THUMP THOUGHTS -- PART 2
SECOND CAUSE - THE TRANSMIT OFFSET
During receive, the frequency of X3 is pulled
slightly lower by C21 (47pf). During transmit, D6
conducts and effectively bypasses C21, which raises
the oscillating frequency. That lowers the transmit
frequency (transmit freq = 22MHz VXO - 12MHz
"BFO".)
I believe Glen Leinweber was first (2/3/97) to
blame the "settling" of U3 (the 2nd NE602) for "Mr. T".
He tried changing the timing of the switching in U2
(4066), to no avail. Others tried delaying the audio
switching (Brad {w8ygg 1/31/97}, Jouni Tuhkio
{2/1/97}). Ori told us (1/31/97) the timing circuits were
added to help reduce "Mr. T", but they might need
tuning.
Dave Fifield observed that "the TX offset
circuitry... imposes a nearly 8V transition to one end of
the product detector/TX mixer crystal (X3). This
transition gets through the NE602 and appears on the
ouput at pin 4."
Now, the crystal has a parallel capacitance of only
2.5 pf (I measured it at about 300kHz with my AADE
L/C II meter). That's a pretty high impedance at 600
Hz (100 megohms), and I was doubtful any audio
would get through. So I modeled the circuitry of the
TX offset on the computer, and drove it with a
simulated 8v square wave. It put out nice big spikes of
about 15mv (seen by pin 6 of U3), lots bigger than the
ground loop spike. Is this the cause of "Mr. T"? This
spike should be eliminated by the balanced nature of the
602 mixer. Is it?
Dave Fifield got rid of his thump, but he cheated.
He not only removed the transmit offset circuitry (D6
and 10K current-limiting resistor R9), but he also
bypassed the 4066 switch and C14 altogether! This
may be sensible engineering, but it is lousy science. If
you want to investigate something, only change ONE
variable at a time! It would have been interesting to
know if merely removing D6 would have eliminated
"Mr. T".
TX OFFSET TRANSIENT- MEASUREMENT
I went to my "lab", and breadboarded the U3
oscillator circuit. I looked at the output from pin 4 of
the NE602 via a 0.22uf capacitor. I drove the switch
with a square wave from my function generator. Now
you have to understand that I live 1.2 miles from
THREE huge transmitting towers. I have 5-15 mv of
VHF RF on everything in the shack longer than 1". So
what I see at pin 4 with D6 conducting is about 10mv
of rf junk, mostly 12 MHz leak-through plus about 3-
4mv of FM broadcast. Switching D6 off increases the
rf leak-through to about 15mv (why?), and moves the
baseline by about 3 or 4 mv. That is a big audio
transient at this point in the circuit! Remember, we still
have an audio voltage gain of >1000 to follow.
TRANSMIT OFFSET- POSSIBLE CURES
I see two ways to eliminate the transmit-shift
transient. The first is to use a balanced output for the
audio from U3. This has already been suggested (can't
find the reference in my thump-file, sorry). The second
is to eliminate or reduce the transient itself.
Most designs, including the NC 38S, do not use the
full potential of the NE602. In the 38S, the 602 output
is not balanced. If the internal transistors are not
perfectly matched, some leak-through of the input signals
can be expected. Utilizing the balanced output will
minimize this effect. That will decrease the leak-
through of audio thump caused by the switching voltage
at X3/D6. A balanced output will also give twice the
output power- a 3db gain. However, it is an awkward
fix for an elegant board, involving rf chokes and another
op-amp or an audio transformer. Perhaps I will do it
for another version of the 38S.
I should note that I tried a quickie breadboarding of
a transformer output. The transient is gone, but there
is still a difference in the output. I estimate about 7mv
of 12 MHz rf coming through when the diode is
conducting, and 3mv when it is not. In addition I get
about three times that much FM broadcast stuff, on top
of the 12MHz signal, as I had with the single-ended
output.
The transient could also be reduced by lowering the
voltage on R9, the 10k current-supply resistor for D6.
We would still need to run a milliamp or so through D6
to have it conduct well at 12MHz. We could use an rf
choke of 47 to 100 uh in place of R9, and reduce the
voltage supply at the other end of this choke to one volt
or so. A neat way to do this would be to use an LED
as a pilot light (you'd need a current-limiting series
resistor of about 4.7k), and use the 1.8v drop across the
LED to supply this choke. Now the transient is only
1.2 volts (1.8 - 0.6) instead of 7.4v (8 - 0.6). You
could reduce the transient even more by using a red
LED in place of D6, and a green one for the pilot light.
The green LED "turns on" at a slightly higher voltage
than the red one; presto, a transient of only .2 volts or
so!
Sounds great, but I tried a quickie breadboarding
of it (a "quickie" means it only took me an hour), and
it didn't work. For some reason I get about 30mv of
FM broadcast rf coming out, which almost hides the
5+mv transient which still exists. That was with the
1N914. With a red LED for D6, it looks even worse,
with 40-50mv of rf and a 10mv transient. The red and
green blinking LEDs are kinda cute, though (blinking
when I slow the function generator down- to view it on
the scope conveniently I was running it at about 2KHz,
so the LEDs just glowed at half brightness).
TRANSMIT OFFSET - CONCLUSIONS
Clearly there is a big transient produced by the transmit
offset circuitry. Is that the cause of "Mr. T"? On
3/19/97 Jim Kortge reported that he had "bent pins 9
and 10 away from the chip body, and put it (U2, the
4066 switch) back into its socket." His thump remained,
despite a lack of input from U3. He did not try bending
out pin 11 of U2 (output side of the switch), and he
didn't say what the status of his C14 ground pin was
(stock, or moved to be nearer U5, the audio chip.)
I think the transient at D6 produces part of "Mr. T", but
not all of him. Balanced audio output would eliminate
this cause of "Mr. T".
****************************************
Stay tuned for part 3, and some possible solutions.
Dan Winkler N7IVR Seattle WA
< GDWinkler@compuserve.com >
< DWink@Juno.com >
[
QRP-L Archive |
]
[
1993 |
1994 |
1995 |
1996 |
1997 |
1998 |
1999 |
2000
]