Hi folks. Here's a problem I've encountered with amateur radio's most
complex amp. Hopefully, I can resolve this tomorrow with some nominal test
voltages from Alpha Power, but I thought perhaps the experts would like to
speculate a bit (not that they ever do that...<g>)
I've had my 87A for about 4 years and it's always done a superb job. I know
there's controversy on this, but I think it's the best amp made today. I've
had a few minor problems with it, but have slowly managed to track down
explanations and/or corrections for them (found the plate choke relay
firmware bug, discovered that input coax length really makes a difference to
input SWR and discovered that my TS950SDX doesn't have the best ALC in the
world.) Here's a problem that's been bugging me for a while:
A couple of years ago, the lowest LED segment on the plate current meter
began to flicker intermittently. At first it was so infrequent that I
wasn't really sure it was happening -- I would see a flash in my peripheral
vision but never caught the flicker itself. Over time, the frequency of the
flicker slowly increased, until I could finally confirm that it was actually
happening. It began as soon as the amp was turned on, even before the warmup
period had completed. After warmup, the LED flickered whether the amp was in
standby or operate mode. As still more time passed, the flicker became
increasingly more frequent and increasingly more annoying. The amp was
working fine, so I speculated that it might be the reference voltage
adjustment on the LM3419 LED bar/dot drivers used to run the display
(there's a mini pot for this, but it's got a drop of nail polish on it.)
I happened to mention the problem to Carl at Alpha Power when calling about
something else, and he had me run some voltage checks on the D/A converter
chip used to drive the display. We found nothing wrong with the chip, so he
speculated that perhaps I had a leaky tube and that the plate current
indication was real. But he said the leakage would probably be accompanied
by the tube getting soft, and I had noticed no such problem (still 35-50W
for 1500W output, with plenty left over.)
Now the LED is on steady. I figure I'd better start treating this more
seriously in case it really is a leaky tube or some gradually deteriorating
component that might do some damage when it fails.
I've been conducting a series of measurements and tests. Here are the
results:
1. The CPU and the bar graph agree on the thee amount of plate current while
in receive (standby or operate). The RS232 interface says it's .070A, and I
compute each LED segment at about 75 mA. Certainly, this is too much plate
current for receive mode.
2. I tried running with one tube, then the other. As I expected, the amp
tripped out with the "filament open" status, but the first LED was still lit
and the CPU still reported 70 mA plate current. The reading was the same
regardless of which tube was in the amp.
3. I found a better way to test for tube leakage -- I disconnected the HV
lead from the HV power supply board, completely disconnecting the plates and
all associated circuitry from the HV supply. This time the amp didn't trip
out, but the first LED segment was still lit and the CPU still reported 70
mA plate current.
4. With the HV still disconnected, I measured the voltage drop across the 10
ohm shunt resistor, the same place that the 87A gets its plate current
reading (some things never change...) Here, I found a drop of about 15 mV,
which I believe translates to about 1.5 mA. I can't quite figure out why any
current is flowing in the resistor at all with the plates disconnected, but
perhaps it has something to do with how the cathode bias supply is arranged.
Maybe somebody can explain this to me.
5. Just to make absolutely sure it wasn't related to the tubes, I
disconnected the cathode bias voltage at the HV power supply board and got
exactly the same 70 mA Ip reading from the LED meter and CPU. I don't think
I have a leaky tube.
6. With the HV reconnected to the plates, I measured the voltage drop across
the shunt resistor at about 30 mV, indicating a plate current of about 3 mA.
Deducting the 1.5 mA that was already there, perhaps the tubes are each
leaking about 750 uA. Rich says it should be 500 uA max, so it's high, but
this is a lot closer to nominal than the 35 mA per tube the LED Ip meter and
CPU would have me believe!
7. FYI, cathode bias is about 33 volts in standby and 8.8-9.1 volts in
transmit (0 to 1000W out.) The cathode bias is electronically switched, from
a key-down signal traceable back to the CPU. This signal cuts off a
transistor that connects the cathode bias directly to the 36V supply. With
the transistor switch turned off, the cathode bias is connected to the 36V
supply through a resistor, which I believe acts as a current limiting
resistor for a 8.2V zener diode, the cathode of which is connected to the
cathode bias. The anode of the zener diode connects to a 30V zener which is
connected to ground. The CPU transmit signal goes through some
transformations and eventually becomes the UNBLANK signal, which is attached
to the junction of the two zeners. When UNBLANK goes low, the 8.2V zener
kicks in and sets the transmit cathode bias. At least, this is how I
interpret the schematic. There's no circuit description and I'm not a
professional electronics engineer.
8. The meter circuit is composed of an op amp in what I believe is a
differential amplifier configuration, with the inputs connected across the
shunt resistor through some 1M resistors. There are a bunch of 1M and 10K
resistors on the input side, but I don't know enough about op amps to know
their function. The output from the op amp goes through a resistor divider
pair and out to the CPU's monitoring circuits. There's a 5.1V zener
connected across the output and ground, but I think that's just for
overvoltage protection. The output from the meter runs through an analog
multiplexer, then an A/D converter, and is fed to the CPU. It looks like the
CPU's digital version is then converted back to analog by a D/A converter
and fed to the LM3914 bar/dot drivers for the meter. Kinda complicated, huh?
9. The output from the op amp is kinda suspicious. Here are some
measurements for different operating states and power levels:
STATE INPUT OUTPUT Ip (CPU) LED seg Shut R op amp
STBY 0W 0W .07A 1 30mV 95mV
OPER 0 0 .07 1 33 94
KEY 0 0 .11 1 680 138
KEY <1 5 .12 1 1.00V 162
KEY <1 14 .15 2 1.20 199
KEY 1 56 .23 3 2.05 304
KEY 4 163 .36 4 3.29 457
KEY 5 204 .41 5 3.69 508
KEY 7 274 .46 6 4.28 581
KEY 11 412 .55 7 5.17 691
KEY 15 526 .62 8 5.91 784
KEY 27 975 .84 11 8.11 1.06V
The key-down readings for Ip are always about 20-40mA above the shunt
resistor readings. Some of the discrepancy could be due to error inherent in
the A/D and D/A conversions, plus the amount of time the amp had to
stabilize after each power increase (the CPU's Ip readings bounce around a
bit.) The op amp output voltage clearly follows plate current, although I
can't tell what the relationship is supposed to be. The delta between any
two readings is about the same as the delta between any two Ip readings,
plus anywhere from about 1 to 25. All in all, it looks like the op amp
voltage is always about 40 or so counts above the current measured through
the shunt resistor. That would only account for less than half the idle
voltage being put out by the op amp when the 87A is in standby. Then again,
maybe I don't quite have the formula for the A/D plus D/A conversion.
The op amp supply voltages are taken from the same 36V supply used for the
cathode bias. This supply originates on the low voltage board as +36V
and -37.4V. Both voltages are dropped through a 1K resistor before getting
to the op amp supply pins. In standby, the op amp gets +34V and -35.4V. In
operate, the op amp gets +30.5V and -35V. Key down, the op amp gets +29V
and -32V. I'm a little suspicious about the voltage discrepancy between the
plus and minus supply rails, but remember that the op amp's output seems
pretty high even in in standby mode, where the voltage difference isn't as
great. Again, I don't know squat about op amps -- is that discrepancy
between the plus and minus supply rails important to a differential
amplifier? Each supply pin is has a 10uf electrolytic shunt to ground. I
thought maybe one of them was leaking, but they measure OK on a capacitance
meter (in circuit, though, and not at operating voltage.)
Bad op amp? Bad zener diode? Power supply problem? Bad caps? Any ideas?
73, Dick, WC1M
--
FAQ on WWW: http://www.contesting.com/ampfaq.html
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
Search: http://www.contesting.com/km9p/search.htm
|