So ... back from experimenting. What I did:

1.) I connected a very strongl regulated powersupply with thick wires direct to the U3S mainboard. Hmm ... there was no visible change at all. I played with the supply-voltage, reducing the voltage to 4.8V changed the frequency only for 10Hz. But at start of transmission the "chirp" seems to come from "infinity"...

2.) When I put my finger on the crystal I see the frequency jumping for 100Hz or more. It seems to be very temperature-critical. Somewhere I read that it is a good idea to put a heatsink to the crystal to make the temperature-changing lazy. So I soldered a 2 Cent coin direct on the crystal. But again, no influence to the chirp at the beginning of transmission.

3.) Now I tried to measure the 27 MHz Carrier from the crystal. I connected my Perseus antenna-input direct with the 2 Cent coin and I found the crystal oscillating at 27.005 MHz. Then I tuned the Perseus into the wspr window and made a recording with Argo. This was very interesting:


 
During the pause between the transmissions the crystal frequency makes a big jump. But the crystal should oscillate accurate on one fixed frequency! What is the reason for this behaviour ? Currently I have no idea and have to think about it. Possibly somebody else has an idea for the reason ?

However ... I think it`s a good idea to order the oven-controlled version ... not primary for the oven but for the decoupled 27 MHz oscilator.

Correct,

Perhaps if you take the OXCO route as you have documented characteristics for the Si5351 kit at VHF, it would be a useful half way step to repeat the testing with the OCXO
but with the heater unpowered to purely to observe the improvement brought about by alternative Osc circuit.Before employing the heater option.
Alternatively as others have stated the cheap TCXO's generally have a robust output not to be affected by loading changes.
The FOX has being used in the past, if you search the complete forum thread, another device characterised is :-

https://groups.io/g/QRPLabs/topic/28195580#28671
https://ka7oei.blogspot.com/2018/02/better-frequency-stability-for-qrp-labs.html

The page title is for the ProgRock, but in actual fact the improvements are made solely to the Si5351 module and is applicable to all QRP-Labs kits using the Si5351 kit.
It's smaller than the Fox alternative, but may be worth consideration if you can work with the device?


Alan

73, Henning

merci beaucoup Gilles, many thanks Alan,

i knew, ther was something written somewhere....
in the next weeks i'll try different clocks with
my Arduino-based transmitters....
I'll report my experiences....

73 de DC1MAK micha

This is explain in assembly manual of Si5351A of QRPLabs.
In this module, a
27MHz crystal is used. This frequency is chosen because it is possible to
configure the chip to produce the exact 1.46Hz tone spacing for WSPR, on
any amateur radio band from 2200m (136kHz) to 2m (145MHz).

What we do not know is just how sensitive the WSPR decoder is to frequency accuracy.
I use 26MHz TCXOs and have decoded my 2m transmissions with no problem.
There was some discussion a few months ago about using even a 10MHz reference for the Si5351 but nobody thought to try WSPR using that.
WSPR has to cope with borderline HF propagation, TX and RX drift.
Without knowing the tolerances of the WSPR decoder it is wise to keep as close to the specifications as possible but I suspect the tolerances are wider than many think.

73 Alan G4ZFQ

--

Hi Hans, hi all,

somwhere i read about, that the reference clock of the Silab synth in a U3S MUST be at 27 MHz.
Of course this depends on the firmware of the U3S.
But i think, i heard that there is another cause.
Something with the 1.4648 Hz spacing of the wspr tones ?
Does anybody know about it and can explain that
to me ?

Hi,

If i look on the diagram, R36 has one end grounded. So putting the pot on this side SHOULD produce zero output !
The pot is also a logarithmic version and sound level SHOULD follow a smooth curve.

I suspect that you have a defective pot or something is not correct in your wiring. Check, visualy and with an Ohm-meter, if the ground side of the pot is really connected to ground.

73, Patrick.

Hi all,

This afternoon i wanted to bridge the reverse protection diode on my 40 m QCX, to save some power drop.

After doing this and putting the board back into the enclosure, i measured the power and noticed that it dropped from 5 W before mod. to 4 W !

I've checked for any problem that could happen during my intervention, but couldn't find anything...

Can the higher voltage (14 V) destroy something ?  I suppose that all 3 x BS170 are still alive, 2 can't probably deliver 4 W ?

Before i go more deeply into the beast, does anyone have any guess ?

73, Patrick.

Hi Henning,
In that case, what you're seeing is very much like what I saw before I installed the TCXO which changed chirp and drift to discrete frequency jumps.  Eventually, I lifted the TCXO a couple cm off the PCB, which helped reduce my frequency jumps.  That lead me to wonder whether the TCXO was necessary at all.  But I didn't go back to the crystal to find out.  If the chirp is due to the crystal temperature changing, then lifting it a couple cm off the PCB, connecting it with thin strands of wire (enough to conduct the small current but not enough to conduct much heat), and wrapping it and its coin with foam might help.  Using thin wire would reduce the heat coming in from the foil on the PCB and the PCB itself, and wrapping it with foam would reduce the heat coming in from the PA transistor which warms up at key-down.
Before lifting the crystal, you could back off on the PA's bias and see if the chirp magnitude changes.  This might indicate whether heat from the PA transistor is part of the equation,.
Cheers
Halden

Hello Andy,
I`m using parkmode 0. I have only one enabled mode screen so the parking frequency should be the 2m frequency itself?
Yes, I tried transmitting every 4 Minutes but this didn`t changed the "chirp"-behaviour. Later I will make Tests with all park modes.
I`m not sure yet, but I have the feeling that the park-carrier is only replacing the WSPR-transmission but it is off during the seconds between two tranmissions. I will check this detailed ...

73, DF5FH
Henning

Hi Curt, 
App note 4 states use a 25 turns inductor not bifilar wound at 200mw. 
Either set the u3s with a power meter or use the current method of biasing, that should put you somewhere close. Do this before connecting the amp
Andy
--
The universe is made up of Protons, Neutrons, Electrons but contains only one M0RON.

Hello Halden,
sorry, my fault, I scambled the definitions. I used the word "txco" without thinking. I`m talking about the original Si5351 synthesizer  using the kit-supplied 27 MHz crystal. I have no optional "SMD 27MHz TCXO" in use .

Sorry,
73, DF5FH
Henning

Hello Henning,
Looking at your first screenshot and the second it looks as though you are running from one wspr transmission straight into the next. Have you tried leaving a gap between transmissions? Also what park mode and frequency are you using? 
Andy
--
The universe is made up of Protons, Neutrons, Electrons but contains only one M0RON.

Thanks for your reply Curt. Anyone have any idea about setting the power output of on the U3S either before or after connecting the U3S to the amplifier?
Curt M

Hi Henning,
In your first post, you wrote that you were using a txco.  I thought you meant TCXO, so my comments were based on that.  But your recent post refers to a 27 MHz crystal oscillating at 27.005 MHz.  Could you clarify whether you're using a crystal or TCXO in your U3S?
Cheers
Halden

Finally got the 27 MHz oscillator working.  Before proceeding with the assembly, I am a bit concerned by the non symmetrical waveform as shown in the attached .jpg file.  Could this be a biassing problem?
Thanks for any advice.
73 Kevin VK3DAP / ZL2DAP
--
Kevington

Curt

I like your idea of biasing and testing the low power U3S first.

Page 2 and 3 clarify how to route RF to the PA and back to the main board. Read carefully, including details to install the bs170 circuit for this app. Yes the LPF goes between the PA and the antenna.

Enjoy your radiations.

The other Curt M

So ... back from experimenting. What I did:

1.) I connected a very strongl regulated powersupply with thick wires direct to the U3S mainboard. Hmm ... there was no visible change at all. I played with the supply-voltage, reducing the voltage to 4.8V changed the frequency only for 10Hz. But at start of transmission the "chirp" seems to come from "infinity"...

2.) When I put my finger on the crystal I see the frequency jumping for 100Hz or more. It seems to be very temperature-critical. Somewhere I read that it is a good idea to put a heatsink to the crystal to make the temperature-changing lazy. So I soldered a 2 Cent coin direct on the crystal. But again, no influence to the chirp at the beginning of transmission.

3.) Now I tried to measure the 27 MHz Carrier from the crystal. I connected my Perseus antenna-input direct with the 2 Cent coin and I found the crystal oscillating at 27.005 MHz. Then I tuned the Perseus into the wspr window and made a recording with Argo. This was very interesting:


 
During the pause between the transmissions the crystal frequency makes a big jump. But the crystal should oscillate accurate on one fixed frequency! What is the reason for this behaviour ? Currently I have no idea and have to think about it. Possibly somebody else has an idea for the reason ?

However ... I think it`s a good idea to order the oven-controlled version ... not primary for the oven but for the decoupled 27 MHz oscilator.

73, Henning