Re: RF power chain mods and improvements..

Howard Fidel

I just got to retest inserting a 1 KHz tone and also with speech. I tested on 21 MHz and drove the signal until I got my full output, 7 watts. The spurs were around -46 dB. With my equipment, I need to eyeball it, it doesn't make measurements, its a 40+year old analyzer I bought in the 70s. IT seems my rig is compliant, but not by a huge amount.
On 10 meters I get 4.5 watts. I have done a lot of simulations using IRF510s and believe that I can get them to produce 10 watts on 10 meters. I am documenting the work, and will share it later.
Oh, my power supply is 12 volts, not 13.5.


On 6/4/2018 5:20 PM, Arv Evans wrote:
Allison is correct that things are much cleaner when using CW. 

For my tests I inserted 1KHz at variable levels and adjusted the audio level for
cleanest RF on the oscilloscope.  Then I performed a frequency sweep to look for
spurs.  As long as the microphone port audio input is below 45 millivolts the spurs seem
to be down at least 50db, but can be pushed even further down by reducing the audio
level to around 32 millivolts.  This seems to be where oscilloscope peak-to-peak RF level
is about the same as when using CW mode. 

I could not duplicate this using an averaging type RF voltmeter as the signal level detector.

It is interesting to vary the microphone port input frequency and watch each spur to see
if it moves up or down.  This could be an aid in figuring out how each spur is being



On Mon, Jun 4, 2018 at 10:32 AM Howard Fidel <sonic1@...> wrote:
I just got an explanation from Allison, which I will confirm later. I measured the spurs in CW which I thought was a side tone, but isn't. I need to retest with actual audio.


On 6/4/2018 12:21 PM, Arv Evans wrote:

I'm seeing about the same.


On Mon, Jun 4, 2018 at 10:18 AM Howard Fidel <sonic1@...> wrote:
When properly driven, I see the spurs almost at -60dB.

On 6/4/2018 11:37 AM, Arv Evans wrote:
The uBITX does actually work on 15M, 12M, and 10M (and yes it does work on 11M as
evidenced by the number of CB'ers who have purchased it).  Power output is much lower
on the higher bands, but is still enough to make QRP contacts.  Just do not try to increase
power output on upper bands by adding microphone gain.  Too much microphone audio
will cause spurs, distortion, and QRM. 

Arv  K7HKL

On Sun, Jun 3, 2018 at 9:01 PM Jerry Gaffke via Groups.Io <> wrote:
To keep the price down, I think hfsignals may as well continue to use the 45mhz filter.
Perhaps claim 80m through 17m, beyond that is experimental, disable 15m,12m, and 10m
in the stock firmware.  Different transistors for more consistent gain should be considered.

However, if we can instrument drive level into the mixers, sounds like we may be able to have
a clean signal on up to 30mhz.   Perhaps replace transistors to get consistent gain through the 45mhz IF
and the Q90 stage, then monitor the top of RV1 with a diode RF probe into a Nano analog pin.

Better yet, add another 10dB of gain after the mixers so the IRF510's show trouble (much more obvious)
long before the mixers do.  So maybe add an extra gain stage between Q90 and RV1,
existing rigs could easily patch this in with an MMIC. 

If we do decide to go to 70mhz and beyond for that first IF, there are viable filters available.

In this post:
Farhan explains that the 45mhz filter must be narrow enough to reject signals 2 mhz away from center.

Here's the filters on Mouser that are between 70 and 90 mhz, and have a bandwidth of less than 4mhz:|0

The PX1002 from Murata looks good to me, center frequency of 86.85mhz.
25khz wide, 3dB insertion loss, over 60dB of rejection at 1mhz out,
datasheet shows how to use it in 50 ohm environment.
Not exactly cheap at $12 single unit, $6 if buying hundreds from Mouser.
The similar PX1004 at 82.2mhz is harder to get, especially in low quantities,
but might be preferred as it allows a lower vfo frequency. 

To operate at 30mhz with an 86.85mhz first IF, the vfo should be 86.85+30 = 116.85mhz.
The Si5351's internal vco is spec'd to a max of 900mhz, and our fractional output dividers can divide down
to a minimum of 8.0, so 900/8=112.5mhz max using the current si5351bx routines with the vco moved to 900mhz.
Hans, G0UPL, has found that the vco can be pressed to go much higher, beyond 1100mhz,
so we could just cheat on that, perhaps 935mhz for the vco giving 935/8 = 116.875mhz max.
I'd try that first.
The other possibility is to use the second Si5351 internal vco with fractional pll feedback and an
integer output divider on clk1, giving fine grained frequency control up to 200mhz (290mhz, according to Hans).
This second method would roughly double the size of the si5351bx routines, but that's not a major hit.
The other two si5351 outputs would continue to use the first vco, and be restricted to 112.5mhz max.

If using the equations of post
the only needed change to the uBitx code outside the Si5351bx routines would be to change this
uint32_t  f45c  = 44995000;     // center of 45mhz filter
  to this:
uint32_t  f45c  = 86850000;     // center of 87mhz filter

The filter and 50 ohm matching networks could be on a very small PC board
glued to the back of the uBitx main board.
This daughterboard could include a new BiDi amp with appropriate transistors.
If the uBitx mixers are problematic at 86.85mhz, then perhaps a couple ADE-1's also,
replacing everything from T2 to T4 inclusive.  Parts cost of around $10 if building hundreds,
$15 bucks if adding the two ADE-1's. 

Then everybody will get fixated on making the uBitx work on 6 meters
and we get to do this all over again.

Jerry, KE7ER

On Sun, Jun 3, 2018 at 03:44 pm, ajparent1/KB1GMX wrote:
Its not a drop in.  The filter would likely be hard to find and costly or for a one off salvage from 
an old junker commercial radio. 

I think Kenwood radios used a 63mhz filter or maybe ICOM.  so lets do a thought
experiment on what needs to be done.

The filter would need to be matched and like data for it will require experimental testing.
The T30-2 toroids could be reused and rewould and the cap for the L-networks changed
as needed.  The entire string of 3904s [all 6 of them] in the 45mhz section would have
to be some thing like 2n2369, or better BFR106s. 

Then the firmware has to be revised as the first lo needs to be higher to start maybe
65mhz and go up from there.  

So yes it can be done.  I don't have such a part so thats ruled out.  So for a one off
maybe, doesn't help everyone though. and even if the filter was not too expensive 
thats a lot of SMT and through hole work plus a new firmware.  At the production
level it could add significant cost and interrupt the product flow as you have all the
material in place and likely paid for.

Right now the most feasible option is bandpass filters and switching.  Cost wise the
parts are cheap is SMT inductors are used.  The trick there is drop in and play no
manufacturing tuning.


Join to automatically receive all group messages.