15037	2018-09-05 12:04:08	kb1gmx	Si5351 crosstalk between outputs
Message	Date	From	Subject
Did some testing today as I was going to put a 5351 in an Atlas 210 and use software to switch BFO and VFO rather than the existing FET switch.... Hard stop on that. I"m gong back to the crystal oscillator. Reason THe SI5351 if you run two or more VFOs the crosstalk is about -58DB. Again differently, if you use a 5351 for two or more oscillators the signal from the other two will appear as a signal -58db down from the oscillator in use. I first saw this in uBitx. Figured it was layout issue but testing it with several obtainable 5351 boards I got the same result hooked to nothing. Test is simple pick three frequencies that are distinct and non-harmonic look at any one with SA. note the spectra clean and has harmonics (its a square wave so ok). Now load any of the reamaining two oscillators and bingo you now have a spur at that frequency about 58DB down. Worse for all 3. Also I tried a 100ohm load, slightly better (2db) but threre. At 240ohms its nearly -70db down the measure output. This means for best crosstalk one must very lightly load the output. That makes its output about as good as a 9851, and people moan about the SFDR being 60 or 70db. Allison
15038	2018-09-05 12:22:39	Dana Myers	Re: Si5351 crosstalk between outputs

15039	2018-09-05 12:52:35	Tayloe, Dan (Noki...	Re: Si5351 crosstalk between outputs
Since a 5351 only cost $1 each, why not just use two? Dan
15040	2018-09-05 13:06:59	Dana Myers	Re: Si5351 crosstalk between outputs

15041	2018-09-05 13:18:02	Henning Weddig	Re: Si5351 crosstalk between outputs
Allison, Your observation is what IO always feared i.e. the internal crosstalk between the oscillator outputs. In general these chips are designed for some clocks in digital circuitry, so crosstalk in the range You measrued will probably not hurt the digital circuits. BUT for an "analog" radio these spurs do matter! So the conclusion is: use separate SI5351´s at the place where it is needed. This will ease the routing of criticla analog lines (from the Raduino to the main board). I n addtiion I really dislike the idea to power the SI5351 from a "distroted" supply voltage from the Nano. Each SI5351 should have its own filtered 3.3 V supply, and even use these modern low noise but expensive LDDO´s from LT.. eg. LT3045. Best would then be a common 25 MHz oscillator driving all SI5351 in order not the have problems with drift of the different clocks, but this can caus ohter problems in crosstalk between the SI5351´s. There is a lot of design work to be done!! Henning WEddig DK5LV Am 05.09.2018 um 21:02 schrieb kb1gmx@arrl.net [emrfd]: Did some testing today as I was going to put a 5351 in an Atlas 210 and use software to switch BFO and VFO rather than the existing FET switch.... Hard stop on that. I"m gong back to the crystal oscillator. Reason THe SI5351 if you run two or more VFOs the crosstalk is about -58DB. Again differently, if you use a 5351 for two or more oscillators the signal from the other two will appear as a signal -58db down from the oscillator in use. I first saw this in uBitx. Figured it was layout issue but testing it with several obtainable 5351 boards I got the same result hooked to nothing. Test is simple pick t hree frequencies that are distinct and non-harmonic look at any one with SA. note the spectra clean and has harmonics (its a square wave so ok). Now load any of the reamaining two oscillators and bingo you now have a spur at that frequency about 58DB down. Worse for all 3. Also I tried a 100ohm load, slightly better (2db) but threre. At 240ohms its nearly -70db down the measure output. This means for best crosstalk one must very lightly load the output. That makes its output about as good as a 9851, and people moan about the SFDR being 60 or 70db. Allison
15042	2018-09-05 16:18:00	Clint	Re: Si5351 crosstalk between outputs
In making mods to an Atlas 210X transceiver, everything is relative. While a given mod may not look good on paper, compared with industry standards, the mod may make a big improvement in the operation of the 210X. Here is an example: The 2^nd and 3^rd harmonics from the factory VFO and the Carrier Oscillator are very high: VFO 2^nd harmonic is down 8 db 3^rd harmonic is down 22 db Carrier Oscillator 2^nd harmonic is down 16 db 3^rd harmonic is down 17 db It is assumed that a low pass filter will be used on the output of the Si5351 in order to get a sine wave drive. Driving the Atlas first mixer with a square wave is a controversial subject. So having spurious signals down 58 db might not create a problem. 73 Clint W7KEC
15043	2018-09-05 16:47:43	Clint	Re: Si5351 crosstalk between outputs
Here is what I measured on the output of the FET switches on the PC-300 board in a factory stock Atlas 210X: Pin 4 5645 Khz carrier oscillator down 68 db from VFO signal Pin 8 8525 Khz VFO signal (20M) down 52 db from Carrier oscillator signal 73 Clint W7KEC
15044	2018-09-05 21:49:30	John Maxwell	Re: Si5351 crosstalk between outputs

15045	2018-09-05 21:50:39	augustinetez	Re: Si5351 crosstalk between outputs
I have always said these things are only good for one output at a time because of the non-isolated nature of the (three output chip's) output structure. The 8 or more output versions are better because their outputs run on individual Vcc rails and you can arrange frequency selective filtering on each rail. Terry VK5TM
15046	2018-09-05 23:40:40	Dana Myers	Re: Si5351 crosstalk between outputs

15049	2018-09-06 09:12:10	kb1gmx	Re: Si5351 crosstalk between outputs
All, I know of prior work but it was from my perspective incomplete. I also did the measurement with higher than 50 ohm load and got remarkably better result. Hint: drive the load using a 74LVT04 or one of the friends of that part or any cmos buffer. The output is loaded with CMOS gate and the CMOS inverter can be loaded heavier. The problem is likely greater for bitx/ubitx with multiple clocks and no buffering. Yes Crystal is nice until you want to use Arduino to reverse the side band and then must use a relay and driver. Much of this is not new to me. Been at it for many decades. I just had time enough now to document that experiment. I was not surprised at the result but I was shocked at the number of designs using it without consideration. As to diode (DBM) mixers and square waves. If you feed it a sinewave of sufficient amplitude to work well and look at the signals internally they end up square as the diodes have fairly hard turn on. That and most should understand DBMs are a conundrum and headaches in three ports. Most of what I've read for the Atlas is guess, speculation, did this it helped. That and filters made with molded mud for caps and worse. Fix that with decent caps many issues just go away. Allison
15050	2018-09-06 09:37:08	jim	Re: Si5351 crosstalk between outputs

15051	2018-09-06 13:08:18	Dana Myers	Re: Si5351 crosstalk between outputs

15052	2018-09-08 05:18:30	kb1gmx	Re: Si5351 crosstalk between outputs
Any load that causes increased current flow on the internal power buses of the chip impact cross talk. So 80 ohms is better but only a tiny bit and 240 ohms is about 8db better and using a cmos input inverter or high input impedance buffer is far better (>20DB). How do i know? Rather than "I read it somewhere" I went and measured it. Why Because every thing I read was mostly vague hand waving. I found nothing to say what load impedance would be better. Even those that mention it didn't ask how does one do better? Allison
15056	2018-09-08 11:30:46	AD7ZU	Re: Si5351 crosstalk between outputs
Thanks Allison, good report, when in doubt, measure! I have a few Si5351C 8 output devices that (were or are) destined for a receiver / transmitter project 2 rx, 2 tx, 1 ADC. The 8 output version has 4 separate output voltage supply pins, I suspect this arrangement may minimize crosstalk if the output supply pins were separately and sufficiently bypassed? Though per the Si datasheet the purpose of the separate output supply lines is to drive different outputs to different levels On another note the Si5351C part uses an external clock input.. the thought here is that CLKIN could be driven with a decent quality TCXO reducing temp drift and possibly achieving better phase noise?. The CLKIN (external osc input) is divided from up to 100mhz to 30 Mhz max prior to the stage 1 synth block. I have yet to hook these up much less made any measurements. Randy AD7ZU
15057	2018-09-08 13:13:27	iq_rx	Re: Si5351 crosstalk between outputs
Cool, thanks--a much needed set of measurements. I'd like to see some more, as different multi-output ICs will have different internal bus structures, and I'd expect to see significant differences in cross-talk with different supply bypassing and layout of the ground and supply lines. Best Regards, Rick KK7B
15058	2018-09-08 14:05:08	jim	Re: Si5351 crosstalk between outputs
I felt I needed to start at the "back end" of the pa to get some validity as to "how much is too much" proceeding backwards towards the clk 2 mixer ...Have some photos and remarks soon Jim PS I agree on measure ..Was not particularly interested in the Si confabulation at that point, just noted it in passing.
15059	2018-09-08 15:00:17	Dana Myers	Re: Si5351 crosstalk between outputs

15067	2018-09-10 12:26:08	kb1gmx	Re: Si5351 crosstalk between outputs
>>>I have a few Si5351C 8 output devices that (were or are) destined for a receiver / transmitter project 2 rx, 2 tx, 1 ADC. The 8 output version has 4 separate output voltage supply pins, I suspect this arrangement may minimize crosstalk if the output supply pins were separately and sufficiently bypassed? <<< Measure! The datasheet may give clues but likely you can get crosstalk on the ground system on chip and on the board as well. Always watch the total chip current. I've seen specs in the past where currinto the chip had a max that was higher than the chips other pin (current out). Remember the chip may have to source current, sink it, or both! >> Though per the Si datasheet the purpose of the separate output supply lines is to drive different outputs to different levels<< Multiple supply but how many return? Again goes inta, and comes outta. Kirchhoff can turn into a angry problem. >>On another note the Si5351C part uses an external clock input.. the thought here is that CLKIN could be driven with a decent quality TCXO reducing temp drift and possibly achieving better phase noise?. The CLKIN (external osc input) is divided from up to 100mhz to 30 Mhz max prior to the stage 1 synth block. Phase noise and drift are independent. Check both! I have yet to have issues on the bands with 1hz error, or for that fact 10hz. My idea of fun (yes I'm nuts) is get on 40 and pick a frequency like 7.181.455 and listen to the howling I'm of frequency and explain exactly where I am to drive them nuts. Or the "you gotta be on .xx00" guys. VFO radios does cause them great pain. Sorry for the humor, as just because its digital better than 1hz to me is competing with WWV. I have yet to hook these up much less made any measurements. I would start now! Allison