(phase?) noise around carrier

[in3otd]

Hello,
while doing some measurements on the H-L TX output I realized there seems to be some noise sidebands around the carrier; at 10 MHz the output spectrum here looks like this (TxDAC output)

I don't think this is actually the noise from the spectrum analyzer LO, since measuring a low-noise OCXO gives a cleaner spectrum (and the SA specs are better than the noise shown above):

Measuring the spectrum of the H-L at higher frequencies, gives higher noise sidebands and additional/higher spurs (here taken from TxDAC after a 15 dB buffer)

Can someone measure the TX output spectrum, with a SA or a good SDR to see if the same noise sidebands are present?

The noise could come from the Si510 (but it has much better specs) or could be the result of power supply noise on the AD9866 (did not yet try to play with the decouplings) or... a measurement error on my side...

73 de Claudio, IN3OTD / DK1CG

[Alan Hopper]

Claudio

I see similar sidebands just using full duplex to monitor. They have nagged me for a long time as they make me worry that my software is creating them but powersdr shows the same thing. Some power supplies create symmetrical tx spurs for me and can change the shape of close in sidebands. Sorry I can't put any real numbers to any of this.

73 Alan M6NNB

[John Laur]

This noise is jitter but the shape is not characteristic of pure phase
noise. It is however indicative of the jitter of a PLL with high loop
bandwidth and inadequate damping. I would assume in this case it is an
artifact of the DSPLL synthesizer in the Si510. If you would like to
confirm it you can inspect the clock's fundamental on your SA. I
searched around for similar plots of Si510 and I found little to
directly compare. There are some charts here
https://www.silabs.com/Support%20Documents/TechnicalDocs/How-to-Select-the-Right-PLL-based-Oscillator-for-Your-Timing-Application.pdf
that indicate its performance should normally be much better than that
as you say. It shows to be well damped and have a loop bandwidth of
300KHz. There could be a lot of things at play and we would need
additional measurements. I may borrow a SA and check things out if I
am able. I have several Hermes-Lites to compare.

73, John K5IT

> --
> You received this message because you are subscribed to the Google Groups
> "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to hermes-lite...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[James Ahlstrom]

Hello Claudio,

I measured the TxDAC output spectrum at 10 MHz and did not see the noise you identified.  But my dynamic range is only 70 dB, and your noise is below that level.

But I did see the noise while transmitting at 24.895 MHz.  See photo.  The noise is 60 dB down instead of your 80 dB.  But it occurs at +/-  1 MHz, the same as yours.  I used the Quisk Spot button as a source.  When reducing the level to 0.500, the output dropped by 6 dB as expected, and the noise also dropped by 6 dB as best I can judge.  I also changed to CW mode and closed the key as a source (no Quisk Spot).  The result was the same.

The most interesting thing is that I am using an 80 MHz Connor-Winfield crystal oscillator, not an Si570.  This was still mounted from my 80 MHz work, and I was about to replace it with the old Si570.

Jim

N2ADR

[in3otd]

Thanks Jim and all,
this rules out the Si510 as possible culprit; indeed measuring the Si510 output clock on the SA I see a clean spectrum, without any sidebands (just a few weak lines here and there, but those may be just supply noise).
I quickly tried to add a 470uF (electrolytic) on CLKVDD but I saw absolutely no change; while it's not the best HF decoupling I think it should have had some effect if the supply noise was the cause.

With these noise levels around the carrier, the H-L will be amongst the less clean TX, comparing with the data here. :-(

73 de Claudio, IN3OTD / DK1CG

[Glenn P]

This is an older plot thru the PA also and sees similar frequency offset 'spurs' but not in the' format' of side band noise.
My plot is wider BW and fairly narrow RBW.

glenn
vk3pe

[Glenn P]

Just took a quick measurement also at 10MHz (10.13MHz actually.)

Can confirm much the same as others although absolute levels seem to vary between us. But I also see strong signals at ~3MHz & 4.76MHz

My 10MHz out put level is not +10dBm, more like +2dBm. At 29MHz I get +10.    I must check power on all bands come to think of things to do.......

glenn
vk3pe

[Glenn P]

I dragged out my other SA, an HP8560A to see if those lower frequency spurs are 'real'. Appears they are not.

Two screen pics of screen, one is zoomed in around carrier.

H-L output is via the Wiki re construction filter, PowerSDR used, and drive at 100%. Power out is about +7dBm

Glenn
vk3pe

[ZL2APV]

Hi Glen,

Try inserting a parallel resonant circuit in series with the signal. It will of course affect the fundamental and close in sidebands but won't affect the 3 MHz signals unless they are a product from your instrument so you can eliminate that point.

Graeme
zl2apv

[ZL2APV]

Our posts crossed. I see you have an even better method at your disposal. I didn't expect you to have that luxury :-) I have also used crystals on the fundamental to increase the dynamic range of the spec an but still see (fairly) close in phase noise and sidebands.

73, Graeme

[Glenn P]

Hi Graham, I mainly use the Rigol as its small and fits in my tiny shack. Plus it has screen capture and tracking gen.   It's so easy to be fooled though by an SA though. The Rigol has a max input of +20dBm, the HP is +30dBm.
Adding a 20dB pad to the Rigol input removes those spurious signals.
The signals +- 900KHz either side of carrier remain so they are real. Phase noise is present but that is from the SA I am sure.  The HP is a nice gadgetfor it's time, but is limited by it's phase noise of course.

glenn

[John Williams]

Graeme,

It is so good to see you back on the list. Hope all is going well...

John

[ZL2APV]

Hi john,

Thanks for your inquiry, much improved and spent morning in workshop. There has been some great work gone on and I really enjoyed Steve's summary. The work on spurs and noise is outstanding and I would love to see K3 equivalent noise products from HL. I hope Jim gets his board going before skiing starts but life does not start and end in the hamshack. You and Glenn are doing so well on the Amp side of things, it is a great collaboration and everyone who adds to it is so positive. Its pretty rare in the internet to find a group where you feel helped rather than criticised by someone who disagrees with you.

73, Graeme

[Glenn P]

Hi Graeme,

I think i trust the HP a bit more, which does not show the 3MHz sigs.   So i assume they some sort of artifact of the Rigol. All other sigs appear about the same between the two instruments.

glenn

[Steve Haynal]

Hi Claudio,

That sure looks like some sort of phase noise. For those interested, there is actually a good plot of phase noise in the ad9866 datasheet, figure 77. I went through all the raw data I saved from my scope/analyzer when I was investigating the spurs for TX=24.895 MHz. I could see these sidebands, although they were not as bad as others report. With the cheaper LO in my scope, I was expecting the sidebands to be worse or lost in the noise so am a bit surprised. In fact, what I measure is at a level low enough (>85 dBc) I wouldn't have worried about it. There are possible real spurs of more concern. I did find two sets that appear better than the others: interpolation filter off, and different BeMicro/Hermes-Lite. The plots are below. Here are a few other ideas:

• The TX clock always goes through a multiplier internal to the AD9866 as seen in figure 76. This multiplier uses a PLL. Perhaps phase noise is added at this point. But there is no phase noise visible in figure 30. Fortunately, the RX can and is connected directly to the oscillator in. The reduced noise in my interpolation filter off configuratio supports this idea as maybe there is less phase noise introduced when multiplying by 1. Unfortunately, disabling the interpolation filter excludes full-duplex. We should probably consider an easier switch between half and full duplex than switching firmware, but this involves defining some FPGA pins that need to be fast as bidirectional and can be tricky to achieve timing closure.
• The phase noise is introduced from the BeMicro's power supply via one of the signal pins. The reduced noise I see with the CVA9 and alternate Hermes-Lite support this. How are you powering your Hermes-Lite, via the BeMicro? Do you have J7 open or closed?   

Regarding spurious emissions, figure 32 in the datasheet indicates that as frequency goes up SFDR goes down as we've seen.

73,

Steve

KF7O

This is what most of my measurements look like, sidebands at ~85 dBc, some spurs may be real others, especially the close in ones, are probably artifacts.

Below is with a CVA9 and different Hermes-Lite board. Sidebands now around 88 dBc.

Here is with the interpolation filter off. Sidebands now at -91 dBc. I am surprised that the 2*Fs - 5*Ftx spur still exists...

[in3otd]

Thanks Steve,
I overlooked that we have the 2x interpolation in the TX path, very likely the noise sidebands are from the internal VCO, as you mentioned. Figure 77 looks very similar to what we measured.
The noise may not appear in fig. 30 of the datasheet depending on the resolution bandwidth used, this could explain also some of the differences seen in the various measurements here (BTW, what's the equivalent RBW of your FFT plots?)
I too would have expected that the scope timebase was too noisy to measure this; maybe your last plot with TX interpolation disabled is actually showing the scope phase noise?

I am powering the H-L from the BeMicro, will try a separate power supply; the differences you see between boards may be just typical variation from the integrated VCO between different parts?

73 de Claudio, IN3OTD / DK1CG

[Joe]

Claduio,

  I attached some spectrum analyzer screenshots showing 20,12&10 meters of noise. At my QTH the noise

 gets worse at higher frequencies. I'm using the firmware with direct DAC output. All done at 2MHZ/ DIV 100K BW.

  Joe  wa9cgz

[Steve Haynal]

Hi Claudio,

I've been wondering too why there is so much variation in these sidebands. Perhaps it is variation in the VCO as you suggest. Glenn and Joe, are you using the IAMP or not?

Yes, I'd expect some phase noise form my scope timebase somewhere. I already get lots of "bonus" artifacts. :)

For those plots I was using 24 averages for 24 million points with max frequency of 500 MHz, so around 1 kHz wide FFT bins.

73,

Steve

KF7O

[Steve Haynal]

Hi Joe,

Thanks for the plots. You output always looks so clean! I'm puzzled by why there is so much difference among radios for this phase noise. Are you using the IAMP or is it disconnected and you are using the raw TxDAC output?

73,

Steve

KF7O

[Glenn P]

Steve I am using the standard output.   For testing I am using HP8640B as the signal source.

glenn

[Glenn P]

sorry, I misread the post; thought it was for me...........

vk3pe

[Joe]

Steve,

I'm using Claudio's firmware without the IAMP a 1:1 transformer with grounded center tap connected to CN3.

Version 1.21 SDK 73MHZ for test. Please disregard analyzer data shown on right side of screen shots 

these are due to software used to capture the correct data is shown in picture label.

the peak of the noise is about 900KHZ above and below the CXR it's at the limit of my analyzer and

the noise gets worse as the frequency increases for most ham bands the noise not in band.

Joe  wa9cgz