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
Test for solving Spurs
#ubitx
|
I learned Spurs while reading the group 's mail today.
It seems to be difficult to control with LPF or BPF because it occurs near the main transmission frequency. And I understood that it was caused by the frequency generated to pass the 45Mhz Filter. I read that the Spurs were slightly beyond specification. I have experience certifying commercial electronic devices, but I am not familiar with FCC amateur radio test specifications. I modified the code slightly. I have figured out a 10dB reduction in my instrument, but I can not be confident.?If you have someone who can test, please test. You do not need to modify the hardware. Just upload the firmware and test it. The .HEX file is located below. For how to upload firmware, please refer to the link below. If you are using stock LCD (16x02 Parallel), upload the file 'UBITX_CEC_V1.09TEST_16P.hex' If you are using Nextion LCD, please use 'UBITX_CEC_V1.09TEST_NX.hex' I would also appreciate if you test the Spurs test as well as the issues that appear when you first press a PTT or CW Key. Source code :? The source code below is not versioned (include wrong version tag). Please use for reference only. https://drive.google.com/open?id=1Z-oKnVibYRkHQKoas6zLCkNqFbbOnu-1 Thank you for reading the mail. |
|
IN THE SI5351.INO i FOUND: // User program may have reason to poke new values into these 3 RAM variables void i2cWrite(uint8_t reg, uint8_t val) { // write reg via i2c void i2cWriten(uint8_t reg, uint8_t *vals, uint8_t vcnt) { // write array uint8_t si5351Val[8] = {0, 1, 0, 0, 0, 0, 0, 0}; //for reduce program memory size void si5351bx_init() { // Call once at power-up, start PLLA extern char NowTXMode; void si5351bx_setfreq(uint8_t clknum, uint32_t fout) { // Set a CLK to fout Hz //if TX Mode and SSB On August 11, 2018 at 10:37 AM ajparent1/KB1GMX <kb1gmx@...> wrote: |
|
On Sat, Aug 11, 2018 at 06:54 AM, Ian Lee wrote:
learned Spurs while reading the group 's mail today.The spurs development is a side effect of DBM (MIXERs) operation.? The 45mhz I refer to is the IF output? that feeds the last DBM the other signal is the local oscillator (raduino osc-2).? ? The simple explanation is any time you mix two signals you get two more and those mixed with the first two give you 8 more.... IF you add the fact that those first two signals have harmonics each possible pair do as above creating more pairs of many pairs... Some of those will be desired, others are a side effect of the process and those will have varying strength according to the? their harmonic order.? The specific one that is most troublesome is the 2IF-LO (the 45mhz if *2) and the LO is clk-2..? So at 28mhz that is a LO nominally a LO of 73mhz to convert 45mhz to 28mhz.? It happens that 90mhz (2IF)- LO is 17mhz. ( you get the same this of you say 45(if)-dial frequency of 28mhz =17mhz).? ?It is a problem because the radio uses very few filters to eliminate undesired results so the 17mhz signal is amplified in the power amp and escapes along with the desired 28mhz. The presence is not the problem, it is the intensity of it (strong).? Because the dial frequency changes? so does the spur frequency. There is no way to control that in software as ubitx is an analog radio (in SDR its possible) so the analog solution is add the missing filters. **** for those looking to pick and ding I used round numbers to keep the concept clear.? It makes zero ? ? ? difference if I go to the nearest 1 Hz as a 16.983... hz, the spur is as unacceptable.? We do not ? ? ?need to be be precise to say its not allowed as its outside any Amateur Radio band. I read that the Spurs were slightly beyond specification.The US spec is -43dbc for all harmonics and spurs.? ?FCC years back did this to make USA in line with the ITU standards. In some cases its an improvement and other cases it was a relaxed number.? The one I'm not certain on is if the FCC PART-15 standards below 30mhz? also apply for out of amateur band signals as they do for most every other device.? If so that is stricter as its a radiated signal standard. I also have worked in EMI/RFI engineering prior to entering the antenna engineering and development field about 9 years ago. Allison |
¿ªÔÆÌåÓýAllison, I was wondering why the first IF could cause problems! But due to Your explanation I now understand the problem! It is not caused by the If filter itself but in TX mode by the amp beween the 45 MHz filter and the mixer, producing a second harmonic and even higher harmonics if driven too hard. So the effect must be output power related. As the transistors within this amp are limited in gain? due to their limited gain bandwidth product (my LTSPICE simulation showed a 3 dB gain reduction) higher harmonics must be much lower.? Using e.g. BFR106 insteead could worsen the spur problem. A simple remedy could either be a lowpass filter between the
output of the 45 MHz TX am and the mixer input or simply a 90 MHz
trap (bandstop filter)? Maybe a 45 MHz LC bandpass filter also
could help. ? Henning Weddig DK5LV Am 11.08.2018 um 17:29 schrieb
ajparent1/KB1GMX:
On Sat, Aug 11, 2018 at 06:54 AM, Ian Lee wrote: |
|
Allison Thank you for your feedback and test results.?I am not an RF expert like you. My major is not RF. I just wondered how Harmonics and Spurs can be easily tailored to the level that users want. Looking at the circuit and the source code, I figured out a couple of things to solve, but it seems like it was my mistake. I have one or two experiments left for this issue, but?I knew my experiment was wrong, so I do not think I need to go any further. Once this problem is solved by experts, I can expect to learn about it. Thank you again for testing. Ian KD8CEC 2018? 8? 12? (?) ?? 12:29, ajparent1/KB1GMX <kb1gmx@...>?? ??: On Sat, Aug 11, 2018 at 06:54 AM, Ian Lee wrote: |
|
Ian.
In the analog realm code code for the Raduino can only assure the oscillators (SI5351) are on the right frequency.? Not many levers we can apply there. Since we are analog for the RF the solutions are improve or add filters. I'm old and computers then were big and insanely expensive.? I'm glad I studied both! Computers got smaller, cheaper, and analog keeps going. Allison |
|
My understanding is that these spurs arise primarily in the mixer, not in the 45mhz IF amp itself.
toggle quoted message
Show quoted text
First off, the Si5351 with the 6dB pad is not giving anywhere near 7dBm into the mixer LO port, more like 0dBm. The 45mhz signal entering the mixer should be 10dB below the local oscillator, so roughly -10dBm, anything more than that and the mixer starts misbehaving.? ? These are some very round numbers, but should give an idea. Gain distribution in the transmitter is such that we can have too much 45mhz signal entering the first mixer, even if we did have 7dBm at the LO port. Especially true if you crank up the audio to get maximum?power out to the antenna. Farhan mentioned ALC as a possible solution to this. Seems simpler to design the 45mhz amp to give constant gain regardless the actual performance of the transistor in that particular rig, this means we need e a transistor with a higher Ft than the 2n3904. Also, a procedure for measuring the maximum audio signal out of the mike amp, a DVM set to AC volts? might be sufficient. If ALC is desired, John VK2ETA's scheme of moving the 45mhz IF signal around within the 45mhz filter passband is a zero additional hardware solution that gives good results. ? ??/g/BITX20/topic/16737180 Allison also had a solution involving 1n4007 pin diodes in the emitters of the IF amps, a $0.25 hack. But ALC for transmit seems an unnecessary complication for a simple rig like the uBitx. Jerry, KE7ER ? On Sat, Aug 11, 2018 at 08:49 AM, Henning Weddig wrote:
|
|
Jerry,
They are!? The two sources are the 45mhz IF and the LO...? ? Thanks for repeating some of the posting on the 5351,? The output is too low to attenuate (for better matching) and that amkes issues worse but even if the mixers where driven fully like I did and reported on the bench way back your still get all of the mixer products. I did this back in early June, your even had comments on it.? Nothing has changed save for Warren looked at CW, something I'd only glanced at and yes we problems. I did that work then to verify I was not nuts...? Not one part of this test used ubitx as its a test of what DBMs as used in uBITX do best possible case and degrade conditions from there. To rehash: Again the bench test was a MiniCircutis ZLW-2 packed mixer with SMA connectors. Mixer is then terminated on all ports with 6db pads (minicircuits SMA attenuators) its fed with not anemic LO from a signal generator at 73 mhz and 13dbm (assures that we have 7dbm, ideal). A second clean generator provides the 45mhz signal. What at the output: LO and all its harmonics (not all the same level) The "IF" 45mhz and all its harmonics And with a reasonable level for the 45mhz "IF" we get: All the possible sums and differences of any combination of those and then at lower levels any sum or difference of their results and so on out to about 5th order as then they are a bit weak and not a problem. So that is a lot of signals and only the low order ones are significant. IF we push the IF signal up to more than -10dbm? approaching 1db compression? all of those signals are still there but the relative levels grow much bigger. Bigger than what?? The desired 28mhz.?? Since I have them filter the output of the mixer after the 6db pad with a 30mhz Werlatone T2534 low pass filter. What do we get... well all the sums and differences above 30mhz are attenuated by the filter. But the differences below 30mhz are still very much there as the filter did nothing to them as we expect it to. Whats there... 28mhz, that pesky spur and and the higher order mixes that add to the? "grass" but are at least 50DBC down. We turn up the 45MHZ to -7dbm we are into overload and some of the 50DB down stuff has grown alarmingly as expected (we are well into overload and very non linear) and the? 17mhz spur is now about 23db down from the LO at 7dbm.? For comparison back with the? nice low 45mhz drive it was almost 41db down.? So it needed to be filtered but even a trivial one would take that out.? at over load we need a decent one to knock it down? by 20DB at least. IF we starve the mixer we get to the wow that nasty level about 10db sooner. So there are a bunch of things going on and the best solution that is available if to filter output for only the desired signal (band pass) as low pass lets a few go that are always there. Allison |
|
Allison,
toggle quoted message
Show quoted text
Yes, most of what I had in my last post was gleaned from your older posts. I posted in response to Henning's warnng about using a better 45mhz transistor: ? ? "Using e.g. BFR106 instead could worsen the spur problem." Here's Warren's plot for 15m: ? ??/g/BITX20/photo/65861/2 There's a fair bit of dirt showing that's lower in freq than the 15m fundamental. Is that stuff real, or an artifact of the test setup? Warren used a 1mhz RBW and 1mhz VBW, are the results accurate enough for our purposes? I'm still chewing on this from your post? ?56613, would be less ambiguous if I knew something about spectrum analyzers. "FYI: if you run video bandwidth down far enough you can get a 10db? error and also clean up the baseline, save for it will be wrong.?? Anything less than 30khz (for 3-30mhz span) on the Rigol and a less? than 1:3 radio of RBW and VBW will yield low readings.? With the? HP8568B less than 10khz and 1:3 give errors.? Also sweep rates less? than 300mS give sampling errors. The test in my case was radio, Bird model 43 wattmeter, NArda 30db attenuator,? Telonics step attenuator, Spectrum analyzer.? ?With 10W indicated on the Bird BOTH machines of known calibration will read low if set up wrong.? For the? Rigol I could not get an accurate 10W (40DBM) indication with 3khz RBW and 1Khz VBW (is was short by 11db) at 311mS sweep and 3-30mhz span. If I lowered the span to 10mhz it was off by 3db.? At 1mhz the indication? was under 1db of correct.? ?For the setup you used I add 10db.? If I make the? VBW 30khz it gets better but its still off." Jerry On Sat, Aug 11, 2018 at 12:11 PM, ajparent1/KB1GMX wrote:
?I did this back in early June, your even had comments on it.?? |
¿ªÔÆÌåÓýAll, When looking on the schematic of the ?BITX I am wondering why the
LO inputs of the mixers must have a 6 dB attenautor.? Is there a
reasonable explanation? As we already know the SI5351 is limited in output power. My fear
is taht the diodes within the mixer are nto driven hard enough so
that their on and off state is not a function of the LO signal but
also depends on the RF signal applied to it.? From my
understanding a balanced unit will always attenuate even
harmonics? (here 2* IF = 90 MHz) better than any odd harmonics. Also the SI5351 outputs rectangular drive signals which in my opinion is perfect for fast switching action within the diodes. The transformers will limit the harmonics of the driving signal, so it might be more or less a sinosoidal.? It then only the peaks of thsi signal will switch teh associated dioe on, any RF signal of significance will shift the "on-time" meaning distortion and unbalancing. Could thsi be a soruce of the second harmonic within the mixer? Henning DK5LV
Am 11.08.2018 um 21:56 schrieb Jerry
Gaffke via Groups.Io:
Allison, |
|
On Sat, Aug 11, 2018 at 12:56 PM, Jerry Gaffke wrote:
Yes, most of what I had in my last post was gleaned from your older posts.Yes, it improve the gain but we really don't need that as it just makes over drive worse. Here's Warren's plot for 15m:Using 1mhz/1mhz is fair and resonable.? The whole of comments was for a different posting. An very yes, all that junk is real thing the wide bandwidth maybe the "zero frequency" of the SA more prominent (extreme left side.) Warren used a 1mhz RBW and 1mhz VBW, are the results accurate enough for our purposes?Absolutely. I'm still chewing on this from your post? ?56613,Spectrum analyzers are graphical receivers (very much like panaramic adaptor) with? variable bandwith IF (RBW) and variable bandwith detector (VBW).? ?We use them to measure signals which in general we can do on most any radio.? They are calibrated and the detector has a log (linear is selectable) output.? Then we have the display and control (what we see and knobs and levers).? For example the screen is likely something like 800x640 (size doesn't matter more is better but it works the same) so a scan may be only 800 discrete frequencies for a screen width as more data cannot be displayed anyway.? Hence the statement of finite samples. So for a 27mhz span (3-30mhz sweep) we have say 27,000,000 or a sample every 33750hz if the exact frequency is between them and you band width is say 1khz you may miss it or the SA may over sample and interpolate to present a a find by its accuracy is lower.? Tracking generator is only a synchronous RF source of known frequency. Very handy for measuring how a black box responds to RF signals. Lots of other things go on too..? For example a very narrow filter like the crystal filter. To get a good look we use the tracking generator to generate a signal and we look at it on a point by point basis.? Filters have delay, so if we do it too fast what we sent is measured later in time and we get and image that is tilted to the right (later in time). We and not only in the frequency domain but also time. So to do a filter with a high group delay we use say 10hz resolution, narrow span maybe 10khz, and a very slow sweep so that each point (frequency step) has time to propagate though the filter and get measured before we move on. There is much more but the basic idea is measure RF amplitude at a given frequency. Its a project that to build one you have ot know how they work and to do that you need to understand both measurement and receivers. The most rudimentery would be a simple superhet with a say 1Khz crystal filter feeding a 8307 log detector and tuned by a 5351 with the arduino also displaying the 8307? result on a graphical (240x128) lcd as a vertical points and the horizontal points of?one per frequency sample.? Primitive, and somewhat limited but not terribly hard to build.? ?It does not even have to be sensitive to prove the point as the 8307 can measure down to -90dbm (with care) and up to about 0dbm without overload. Maybe more? than you wanted to know about SA. Allison |
|
Henning,
The 6db pads are there to try and tame the wily DBM.? They provide a optimized match to the ports for best possible performance.? Also because the 5351 is not? a 50ohm output devices as it was initially designed to drive digital things. Yes that means the diodes are not turned on "hard". That does two things, makes the mixing process less linear and more subject to having the other signal also interact with the diodes to alter their instantaneous current (undesired). Now the signal happen to be very square as in on time is equal to off time. In the frequency domain it means even harmonics are weaker than the odd harmonics.?? However the signals seen at the diodes are square as the leakage inductance of the transformers is small as a result of tight coupling (twisted wire) so the? extrapolating the wave?shape?at the diode being a factor is not true.? its limited? current flowing?in the?diodes interacting with the fairly high current of the signal that causes some of?the havoc. Combine that with the very nature of sampled signals (yes nyquist) we get a lot of mush at the output.? Its very simple looking device but anything but simple in operation. Allison |
|
I was figuring we're best off looking hard at everything between the mike and that first mixer,
toggle quoted message
Show quoted text
adding sufficient feedback in each amp such that we get consistent results. Gain distribution needs attention, giving lower level signals going into the mixers, more gain in the power amp. But sufficiently high Ft's in the 45mhz amp to get the same target gain each time we stamp one out. Thanks for the Spectrum Analyzer notes, I'll go over that thoroughly. Jerry On Sat, Aug 11, 2018 at 01:49 PM, ajparent1/KB1GMX wrote: On Sat, Aug 11, 2018 at 12:56 PM, Jerry Gaffke wrote: |
|
Regarding Spectrum Analyzer basics:
toggle quoted message
Show quoted text
Well, that was easy enough. But need to study up on VBW in particular, haven't yet found a full description. I assume the name "video bandwidth" comes from analog spectrum analyzers, the filter used to smoothed the Y beam deflection when showing a trace on the CRT.? Exactly how VBW might affect the measurement of signal peaks as seen on the screen and how it fits in with the selection of RBW still evades me. On Sat, Aug 11, 2018 at 02:02 PM, Jerry Gaffke wrote: Thanks for the Spectrum Analyzer notes, I'll go over that thoroughly. |
|
The diode ring mixers have stuff going in and out of all three ports,
toggle quoted message
Show quoted text
terminating all three ports into 50 ohm pads helps to dampen undesirable products. But if you want a 6dB pad between the si5351 and a 7dBm level mixer, you had best have an amp in there too. The si5351 can't drive hard enough at 7+6=13dBm? for this to work very well.? See posts 33902 35206, 46122 Jerry On Sat, Aug 11, 2018 at 01:11 PM, Henning Weddig wrote:
|
|
Jerry,
VBW in the simplest form is a low pass filter after the detector to take out high frequency noise.? If overdone you start filtering useful information as we are dealing with signal over time. Close analog, Take a simple half wave power supply look at it with a scope. your see DC, make the filter cap smaller so that it stops filtering the 60hz and you start to see bumps then humps and half sine waves. Same happens with the log detector in SAs.? You measure a point frequency and if the? filter is narrow it doesn't register it because the filter saw it as a fast event and its eliminating them. You also get that with noise as averaging noise tends to make an average of zero (if the noise is truly random). Allison |
to navigate to use esc to dismiss