|
Hi, ? I hacked the standalone mode by adding another trace showing the combined gain of the LNA and PGA during a sweep.? If the AGC is enabled then the gain can be read in Register 0x69 bits 4-0 The vertical scale for the gain is 5dB/division.? T he yellow trace is the 0-255 RSSI value from the SI4432,? I was too lazy to change the scale at the side for the quick hack.This is how the gain varies with agc enabled (x69 = x60).? The dBm value at the peak includes the correction for the filter/mixer loss, and yes I know it should say dBm not dB - fixed after the photos were taken.  Interesting how the gain drops early, and the spike at each change of gain.? Also interesting how the gain starts increasing before we get to the main signal spike.? Presumably the gain algorithm is working on the signal strength before the FIR filter? Fixing the gain at the value the ALC chose, 37dB we get this:  go up to 43dB and we see some spurious signals and the peak RSSI drops - ADC overload?:  Drop down to 25dB and the peak RSSI stays at -13dBm but the minimum detectable signal level increases:  So it looks like the AGC does a great job of setting the best gain level, the RSSI value is compensated for the gain change, but there are unwelcome spikes at the changes in gain.? Perhaps some sort of on-demand "Optimise gain" button is needed where the gain at the peak (or perhaps the minimum through the sweep) is recorded and then used to keep a fixed gain throughout the sweep. I tried detecting a change in gain then delaying and retaking the reading, but that did not offer an improvement.? Perhaps there is some instability and the gain is swapping between different levels? 73 M0WID
|
Interesting, This is what I get with 0x69 set to 0x60 and a -5dBm and -30dBm signals and a 20dBm/div scale. As you can see my AGC acts much earlier. The AGC is multiplied by 4 before display   Can you set the RBW a bit wider and see what that does? -- HBTE Files section:? /g/HBTE/filesErik, PD0EK
|
Interesting indeed. Some more pictures.?? First with x69 = 60, agc enabled  And with the gain fixed:  Perhaps some explanation - the response of? my SAW filters as measured with a NanoVNA:  I tried with and without matching for the filters.? With matching (pink trace) the response is much flatter but has the feature around 438MHz.? I figured, perhaps wrongly, that the narrow response at the IF without the matching (brown trace) would be an advantage, and the other peaks far enough away to not be an issue, and no worse than having a flat passband.? My filter is on a double sided home etched PCB.? With the matching components fitted having enough vias in place proved important. This filter response explains I think the odd agc action, but not the spurious signals present when the agc changes the gain. Dave
|
You guys are a lot smarter than me, but, I have been seeing something that I can't understand, but maybe this is the AGC action that Dave is seeing. When looking at my AD9851 (DDS-60 kit) I always have a funny? response on the low frequency side only. My 9851 is generating SSB?? It is always on the low side and it tracks with frequency changes.? Any ideas? Gary WB6OGD 
|
Looks like the image created inside the SI4432 due to insufficient selectivity of the SAW BPF. What SAW filters are you using? I did a post some time ago on this issue -- HBTE Files section:? /g/HBTE/filesErik, PD0EK
|
Ahh ha! ?Yes I am using a temporary sub-standard one SAW filter. I will be improving that, thank you Erik. 73, Gary WB6OGD?
toggle quoted message
Show quoted text
On Feb 12, 2020, at 10:08 PM, erik@... wrote:
?Looks like the image created inside the SI4432 due to insufficient selectivity of the SAW BPF. What SAW filters are you using? I did a post some time ago on this issue -- HBTE Files section:? /g/HBTE/filesErik, PD0EK
|
I have been away, and will be the rest of the week, so no time to play, however last night I did add back in the matching components for the SAW filter.? It eventually dawned on me that a highly selective SAW filter at the IF was not ideal when doing a pass with wider RBW.? Apologies for being stupid. This is the trace now - still the artifacts from agc changes.? Is this unique to my module, perhaps I have a dodgy one?? The artifacts do not appear, or change position, with different signal levels, and do not appear if the gain value is fixed.  I was wondering whether to try a fast scan at max RBW to identify the optimum gain at different points of the sweep and then fix the gains at possibly different values through the? actual sweep hopefully before the ADC gets overloaded and produces the artifacts (if that is what is happening).? Is this also a stupid idea? 73 Dave
|
I don't yet have working hardware, so can't verify your results.?
I assume the blue line represents both PGA and LNA gain.
And that the single step down and back up near the middle represents the LNA gain.
The PGA gain initially staircases down in 8 steps over a period of roughly half
a horizontal division as the four bits of PGA gain are reduced from max gain of 24dB
to min gain of 0dB.? (From page 50 of SiLabs AN440)
Your artifact is also roughly half a horizontal division, somewhat delayed from that
reduction in PGA gain.? I think it's just that the PGA gain register and the RSSI register
are not in any way synchronized.? A quirk of the Si4432 design.
I think the solution is to leave the AGC off.
Each time the RSSI register is read, check the value to see if it is getting close to
either the upper or lower limits.? If it is, then adjust the PGA and/or LNA gain manually,
wait for a millisecond or two, and then re-read the RSSI register
before advancing to the next frequency of the scan.
A couple curiosities I don't really have an answer for:
1)? Why is the 20dB of LNA gain shifting showing up as only about a third
of the scale of the 24dB of PGA gain shifting in the blue trace???
2)? If PGA gain changes are not synchronized with the RSSI register,
how is it that changing the LNA gain bit is synchronized with the RSSI register?
The LNA is positioned before the PGA in the Si4432 receiver architecture,
the RSSI is way back toward the end of the receive process.
Jerry, KE7ER?
toggle quoted message
Show quoted text
On Tue, Feb 18, 2020 at 02:15 PM, m0wid wrote:
I have been away, and will be the rest of the week, so no time to play, however last night I did add back in the matching components for the SAW filter.? It eventually dawned on me that a highly selective SAW filter at the IF was not ideal when doing a pass with wider RBW.? Apologies for being stupid.
This is the trace now - still the artifacts from agc changes.? Is this unique to my module, perhaps I have a dodgy one?? The artifacts do not appear, or change position, with different signal levels, and do not appear if the gain value is fixed.
|
Dave,
One other thing, that sgin bit in register 0x69.
Not totally obvious to me exactly what the devil it does.
If you have it set, I suggest you clear it.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Tue, Feb 18, 2020 at 03:15 PM, Jerry Gaffke wrote:
I don't yet have working hardware, so can't verify your results.?
I assume the blue line represents both PGA and LNA gain.
And that the single step down and back up near the middle represents the LNA gain.
The PGA gain initially staircases down in 8 steps over a period of roughly half
a horizontal division as the four bits of PGA gain are reduced from max gain of 24dB
to min gain of 0dB.? (From page 50 of SiLabs AN440)
Your artifact is also roughly half a horizontal division, somewhat delayed from that
reduction in PGA gain.? I think it's just that the PGA gain register and the RSSI register
are not in any way synchronized.? A quirk of the Si4432 design.
I think the solution is to leave the AGC off.
Each time the RSSI register is read, check the value to see if it is getting close to
either the upper or lower limits.? If it is, then adjust the PGA and/or LNA gain manually,
wait for a millisecond or two, and then re-read the RSSI register
before advancing to the next frequency of the scan.
A couple curiosities I don't really have an answer for:
1)? Why is the 20dB of LNA gain shifting showing up as only about a third
of the scale of the 24dB of PGA gain shifting in the blue trace???
2)? If PGA gain changes are not synchronized with the RSSI register,
how is it that changing the LNA gain bit is synchronized with the RSSI register?
The LNA is positioned before the PGA in the Si4432 receiver architecture,
the RSSI is way back toward the end of the receive process.
Jerry, KE7ER?
Hide quoted text
?
On Tue, Feb 18, 2020 at 02:15 PM, m0wid wrote:
I have been away, and will be the rest of the week, so no time to play, however last night I did add back in the matching components for the SAW filter.? It eventually dawned on me that a highly selective SAW filter at the IF was not ideal when doing a pass with wider RBW.? Apologies for being stupid.
This is the trace now - still the artifacts from agc changes.? Is this unique to my module, perhaps I have a dodgy one?? The artifacts do not appear, or change position, with different signal levels, and do not appear if the gain value is fixed.
|
So perhaps this is a naive perspective, but I would think that having AGC enabled in a spectrum analyzer would always be sub-optimal during measurement sweeps - because the changing gain/attenuation adds yet more variables to the calibration between samples within a single sweep.? Having the ability to do a "pre-sweep" of sorts to define the maximum/optimum gain settings for a given input could be valuable (along the lines of an "Auto" button for the analyzer, to pre-set gain balance), but throughout a single sweep I would expect the most consistent results to come from a fixed gain setting.
Obviously a fixed gain setting throughout the sweep may result in a lower overall dynamic range, but that seems like a tradeoff that makes sense.? Of course, all of my spectrum analyzer experience is with older units that are happy to have 70-80dB of dynamic range under ideal conditions!
Josh, KB8NYP
On Tue, Feb 18, 2020 at 6:15 PM Jerry Gaffke via Groups.Io <jgaffke= [email protected]> wrote: I don't yet have working hardware, so can't verify your results.?
I assume the blue line represents both PGA and LNA gain.
And that the single step down and back up near the middle represents the LNA gain.
The PGA gain initially staircases down in 8 steps over a period of roughly half
a horizontal division as the four bits of PGA gain are reduced from max gain of 24dB
to min gain of 0dB.? (From page 50 of SiLabs AN440)
Your artifact is also roughly half a horizontal division, somewhat delayed from that
reduction in PGA gain.? I think it's just that the PGA gain register and the RSSI register
are not in any way synchronized.? A quirk of the Si4432 design.
I think the solution is to leave the AGC off.
Each time the RSSI register is read, check the value to see if it is getting close to
either the upper or lower limits.? If it is, then adjust the PGA and/or LNA gain manually,
wait for a millisecond or two, and then re-read the RSSI register
before advancing to the next frequency of the scan.
A couple curiosities I don't really have an answer for:
1)? Why is the 20dB of LNA gain shifting showing up as only about a third
of the scale of the 24dB of PGA gain shifting in the blue trace???
2)? If PGA gain changes are not synchronized with the RSSI register,
how is it that changing the LNA gain bit is synchronized with the RSSI register?
The LNA is positioned before the PGA in the Si4432 receiver architecture,
the RSSI is way back toward the end of the receive process.
Jerry, KE7ER?
On Tue, Feb 18, 2020 at 02:15 PM, m0wid wrote:
I have been away, and will be the rest of the week, so no time to play, however last night I did add back in the matching components for the SAW filter.? It eventually dawned on me that a highly selective SAW filter at the IF was not ideal when doing a pass with wider RBW.? Apologies for being stupid.
This is the trace now - still the artifacts from agc changes.? Is this unique to my module, perhaps I have a dodgy one?? The artifacts do not appear, or change position, with different signal levels, and do not appear if the gain value is fixed.
|
Not naive.
Sometimes measurement accuracy can be more important than dynamic range.
But not always.
AGC could be a menu option if it works well enough.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Tue, Feb 18, 2020 at 06:38 PM, Joshua Blanton wrote:
So perhaps this is a naive perspective, but I would think that having AGC enabled in a spectrum analyzer would always be sub-optimal during measurement sweeps - because the changing gain/attenuation adds yet more variables to the calibration between samples within a single sweep.? Having the ability to do a "pre-sweep" of sorts to define the maximum/optimum gain settings for a given input could be valuable (along the lines of an "Auto" button for the analyzer, to pre-set gain balance), but throughout a single sweep I would expect the most consistent results to come from a fixed gain setting.
?
Obviously a fixed gain setting throughout the sweep may result in a lower overall dynamic range, but that seems like a tradeoff that makes sense.? Of course, all of my spectrum analyzer experience is with older units that are happy to have 70-80dB of dynamic range under ideal conditions!
?
|
