

:-)   Gerhard

Nothing with a cooling fan (in front or on top) is better than a glider Plastic gliders are better in cow fields: traditional problem with cows and older gliders is that cows have a taste (literally) for licking dope-covered fabric, with expensive consequences!

A very common occurance on a big field near where I used to glide. Little big buggers lie contentedly on the hot tarmac when chewing the cud - and completely ignore car horns Traffic queues form.

Useless for landing even when the cows weren't there, since it is fully of 80 year old anti-glider trenches specifically designed to stop gliders landing there.

[Gerhard_dk4xp]

But in Phokara, Nepals, you cannot even pull the horn.
Not even for these sorry specimens of a cow. You would
like to feed them.
But a test flight is still allowed. Annapurna group to the
left in the dust.   I prefer REAL gliders.

to be more on-topic:
there are left-over boards from my experiments. They are
available for free in EU. Brazil  and US were a customs
disaster in previous tries.   Gerhard

[Verticon]

@ mhedi

FYI and for comparison with your work, some spectrum analyzer plots of my signal generator.

The 1st plot is to compare with yours from your previous plot: span 20MHz, RBW 100Hz. The tiny spurs in the MHz region around the center are yet to be eliminated.

The 2nd plot shows the region around the carrier in more detail: span 20kHz, RBW 1Hz

The 3rd picture shows the phase noise behaviour vs offset frequency. It is made with an 8561 (data sheet says -116dBc/Hz @ 5GHz typical

Perhaps it's useful for you

Cheers

[mehdi]

Thanks for the pictures.
It's definitely useful to compare the results.
I have also attached my phase noise measurements result, but it's not very dependable, as my spectrum analyzer's phase noise is not that great itself (I'm using SignalHound BB60D)
Here, I am feeding the eval kit with another signal generator as reference input (Signal Hound VSG60A)

[mehdi]

BTW a customer of mine bought a Signal Hound and it was off by > 10 dB fresh
out of the box against URV-35 level meter & SMIQ04. Check that.

The spectrum analyzer or the signal generator?

[mehdi]

Ok, I just compared my Signal Hound spectrum analyzer with a calibrated RF power meter. It's off by 1.1 dB, which is acceptable.
Here's the 0dBm 5.8GHZ signal measurement comparison:

[Gerhard_dk4xp]

Spectrum analyzer. SMIQ only for initial tests. We were more interested in 8 GHz.
There, the signal source was an ADF5356 eval board.
Thermal power head Z51 on URV-35, DC to 18 GHz.
A thermal head that works from DC is easy to check itself.  :-)

BTW, the ADF-5356 eval software allows weird register programming, and that
can generate funny spectra. Also, the ADF-5356 output doubler has just so some
dB of fundamental suppression, what drove me to the LMX2594.

<      https://www.flickr.com/photos/137684711@N07/50403778976/in/datetaken/         >
Gerhard

[mehdi]

I got an up/down converter to be able to see signals beyond 6GHZ.
Although it's got bad phase noise, it's helpful to check the signal generator's output.
Here's a 12GHZ signal shown on 5GHZ.
Also, got an ultra low phase noise OCXO to use as reference input.