Keyboard Shortcuts
ctrl + shift + ? :
Show all keyboard shortcuts
ctrl + g :
Navigate to a group
ctrl + shift + f :
Find
ctrl + / :
Quick actions
esc to dismiss
Likes
Search
a simple spur fix
|
so, there is one spur fix that is now consistently working for me. among rhe solutions, i was looking for one that would be easy to apply. first, the distortion in the 45 mhz amp is creating harmonic distortion. so a 0.3uh in series with a 10 pf is used across TP13 (output of the 45 MHz IF amp going to the front end mixer). 8 turns on T30-6 gives you 300nh. second, there is a cross talk inside the Si5351 where the 12 mhz from the bfo leaks into the local oscillator. A 5uh in series with a 47 pf across C202 of the LPF will suppress it. 35 turns on T30-6 gives you 5uh. - f 
IMG_20181109_202634.jpg
IMG_20181109_132739.jpg
IMG_20181109_132244.jpg
IMG_20181109_132549.jpg
IMG_20181109_132100.jpg
|
|
Excellent!
toggle quoted message
Show quoted text
So no second 45mhz filter, no surface mount inductors at L5,L7,? L1,L2,L3,L4 ? TP13 is not shown in the v3 schematics, only in the v4 schematics. I assume these mods work will work on the v3 boards too. TP13 is the node at T2-pin3 in the first mixer. I assume "across TP13" means that we add the cap and inductor in series from that node to ground. Will be interesting to see if it matters where we physically place it, at T2-pin3 would be nice because it could be added to the bottom of the board directly to the ground plane. The second series cap and inductor get placed across C202, in the middle of the 30mhz low pass filter on the far left side of the schematic. So just leave C202 as it was, but add the cap and inductor from the through hole pin for L2 (or L3) to the ground plane. Will be interesting to see how picky we have to be about those inductors. If we're lucky, they can be most any small surface mount inductor. Jerry, KE7ER On Fri, Nov 9, 2018 at 07:35 AM, Ashhar Farhan wrote:
|
¿ªÔÆÌåÓýI wondered about the cross-talk in the Si5351.? There have been a number of papers written on it being problematic¡ to the point of suggesting using only one channel of the Si5351 but multiple parts if you need more frequencies.? Channel separation in some cases is only in the 30ish dB range¡ ? ? From: [email protected] [mailto:[email protected]] On Behalf Of Ashhar Farhan ? so, there is one spur fix that is now consistently working for me. among rhe solutions, i was looking for one that would be easy to apply. first, the distortion in the 45 mhz amp is creating harmonic distortion. so a 0.3uh in series with a 10 pf is used across TP13 (output of the 45 MHz IF amp going to the front end mixer). 8 turns on T30-6 gives you 300nh. second, there is a cross talk inside the Si5351 where the 12 mhz from the bfo leaks into the local oscillator. A 5uh in series with a 47 pf across C202 of the LPF will suppress it. 35 turns on T30-6 gives you 5uh. - f ? -- ¡_. _._ |
¿ªÔÆÌåÓýAshhar: ? You give the number of turns necessary, and the core IDs, but what is the gauge of wire used??? ? David A Posthuma, WD8PUO 1 (616) 283-7703 ? From: [email protected] <[email protected]> On Behalf Of Ashhar Farhan ? so, there is one spur fix that is now consistently working for me. among rhe solutions, i was looking for one that would be easy to apply. first, the distortion in the 45 mhz amp is creating harmonic distortion. so a 0.3uh in series with a 10 pf is used across TP13 (output of the 45 MHz IF amp going to the front end mixer). 8 turns on T30-6 gives you 300nh. second, there is a cross talk inside the Si5351 where the 12 mhz from the bfo leaks into the local oscillator. A 5uh in series with a 47 pf across C202 of the LPF will suppress it. 35 turns on T30-6 gives you 5uh. - f ? |
|
Timothy Fidler
Yup.? it is correct.? The easiest way to get the units right is? L in uH , C in uF and result will be in Mhz? ie recip of sqrt of micro x micro .
toggle quoted message
Show quoted text
But remember when you are working at VHF with 5 percent capacitors and air wound coils,? even Tiny errors or tolerances? will put you off by 100s of KHz.? Ditto temperature effects on components. ? Timothy E. Fidler : Engineer BE Mech(1) Auckland , NDT specialist AINDT UT /RT3 , MT2 CB #2885,?
|
|
Farhan,
Your doing what I tried back in June/July time frame, it failed then. Reason first series traps good enough to suck out the spur at a given frequency are narrow enough? to be only a spot frequency fix.? The 2IF-LO spur for 28mhz BAND is between 17 to 16 mhz for 10M alone.? So you need a series trap that can not do a band. The other issue is the power amp.? At 80M and 10W out the gain is about 60db but as you go up, at 20M the gain is down to ~56db, and at 10M the gain is maybe ~52DB.? ?So at the spur frequency you have more gain available than the desired output frequency and the differential gain is enough to put that spur as shown at -45db at more like -37db below 2W. The fix has to be more comprehensive.? Also it must be field verifiable or factory set to allow for 1 or 5% parts. The 45mhz section needs: ? 45mhz filter inductors shielded or better that section relocated. ? Amplifier distortion clean up. ? Swap the 45mhz port and LO ports as IF ports are most sensitive to mismatch. A thought to look into... ? Replace the TIA amp at 45mhz with a more common resistive feedback amp (typical 2n5109 amp) ? with 20db gain.? Here is the hint:? Often when that amp is shown due to the fact that changes at either ? port impact the opposing port they often have a 6db pad at the output.? However for the TX case that ? same pad can be at the input and it insures the 45mhz filter will have a good match for the network ? and the transistor with its input impedance stable will have output stability going to the mixer. Allison |
|
I installed the series trap across C202 in the LPF and it reduced the 30 meter sensitivity drastically. On Fri, Nov 9, 2018, 5:46 PM ajparent1/KB1GMX <kb1gmx@... wrote: Farhan, |
|
Good!? Sounds like it's working!? ;-)
toggle quoted message
Show quoted text
I for one am willing to give up 30m on what is mostly an SSB rig. If you were planning to operate CW on 30m, you could put a couple pins and a jumper in series with that C+L.? So with the jumper out, the mod is undone. Transmitting CW we don't have trouble with spurs, since the BFO and clk1 are both shut down. The only RF floating about the rig is clk2 at the transmit frequency. Jerry, KE7ER On Fri, Nov 9, 2018 at 03:03 PM, barry halterman wrote: I installed the series trap across C202 in the LPF and it reduced the 30 meter sensitivity drastically. |
|
Allison,
toggle quoted message
Show quoted text
I'm not 100% sure what his LC traps are doing, but they may not have to move around. My best guess is that the one at TP13 (tuned to around 90mhz) just removes the 2'd harmonic of the 45mhz IF that gets created by those 2n3904's running flat out.? It doesn't have to move. Example:? Transmitting at 21mhz, clk2 is at 45+21=66mhz. If any 2*45=90mhz energy is present, it mixes with the 66mhz clk1 at D1,D2 to create a 90-66 = 24mhz spur. Farhan suggests the second LC trap is to suppress the 12mhz BFO energy due to crosstalk between clk0 and clk2. Though we have a report that this also sucks out 10mhz reception, so either that particular example is tuned differently than Farhan's, or that's an awfully wide trap, or I don't know what's going on there.?? If these mods are easily reproducible for acceptable results, good enough by me. We'll see. Jerry On Fri, Nov 9, 2018 at 02:45 PM, ajparent1/KB1GMX wrote:
Your doing what I tried back in June/July time frame, it failed then. |
¿ªÔÆÌåÓýMy guess is that the two intended
filter frequencies are:? 90Mhz (second harmonic of 45MHz 1st LO)
and 12MHz (2nd LO).? The parts represent the right values for a
filter at 90MHz, but it seems that 30M band would be wiped out by
the second of the mods (which somebody has already observed).??
12MHz should be being wiped out instead, so I suspect a typo.
Perhaps? either a 35pF capacitor and the 5?H inductor as stated or a 3.8?H inductor and the 47pF capacitor as stated being about right for getting the filter resonant just on the low side of 12MHz.? Mike ZL1AXG On 10/11/18 9:49 AM, Brian L. Davis wrote: 10pf series with .3uhy - resonant about 91 MHz
|
|
allison and others, the traps are capable of adding a notch of -20 db. so even if you don't hit the bottom precisely, a 10 db reduction is all that is necessary to clean it up to the required level. these are meant to be fixes that work with existing boards : from v1 onwards. i tried to choose a method that was easy to do. we will need a different post mix if amp design going forward yes the traps are for 90 mhz and 12 mhz. the 10 mhz insensitivity is something i will look into today, I havent experienced it. perhaps we should move the trap to tx filters for the 21 mhz to 28 mhz band? the 12 mhz is filtered for the lower bands in any case. the other place to apply the 12 mhz trap would be the local oscillator output? - f On Sat, 10 Nov 2018, 05:48 Mike Woods <mhwoods@... wrote:
|
|
How did you arrive at 47pF and 5uH for 12mhz?
toggle quoted message
Show quoted text
freq = 1/(2*pi*sqrt(L*C)) = 1/(2*pi*sqrt(47e-12 * 5e-6)) =?10382123 Hz Though that 30mhz LPF is a fairly complicated network to hang an LC trap on, there may be more going on here.? The 12mhz trap my not be needed if you reduce the loading on the si5351 clocks since this would reduce crosstalk between clocks within the si5351. ? Table 3, "DC Characteristics", in the si5351 datasheet says the output impedance is 50 ohms. My best guess is that this means there is effectively a series 50 ohm resistor between the? CMOS driver and the output pin.? If you then put a 50 ohm external load to ground on it and? the driver is trying to drive high to 3v, then the expected current would be 3v/(50+50ohms) = 30ma. But the output driver is limited to 8ma, so that output will be current starved. And with three clocks all working hard as they possibly can, the crosstalk is pretty bad on the si5351. So we want an external load that's more than 50 ohms on the si5351 clock pin, but we want the mixer to see a 50 ohm source impedance to avoid reflections. A pi network resistive attenuator is not a good choice. I suggest each si5351 clock pin drive a series DC blocking cap and a series 100 ohm resistor into the mixer port.? Assuming the si5351 has not hit its 8ma current limit, the mixer will see a source impedance of 50 ohms (internal to the si5351) plus that 100 ohm resistor, so 150 ohms. Add a 75 ohm resistor from mixer port to ground, that in parallel with 150 ohms makes it a 50 ohm source impedance for the mixer. Jerry, KE7ER it? On Fri, Nov 9, 2018 at 05:53 PM, Ashhar Farhan wrote:
|
|
you are right jerry. i remember encountering this ... back to the bench to try this out... - f On Sat, 10 Nov 2018, 10:07 Jerry Gaffke via Groups.Io <jgaffke=[email protected] wrote: How did you arrive at 47pF and 5uH for 12mhz? |
|
Right Brian, I don't really understand that as well. Maybe it might make sense but I should see how the schematics would be. The 91mhz might make sense to notch out the harmonics of 45mhz if this is the cause of the spur. Apart from the resonances I would look at the Q factor as well. Il 09/nov/2018 21:49, "Brian L. Davis" <brianldavis@...> ha scritto: 10pf series with .3uhy - resonant about 91 MHz |
|
To drastically increase the Q and thus the notch depending on whether the circuit is in series or in parallel one can lower or increase the impedance of the filters with impedance transformers at the input and the output of the filter. It would add complexity and likely lower the TX output. Il 10/nov/2018 07:20, "iz oos" <and2oosiz2@...> ha scritto:
|
|
Unless we add an amp to each of the three clocks between the si5351 and the mixers,
toggle quoted message
Show quoted text
we are trading off si5351 crosstalk, local oscillator injection level, and reflections due to? driving the mixer from a source impedance higher than 50 ohms. All three of these factors will contribute to spurs. One might replace the 100 ohm and 75 ohm resistors of my previous email with small pots, then try various settings to find the best compromise. ? On Fri, Nov 9, 2018 at 08:37 PM, Jerry Gaffke wrote:
I suggest each si5351 clock pin drive a series DC blocking cap and a series 100 ohm resistor |
|
This might be an opportunity for somebody to create yet another new Raduino
toggle quoted message
Show quoted text
Add buffer amps to all three clocks from the si5351 to help solve the spur problem. Choose a processor clock that won't beat with the BFO, creating audio tones cases where the Nano's 16mhz ceramic resonator happens to fall on an unlucky frequency . Communicate with a host processor via well filtered 3v UART lines that go off to a well shielded USB-to-UART dongle. Processor is a little beefier, something like an STM32F*, not pressed for memory space, more IO pins, and with the horsepower for a few simple digital modes such as PSK31. This could easily be used to upgrade old v3 and v4 uBitx's. Jerry, KE7ER On Sat, Nov 10, 2018 at 06:41 AM, Jerry Gaffke wrote: Unless we add an amp to each of the three clocks between the si5351 and the mixers, |
|
Farhan's 90mhz filter is clearly meant to take out the 2'd harmonic of the 45mhz IF signal.
toggle quoted message
Show quoted text
It's perhaps 90mhz and not the calculated 91mhz due to some capacitance in the node it connects to. May need a trimmer cap to tune it for exactly 90mhz unless low tolerance components are used. The user could have Nano code that sets clk2 to 90mhz and unbalances the mixer with CW-KEY, then adjusts the trimmer cap for a null while monitoring the 90mhz clk2 with a diode RF probe. Perhaps no diode RF probe is needed for this if we also set clk1 to 90mhz+12mhz+500hz, and we monitor audio in the receiver for a null in the resultant 500hz tone while adjusting the trimmer. I'm assuming some of that 90mhz energy can sneak by the 45mhz filter somehow. The? proposed 12mhz (10.4mhz?) trap in the 30mhz LPF might be more problematic. Perhaps easier to add a high pass filter to clk2 instead, removes any 12mhz energy present on clk2 before it gets to the D1,D2 mixer.? No critical tuning required, and does not trash 12mhz (30m?) reception. That in addition to monkeying around with the 6dB pads on clk0,1,2 to reduce crosstalk. Jerry, KE7ER On Fri, Nov 9, 2018 at 10:20 PM, iz oos wrote: The 91mhz might make sense to notch out the harmonics of 45mhz if this is the cause of the spur. |
|
I have no doubt Farhan solution might work but I guess that the Q of the 90Mhz and the null could be higher using a quarter wave open coax stub cut for the 90Mhz (likely a cap between 20 and 30pf in parallel would be needed). Il 10/nov/2018 17:38, "Jerry Gaffke via Groups.Io" <jgaffke=[email protected]> ha scritto: Farhan's 90mhz filter is clearly meant to take out the 2'd harmonic of the 45mhz IF signal. |
MadRadioModder
David Posthuma
to navigate to use esc to dismiss