Need reasonably priced (i.e. cheap) signal generator with low phase noise

[Leo Bodnar]

I am looking for low phase noise benchtop signal generator for use in day-to-day phase noise measurements.  I still haven't found what I'm looking for.
Here are my rough criteria:

10MHz..1GHz range. More is better
+7..+15dBm output
External frequency control (FM with DC coupling) or internal reference tuning input.  10Hz BW and few ppm is enough
10MHz carrier phase noise: 10dB better than attached chart
1GHz carrier phase noise:–90dBc/Hz at 10Hz, –120dBc/Hz at 1kHz offset, –150dBc/Hz at 100kHz offset
Higher carrier frequencies are OK to follow 20dB/decade degradation rule.
Not more than £2k from eBay in fully working shape (to be able to use two for cross-check tests)

I have a few Agilent ESG-AP and ESG-DP series but they are not very impressive below 250MHz where they start heterodyning the output. R&S SM300 and few Marconis that I have tested are way off too.
Would  Aeroflex/IFR/Marconi 2040, 2041, 2042 or R&S be anywhere close?

If you have any generator phase noise graphs you took please drop it in this thread - even if they are noisy.  This makes selection process much easier.

Cheers
Leo

[tautech]

The SSG3021X (2.1 GHz) is 2k Euro new.
Phase noise graph on P8 in the datasheet:
https://www.siglenteu.com/wp-content/uploads/dlm_uploads/2018/11/SSG3000X_DataSheet.pdf

[Ice-Tea]

HP8662A? 

EDIT: not easy to reach, your requirements...

[Leo Bodnar]

HP8662A is good for close-in noise but a bit rubbish at 100kHz. 

Siglent is a bit rubbish throughout for XXI century instrument. It also does not seem to have DC coupling on FM mod input. Datasheet says "10 Hz~100 kHz" which makes it unusable as controlled signal source in PLL loop.

I'd like to test HP8644A someday but they weigh a tonne and still too expensive to impulse-buy.

[Ice-Tea]

What about a 8642A? Got one of those inbound...

[Leo Bodnar]

not easy to reach, your requirements...

Flukes are close but they are very rare and very expensive - even used.

[MadTux]

HP 8662A/8663A is best at low phase noise close to carrier.
HP 8642B is better at low phase noise far away.
I have a couple of HP8662As and they are far superior to other HP instruments such as HP 8656B/8657B (sold them already) and nowadays digital DDS rubbish from china.
http://hpmemoryproject.org/technics/bench/3048/config_hard_01.htm

Apparently, HP8642B has lots of SAW filter magic inside and I hate HP LCDs, (particularity the HP3457 display) so I didn't get one already. Also weights a ton (I think even more than HP8663A), so it's expensive to ship

[capt bullshot]

Doesn't fit your requirements, but there's a chart in the spec sheet:
http://wunderkis.de/mg3633a/manual/

[Leo Bodnar]

What about a 8642A? Got one of those inbound...

8642A is good at 1kHz and further out but rubbish at close-in.  I have seen HP recommending using both 8642A and 8662 and patching together measurements.  It's a bit too much for a small lab.

[Leo Bodnar]

Thanks.  Close in noise is not the best but otherwise not bad.

Doesn't fit your requirements, but there's a chart in the spec sheet:
http://wunderkis.de/mg3633a/manual/

[hendorog]

Marconi 2041

[Ice-Tea]

What about a 8642A? Got one of those inbound...

8642A is good at 1kHz and further out but rubbish at close-in.  I have seen HP recommending using both 8642A and 8662 and patching together measurements.  It's a bit too much for a small lab.

Too much gear? What have you been smokin' son?

[hendorog]

Also these might be worth a look:

http://www.crystek.com/home/source/source.aspx

[jpb]

I have a Hameg-HM8135 which is about 30dB above your required level! (it was cheap though at £400 for new-old stock):
https://www.testwall.com/wp-content/uploads/2016/10/Hameg-HM8135.pdf
Your requirements are very demanding especially at 1GHz. Looking at SAW oscillators such as the crystek ones mentioned above :
https://www.digikey.co.uk/product-detail/en/crystek-corporation/CRBSCS-01-1000.000/744-1659-ND/4741148
its output is too noisy and it is designed as a low phase noise test source!

Your 10MHz requirements are more reasonable at least compared with low noise OCXOs.

If you use the cross-correlation approach to phase noise measurement then this reduces the noise requirements on the references:
http://www.holzworth.com/Aux_docs/Holz_MWJ_TechFeat_Feb2011.pdf

[Performa01]

Thanks.  Close in noise is not the best but otherwise not bad.

Doesn't fit your requirements, but there's a chart in the spec sheet:
http://wunderkis.de/mg3633a/manual/

I don't have a proper phase noise analyzer, yet I'm pretty confident the MG3633A is about the best you can get and its close-in phase noise is actually much better than indicated by the graph in the data sheet. Here's a phase noise measurement at 1GHz, actually showing the phase noise of the analyzer rather than the MG3633A, but still indicating a much lower phase noise below 60Hz than what would be expected from the data sheet graph.

Also at 10MHz, the MG3633A should be at least 5dB better than the plot you've shown. I know, you want 10dB and it might well be that the MG3633A can deliver that, I'm just unable to verify it.

[DaJMasta]

Gigatronic's 2400 series generators advertise low phase noise but don't even come close to some of your specifications (a table in the datasheet)... maybe their newer stuff fares better, but it's not going to qualify as cheap and they've essentially ended any kind of support for their non-current model generators, which is frustrating.

I think you can find low phase noise in that budget to 1GHz, I don't think you can find phase noise class you're looking for in that range unless you find a steal.

[dkozel]

I'm fairly convinced that there are no signal generators that meet your requirements. The Phase Matrix QuickSyns are generally lauded for their phase noise performance but are 10 dB short of your 1 GHz 100kHz offset number, even used they'd probably be >$2k.



The Holtzworth HSX are the same.

Though the just "scrape by" at 10 MHz.


It sounds like you need a cross-correlation based phase noise system. That way the reference sources can be above your DUT's phase noise by a few dB. The Holtzworth analyzers do this and can measure far below your requirements.
http://www.holzworth.com/Spec_sheets/HA7062C_Web_Datasheet.pdf

The Erasynth generators meet your price requirement and are solidly competitive with the above signal generators in performance. I've wanted to build a phase noise analyzer using cross correlation for some time but it keeps slipping on the project list. I'd use Erasynths or similar in that case.

[Ice-Tea]

Does anyone have a full DS for an R&S SMHU? Seems like it should be pretty good, but I only have the two-pager on the internets.

I have one. Albeit with a PSU that I've blown. Twice.

[KE5FX]

-90 dBc/Hz at 10 Hz from a 1 GHz carrier is a tall order.  To this day I'm unaware of anything on the market with lower close-in noise than an 8662A, and it will just barely meet that requirement if you get a good one that outperforms its 'typical' specs by a couple dBc/Hz. 

Pro tip: spend a bit more for an 8663A, and buy two of them, and then buy another for a parts mule... and as you note, this will provide good close-in noise at the expense of poor broadband noise.  An 8663A and 8642B make a good combination in that respect. 

Basically, getting all of that stuff from a single synthesizer is going to cost about as much as a new Jaguar.  You're much better off using separate sources for different measurements.  Or put together a cross-correlating rig; see this thesis and attached articles by Simon Bale for some good pointers there.

[Gerhard_dk4xp]

The idea of having a spare 866x for replacement parts is a good one.
When I compare my 8662A to my SMPD and SMIQ4, there is
one  clear loser: the 8662A. Most time, it has some "issues".
currently, it works from 0 to 160 MHz, works not from 160 to 320,
works again for 160 MHz and so on. Sick of repairing it; I have
alternatives and it collects dust in a corner.

For close-in noise the choice is crystal oscillators anyway.

[Gerhard_dk4xp]

to complete the two pages, taken from U. Rohde: Microwave & wireless synthesizers, J.Wiley

I think that U.Rohde is preparing an update with E.Rubiola; at this price level you might want to avoid
the double expense.

[chris_leyson]

Adret 7100A won't meet your phase noise spec but it wasn't too bad for back in the day, 1978. I think Marconi aquired Adret and one family of Marconi synthesizers is based on the Adret design but I can't remember which one.

Adret 7100A phase noise at 160MHz

100Hz   -87dBc/Hz
1kHz     -117dBc/Hz
10kHz   -136dBc/Hz
100kHz -136dBc/Hz
1MHz    -145dBc/Hz

I will have to go and dig out my copy Frequency Synthesizers: Theory and Design by Vadim Manassewitsch and have a look at how double or triple mix/divide synthesizers performed back in the day. It's a brute force synthesis technique that was replaced by PLL synthesisizers but potentially has much better phase noise performance. Even with todays improved techniques the double or triple mix/divide synthesier will still cost a lot more than a PLL based synthesier. Just my 2 cents worth.

edit: DDS synthesis has improved a lot since the days of double or triple mix/divide techniques and high speed D-A converters are a lot smaller and cheaper these days. If you summed the signals from N D-A converters and assume the D-A converter noise is incoherent do you still get an N/root(N) SNR improvement. Paralleled amplifiers were used back in the day to get very low noise floors and in principle I don't see why you can't use the same technique for oscillators.

Now you've got me wondering if an injection locked oscillator has a lower phase noise than a free running oscillator, damn it.

Chris

[Leo Bodnar]

Now you've got me wondering if an injection locked oscillator has a lower phase noise than a free running oscillator, damn it.

Do you mean self-injection-locked via delay line?

[Gerhard_dk4xp]

In Rohde, Poddar, Böck: The Design of modern microwave oscillators  for wireless applications,
there is a whole section on the noise of n coupled oscillators.
More math than I'm willing to digest unless someone pays me for that.

Methinks Rohde has also written sth. on self-locking in Microwave journal last year.

cheers, Gerhard

(who would like to couple 16 MTI-260 crystal ovens, but forced by a slooow PLL. The oscillators
are here, time is mising.  More like locking the 16 pcs. to a 17th and wilkinson the outputs together.)

[Gerhard_dk4xp]

I've wanted to build a phase noise analyzer using cross correlation for some time but it keeps slipping on the project list. I'd use Erasynths or similar in that case.

I have the same fantasies here. The idea is to use some AD JESDI-204B 2GSPS ADCs,
an Artix FPGA and a BeagleboneBlack.

In a former project I had easy everyday access to to an E5052B, but now I'm hacking something
together centered around my Agilent 89441A FFT ( can do Xcorr...), ring mixers and a delay line
made from 6 1:6 coax relays and lots of SMA-Semi-Rigid that can enforce 90° delay for any
frequency between 5 and 100+ MHz. That SMA-soldering orgy is of great meditation value.

<    https://www.flickr.com/photos/137684711@N07/40639478472/in/album-72157662535945536/lightbox/    >

Some of the coax relays had TTL drivers and I quickly found out why they did end up on ebay. :-(
I did grind them open, removed the drivers and switch the coils myself. They do not look pretty
any more.

regards, Gerhard

[suj]

Your requirements are very high and with price restrictions in my opinion it's difficult to meet. A good solution for frequencies >1 GHz is R&S SMF100A, but very far from your price point  . With the SMF-B22 option installed, it's not able to meet your requirements, but it is close (SMF100A_1.PNG). Without this option, it's a bit worse (SMF100A_2.PNG).
The main source is the YIG 3..11 GHz oscillator (Micro Lambda with R&S numbers). The 1..3 GHz range is obtained from the divider :2 and :4. Optionally <1 GHz band with downconversion. Range 11..22 GHz multiplication x2 and optional 22..44 GHz next multiplication x2. A large part of the magic with low level of phase noise is contained in the synthesizer, the structure of which I don't know (only a few sentences of the description in the service manual and no block diagram).
The frequency modulator and phase modulator as a source can use the built-in LF generator, external inputs (AC/DC coupling) and the built-in noise generator. This allows you to generate a signal with a pre-set phase noise, for example to check your measuring system.

[Leo Bodnar]

The specs are the target, not a must.  A lot of suggestions here are very good, thank you, guys.
Nothing beats own experience - this is why I don't want to go purely by datasheets.
I have measured PN of my three ESG-AD sig gens and two of them were within 1-2dB from Agilent datasheet but the third is about 10dB worse.  I am tempted to trace and fix the reason but it's a distraction from what is already a distraction.

Commercial cross-correlation PN kit is outside of my money budget and designing one blows the time budget so for now I am staying with classic measurement methods.
Leo

P.S. A lot of my needs are specific to PN testing so they relax overall sig gen criteria.
What I don't need or care too much:

Don't need accurate small step attenuator.
No need for anything below +0dBm
No need for reverse power protection
Amplitude noise is not critical as long as it is not awful.
Modulation is not necessary as long as it outputs CW and can be gently steered with external DC input
Dwell time, sweep speed or range, I/Q modulation, AWG, multi-tone output are not important
Can live with high harmonics levels
Don't care about lower range below 10MHz

Your requirements are very high and with price restrictions in my opinion it's difficult to meet.

Here are my rough criteria:

10MHz..1GHz range. More is better
+7..+15dBm output
External frequency control (FM with DC coupling) or internal reference tuning input.  10Hz BW and few ppm is enough
10MHz carrier phase noise: 10dB better than attached chart
1GHz carrier phase noise:–90dBc/Hz at 10Hz, –120dBc/Hz at 1kHz offset, –150dBc/Hz at 100kHz offset
Higher carrier frequencies are OK to follow 20dB/decade degradation rule.
Not more than £2k from eBay in fully working shape (to be able to use two for cross-check tests)

[Ice-Tea]

I really wonder how good the R&S SMHU is. Pretty sure it was set up to compete or best HP contemporaries but I can't find a comprehesive datasheet on it.

Perhaps you might be interested in a test run if I manage to bring her back?

[Gerhard_dk4xp]

I have it on paper in German language. You force me to explore how my scanner works.   

[Ice-Tea]

Hah! Uh... you'll thank me one day?

If it's any consolation if it's any good you'll force me to put time in repairing it 

[Gerhard_dk4xp]

I have put it on  ftp://bama.edebris.com/uploads/           smhu.pdf  21.3 MB

( not yet approved for download by admin. )

Gerhard

[jpb]

The low-noise option of the anapico-apuln40 seems to meet your spec at 1GHz but it is POA so I suspect it is not in budget:
http://www.instruments4engineers.com/anapico-apuln40-100khz-40ghz-ultra-low-phase-noise-portable-signal-generator
SSB Phase noise at 1 GHz
10 Hz offset Option LN: -100 dBc/Hz
1 kHz offset -130 dBc/Hz
100 kHz offset -150 dBc/Hz

[KE5FX]

The low-noise option of the anapico-apuln40 seems to meet your spec at 1GHz but it is POA so I suspect it is not in budget:
http://www.instruments4engineers.com/anapico-apuln40-100khz-40ghz-ultra-low-phase-noise-portable-signal-generator
SSB Phase noise at 1 GHz
10 Hz offset Option LN: -100 dBc/Hz

That's pretty awesome.  A very good 100 MHz OCXO can do about -110 dBc/Hz at 10 Hz, which turns into -90 at 1 GHz if the signal never goes through anything else with its own 1/f^n contribution.  It is hard as heck to improve the PN at those offsets once it has been introduced into the signal path.  So they are either starting with a very good 5 or 10 MHz OCXO and multiplying it, which is expensive compared to a traditional VHF OCXO, or else using a cancellation loop that's much better than anything in the HP or Fluke lineup.

[Ice-Tea]

I have put it on  ftp://bama.edebris.com/uploads/           smhu.pdf  21.3 MB

( not yet approved for download by admin. )

Gerhard

Really appreciate you taking the time for that. Can't download it yet, though..

[Leo Bodnar]

It does look like LN version adds an additional loop below 200Hz.

SSB Phase noise at 1 GHz
10 Hz offset Option LN: -100 dBc/Hz

That's pretty awesome.  A very good 100 MHz OCXO can do about -110 dBc/Hz at 10 Hz, which turns into -90 at 1 GHz if the signal never goes through anything else with its own 1/f^n contribution.  It is hard as heck to improve the PN at those offsets once it has been introduced into the signal path.  So they are either starting with a very good 5 or 10 MHz OCXO and multiplying it, which is expensive compared to a traditional VHF OCXO, or else using a cancellation loop that's much better than anything in the HP or Fluke lineup.

[Ice-Tea]

R&S SMHU datasheet:

http://www.ftrnf.com/EEVBLOG/smhu.pdf

Not shabby, not what Leo is looking for either

Thanks for sharing Gerhard!

[Leo Bodnar]

Is there a trick for downloading manuals from this site?
Nothing in the /upload and /uploads is downloadable.
Leo

I have put it on  ftp://bama.edebris.com/uploads/           smhu.pdf  21.3 MB

( not yet approved for download by admin. )

Gerhard

[Leo Bodnar]

Here is Marconi/Aeroflex/IFR 2041 PN plot for 10MHz +13dBm output with internal reference
Leo

[analogRF]

wonder why nobody looked at Aeroflex SGD/SGA?
they are hard to find though...or maybe Leo actually bought one of them that were sold recently at a great price (6GHz version)? 

[rf-messkopf]

Rohde & Schwarz SMA100B with Option B711. Pretty impressive, but far from your price target.

[analogRF]

Rohde & Schwarz SMA100B with Option B711. Pretty impressive, but far from your price target.

yes, that's amazing