Listen to Latest SolderSmoke Podcast

Thursday, August 20, 2015

Si5351 Phase Noise? A Tale of 3 Oscillators

Si5351 at 16 MHz
 
There is still a lot of talk about the supposedly horrible phase noise of the Si5351 chip. In a recent episode of a popular (and very good!) podcast about homebrewing, the podcasters talked about this in the context of some megawatt AM shortwave broadcast stations that had oscillator phase noise problems and were wiping out large portions of the HF spectrum.   I don't think those stations were running Si5351s, but the listener was left with the impression that these handy little chips are very noisy with lots of spurs and will inevitably produce horrible dirty, spectrally impure signals. 
 
This has not been our experience.   Following Pete's lead, several of us are using the Si5351 to generate both VFO and BFO signals in our transceivers, with good results. The receivers sound very good and we have not heard complaints of "broad" or "noisy" transmitted signals.
 
I decided to dig into this a bit.  This was also an excuse for me to use the FFT and screen capture features on my Rigol 'scope.  
 
I now have THREE BITX transceivers in the shack.   My BITX17 uses a VXO at round 23 MHz (IF at 5 MHz)/  My BITX20 uses a classic LC VFO running around 3.5 MHz (IF at 11 MHz).  Finally, my BITX40 (DIGI-TIA) uses the dreaded and much reviled Si5351 running at around 16 MHz (IF at 9 MHz).   I thought that these three rigs would provide a good opportunity to test the scurrilous claims about the Si5351.   
 
As a simple first test, I put my Rigol scope in FFT mode and just put the probe at the VFO Mixer's LO input.  The screenshot above is the FFT for the Si5351.  It looks pretty clean to me.  The 'scope is looking at 15 Mhz above and below the VFO signal.    
 


VFO at 3.5 MHz
 
Next I measured the output of the BITX20 VFO at the same point (input to the VFO mixer).  (I had to change the vertical range, but the horizontal was unchanged.)  Here you can see the second harmonic (just because at this low freq it is within the freq range setting of the 'scope).  It doesn't look much different than the Si5351.   


VXO at 23 MHz
 
Finally, here is the BITX17 VXO at 23 MHz, again at the input to the VFO mixer.   It looks remarkably similar to the Si5351, don't you think? 
 
More on this to come.  The ARRL Handbook (2002) has a good discussion on phase noise. I am digging into this and hope to do some more tests.   For now, I think we should reserve judgment on the utility (for us) of the Si5351.  


Our book: "SolderSmoke -- Global Adventures in Wireless Electronics" http://soldersmoke.com/book.htm Our coffee mugs, T-Shirts, bumper stickers: http://www.cafepress.com/SolderSmoke Our Book Store: http://astore.amazon.com/contracross-20

6 comments:

  1. Hi Bill,

    Thanks for sharing the results of your data collection. Very telling and that is what is really needed --solid data and not hearsay.

    For instance did the individuals who conducted the podcast actually build a radio and test the Si5351. Have they made a comparative analysis (such as you have)? Or have they simply parroted what is on various reflectors.

    In the SS Podcast 169 we talk about this very issue of the passing along of 'the phase noise' is bad comment without having a sound basis for that statement.

    Two points came out of that Podcast: 1) Actually test the device and then make your statements and 2) It is a relative matter -- if the phase noise of the Si5351 is 1 DB worse than that of a Si570 then you can accurately claim it is worse; BUT will your ears tell the difference? [For those who want to pounce --it is not 1 DB but the matter of relativity is the point.]

    Is the Si5351 a laboratory grade device suitable for deep space weak signal tracking -- probably not. BUT is it good enough for use in homebrew SSB and CW transceivers and for that I would give a resounding Yes!

    Those individuals conducting the podcast should be asked "Have you personally ever used a Si5351 and where is your data?"

    I currently have four SSB transceivers with an Si5351 as the principal frequency generating device and I do not find phase noise as an issue. I might add that the Spectral Purity Police who lurk on 40M and all run Flex 6000 Series Radios have never given me a report "well your radio sounds like it has phase noise" You and I both know if it did sound "phase noisy" they would let you know. Admittedly the phase noise is more of an issue on receive but for the HF radios I have The Si5351 works very well. Get a grip this is an $8 board and now $1.16 cents for the device itself --that is a lot of bang for the buck!

    73's
    Pete N6QW

    ReplyDelete
  2. Do you plan on revealing the name of this podcast? Always looking for more podcasts on electronics and homebrewing. (even if they are flawed)

    ReplyDelete
  3. Hi Bill.....There are folks that know much more about this than I but I don't think that the scope pix you show actually speak to the phase noise problem. The shape you see is more a function of the FFT processing than anything. To depict phase noise you need a spectrum analyzer type display where the vertical axis is logrithmic and shows signals down to maybe 100 dB below the signal power. (In a straight voltage type display you are lucky to see sigs down 30 or so dB) Doing 100dB at close frequency separations is not an easy thing and I don't thing very many amateurs can do it. The equipment is in the professional class ($$$). As to on-the-air experience, it is one thing to use the Si5351 in a qrp rig and another to do it in a 500W or legal limit station. The phase noise in the high power rig will be proportionally higher on an absolute basis and thus potentially more troublesome. I have no experience with the circuit in question but given what I think it is, it is hard to believe that its phase noise is in any way comparable to say a crystal oscillator or even an HP 8640 sig gen. Cheers and 73 Dan K9EA

    ReplyDelete
  4. There are two sides to this debate; the factual/science side and the practical side.

    On the factual side, Bill's test is an invalid one and a good example of why FFT functions on a scope, or SDR receivers, etc are not Spectrum Analyzers. In order to see the phase noise of the output of a si5351, you need to have the special high-speed processing capabilities of a Spectrum Analyzer. I have an Agilent E4402B with the Phase Noise personality and will post a measurement to my Blog as soon as I can make the time to do it and am not at work.

    The practical side of the debate is Bill and Pete's experience using rigs built with si5351's. You might call this side the "proof in the pudding" part because they are not imagining what others (some very picky others) tell them as they get reports about their rigs from other hams. I'm an EE and I learned a long time ago not to argue with practical, proven results. This doesn't mean there are other sides or dimensions to the discussion, but that people are simply repeating what others (maybe many others) have told them. These are also a set of facts when you think about it. Analog circuits, especially those used throughout communications equipment, are very tolerant of slop and what appear to be non-ideal operating conditions. This is one reason we use Monte Carlo simulations to ensure that the component tolerances do not create instability in circuits less tolerant of change; especially temperature drift.

    I also agree and would echo K9EA's comments of QRO operation versus QRP. The harmonic distortion created by SSB Phase Noise is enough that you could in fact be out of the bounds of legal station operation and cause unintentional interference to other services; a big no-no with the FCC.

    Russ
    K0WFS

    ReplyDelete
  5. As promised:

    http://k0wfs.com/2015/08/21/si5351-phase-noise-and-thd-tests-using-an-agilent-e4402b-spectrum-analyzer/

    Russ

    ReplyDelete
    Replies
    1. I'm sorry, but what you are measuring is the windowing function of the sweeping frequency FFT routine that Agilent uses in the spectrum analyzers. You have not run the Phase noise routine.
      This measurement is in effect just the same as the one Bill did run.

      In my opinion, the THD measurements is flawed, its a square wave, and mr. Fourier tells us that it must have some odd harmonics.

      The Phase noise of the Si5351 is around -130dBc/Hz at 10KHz. This is quite decent, If compared to a Hartley or collpits you would see little or no difference. Some of my measurements of published free running oscilators show phase noise in the -110dBc/Hz range!

      Phase noise is the hardest of the noises both to measure and to explain. Its effect in a receiver is for most part not a reduction in MDS ("sensitivity") but in IMD. If you have a receiver with no gain before the bandwith defining filters then you may see a reduction in IMD preformance. But since we still are at a place with SA612 mixers and post mixer amps in receivers, it does not matter.
      In a transmitter, audio noise injected with the microphone will make more probems than the phase noise.

      As for the Si5351, here are good measurements done by KE5FX, probably the ham with most knowledge of phase noise outside of NIST:
      http://nt7s.com/2014/11/si5351a-investigations-part-7/

      73 de Thomas LA3PNA.

      Delete

Designer: Douglas Bowman | Dimodifikasi oleh Abdul Munir Original Posting Rounders 3 Column