On Fri, 30 Jun 2006 23:46:14 GMT, David
wrote:
Thanks all for your input into this issue. I appreciate all the feedback
I have received.
I do like to discover and understand the mechanism for this issue so I
can avoid it next time.
"Sprogging" was probably not the best term to use afterall.
The unwanted response is a clean carrier with no sidebands.
This carrier appears within the tuning band of the amp tuned circuits
and moves as either the input tuning
or Drain tuning is adjusted.
The Source is connected to ground plane.
The circuit now only just starts to produce a very low level signal
(maybe around 60 dB down) when I unplug the input cable from
the signal generator. If I terminate the floating cable with a 50 Ohm
attenuator, the oscillation stops.
The most improvement has been achieved by addition of shielding between
Input circuit and drain circuit.
I suspect that I am getting radiated signal fed back. I also suppose
that as long as the path loss between drain and input is greater than
the gain of the amp then the circuit should not be able to oscillate.
You may try changing the input coil (less L and more C) and see if
that helps.
One other issue I were considering was that even though the circuit is
built with very short leads directly on a continuous ground plane (PCB),
Some "Pads" have been made by cutting away copper. But because the PCB
is double sided, I wonder if the interaction between these pads make the
lower ground plane and pads act like a capacitor (ie. energy coupled
from the pads into the lower ground plane and up into other pads through
capacitive coupling).
Grount the top to the bottom of two sided boards. I use foil or braid
to wrap the edges.
Also add a few holes and put wires from top to bottom near the source
lead.
Next time rather than cutting pads in the board try this, Cut tiny
islands of PCB material and super glue them down. Then you have
isolated pads for SMT parts. For devices like BF998 in SMT I've
used board "chips" that are 0.1x0.2 inch and that works fine.
FYI when I used that device the source has a 100 ohm resistor to
ground and the bypass cap is .001(for 6m) with a ferrite bead
on the non grounded lead. That is a "stock" circuit for DG fets
and it seems well behaved to me.
Maybe if it was single sided board or I add heaps of track pins from top
to bottom layer would prove if this was the case.
Do the latter. If you wrap the edges and put pins through near
grounded compnents that's enough (likely only a few needed).
Another thing is the coils do not have their own shields so they
probably act like antennas. I think it would be prudent to use only
shielded cans as I cannot always ensure large physical distance between
them.
That helps.
Also make sure the output side has a valid load on it too.
Do not load or resistive pad the input as that degrades the NF and
gain.
Allison
Regards
David
Steve N. wrote:
"Eamon Skelton" wrote in message
...
On Fri, 30 Jun 2006 03:36:28 +0000, David wrote:
Thanks for the feedback (no pun intended I appreciate it.
I will add a shield between input and output circuit. I do remember
seeing this in other designs. In my prototype construction I do have the
input and output coils fairly close and in the same orientation so I
imagine there very well could be coupling between them.
I had a similar problem with a 2M RF amp. I cured it by
reversing the winding direction of one of the inductors.
Can anyone suggest a reasonably simple method to determine noise figure
of this stage ?
I'm a bit reluctant to suggest a method that I haven't tried
mysef, but you should be able to calculate NF by measuring the
receiver noise output with a hot and cold resistor connected
to the input. The usual method involves boiling water and
melting ice. http://tinyurl.com/psdfo
73, Ed. EI9GQ.
David,
I don't know your technical level, but here goes. I thought the dual
gate is a type of cascode arrangement and should be stable owing to the
isolation of the "top" device... Anyway...
You are very correct about trying to "band-aid" it with de-Q'ing resistors.
You need to find the root cause, not hope a patch works. You are well aware
of and trying to solve a possible proximity feedback issue, so I won't go
there.
It is not clear what your term "sprogging" may mean. There was a term
"Squeeging" which referred to oscillation or what is also called
regen(regeneration) which is typically used for dirty (sideband-type),
rather than clean (single frequency) oscillation.
Is the oscillation a clean, oscillator signal (pure carrier), or does it
have sidebands?
If sidebands, are they clear and distinct or is it more of a wideband
grunge/garbage?
Are the sidebands mirrored on both sides or are they lopsided?
These are all clues.
If you have regular sidebands or grunge, you have a low frequency
oscillation component and need to also look at bypassing. 33-200 ohms or a
ferrite bead in series with a bypass can help diagnose it.
If the sidebands are unequal, there are most likely both AM and FM
components to the oscillation. Putting a regular scope on the supply or
other places which are supposed to be bypassed might see the signal, helping
to ID the cause.
Be aware that two bypasses in parallel, of different size, can become a
tuned circuit and therefore a high impedance at frequencies where the larger
one is above self resonance and appears inductive.
Pay extreemly close attention to the ground currents for the input and
output circuitry. You want to have an absolute minimum of common path for
these. I didn't see anything in your circuit description about the FET
source. If it is self biased with a resistor & cap combination or what, but
from the source you should have two distinct paths in the ground you have to
the respectice components tith no length of this "ground current path" in
common - where both currents share a common path. you can even cut the PCB
so there are two distinct ground planes which only meet at the FET source or
its resistor/bypass cap.
Watch the power supply side pretty much the same. Where does the Drain
bypass connect to the plane?
Hope this helps.
73, Steve, K9DCI