THUMP THOUGHTS --- PART ONE
with a tutorial
DISCLAIMER/QUALIFICATIONS
See Appendix
DEFINITION
The "Thump" is a "feature" of the NorCal 38 Special transceiver.
It is "a rather large, exponentially decreasing envelope of 500 Hz
sine waves" (Ron Reder, 3/11/97), heard as an audio "thump" on
key-up. There is no associated RF distortion of the transmitted
signal. It was noted in the original prototypes (Dave Fifield,
2/11/97). Altering the timing of the TR switching helped reduce the
thump, but didn't eliminate it. I call the thump "Mr. T."
VARIABILITY / MULTIPLE CAUSES
Not all 38S's have thumps. Some thumps are not "Mr. T."
Paul Hardin had "a woozie, anemic, psychopathic start on key
down", which he cured. He apparently did not have "Mr. T", who is
a key-UP critter. Bill Jones had "a tiny little bit of a thump" (kd7s,
1/30/97). Thump cures on some rigs don't cure thumps on other
rigs (see below). There may be more than one thump. "Mr. T" may
be a composite of several thumps.
FIRST CAUSE - THE GROUND LOOP
Ron Reder noticed that "the ground connection for C14 goes to
the ground plane that serves U4." He reasoned that "this chip
conducts a lot of current. That causes the voltage on that ground
plane to rise on transmit and drop on receive relative to the ground
point for C38, the audio amplifier reference voltage. This causes the
thump when the transmitter is unkeyed." (Reder 3/11/97)
I was skeptical of this explanation until I did a BOTE (back of the
envelope) calculation (see appendix), which led me to think I might
expect as much as 200 uv of signal at the ground connection of C14
due to the voltage drop across the ground path common to both C14
and U4.
However, subsequent posts refuted this theory (Dave Fifield,
3/18/97)(Jim Kortge 3/19/97)(wb8ygg 3/19/97)(Reder 3/19/97).
GROUND LOOP - MEASUREMENT
On the basis of my calculations I was interested in examining
this theory further. So I measured it. This was easy, as I have not
yet built up my board.
I connected the (-) terminal of a variable power supply to the (-)
input on the board at connector hole 3. I connected the (+) lead in
series with a curent-limiting resistor to pin 10 (ground) of U4. Then I
connected my DVM between the ground connections for C14 and
C38. I adjusted the power supply for 500ma current. DVM reads
1.6mv. That's MILLIVOLTS ! Now, U4 only draws about 75ma
extra on transmit (BOTE calculation: 300 mw out @ 50% efficiency
= 600mw input; 600mw from 8volt supply = 600/8 = 75ma). So the
expected shift in C14's ground reference would be 75/500 * 1.6 =
..24mv = 240uv.
The audio amplifier chain (all 2 links of it!) has a voltage gain of
2000 at 580 Hz, according to my computer model of it. So an input
of 240uv will produce 480mv in my earphones. Since a peak-to-
peak signal of 50mv is comfortable listening, and 200mv p-p is
almost painful, I can imagine a thump with initial amplitude of
480mv might be quite noticable. (See Note 1)
GROUND LOOP - CONCLUSION
The ground loop may be a significant contributer to "Mr. T."
Certainly the ground connection for C14 needs to be brought closer
to the reference ground for the audio amplifier.
*************************************************************************
GROUND LOOP - TUTORIAL
This is the best example of that infamous and nefarious badger,
the ground loop, that I have seen since I first read of them 30 years
ago. Do you Novice/Techs understand what is happening? C14 is
the first low-pass filter element in the audio signal path. It is
supposed to short higher frequencies (>1000 Hz, up through RF) to
ground. It follows the product detector U3, and preceeds the audio
amplifier sections U5 A and B.
Unfortunately, the trace from C14's ground end to the audio
reference ground near U5 is long (about 4 inches), and the ground
path for the power supplied to U4 shares part of that long path from
C14 to U5. Current flowing along the ground to U4 raises the
"ground" end of C14 by 240uv during transmit. On both key-down
and key-up there will be a 240uv transient coupled to the input of the
audio chain by C14. That transient will make the active audio filter
(U5B) ring like a bell. (See note 1, below)
Note that I said ground reference point NEAR U5. The audio
amp ground reference is really the (+) input to both of the audio amp
sections, U5 A and B. That point is held at one half the supply
voltage by the resistor divider pair, R25 and R26 (30K each). This
reference point is shorted to ground (but see note 2, below) at audio
frequencies by C38, a 22uf capacitor, which has a very low
reactance (72 ohms at 100 Hz.) compared to the 30K divider
resistors, or to the >100K inputs to the op amps in U5.
All you new folks, take a look at the schematic. See all those
"ground" symbols ( /// with a line on top). They are all supposed to
be the ultimate reference for the circuit. They are all supposed to
be equivalent. There isn't supposed to be any difference between
them. But in the real world, they can mess you up. Some are
more equal than others.
I hope this gives you some insight into one of the pitfalls in
building equipment.
---------------------------------------------------------------------------
-----------------
Note 1 Actually the filter action of the audio chain will keep it from
responding with the full x2000 gain. My software shows a nice
10ms long (6+ cycles) "thump" with the initial amplitude 300x that of
the exciting transient. That is still 72 mv pk-pk, quite loud in my
earphones.
Note 2 Glen Leinweber has a mod (2/12/97) for reducing audio
noise caused by the voltage regulator. He AC "grounds" the
junction of R25 and R26 to the +8 volt line, that is, C38 is placed in
parallel across R25 instead of R26. This lets the excellent noise
reduction circuitry in U5 work on the low frequency audio noise
riding on the +8 volt power supply line. Note that the AC ground
path from the input of the audio chain back to the ground of U3 (the
NE602 product detector) must then go through C40 (in the middle of
the board). Aren't grounds fun?!
---------------------------------------------------------------------------
-----------------
APPENDIX
BOTE CALCULATION OF GROUND LOOP
Look at the 38S board. C14 is about in the middle, between U4
above, and U2 below (board viewed so that silk-screening reads
properly, with hole connections 1-10 on the top edge of the board,
and 11-20 on the bottom edge). The right hand lead of C14 is the
"ground" end. It makes a 10mm run up to pin 10 of U4. Past U4
(and underneath it, so you can't see this if your board is already built
up) the trace is 2mm wide for about 5mm before it widens to 4mm
while making a right angle turn to the right. In this area, just to the
right of U4, the ground on the top of the board makes contact with
the ground on the bottom via 5 plated through holes. C26, R13, and
C24 all have their top (from board orientation ppoint of view) leads
grounded through three of these plated through holes. Turn the
board over and you can see these 5 plated through holes in a little
square of ground trace.
Ok. I only worried about that 2mm wide by 5mm long section.
I guessed the trace was about as thick as a 36 gauge wire. #36
wire yields 175 turns per inch. The trace is 2mm = 2/25" =.08";
..08*175 = 14 wires wide. Resistance of #36 is 423 ohms / 1000 ft.
(wire data are from the handbook), or about 0.035 ohms/inch.
Trace is then about 0.035/14 (trace is about like 14 wires in parallel)
= 0.0025 ohms/inch. Trace is 5mm long = 3/25 = .12". Resistance
of trace is about 0.12 x .0025 = .0003 ohms. TRIVIAL, you say.
But I estimated 100ma current draw by U4, so that's a voltage drop
of E=IR = (.0003)(.1) = 30uv. Funny. Somehow I came up with
200 the first time. I wonder if I would have bothered to measure it if
I had only guessed 30uv?
DISCLAIMER / QUALIFICATIONS
I am not an EE. I am entirely self-taught, primarily from the main
ham books and magazines, and from futzing about on and off for 35
years. I have had no elmer and no formal classes. My knowledge
is a little spotty, but my logic, in general, is good.
I rarely operate. My code is wretched. Designing, building and
modifying circuits are my favorite play. However, I have not yet built
my 38S; I have been spending too much time on the list!
I enjoy teaching, for it forces me to think more clearly about
problems. I sometimes use extreme examples to illustrate a point.
Until last week I used Juno for my email. Juno cheerfully
reformatted my postings, deleting carriage returns, extra spaces and
tabs with gay abandon, eliciting at least one flame from a reader.
My Juno crashed last Sunday, taking with it my first version of this
epistole. I have no confidence that Compuserve will preserve my
original formating any better than Juno did.
If my writing offends you, please just delete it.
There, that ought to lower the flamability of these posts just a
smidge. My monitor is still smoking from a few responses to some
of my previous posts.
72 Dan Winkler N7IVR Seattle, WA
< GDWinkler@compuserve.com >
[
QRP-L Archive |
]
[
1993 |
1994 |
1995 |
1996 |
1997 |
1998 |
1999 |
2000
]