Re: Many uBitx's have reduced power on 40m CW
Actually 80,40,20 are all fairly even.
So adjust RV1 or R83 as appropriate for SSB use.
toggle quoted message
Show quoted text
On Thu, Apr 8, 2021 at 10:24 AM, Jerry Gaffke wrote:
Evan,
I have no idea if you will have trouble with IMD on 80m.
Increasing R83 would cure IMD and also further improves the impedance seen by the filter,
but reduces power on all bands.
Could just rotate RV1 for suitable 80m power levels.
Could use TXC (turns on the 80m LFP) to turn on a FET which attenuates the signal at RV1
Could add something at Q911,912 that reduces gain at low frequencies similar to C81,R83
Having Q90 as a buffer should give the filter and mixer a proper load impedance.
Jerry, KE7ER
|
Re: Many uBitx's have reduced power on 40m CW
Evan,
I have no idea if you will have trouble with IMD on 80m.
Increasing R83 would cure IMD and also further improves the impedance seen by the filter,
but reduces power on all bands.
Could just rotate RV1 for suitable 80m power levels.
Could use TXC (turns on the 80m LFP) to turn on a FET which attenuates the signal at RV1
Could add something at Q911,912 that reduces gain at low frequencies similar to C81,R83
Having Q90 as a buffer should give the filter and mixer a proper load impedance.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Thu, Apr 8, 2021 at 09:41 AM, Evan Hand wrote:
Jerry,
I did not think of the LPF causing the dip on 30meters.? That makes sense.
I will check the SSB signal on 80meters with my RSP1A and the SA software.? The choice would then be to change R83 to a higher value like in the v3 10ohms.? Maybe 6.8ohms?
Guess I am not finished yet.
73
Evan
AC9TU
|
Re: Many uBitx's have reduced power on 40m CW
Jerry,
I did not think of the LPF causing the dip on 30meters.? That makes sense.
I will check the SSB signal on 80meters with my RSP1A and the SA software.? The choice would then be to change R83 to a higher value like in the v3 10ohms.? Maybe 6.8ohms?
Guess I am not finished yet.
73
Evan
AC9TU
|
Re: Many uBitx's have reduced power on 40m CW
Evan,
The dip on 30m is because it is using the 40m LPF, which must attenuate a 20m second harmonic.
But hey, it's primarily an SSB rig.
My only concern would be that the big C81 causes too much power out on 80m.
Shouldn't be a problem for CW where the PA can be non-linear,
but for 80m SSB there may be some IMD popping up unless you back off some on RV1.
Or increase R83.
Jerry, KE7ER
On Thu, Apr 8, 2021 at 02:48 AM, Evan Hand wrote:
The only abnormality with that mod is the dip for 30meters.? I am not going to pursue that issue as I consider it a good tradeoff for the 40meter increase.
|
Re: ArduinoShrink - reduced size Arduino AVR Core
If it's at least a 16 bit machine and either has lots of registers or an easy way to
access words in RAM as an offset from the stack pointer,?C is the way to go.?
This applies to pretty much all modern processors.
The instruction set for the MSP430 is designed for C and is very simple,
worth looking over (the peripherals are a tough read though).?
I'd be hard pressed to write assembly language for it that was faster than compiled C.
The old Sun Sparc was pretty good too, especially if access to stack space
in main memory was slow.
The ATMega328P, not so much.? ?
A very small 8 bit machine, and the Harvard architecture limits access to data stored in flash.
https://www.reddit.com/r/linux/comments/73in9/computer_holy_wars_png/
Jerry, KE7ER
toggle quoted message
Show quoted text
On Wed, Apr 7, 2021 at 10:31 PM, Tom, wb6b wrote:
On Wed, Apr 7, 2021 at 09:33 AM, <quad@...> wrote:
ArduinoShrink?replaces Arduino AVR Core functions with smaller and faster versions.
The Arduino Blink example sketch built for the Uno takes 924 bytes of flash, but only needs
196 bytes with ArduinoShrink.
Looks really cool.?
Upon reading his explanation of his code, I need to revisit my assumption that writing assembly is almost never worth the effort now days.?
Although it looks like he put an additional restriction on specifying pin numbers (probably not a bad thing) to do some shrinking in the wiring library.
Tom, wb6b
|
TSW Teensy 4.1 Raduino for the (V6 uBITX) News
#ubitxv6
Even though the sources for the Silicon Labs Si5351a chips have pretty much dried up and the remaining stock at some suppliers is being sold at highly inflated prices, we found, thanks to a post by one of the BITX20 groups members,? that there is a Chinese equivalent (MS5351M) readily available, with reasonably fast (DHL) shipping from at least one supplier out of China (www.chipmall.com) who has very reasonable prices (High DHL shipping cost can be offset by larger volume orders).
TSW placed a volume order and we have enough on hand now that there will not be a shortage of the Teensy 4.1 Raduino Kits. They have been tested and are virtually identical to the Si5351, at least in pinout, appearance? and I2C address (0x60). In the Raduino, they? produce clean output at least as good as the Si Labs chips. Once our small supply of Si5351a's is depleted,? the MS5351M's will be used in our kits.
The currently listed kit prices will not change and they are available for order.
W0EB found a "test socket" that will take the SMD 5351's footprint and is currently designing a test fixture that will be used to test each Si or MS 5351 prior to installing it on the Raduino board to insure (especially with the MS5351M's that the I2C address is in fact 0x60 and the output levels are at least as good as the Si5351 from Silicon Labs.? That way if any substandard ones have made their way into the batch on hand, they can be rejected before causing any problems for the user(s).
Current kit description and pricing information is available on the TSW website at www.w0eb.com .
Jim, W0EB
TSW Project Coordinator
|
Re: ArduinoShrink - reduced size Arduino AVR Core
As someone who started in I.T.? on a Mainframe with 16 k Magnetic core memory no disc drives and only ran a single program at a time there was always a bottleneck. Whenever a routine had to run with minimal memory and as fast as possible it was a good candidate for optimising, rewriting or if that didn't achieve it it was rewritten in machine code. It's funny in some ways not much has changed. There are still bottlenecks.
On Thu, 8 Apr 2021, 3:02 pm Tom, wb6b, < wb6b@...> wrote: On Wed, Apr 7, 2021 at 09:33 AM, <quad@...> wrote:
ArduinoShrink?replaces Arduino AVR Core functions with smaller and faster versions.
The Arduino Blink example sketch built for the Uno takes 924 bytes of flash, but only needs
196 bytes with ArduinoShrink.
Looks really cool.?
Upon reading his explanation of his code, I need to revisit my assumption that writing assembly is almost never worth the effort now days.?
Although it looks like he put an additional restriction on specifying pin numbers (probably not a bad thing) to do some shrinking in the wiring library.
Tom, wb6b
|
Re: Many uBitx's have reduced power on 40m CW
I put my v4 back together with the stock L1-4 toroids.? The only change I made was to C81.? I put a 2.2nF in parallel with the 470pF that was there.? Here are the CW results with 12 volts to the rig:
80meters? ? ?3.510MHz? ? ? 13.4watts
40meters? ? ?7.010MHz? ? ? 14.5watts
30meters? ? ?10.110MHz? ? ? ?7.8watts
20meters? ? ?14.010MHz? ? ?12.6watts
17meters? ? ?18.078MHz? ? ? ?9.9watts
15meters? ? ?21.010MHz? ? ? ?7.0watts
10meters? ? ?20.010MHz? ? ? ?3.6watts
The only abnormality with that mod is the dip for 30meters.? I am not going to pursue that issue as I consider it a good tradeoff for the 40meter increase.
So here is what I think that I have learned:
1 - The production filter is not optimal, but does work
2 - The modification that Curt used works, though through playing with LTspice the optimum value that I found is 2500pF for C81.? The closest I could come was 26700pF which worked.
3 - Probably the biggest impact with changing the C81 value is the impedance impact on the LPF, not really the gain.? The mismatch on the impedance was causing signal distortion that reduced the CW output to 40meters.
4 - Going too high a value for C81 will produce a larger difference in the 80meter vs 10meter output.? Increasing the capacitance does impact the gain of Q90 as well as the impedance.
I still need to do the purity tests for both CW and SSB.? I will do this by modifying my v6 with the above change to C81, test to see that I get the same results, and then do the spectral analysis.? I did change the relays on the v4 and tests with my TinySA confirmed that the harmonic issue was resolved, however, I have not done the spur fix yet.? The v6 should be clean before the mod, so that will be the test to indicate if there is an impact on the signal purity.
This is Fun!? Thank you, Jerry, Curt, and Farhan for the feedback!? Jerry especially thank you for the LTspice link and information.
73
Evan
|
Justin, FYI, There is a Groups.io for that work.? It is here: /g/ucxThe original project was called ?CX, which the group felt was too close to the QCX/QSX names that are used by QRP-Labs.? The official name is now ?SDX or uSDX though the Groups.io page (like this page) has a different net ID. Have fun! 73 Evan AC9TU
|
Re: ArduinoShrink - reduced size Arduino AVR Core
On Wed, Apr 7, 2021 at 09:33 AM, <quad@...> wrote:
ArduinoShrink?replaces Arduino AVR Core functions with smaller and faster versions.
The Arduino Blink example sketch built for the Uno takes 924 bytes of flash, but only needs
196 bytes with ArduinoShrink.
Looks really cool.? Upon reading his explanation of his code, I need to revisit my assumption that writing assembly is almost never worth the effort now days.? Although it looks like he put an additional restriction on specifying pin numbers (probably not a bad thing) to do some shrinking in the wiring library. Tom, wb6b
|
After getting my feet wet with the UBITX I ran across this interesting sounding project:
I sent off the Gerber files and I'm waiting for the boards to arrive by slow-boat as I am in no hurry.? If anybody else might want to play around with this I will have extra boards.? Minimum order was 10 PCB and with shipping was a whole $24!
Hope I'm not breaking any rules.
73,
Justin B.
KI5GKD
|
Re: Many uBitx's have reduced power on 40m CW
The AL of a T30-6 core can vary +/-5% of the nominal value, and inductance is proportional to AL.
? ??
Yet another handy factor to blame bad results on.
Inductance is proportional to the square of the turns count.
Having 8 turns instead of 9 is a 100*(9*9-8*8)/(9*9) = 21% reduction in inductance.
Different spacing around the core and loose vs tight windings can make a sgnificant difference.
As you have noticed.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Wed, Apr 7, 2021 at 04:27 PM, Evan Hand wrote:
The most recent observation is that there is a HUGE variation in the inductance of the toroids depending on the turns AND the spacing.?
|
Re: Many uBitx's have reduced power on 40m CW
The plot thickens.
The most recent observation is that there is a HUGE variation in the inductance of the toroids depending on the turns AND the spacing.? I tested this because my first attempt at winding the "better" filter toroids produced unexpected measurements with my LC meter.? I then remembered the QCX LPF tuning that Hans Summers has documented for the QCX.? I now have another test to perform.? Now I need to see if adjusting the spacing on L1 to L4 can optimize the LPF.
Based on my LC meter, the inductance of 9 turns on a T30-6 can vary from 290nH to 310nH.? It could mean that "tuning" the inductors could balance out the power levels.? That could also explain the production variations, coupled with the gain differences in Q90.
Oh well, more tests to document.
73
Evan
AC9TU
|
Re: Many uBitx's have reduced power on 40m CW
Farhan, Perhaps driving harder will help, worth trying. However, the middle picture of this post shows TP1 when operating 40m CW: ? ? ? /g/BITX20/message/87328The first image here shows TP1 when attempting to operate 40m CW, except Evan has removed L4,C204, and replaced them with a?47 ohm termination to ground: ? ? ? /g/BITX20/message/87415As you can see, the mixer behaves much better when properly terminated. In this post: ? ?? /g/BITX20/message/87369he finds that rotating RV1 fully clockwise to shut down the power amp has no effect on what he sees at TP1. So the dirt at TP1 is not getting coupled in from the high power stages of the final. This is why I'm simulating the input impedance of the filter at L1,2,3,4 while driving the Q90 stage. Going back to the v3 values for C81 and R83 brings it much closer to 50 ohms. Not stated, but I believe those screen shots are taken from a v5. Evan also has a v4 and a brand new v6. Jerry, KE7ER
On Tue, Apr 6, 2021 at 11:40 PM, Ashhar Farhan wrote:
The CW signal is generated by injecting a DC current into the diode mixer, upsetting the balance.?
At present, the R104 and R105 inject approximately (4.4V/ 3.2K =) 1.5 mA of current into the mixer.
The diodes have an on resistance of approximately?(26mV/1.5mA) = 39 ohms. To increase the RF carrier leakage, we can increase this. If we replace the 2.2K of R105 with a lower value (for instance by strapping a 470 ohms across it) we should be able to get the carrier level up back.
- f
|
Re: Many uBitx's have reduced power on 40m CW
As Curt found, increasing C81 seems to be key here.
The filter and mixer see much closer to 50 ohm loads with the v3 values of C81=0.1uF, R83=10ohms.
But v3 had too little power on the high bands.
I wonder if we could go back to the v3 values on C81,R83 to give the filter and mixer closer to 50 ohm loads,
then whack at the Q911,912 stage to increase power on the high bands.??
Jerry, KE7ER
toggle quoted message
Show quoted text
On Wed, Apr 7, 2021 at 01:07 PM, Evan Hand wrote:
All comments are welcome.? I will post the actual values after I have done construction and measurements.
|
Re: Many uBitx's have reduced power on 40m CW
I think I am done playing with LTspice.? Here is the circuit I plan to change my v4 to:  The LTspice file is attached. I wanted to balance out the input impedance issues with the gain issues at higher frequencies.? P{er the model, this seems to be the best compromise. I did change the Chebyshev filter for a 32MHz cutoff to give at least the lower part of 10meters a chance at power (thinking of the new Techs and increasing sun spots). All comments are welcome.? I will post the actual values after I have done construction and measurements. 73 Evan AC9TU
|
ArduinoShrink - reduced size Arduino AVR Core
An interesting new contribution by "NerdRalph" - this may help the uBitx Raduino code size -
ArduinoShrink?replaces Arduino AVR Core functions with smaller and faster versions.
The Arduino Blink example sketch built for the Uno takes 924 bytes of flash, but only needs
196 bytes with ArduinoShrink.
|
Re: A simple question, I hope...
Testing on-air?
There's a tuning device that was very popular a few decades back, and schematics are still available online.? It uses a 555 to generate approx. a 30% duty cycle CW key-down, and you can add a second 555 (or use a 556) to superimpose an audio signal on the CW.? It replaces your mic and just uses your keying line, unless you mix in audio.?
Either way it sounds like crap but it's a quick and easy way to get something to test CW and/or SSB out.? Use of a dummy load is recommended.? This link shows some possible options.
<>
Because it's basically an unfiltered square wave, you'll have enough key clicks to get a signal on SSB.?
When I built mine back around 1990 I labeled the box "N6OTQ Death Ray" because of the horrid noise it made on air.? Thankfully, nowadays it's fallen out of favor among the QRO folks.
73 ?Jim N6OTQ
|
Re: Many uBitx's have reduced power on 40m CW
Jerry, I am still playing with LTspice and have not yet gotten to building the circuits.? In my playing around I did come to the realization that if I did the transformer impedance match I would solve the problem of the 40meter CW drop off.? That would seem to me to then create a receiver sensitivity issue as even with Q90 off, there is still around 320ohms on the C80 side of the transformer that would be transferred to 80ohms on the other side that would be across the antenna input.? Unless switching is added I do not believe that the transformer is viable.? Would you agree with the impedance values??? Attached is the circuit with the transformer.? Here is a screenshot of the transformer analysis in transmit mode.  73 Evan AC9TU
|
Re: Many uBitx's have reduced power on 40m CW
The CW signal is generated by injecting a DC current into the diode mixer, upsetting the balance.? At present, the R104 and R105 inject approximately (4.4V/ 3.2K =) 1.5 mA of current into the mixer. The diodes have an on resistance of approximately?(26mV/1.5mA) = 39 ohms. To increase the RF carrier leakage, we can increase this. If we replace the 2.2K of R105 with a lower value (for instance by strapping a 470 ohms across it) we should be able to get the carrier level up back. - f
Given the variation in impedance as seen in the simulations of ideal components,
it's amazing to me that the output power manages to be as flat as it is.
Fun to figure out what all is going on here, more than it might seem at first.
From a practical standpoint, Curt's bigger C81 strikes me as the best fix,
for those that care about the slightly reduced power when operating 40m CW.
But with C81 too big, the difference in output power between 3.5mhz and 30mhz
may become too extreme.
For the v3 configuration with C81=0.1uF and R83=2.2ohms, my simulation shows that
the Q90 stage has an input impedance of 51 ohms to 45 ohms, at 173 to 165 degrees
Voltage gain ranges from 4.5 to 3.65.
These values are for a range of 3.5mhz to 30mhz, in that order, Q90 stage has a load of 50 ohms.
On v4, R83 was moved to 2.2 ohms for more gain.? ?
Input impedance dropped to a range of 18 to 19 ohms, 170 to 183 degrees.
Voltage gain ranges from 5.5 to 6.7.?
On v5,v6 the cap at C81 was moved to 470pF so high frequencies have more gain than low frequencies,
this compensates for losses at high frequencies in the various power amp stages.
Input impedance now ranges from 200 ohms to 45 ohms at 150 to 100 degrees.
Voltage gain for the Q90 stage runs from 1.0 at 3.5mhz to 5.5 at 30mhz.
The v3 configuration is likely the best as far as giving the filter and mixer a proper 50 ohm load.
But we needed more gain at higher frequencies to get sufficient power out,
without too much drive on 3.5mhz.
I haven't analyzed the effect of the Q90 impedance phase angle
on how the filter or mixer might behave.
Jerry, KE7ER
On Tue, Apr 6, 2021 at 03:25 AM, Evan Hand wrote:
I also looked at changing C81 to .1uf and find that the impedance is better across the spectrum of interest.? I do not know how to see if there is an impact on the rest of the amp without building it.
|
Jim Sheldon