This is all a bit outside of my knowledge base, however... Some years back, I looked into using the Si5351A for various non-radio applications. In my research, I encountered a blog/paper showing the mathematical proofs that the Si5351 could not attain certain radio frequencies using the stock 25MHz reference oscillator that is included on most (all?) Si5351A breakout modules. This is due to the same fixed means of mathematics that you explain. Further, the author proved that the math did work at 27MHz. I would think that the typical cautions regarding fake chips would hold true for the 5351.. The assumption being that the AdaFruit module would have OEM parts. However, it comes with a 25MHz oscillator. Just some off-beat thoughts... Cheers.
Oh I didn't know one can overclock the PLL. Good to know. What I did in the past was underclocking it to 177 Mhz in order to get low frequencies (1.8 MHz and below) with a 90 deg phase shift. The reason why the MultiSynth divider in your case can't divide by values less than 8 is that it works only in so called integer mode and the CLKx_PHOFF register (the one used to get 90 degree phase shift) doesn't work in the integer mode. The details can be found in "AN619 Manually Generating an Si5351 Register Map for 10-MSOP and 20-QFN Devices".
Sorry for asking technical questions WAY after your experiment was completed - but I re-watched this video and continue to have questions :) I'm surprised by the lack of symmetry between the I and Q channels in your mixer. Did you happen to look at the inputs to the mixer, to verify that your input splitter was behaving reasonably as you approached 100MHz? I suppose it's entirely possible that the mux silicon is significantly slower with some of its switches versus others, but the working channel works so well (relatively speaking, of course!) that I'm curious. I would have expected a more consistent failure between both channels, with mixer losses increasing as frequency increased. Also, it might be interesting to look at third-harmonic behavior (using a 50MHz SI5351 input to receive 150MHz) - this would result in higher mixer losses, but should switch cleanly as in your 40-60MHz fundamental tests.
I think the problems were mainly related to the FST3253 simply not being able to switch well at that high a frequency. As a few commenters mentioned I should really put a downconverter in front to get the frequency down to something the switch can handle better
I have wired my detector board so that it connects directly to the SI5351 module via SMA. Perhaps this will improve the situation a little. I think I need to check my version.
It seems the speed limit of analog switch is really a headache when you try to down mix the signal above 100Mhz. I did the same test as you did. It looks the analog switch IC has very limited turn on / turn off speed and doesn't function properly above 75Mhz~85Mhz (I tested a few samples.) I also tested a faster, low voltage (3V) SN74CBTLV3253, which is 10% better in switching time according to datasheet. However, it still won't get me cover the full FM band and airband radio. So it seems to me a pure tayloe detector is not feasible to operate through all Airband. A downconverter is needed. 2 questions: 1. Do you know any alternative (much faster) analog switch IC ou would like to recommend? 2. The SA602/612 gilbert cell mixer IC has quite low IP3. I wondered if you can happen to know (and have tested) some mixer or downconversion IC with better IP3 that can downconvert 100Mhz+ signal to around 10Mhz for easy processing of a tayloe mixer?
I had a look around for a faster switch and I couldn't find anything. Now that I've done my test I am going to have another look around. A few people have mentioned a downconverter. Looking at the SA602 datasheet www.nxp.com/docs/en/data-sheet/SA602A.pdf It says that it can go up to 500MHz input frequency and 200MHz on the oscillator so thats a possibility. Thank you for the comment Sullivan!
@@na5y Thanks a lot for the quick reply! SA602/612 ICs have a high working frequency range, but their IP3's are not ideal. This means that when there is a strong nearby signal tower, the nonlinearity can result in unwanted intermodulation and can weaken the down-mixed weak signal.
It sure looks like it might be an impedance difference between the two channels and it looks like it's the yellow trace is the bad one. just my observation. on another note what if you were to install a circuit, an IF stage of sorts/frequency divider, and bring the frequency into the range of the SI5351? essentially a transverter.
I was wondering about that amplitude difference but it seemed to increase as I went up in LO/RF frequency so I assumed it was more related to the switch, Interesting behavior though. A few people have suggested a downconverter - I may try that, perhaps with an SA602 as was suggested earlier. Thank you for the comment Curt!
@@na5y I wonder if we could sweep those two inputs with a spectrum analyzer? and maybe see some information there? or even the output? it could be caused on either side really. I can't remember the circuit right off-hand just thinking off the cuff. lol. Thanks for the videos I don't know what I would do without you. these types of videos are dwindling away.
@@na5y I didn't realize you were just going straight to the scope inputs with the SI5351 at first. so yes the scope inputs there is no load unless you have one in your scope. I think you should try just the SI5351 with a dummy load to the scope if you change the load value you should see a change in amplitude and I think that is what you are seeing at the detector. then yes the hard part the inputs to the detector. I could see at higher frequencies the input impedance could change maybe the traces are not wide enough or have different lengths. but I would try just the SI5351 first to see what load you need to see for it to be correct. also, your whole problem could just be the measurement and everything is actually ok. I've had that happen too many times. lol
@@na5y The other thing I noticed when hooked up to the populated board, you are using test leads and not your scope probe. the scope probe is a very high impedance which will not interfere with your measurement as much. the test leads don't have the megohm resistor and capacitor as the probe does. If I'm wrong about how I see it I'm sorry it is difficult for me to see exactly how it's connected.
Thank you for showing this it keeps my ancient brain exercised.
Thank you Chris!
Thanks 👍
This is all a bit outside of my knowledge base, however... Some years back, I looked into using the Si5351A for various non-radio applications. In my research, I encountered a blog/paper showing the mathematical proofs that the Si5351 could not attain certain radio frequencies using the stock 25MHz reference oscillator that is included on most (all?) Si5351A breakout modules. This is due to the same fixed means of mathematics that you explain. Further, the author proved that the math did work at 27MHz. I would think that the typical cautions regarding fake chips would hold true for the 5351.. The assumption being that the AdaFruit module would have OEM parts. However, it comes with a 25MHz oscillator. Just some off-beat thoughts... Cheers.
That is why there is reason to counter intuition.
Oh I didn't know one can overclock the PLL. Good to know. What I did in the past was underclocking it to 177 Mhz in order to get low frequencies (1.8 MHz and below) with a 90 deg phase shift. The reason why the MultiSynth divider in your case can't divide by values less than 8 is that it works only in so called integer mode and the CLKx_PHOFF register (the one used to get 90 degree phase shift) doesn't work in the integer mode. The details can be found in "AN619 Manually Generating an Si5351 Register Map for 10-MSOP and 20-QFN Devices".
Thanks for the link Alex - I was looking for a good reference source.
Wait? Arduino C/C++ allows spaces in identifiers? wtf. "SI5351 MULTISYNTH MAX FREQ" How is that possible ? Am I missing something here ?
Its just a video artifact - there are underscores there. Never noticed it before
@@na5y Oh, ok. Lol. That had me scratching my head for sure!
Sorry for asking technical questions WAY after your experiment was completed - but I re-watched this video and continue to have questions :) I'm surprised by the lack of symmetry between the I and Q channels in your mixer. Did you happen to look at the inputs to the mixer, to verify that your input splitter was behaving reasonably as you approached 100MHz? I suppose it's entirely possible that the mux silicon is significantly slower with some of its switches versus others, but the working channel works so well (relatively speaking, of course!) that I'm curious. I would have expected a more consistent failure between both channels, with mixer losses increasing as frequency increased.
Also, it might be interesting to look at third-harmonic behavior (using a 50MHz SI5351 input to receive 150MHz) - this would result in higher mixer losses, but should switch cleanly as in your 40-60MHz fundamental tests.
I think the problems were mainly related to the FST3253 simply not being able to switch well at that high a frequency. As a few commenters mentioned I should really put a downconverter in front to get the frequency down to something the switch can handle better
These make better thermometers than oscillators.
I have heard others also report poor temperature problems with these devices. I haven't really spent any time quantifying that myself though
I have wired my detector board so that it connects directly to the SI5351 module via SMA. Perhaps this will improve the situation a little. I think I need to check my version.
Let me know if that improves the situation Oleg. Having the si5351 mounted above the board as I do can't be helping
It seems the speed limit of analog switch is really a headache when you try to down mix the signal above 100Mhz.
I did the same test as you did. It looks the analog switch IC has very limited turn on / turn off speed and doesn't function properly above 75Mhz~85Mhz (I tested a few samples.)
I also tested a faster, low voltage (3V) SN74CBTLV3253, which is 10% better in switching time according to datasheet. However, it still won't get me cover the full FM band and airband radio. So it seems to me a pure tayloe detector is not feasible to operate through all Airband. A downconverter is needed.
2 questions:
1. Do you know any alternative (much faster) analog switch IC ou would like to recommend?
2. The SA602/612 gilbert cell mixer IC has quite low IP3. I wondered if you can happen to know (and have tested) some mixer or downconversion IC with better IP3 that can downconvert 100Mhz+ signal to around 10Mhz for easy processing of a tayloe mixer?
I had a look around for a faster switch and I couldn't find anything. Now that I've done my test I am going to have another look around.
A few people have mentioned a downconverter. Looking at the SA602 datasheet
www.nxp.com/docs/en/data-sheet/SA602A.pdf
It says that it can go up to 500MHz input frequency and 200MHz on the oscillator so thats a possibility. Thank you for the comment Sullivan!
@@na5y Thanks a lot for the quick reply!
SA602/612 ICs have a high working frequency range, but their IP3's are not ideal. This means that when there is a strong nearby signal tower, the nonlinearity can result in unwanted intermodulation and can weaken the down-mixed weak signal.
It sure looks like it might be an impedance difference between the two channels and it looks like it's the yellow trace is the bad one. just my observation. on another note what if you were to install a circuit, an IF stage of sorts/frequency divider, and bring the frequency into the range of the SI5351? essentially a transverter.
I was wondering about that amplitude difference but it seemed to increase as I went up in LO/RF frequency so I assumed it was more related to the switch, Interesting behavior though. A few people have suggested a downconverter - I may try that, perhaps with an SA602 as was suggested earlier. Thank you for the comment Curt!
@@na5y I wonder if we could sweep those two inputs with a spectrum analyzer? and maybe see some information there? or even the output? it could be caused on either side really. I can't remember the circuit right off-hand just thinking off the cuff. lol. Thanks for the videos I don't know what I would do without you. these types of videos are dwindling away.
Thank you for the nice compliment! Do you mean the inputs to the scope or the audio port on the detector.
@@na5y I didn't realize you were just going straight to the scope inputs with the SI5351 at first. so yes the scope inputs there is no load unless you have one in your scope. I think you should try just the SI5351 with a dummy load to the scope if you change the load value you should see a change in amplitude and I think that is what you are seeing at the detector. then yes the hard part the inputs to the detector. I could see at higher frequencies the input impedance could change maybe the traces are not wide enough or have different lengths. but I would try just the SI5351 first to see what load you need to see for it to be correct. also, your whole problem could just be the measurement and everything is actually ok. I've had that happen too many times. lol
@@na5y The other thing I noticed when hooked up to the populated board, you are using test leads and not your scope probe. the scope probe is a very high impedance which will not interfere with your measurement as much. the test leads don't have the megohm resistor and capacitor as the probe does. If I'm wrong about how I see it I'm sorry it is difficult for me to see exactly how it's connected.