Wavetek Signal Generator
This equipment provides
coverage from 10KHz to 550MHz via step up and step down push
buttons or direct key entry.
It was relatively cheap to buy
(under £200 in 2002 when it was 11 years old) considering
the build quality and performance.
Below is a key to the controls.
It's quite user friendly but it needed a handbook to let me know
most of the functions. When the power is turned on (button 1)
the generator runs through a self test routine which gets more
worrying as the equipment ages. I've had two problems so far.
One, immediately after buying it and turning it on for the first
time, which turned out to be a short-circuit protection diode
and the second quite recently which was again the loss of an
internal power rail. This was easily fixed although the self
test gave a strange fault report which I imagine was because
the guys writing the test firmware must have imagined the power
supplies would never fail?
The two most frequently used functions
are setting the frequency and setting its level.
Direct entry via the keypad "3"
is the most useful way of doing either but the up and down arrows
"4" can make incremental changes, for example to determine
the tuning peak of a circuit.
For example: pressing Freq, 465 then
KHz will set the generator to 465KHz. Pressing Lvl, 5 then uV
will set the output to 5 microVolts. Once this has been done
you can set the output to CW "Mod Off", AM, or FM setting
the modulation depth as required. The button marked "5"
switches the output on or off and "Rate" sets the modulation
to either 400Hz or 1KHz.
All perfectly logical. Also, there's
a menu feature which lets you set global parameters such as the
A store feature lets you keep routinely
used output settings.
Although synthesised generators
are jolly stable and accurate they are not always as useful as
an old fashioned analogue equipment when aligning a radio.
In the case where one needs
to tune a generator to a radio for example an analogue type is
just twiddled until a signal appears but a synthesised type often
is tricky to use. I've often listened for clicking noises which
is the generator locking to its frequency and increasing or decreasing
the settings until the clicking gets louder and louder. Once
you're hearing fairly loud clicks you can increase the frequency
resolution until the clicking changes to a signal, a bit messy.
I was testing my DST100 receiver the Wavetech broke down. It
took me a little time to discover the fact because the receiver
has a few intermittent faults. When AM was selected the received
signal kept cutting out for brief moments before re-appearing
almost instantly, but connecting the generator to an oscilloscope
showed it was the Wavetech not the receiver. RF output was OK
but switching on modulation resulted in an intermittent RF signal.
The manual points to modulation problems being in a specific
area, that of the AM/DIV module so that's where I started. After
removing two dozen screws the top cover can be detached revealing
at the side of the unit the view below.
The AM/DIV module has two circuit
boards mounted in an aluminium sleeve which can be removed after
unplugging various RF plumbing, including one at the rear of
the unit connecting to a power amplifier module. At each end
of the sleeve you need to remove the end plates by taking out
their four fixing screws plus unscrewing the nuts securing the
Below, pictures of the circuit
board carrying a lot of the final RF circuitry. Modulation is
carried out using a mathematical process controlled by one of
the units two microprocessors, but as with all the circuit boards
in the signal generator local power supplies are fitted in order
to provide accurate voltages. The top view does have a clue to
the fault I'm investigating but the underside view reveals clearly
Here's the fault: three
dry joints at the pins of a power regulator transistor Q905 an
MJE253. Although the designers included a hole for fitting a
heatsink, this wasn't fitted so after replacing the solder I
fitted a small brass bush which should help keep the transistor
temperature down a little. In fact, as I tracked down faults
later on I found the overheating was due to chemistry. Everyone
knows about Mullard AF117 problems but I for one hadn't heard
about Motorola transistor problems. Read on....
Back in business after
the Wavetech had gone successfully through its diagnostic tests.
After using the Wavetech on and off for a week or so I noticed
a sizzling noise on the output. I looked again at the repaired
output board and found another dry joint and a discoloured resistor.
I resoldered the joint and fitted a new resistor. After assembly
the sizzling noise was still present. The equipment cannot be
tested by usual means so I decided to detach all the circuit
boards and apply 18 volts to each. This is the standard supply
to all the boards and on each one a 15 volt regulator provided
on-board power. Starting with the output board I noticed the
new 68 ohm resistor I'd fitted was very dark in colour and sure
enough, powering the board showed the resistor was getting very
warm. The 18 volt current was over 400mA and across the 68 ohm
resistor I measured about 11 volts. This represents a power dissipation
of about 1.75watts although the handbook states it's a half watt
component so something is wrong. The chief feed from the resistor
is the collector of an MRF571 RF transistor and this measured
as a pair of diodes on my transistor tester. I substituted a
BFR90 for the MRF571 and the resistor ran cool with 11 volts
on the transistor collector intead of the previous 4 volts, but
after reassembling everything there was still a sizzling noise
on the RF output. Maybe the transistor had failed due to a problem
with its components and this is still giving trouble?
A little more about this
saga. Having diagnosed a faulty drive transistor I eventually
decided to splash out and buy a proper MRF571 (see above) because
the output wasn't quite right. It was difficult to put one's
finger on the problem, but using a scope I determined the RF
was jumping around amplitude-wise and frequency-wise. I found
a source of the rare device and ordered it. In fact it came as
a set of six transistors from Poland although I'll probably never
use the other five. After dismantling the assemblies I fitted
the new transistor and the RF problem seemed to have cleared
up... or had it?
Above right to left...driver
transistor Q900, with Q901 bias set for Q900, Q905 current regulator
for Q902, and RF output Q902. Q903 sets the bias voltage for
Q902. Below Q905 (centre) is a voltage reference diode LM336Z-2.5.
The MRF839F is a variant of an MRF839
fitted to a Type 319-07 base. This is a view of the underside
showing the heatsink.
After replacing the temporary
BFR90 with a new MRF571 it turned
out the problem was still present albeit not as bad as before.
Why was it that sometimes the RF was as clean as a whistle and
at other times infuratingly crackly? By experimenting I discovered
that any RF output greater than 137.5Mz was faultless but below
that I found the RF was sort of wobbly. Time to re-read the repair
manual and study the circuit diagrams. This explained the reason
for the transition frequency. Above the frequency of 137.5MHz
the RF amplifiers, feeding the output socket via a programmable
attenuator, are driven directly from a VFO but below that a mixer
is involved. This mixer combines a second RF signal of precisely
512MHz with the VFO to derive any frequency from 10KHz to 137.5MHz.
This being so, any problem in the mixer circuit or in the generation
of the fixed frequency of 512MHz could be responsible for the
As the board carrying the mixer
was already on the bench I checked this first. Looking at the
mixer circuitry (below with shield removed) showed that it was
possible for a faulty component in the selection process to result
in a bad RF output from the mixer. For example a bad capacitor
in the filtered connections from the microprocessor outputs might
be responsible. For example if the mixer was being turned on/off
by noise on the selector line this might show up as a crackle
on the mixer output. With difficulty I removed a metal cover
from the mixer circuitry but everything seemed to be OK.
Time to study the repair manual
circuits once more. These are poorly scanned in my copy making
it a time consuming job to identify the parts (not to mention
the use of a crossed zero making a nought look like an eight).
Before tackling the 512MHz oscillator which is on another circuit
board I looked at the RF amplifier transistors. I'd already replaced
the driver transistor because it had a leak of a couple of hundred
ohms between its collector and base so I looked at the output
device. This is Q902, an MRF839F fitted on a heatsink. It has
an odd-looking base listed as Type 319-07 and I'd noticed several
years ago when I had a problem with crackly RF output that the
bias circuitry had been running very hot leaving the Q905 power
transistor dry-jointed. I'd resoldered it and for a short time
all seemed well, but why had this happened?
The designers of the equipment
surely would not make the basic error of not fitting a heatsink
to the transistor? Maybe this overheating is related to my current
RF instability problem? My circuit schematic is pretty poor but
I eventually reproduced the circuit on a sheet of paper. The
manual tells me the bias circuit performs two functions... to
set the operating point for the MRF839F and to set its collector
I applied an 18 volt supply
to the board and measured the various points in the circuit.
The RF power transistor collector was sitting at 5.5 volts and
I couldn't figure out why, so began to unsolder components. All
the capacitors were fine and all the resistors were OK. Maybe
the bias setting transistor was duff? I lifted off its connections
to base and emitter and found absolutely no change... very odd.
At this point the penny dropped. The MRF839F was faulty. With
some difficulty I removed the transistor and discovered it had
the same fault as the driver transistor. There was a leak between
the base and collector. It measured 284 ohms and this resistance
didn't look like a diode because its value was exactly the same
in both directions when checked with an ohm-meter. The 284 ohm
leak was turning on the transistor through self-biasing and pulling
down its collector voltage. Q902's resulting collector current
of around 300mA through Q905 together with the latter's increased
emitter-collector voltage was resulting in a dissipation in Q905
of around 3 watts and because there isn't a heatsink Q905 was
getting really hot.
Even though Q902 had a bad leak,
the balancing circuitry in the Wavetek was managing to maintain
the correct RF output to the attenuator, but was the leak responsible
for the crackle in the RF. Perhaps the output from the mixer
was lower than that from the straight-through VFO thus pushing
up the demands on Q902?
Before proceeding further, I
fitted a 2N3866 in place of the MRF839F to test the theory. The
collector voltage correctly measured as 12 volts but that type
of transistor doesn't fit easily due to space considerations
in the metal case for the circuit board and a 2N3866 is anyway
not capable of running enough power output.
A search of the Internet revealed
the MNRF839F was a pretty rare beast with only two far-east suppliers
with stock and a UK source quoting over £55 plus postage
and VAT. I commenced the ordering process for one of the cheaper
options but found that delivery was quoted at 20 to 40 days by
airmail. Maybe it's pigeon post? I didn't place the order as
I would not be home at the delivery date and I didn't fancy supplying
credit card information to a Chinese company.
Vce 16V, Vcb 36V, Veb 3.5V, Ic 0.6A
Dissipation Pd 10W @ 110 deg C
RF out 3W 806 to 960MHz
This and its stablemate the MRF371 both
degraded during their lifetimes by developing a leak between
base and collector. Maybe this is due to tin-whiskers which grow
from the material inside the case over a long period eventually
Back in the workshop I
carefully removed the MRF839F and checked it with a transistor
tester... it read "two diodes". Now.. many years ago
back in the 1970s I had a really expensive RF power transistor
with a similar fault. Our Plessey rep (also a radio ham) had
given me a sample of their latest VHF power transistor... a huge
device worth a huge sum of money. I made a 2m linear amplifier
from it but during experiments I'd killed it. To cut a long story
short I mended it.
Can I repair my MRF839F? As
far as measurements are concerned there's a fixed value resistor
between the base and collector. This had been the exact same
fault with my exotic free sample. My reasoning was that being
an NPN device I could connect a negative voltage between base
and collector and fuse open the short because the device can
tolerate quite a high Vcb without damage. I set up my power supply
and connected the positive output to the collector pin and the
negative supply to the base... set the current limit to 50mA
and gradually increased the supply voltage. All that happened
was an increasing current then constant drain of 49mA from 13
to 18 volts. I adjusted the current limit to 500mA and tried
again. This time, once past about 12 volts the current dropped
then rose again then as I increased the voltage it suddenly dropped
Maybe it worked or maybe I'd
fused a connection to an electrode? I connected my transistor
tester and it said "NPN transistor", "gain 17"...
so I removed the 2N3866 and soldered back the MRF839F. I applied
the 18 volt supply and the collector voltage measured 12 volts.
Feeling fairly confident I reassembled
the Wavetek and turned it on. After a major repair job it's very
comforting to see the self-diagnosis to report all is well. I
turned on my monitor receiver (Icom IC7000) and set the Wavetek
to 137.5MHz. Perfect.. a rock solid carrier at 137.500MHz.
detuned the Wavetek to 137.4MHz and tuned the receiver... drat..
the rock steady carrier was now warbling. Turning from SSB to
AM, I could hear a sort of crackly sound so the duff transistor
wasn't the reason for the fault after all. I suppose the Wavetek
checks the RF level and cranks up the internal RF levels to compensate
for a ropey amplifier transistor so even with a failing part
the output is guaranteed.. obviously up to a point.
The noise on the received signal
seems slightly less than before, but then again there was never
any noise above 137.5MHz, but I'd assumed that the mixing process
increased the stress on the MRF879F maybe because of reduced
RF input level, but no.. the problem must be in the circuitry
associated with the incoming heterodyne frequency (the local
oscillator feeding the mixer)? Checking the manual, I see this
oscillator runs on a fixed frequency of 512MHz and is derived
on the LO/REF board so that's next on the list...
At least the much more complex
VFO RF signal is OK because anything above 137.5MHz, where this
is used without the 512MHz local oscillator, is rock steady.
The local oscillator is not
just a simple crystal oscillator multiplied up because this might
result in lots of spurious images (harmonics etc), but instead
a 512MHz phase-locked loop oscillator whose output is divided
down and referenced to an internal stable source of 10MHz. However,
being a phase-locked loop means any solder joint or dodgy component
in the loop (or its power supply) might result in an intermittent
lock problem. The scratchy crackly noise when listening on AM
suggests a dry solder joint or perhaps a bad capacitor. I might
add that the problem isn't severe enough to result in loss of
lock for long enough for this to be detected, but of course it's
not nice to hear a scratchy crackly test signal when aligning
There seems to be a couple of
options available before removing the circuit board. First I
can unscrew the RF cable carrying the 512MHz local oscillator
and see if what emerges is crackly and secondly check the internal
reference output which is 10MHz and see if that is crackly. So
Well I pulled down the Wavetek
from its shelf and before pulling the LO/REF board, I first checked
the 10MHz reference signal which is supplied via a BNC connector
on the rear panel and found it was rock solid (in fact it is
derived from a 10MHz crystal although it might have been a bad
component in its circuitry) so next I'll check the LO signal
itself. I unscrewed the coaxial copper link carrying the LO signal
and decided first to inject a 512MHz signal from my other decent
signal generator into the mixer board. I turned on my Marconi
TF2008 but I noticed the dial was marked up to only 510MHz or
2MHz short (Sod's law), so instead I set the TF2008 initially
to 170.666MHz and then to 256MHz. Both resulted in a cleanish
512MHz AM signal on my Icom monitor receiver although I suspect
this receiver similarly uses 512MHz in its circuitry so although
I got a clash and some heterodynes the RF output signal was crackle-free.
This then clears the output board with its mixer and points definitely
to a noisy 512MHz local oscillator. I connected the 512MHz output
from the LO/REF board to my monitor receiver and heard similar
heterodynes as before, but I could also hear the dreaded background
crackling so I switched off, unplugged the module and extracted
the LO/REF board shown below. This is neatly laid out but a quick
examination revealed no obvious dry joints or burnt parts.
Above, the LO/REF board:
Top right is the 512MHz RF output and below this the reference
oscillator circuitry. On the left of the board are the phase
lock loop logic chips with some microprocessor-driven control
circuitry. Below these chips are the three on-board power supply
regulators. Turning the board over reveals a large selection
of surface-mounted parts, some quite exotic but including a pair
of tantalum chip electrolytics.
I noticed some discolouration
around the RF output socket.. shown below. Soldering can often
look messy when seen close-up but dry joints, apart from really
bad ones, can only be seen with a strong magnifier. That discolouration
around the 512MHz output socket looks like a candidate for noise
if it's conductive.... but nothing looked unusual when I checked
with an ohm meter so I just cleaned it off.. another red herring.
The next step was to power
the board and see if there were any clues to crackling. I connected
three power supplies, 18V, 8V and -18V. which are stabilized
by three regulator chips to 15V, 5V and -15V. Each supply line
drew a nominal amount and some parts were warm but not hot. I
looked for the phase locked loop and found this on the underside
of the board in the form of a number of tiny chip transistors
of which the oscillator is marked "R33", a microwave
oscillator transistor type NE68133-T1B rated in the upper GHz
plus two others (BSS88C) which are used for biasing and switching
the oscillator on and off. The R33 uses striplines formed by
the printed circuit and various chip components plus a pair of
10uF 25V electrolytics paralleled with low inductance chip capacitors,
presumably for bypassing local noise on the +/- 15 volt power
The 512MHz phase-locked oscillator transistor
NE68133-T1B which is equivalent to a
Pt dissipation 200mW
Marked as "R33"
Listening on 512MHz indicated
the oscillator was turned off, but by adding a 1K resistor connecting
the OSC on/off switch line going to pin 13 of IC604 (a 74HC574N
latch) to 5 volts I was able to turn it on. Again, I had a problem
because the monitor receiver uses an internal 512MHz oscillator,
but at least I was able to detect that the oscillator could be
turned on and off and, more importantly, I could hear that the
signal wasn't too clean so I looked for anything that would give
me a clue to the crackly output.
I tried an oscilloscope at various
points but its bandwidth was miles too poor to see 512MHz although
I could see some intermittency as the trace kept intermittently
bobbing around. Before I fired up my spectrum analyser which
goes up to 1.5GHz I tried my multimeter turned to volts. Nothing
much except I found both emitter and collector connections to
the R33 (at the marked ends of two 10uF capacitors) each intermittently
varied by up to ten millivolts. Measured volts would jump around
say 3.72V.. 3.68V...3.81V.. etc so I tried moving outwards to
the other ends of the emitter and collector feed resistors (R626
and R618) and found these voltages were rock solid. Dabbing R33
with my finger produced much the same effects only worse as the
circuit attempted to maintain lock. During testing the emitter
voltage was about minus 4 volts and the collector about plus
9 volts although the two capacitors, C656 and C649 (the other
ends of which are grounded) were each connected the same way
round. This is a bit odd. As the capacitors are surface-mounted
tantalum types they carry a stripe to indicate polarity (similar
to that on a surface-mounted diode cathode) however, the capacitor
stripe marks the positive connection, so clearly, although C649
is OK, C656 is back-to-front.
Tantalum capacitors are marked
with a stripe for their anode connection so this has to be connected
to a positive voltage.
I measured the capacitor with an ESR
tester which gave 9.4uF with a resistance of around 0.5 ohm although
this figure kept changing by up to 0.2 ohms or so each time I
Next I connected a variable current-limited
voltage across it monitored by my multi-meter switched to mA.
With the striped end positive and a voltage of 1 volt the current
through it read at the limit set by the power supply which is
wrong for a good capacitor. As the voltage and current limit
were increased the capacitor failed (changing to a resistor)
at around 5 volts. Rechecking on the ESR meter showed the capacitor
was now defunct.
This example (C656 on the schematic)
degraded because it was decoupling a negative voltage connected
to its anode.
Checking with a multimeter, with
the oscillator off its emitter is some 12 volts negative and
when on is about 4 volts negative. This means that when the output
frequency is greater than 137.5MHz, although current limited
to an extremely low value, C656 carries 12 volts with the supply
negative connected to the capacitor anode. I removed C656 and
fitted a 10uF electrolytic. The emitter and collector voltages
were now rock steady. Was the back-to-front tantalum capacitor
a fault in design or assembly and for how long did the capacitor
last before starting to fail? Alternatively, could the capacitor
have degraded to the extent it drew DC current resulting in a
change in the operation of the bias transistor circuitry and
polarity reversal across itself?
Because I wasn't sure exactly
what the true voltages would be when the equipment was reassembled
I decided to use two tiny 22uF 16volt electrolytics series-connected
as a non-polarised pair in place of C656. This solves the problem
of the voltage changing from positive to negative and back once
the equipment is fully reassembled and with the mixer in/out/in
service. The clearance between the metal sleeve and the circuit
board dictated the maximum diameter of the new capacitors as
no greater than 4mm.
I fitted the LO/REF board in
its sleeve and reassembled the rest of the Wavetek and turned
it on. I could hear the 512MHz signal turning on when 137MHz
was selected then off when 138MHz was selected. The good news
was that when I listened to both the 137 and the 138MHz signals
at 1mV output they registered at exactly the same level on the
monitor receiver S-meter and both were perfectly stable with
Above you can see from
the button markings the various features. Readings can be swapped
between volts and dB. Changing frequency can be done either by
direct entry eg. Freq, then 110, then a press of the KHz button.
Output by pressing Lvl, then 1000, then pressing the mV button.
Altering frequency is achieved
by pressing Freq which activates a cursor under the frequency
display. The cursor can then be moved by the left and right Cursor
buttons to rest under the desired significant figure which can
then be increased or decreased using the up and down cursor buttons.
The same facility is available for output level or modulation.
Minimum output is 0.1uV and the equipment is screened enough
for this level to be usable.
I must have used the Wavetek loads
of times before it started acting up again. I'd noticed that
when it was turned on, if there was a radio on in the workshop
there would be dreadful crackling while the auto diagnostics
were running which settled down after a short time to be followed
by intermittent lower level crackling.My guess was a bad RF output
transistor so I hunted around and discovered an identically packaged
example coded SRFT3034 that was cheap enough to gamble on it
being OK for replacing the original MRF239F. In the meantime
I invested in a new signal generator. I say "new",
but in fact it was labelled as a "barn find" but the
fact it was (a) clearly going to be cheap and (b) made by Hewlett
Packard convinced me it could be got going despite the sellers
comment about plugging it in and nothing lit up...read
the exploits in commissioning this. I decided to tackle the
Wavetek next... but before fitting the new transistor I carried
out some tests. When I last used it I was aware of a message
popping up in the display and this generally corresponded with
a disturbance in the RF output so I checked and discovered the
clever diagnostic hardware and firmware was certainly picking
up a problem but wasn't keen to tell me what it was.
I rigged up a test, connecting my scope
and spectrum analyser to the Wavetek. The first thing I noticed
was the indicated RF output was miles different to that displayed
on both test equipments. To be sure, I switched everything to
volts to line up with the (high impedance) scope's display. An
indicated output of 0dBm, now reading as 224mV, showed as 146mV
on both the SA and the scope, so the Wavetek is reading high
by around 5 or 6dB. I inserted various fixed attenuators and
saw the same discrepancies. I even swapped coax leads to no avail,
so it appears something is wrong.. perhaps the RF output transistor
is short of gain since its "repair", or even developing
another internal short? I also found during these experiments
that the amplitude of the RF output jumped around and when it
did so the signal was sometimes accompanied by transient signals
either side of its frequency. Little by little I pinned it down
to a similar problem already found and repaired.. see
above. The Wavetek includes some mighty clever self-adjusting
test procedures and I tried one. The RF output at 13dBm is monitored
and adjusted to the correct level using cursor keys, but adding
the maximum adjustment of +6dBm still left the output short by
about 5 or 6dB.
Basically the lower the frequency of
the RF output, below 137.5MHz, the less reliable it became. So,
the problem looks like the 512MHz signal used for mixing down
to the frequencies below 137.5MHz. In summary then, there are
two problems. Not enough RF power into the attenuator and a dodgy
circuit responsible for generating lower frequencies.so I should
be familiar with the circuit boards. It's interesting to compare
this equipment with the HP8640B. The latter uses loads of complicated
mechanical switching to achieve much the same results as the
Wavetek's microprocessors. In theory the Wavetek should last
forever compared with the HP whose mechanical contrivances wear
The first step was to tackle the MRF239F.
Was it leaky? I measured the resistances and found (in-circuit)
2.4Kohm between collector and base so carefully detached it.
It had a leak of 2.4Kohm just as it had measured in-circuit.
Previously the leak had been a few hundred ohms and this had
cleared, but if it's an internal structure, maybe tin-whiskers,
it could recur so a new transistor is the next step.
Checking with a transistor tester
revealed "two diodes" so clearly the device is not
ideal. I applied a current limited DC voltage across the base
to collector with base negative. Vcb can be 36 volts max but
with the current limit gradually increased to 400mA the leak
finally disappeared once the potential across base and collector
reached minus 19 volts with current reacing 220mA. The stud was
fairly warm so odd things must have been going on inside the
case. At this point the transistor tested as an NPN device with
a gain of 8.
As a comparison, the new transistor
checked out as an NPN device having a DC gain of 37.
The question then is whether
the new device will work in place of the original MRF839F?
The type is SRFT3034 which is equivalent to the Motorola TP3034. Click on these links to
compare the ratings. Obviously the new device has a far higher
RF output capability and it's clearly a much beefier chip (35
watts versus 3 watts at 960MHz) with a correspondingly higher
output capacitance, but as the Wavetek is rated up to only 550MHz
the new device might be suitable.
It's worth a try and if it fails
then I can put back the original device and see how that performs...
I fitted it but the RF was down
by circa 10dB or so on the bad 839F.
Having tried the new PA transistor
and failed to improve the low output (in fact making it worse)
I found the intermittent fault was still present (which at least
means it isn't due to the PA transistor). The fault is very similar
to the one I discovered about a year ago when I'd cured by replacing
a 10uF tantalum. To recap.. the RF output drops intermittently
with crackling by about 20dB accompanied by "Status"
appearing on the screen. From the evidence on the spectrum analyser
the frequency of the local oscillator changes slightly as well
as the output changing in amplitude. In fact two things are happening.
The amplitude of the output signal drops by 20dB and the frequency
might (but not always) shift by a MHz or so (perhaps an oscillator
losing lock?). The circuit boards are all inside metal sleeves
and mounted in pairs making it difficult to work on them when
powered up, but by experimenting I found the problem, like previously,
was a fault in the 512MHz local oscillator (LO) circuitry. I
was able to switch between 100MHz, when the LO was running and
200MHz when it was switched off and find the intermittent only
affected the lower frequency.
When I was looking for the fault the
last time it occurred I didn't have a suitable RF source with
which I could interchange with the LO, but having recently repaired
my HP8640B, I used this in place of the LO. With it set to 512MHz
and greater in amplitude than -10dBm it worked perfectly, mixing
the RF to the desired output. No intermittent was visible, but
switching back to the internal LO re-introduced the problem.
I checked one 10uF tantalum (C649)
which proved OK. I also monitored the +15 and -15 volt supply
lines which also appeared to be OK. Access to the powered board
is tricky but eventually, I found some voltages that were varying
when "Status" appeared on the display. The most significant
was the end of L601 which intermittently switched between 6.13
and 8.15 volts (with +15 volts stable). The circuit diagram shows
IC605, an MSA-0204
amplifier whose output uses L602 to block RF. A resistor,
R622 of 475 ohms is used to define the power supply and bias
to the amplifier and a pair of capacitors C640 and C642 are used
as decouplers. The +15 volt supply is used to power IC605. The
two voltages indicate the current drawn by IC605 is either 19mA
(normal=6.13 volts) or 14mA (fault=8.15 volts). The manufacturers
recommended operating range is 18-40mA. Interestingly a second
MSA-0204 (IC614) is used to feed the Internal Reference output
socket so it's fairly straightforward to swap the two amplifiers
if indeed the LO amplifier is bad.
What fault condition could cause
a higher voltage to appear at L601? Besides the amplifier chip
itself, a bad decoupling capacitor C642 is unlikely as a leak
would reduce the voltage not increase it. The most likely problem,
assuming the amplifier current increases with RF drive, is the
circuit to its left (above) has an intermittent fault. This would
include Q603/Q604 and Q605. Of interest is C656 because I'd changed
that because it had failed (see above).
A prime candidate is a second 10uF tantalum C649, but I'd already
removed this and testing proved it was OK. Next, I'll measure
the various points around those three transistors.. awkward because
of the board position.
Not to be overlooked is the rather nasty
possibility that the digital control circuitry is responsible.
There are two connections: an RF sampling connection at R617
and a feedback connection to the varicap diode CR605. A complication
of course is the fact that being a closed loop, a fault in the
oscillator will cause an error correction signal, and a faulty
error correction signal will produce a shift in the oscillator.
I carried out more tests. Monitoring
the LO with both the spectrum analyser and my UHF receiver set
to SSB so I could monitor the stability I checked various voltages.
I found again the when the output was clean IC605 output was
about 6 volts and when bad this rose to about 8 volts. It was
difficult to check the RF input to IC605 but it did appear to
be more stable than the output which led me to believe IC605
was bad. I also managed to check the input to IC605 with some
difficulty because of access. The input voltage read 0.9 volts
but increased to over 2 volts when the fault appeared. This discrepancy
seemed to me to point to a bad amplifier chip because a change
of a volt and a half at its DC blocked input (C644) with a more
or less constant RF input was not logical. I switched off and
removed IC614. This chip provides an output labelled INT REF
which I take to be the 10MHz internal crystal and is only used
in rare circumstances. I can replace this chip in the future
if need be by a modern device such as the BGA622 which gives
15dB rather than the 11dB of the original chip.
It was not easy removing
the two MSA-0204 devices without using excessive heat because
of ground connections, but I managed without damaging either
chip. Checking the DC parameters I found they matched fairly
well with no significant different forward voltages. Once IC614
was fitted in place of IC605, I switched on. The UHF receiver
produced a much stronger signal than before and there were no
crackles during warm-up. The spectrum analyser showed a clean
steady signal at 512MHz with an amplitude of -16.8dBm. I left
the system soaking for ten minutes then connected the LO output
into the PA board. Setting an output that required the local
oscillator, I was able to see a clean 100MHz signal with a clean
LO on the receiver. Changing to 200MHz (where the LO isn't used)
resulted in a clean 200MHz output with the LO correctly off,
then back to 100MHz resulted in the same results as before. No
instability and no crackling and wth the attenuator set to 0dBm
the output read -19.6dBm. The receiver input may be reducing
the output somewhat, but the RF level is low so that now needs
sorting out Note that the Wavetek parts list quotes IC605/IC614
as MSA0304 but they are both marked "2" indicating
The next step is to tidy up the
changes I made to the LO board then reinstall this, remove the
PA board, refit the MRF839F and check the RF output. All done..
the new PA transistor was clearly not correctly biased and had
given -20dB for a setting of 0dBm. I fitted the original (repaired
twice transistor) and the output improved reading -10dBm with
the setting still at 0dBm. I found that increasing the output
from 0dBm raised the output slightly but continuing resulted
in the output dropping, instead of rising, and with more and
more harmonics showing up. I'm putting this down to the PA transistor
which, although now showing as a normal NPN device, had a low
indicated gain of around 7, but as it costs a lot of money and
the fault may be something entirely different, I'm hesitant to
buy a new one.
Basically the Wavetek is now usable
but on the understanding that its output is 10dB less than indicated
and shouldn't be used beyond 0dBm.
There are further tests I need to carry
out. The user manual is unfortunately silent on certain essential
(useful) things such as RF voltages and power levels around the
circuit. My guess is that these are not well enough defined in
terms of what they should be. The reasoning is this... tests
dictate one measures the output power to see it accords with
the attenuator setting, and the internal adjustments (made by
the operator from the front panel) should ensure that throughout
the circuitry the gain setting levels are suitable for providing
the correct output (but with the important consideration that
there are no faulty parts). For example the operator can add
up to +6dB to the output to make it equal to the correct RF output
power. That 6dB could account for loss of nominal power due to
say component ageing but hopefully not bad components.
However, there's a check for harmonic levels to ensure these
are below 26dBc (in other words harmonics from a perfect Wavetek
design are not expected to be much better than say 30dB down).
That means pumping up gain by a significant amount to mask a
bad component would be likely to degrade harmonic suppression
and thus referring the user to look for a fault.
In my example the RF output is about
10dB down on what it should be (and possibly more or less than
this at other frequencies I didn't check). I did increase overall
gain to pump up the output by 6dBm and I did notice harmonics
were then very bad. There are two checks I need to carry out.
Firstly I need to measure the input into the attenuator by disconnecting
the cable and checking the power into 50 ohms. I need to do this
in order to test the attenuator isn't responsible for the loss
in power output. I could also connect my HP8640B into the Wavetek
attenuator to check the attenuator settings. The reason for this
is to confirm there isn't a damaged component in the attenuator
I did this and found the output
from the PA transistor at 0dBm was much the same whether or not
the attenuator is used.
A second test is to monitor the auxiliary
output to see if this is similar to the main RF output in terms
of harmonics. The Aux o/p passes through the two transistor amplifiers
to the main RF output, both of which have previously given trouble
and I wouldn't be surprised to discover yet another fault in
that area. I unscrewed the Aux o/p cable and discovered the RF
level was -11dBm with low level harmonics.
Here's a comparison between the Aux
and Main outputs: 100MHz uses the 512MHz LO and the 150MHz doesn't.
This seems to prove that the problem is in the two RF amplifier
stages, one or both of which must be distorting the sinewave
so, due to limited access, perhaps the best solution is to bench
test the output board using bench power supplies and a signal
generator (knowing that the likely RF input will be something
like -11dBm and the output needs to be OK up to the max output
level of 17dBm).
What next? Well the construction
of the equipment prevents easy fault finding because the circuit
boards are fitted in metal sleeves and RF connections are made
by miniature solid copper coax. My idea therefore is to remove
the output board and test this on the bench. To do this I need
a set of three power supplies, +18V, -18V and +8V, relying on
the on-board stabilisers to produce the required +15V, -15V and
+5V rails. Having identified the RF path I can feed its input
from a signal generator. The RF path comprises no less than seven
amplifiers. The final two being Motorola RF power transistors.
Prior to the first of these I see a feedback path which must
be to stabilise the RF input so that the final RF output fed
to the attenuator is at the maximum RF level of +13dBm (which
is conveniently 1 volt RMS into 50 ohms). The manual as far as
I can see is silent on RF levels but it seems reasonable for
the two power amplifiers to provide a fixed gain over the full
range of the generator. By gradually increasing the RF input
the input to the final two RF amplifiers should follow the input
until the feedback kicks in to stabilise this. Having measured
this RF level I can then check the RF levels at the first and
second transistors. Not only measure the fundamental frequency
level, but also the second harmonic level as this seemed to be
a problem. My guess is that the input to the final amplifier
pair will be free of excessive harmonics but that these will
be getting produced either in the first or second stage.
Below is a rough sketch of the
Output PCB RF path with a spec of the MSA0304
(click to see). The MSA0404 is similar but has a 7.5dB gain.
The MA0304 has a maximum gain
of between 10dB and 12dB so the theoretical maximum gain up to
the feedback loop is between 47.5 and 55.5dB, say 51dB, but in
practice maybe something like 25dB would be the designer's aim?
The RF output measured at the
AUX output was -11dBm and the RF output into the attenuator should
be a little over +13dBm to take account of attenuator losses.
I've indicated on the drawing in red below the sort of levels
I'd expect, but these are pure guesses. The two output transistors
will be operating in a linear fashion so perhaps less than say
Above you can see the
RF PA driver stage on the output board showing the MRF571 and
above its 68 ohm collector resistor. This transistor has a DC
base bias arrangement which includes a small transistor Q901.
One puzzle is the condition of the resistor which is specified
I then carried out some tests
having wired +/-18V and +8V to the board from bench supplies.
With my spectrum analyser
connected to the RF output connector I tried various RF input
levels at 20MHz to IC810. The results look very odd and support
my previous findings re-harmonics. The drivers are producing
decent gain and the PA seems to work well at a very low level
also producing gain which changes to attenuation as the drive
increases. The main reason for this was the observed severe increase
in the amplitude of harmonics. This can be explained by the PA
Because of the condition of
the collector resistor accompanied by a hot resistor smell, the
first step was to look at the MRF571, Q900. For a start
I removed the collector resistor and fitted a physically larger
56 ohm resistor as that's the value given in the manual. Why
was the original resistor burnt? Although the MRF571 base-emitter-collector
junctions looked about right on my multimeter, I removed the
transistor and found it had failed with a bad collector junction.
This transistor was replaced
about a year ago and as the seller was offering them in tens
I fitted a second from my remaining stock of nine. As it gets
pretty warm and possibly this had caused the demise of its predecessor
I glued a brass washer to its top to lower its temperature. I
then continued RF tests. From the table below you can see each
stage produces some gain and between each stage there are some
losses due to components used for coupling the stages and for
matching. The overall gain appears to now be 20dB. I kept the
input signal well below that which produced the previously noted
overloading. IC801 has a gain of 11dB, Q900, the MRF571 has a
gain of 8dB and the PA has a gain of 8dB. Losses total 7dB.
The next step is to figure out
why the PA is overloading. I measured its base voltage at 0.7V
which seems low. The emitter resistor is shown as a pair of parallel
17.6 ohms so given a Vbe of say 0.6V the collector current must
be very low at circa 10-12mA. To develop 1 volt RMS at 50 ohms
which represents 20mW, given the transistor is operating in Class
A (say 30% efficient) you would need a supply voltage swing of
more than 20 volts. The collector is sitting at 7 volts hence
the transistor will be saturated once the drive exceeds say 0dBm.
Saturation=square wave rather than sine-wave, hence harmonics.
Clearly the PA circuit isn't working
as it should. My first suspect is the RF power transistor itself
which had only a single-figure gain when I tested it out of circuit,
most likely because I've repaired it twice!... look back in this
But.. what else could it be? I checked
the various voltages, in particular around the PA bias circuit.
I found several anomalies which of course could be due to a bad
PA transistor, but my first alternative suspect was a tiny high-precision
2.5 volt zener diode (LM336 Z-2.5) as this had only 1 volt across
it. I detached it and tested it, using my Peak
zener diode tester, finding much to my surprise it was a
"2.5 volt zener diode", so back it went.
Next I looked at the power
transistor used as a series pass regulator for the PA base bias.
Bearing in mind it was a PNP device, its electrode voltages didn't
entirely look right so I detached and tested it. The test meter
reported "Two diodes" so I looked around and found
a suitable alternative, a TIP125, and fitted that. Now it's base,
emitter and collector voltages made sense (ie. base and emitter
now differ by around 0.6 volts) and the adjacent precision zener
voltage correctly measured 2.5 volts. The PA collector now measured
around 13 volts and an RF input at IC801 of -20dBm gave me an
output of -5dBm. Varying the input to IC801 produced a linear
response at the output socket (see below) so hopefully the problem
is solved. The maximum rated output is +13dBm so the final test
giving +14.4dBm is fine.
The twice repaired MRF839F seems
to be OK after all...
Above, the faulty
PA bias regulator transistor. This is a PNP MJE253 rated at 100V
After fitting the circuit board
back in the chassis I checked the output. The RF is now pretty
clean with harmonics more than 35dB down and also fairly accurate
in terms of the attenuator setting versus spectrum analyser display.
Most signals are within 2dB, many less than 1dB, and with only
very low frequencies, for example 10kHz, showing more than 2dB
discrepancy. Switching from +10dBm to the maximum output of 13dBm
actually gives 16dBm. My guess is that, like the HP8640B, the
Wavetek has extra circuitry for outputs at this level. Of course,
I've carried out lots of repairs to the RF path but haven't touched
any of the preset adjustments, indeed if there are such things.
There is a strong possibility that there are pots or even firmware
settings, for the feedback circuitry to maintain a flat response
across the frequency range.
I looked in the Wavetek manual
and then remembered that, instead of pots and internal twiddling,
one can use the firmware to carry out calibration. During my
first attempt to correct the low output level when set to +13dBm
I'd cranked up the power level as high as it would go and failed.
That's why, now the RF amplifiers are working correctly I'm getting
+16dBm. The power level is faithfuly being increased to the power
I'd set earlier.
All I now need to do is follow
the complete procedure and the RF output should be dead flat
from 10kHz to 550MHz. It should be easier than described in the
handbook (using a power meter) because I can use my DSA815 to
monitor the output. In fact the latter method will be more accurate
because the power meter method unavoidably adds together the
fundamental and harmonic powers. The manual describes what is
called "Autocal" which turns out to be not exactly
"auto" because no less than thirty-one manual adjustments
are required. Basically, you need a power meter (or with some
extra fiddling a spectrum analyser) to check each of the 31 frequencies
determined by the firmware. At each frequency you have to use
push buttons on the front panel to set the power output to precisely
0dBm. Say you use 6 button presses per frequency.. you'll need
186 button presses to set the output level, plus 31 further presses
to move across the list. "Auto" becomes therefore 217
I just completed the level and
linearity "auto" test. It probably took at least 20
button presses per each of the 31 pre-assigned test frequencies
then a further eight button presses for around 21 pre-assigned
power settings. My count is therefore 788 button presses plus
the date.. a further 12 because I mixed up the month and day.
That's 800, but the thing is now well nigh perfect to within
0.5dB for any frequency at any RF level!
The way it works is roughly
as follows...Level Setting: a frequency is shown on the
SG display, say 8MHz. The level is supposed to be 0dBm. If the
spectrum analyser says the level is -2.6dBm you press the SG
up button 0.1dB per press till the display reads 2.6, at which
point the spectrum analyser will show 0dBm. You then press the
right arrow to move the frequency to the next one, say 16MHz
and so on. Linearity Setting: the first SG display shows
+13dBm at 300MHz so you check the SA level, say +14dBm and press
the down arrow till the display shows 1.0dB at which point the
SA level will be +13dBm, then the SG right arrow which brings
up +12dBm etc etc....When you're using the signal generator you
see the raw output which is followed within a second or two by
the modified level which has been corrected by the look-up table
generated by "auto" calibration. What would we do without