Rohde & Schwarz Test Set SMDU

Model 249.3011.04

Selected drawings for the SMDU

Go to second page of commissioning

One of the chief reasons for buying this monster is that it pre-dates the use of microprocessors. Not that these are especially unreliable, but often the features of the equipment are hidden away and rely on complicated button presses to use them. In addition, once an equipment is getting on in years I've discovered a "self test" will stop one dead in the water even if it's due to an innocuous tarnished contact buried in the circuitry. This has happened recently with both my later model R&S signal generator and my Wavetek.

My first inspection of this SMDU suggests the on/off switch is bad (it feels rough and not latching). The design of the on/off switch in my other R&S equipments is not ideal, depending on precise movement of a metal tube connecting the front panel switch button to a double-pole switch which is very similar to those used in TV sets from the 1980s to 1990s. It remains to be seen whether this model has the same mechanical arrangements to its later cousin. .....And here's a second example.

The long tube isn't fitted. The switch is mounted close to the front panel in a very inconvenient space under a pair of solid coax links. I guess this was seen to be bad practice so the tube was incorporated in later models to allow the on/off switch to be fitted in a more convenient position close to the mains socket. Alas, the choice of switch is much the same as in later models, having very bad reliability. Mine has failed with both poles open circuit. I'm hoping that this is the only fault...

Professional signal generators and similar items, back in the 1970s were sold at inordinately high prices. I recall a sampling scope made by HP in the mid 1960s cost as much as a 3 bedroom semi, and if you wanted a really up to date R&S signal generator today it would cost you upwards of £30,000. I repaired a LeCroy scope a few years back.. it's big brother tipped the monetary scales at $1,000,000. I recall the chap I got the scope from worked at a public school. He said all their scopes were being put in a skip because they'd got a good deal from a "schools supplier". I quizzed him on their new scopes and heard with some dismay they were ancient CRT scopes. The school was being diddled!

Anyway, I bought this equipment described as a "Signal Generator", but it turns out, when one studies the front panel, that it has more functions. Click the picture above. It includes a frequency counter and voltmeter as well as providing both audio (15Hz to 150KHz) and RF outputs 140KHz to 525MHz (this example is missing the range extension).

Having detached the lower panel to see the on/off switch I became more than a little concerned as I'd seen another power supply board for this equipment and at the side was a slot for a circuit board with dangling connectors. I immediately feared the worst and imagined the PSU board had been removed so looked at the circuit diagram to investigate the practicality of making a replacement. In fact this seemed perfectly feasible and the design could also be very much simplified by using extremely cheap modules from China.

I then checked further. At this point I should say that, because of its weight, it's very awkward to move the thing around, but I did spot an array of large electrolytics (below.. top left) and looking behind these I spotted the PSU board in place. It turned out I'd seen the space for an optional extra.. not at all important for my intended use.

Click for larger picture

Now to tackle that failed on/off switch (above.. bottom left corner). I gathered together several old TV switches and found one that was potentially going to be a good fit. The two screws holding the switch are almost impossible to get at but eventually I got these loose and was able to pull out the old switch (shown here).

Note that when fitting a new switch the securing screws appear to be 2.5mm so 3mm are too big. This isn't obvious until one finds it impossible to get the 3mm screws started in their really awkward locations.

Note the connections shown here. Blue + Black are mains and Purple + Black go to the transformer. The old switch links along the body.

A replacement such as the one I fitted also has its links along the switch body but some go across so great care must be taken to swap a switch.

Another point of interest is the mounting hole spacing. This can vary by around half a millimetre and due to the tight tolerancing the holes might need filing to increase or decrease their spacing.

The proximity of the solid copper coax leads passing over the switch solder tags means a plastic cover must be used to prevent potential shorting mains connections to ground.

Yet another point is the coloured oblong cap is glued in place and must be snapped off in order to remove the old switch. The switch cap then needs to be very accurately drilled to accommodate the new switch.

Drill to match the diagonal of the new switch push spindle.

This picture shows the very awkward position of the mains switch (with plastic cover). In fact I had to reposition the straight coax and temporarily unscrew the looped coax in order to fit the plastic cover.

I initially used a tiewrap around the plastic to keep it in place but found this pressed against the return mechanism jamming the switch on.

After checking continuity I found that the DC resistance across the mains socket live and neutral was around 8 or 9 ohms with the switch on.

I now need to confirm the mains settings, if any are fitted, are set to 230 or ideally 240 volts.

Below is the top view of the signal generator.

Click above for larger picture

PIN NUMBER	VOLTAGE	TYPE
1	+21V	SET BY R4
2	+15V	REGULATED +/-0.6V
3	+15V	REGULATED +/-0.6V
4	+65V	UNREGULATED
5	+28V	UNREGULATED
6	-24V	UNREGULATED
7	0V	GROUND
8	0V	GROUND
9	-15V	SET BY R14
10	+5.2V	REGULATED +/-5%
11	+5.2V	REGULATED +/-5%
12	+12V	UNREGULATED

Click the circuit to see it full size

Here's a picture showing the location of the 12-way connector (top left), which carries the various voltages to the equipment, at the rear of the chassis.

The blue connector is ST91 carrying mains supply to the transformer TR1.

I guess, now that the mains switch has been replaced, it would be prudent that after checking the correct mains transformer primary winding is selected, the power supply is producing correct voltages. One problem I've met previously is once a mains switch starts to fail, repeated switching it on and off, will damage something. The plan is to just unplug the 12-way connector shown above, apply mains power and check each output pin to see if its voltage conforms to the table above. Because of the slightly awkward position of the connector plug I might find wire suitable mating socket test cable. This has the typical Euro-Imperial pin-spacing of 2.54mm.

The mains selection is carried out by rotating that pale blue square cover and inserting the fuse associated with the required voltage into the screw-in fuse holder. Below, the correct 235V position is lower right as shown.

MEASURED VOLTAGE

+21.01

+15.23

+14.75

+80.00

+32.40

-24.50

0.00

-15.01

+0.17

+16.25

EXPECTED VOLTAGE

+21

+15

+65

+28

-24

-15

+5.2

+12

NOTES

Set

Reg

Unreg

Gnd

Set

Reg

Unreg

TEST RESULT

Good

Gnd

Good

Bad

The table above gives the result of the initial test of the power supply. From the tests the only problem is the missing 5 volt rail. The rectified voltage from which the 5 volts is produced is present so either the rail is switched off because there's no load, or there's a short withing the circuitry fed by the 5 volt supply, or of course a component in the 5 volt regulator circuit has failed. The 12V output uses a common ground to the 5V circuit and the same transformer winding so both these parts of the circuit will be OK.

I removed the twin MC1569 devices (B1 & B2) plus the TO3 2N3772 transistor and found the latter had no shorts, reading as an NPN device, and comparing the pair of MC1569s both seemed almost identical from resistance checks at their pins. Maybe the output needs to be loaded? I thought this because the Motorola circuit shows a 10mA drain and the circuit above has only the 1K resistor R9 (=5mA). But adding a 200 ohm to ground proved this wasn't the problem as, with the 200 ohm load the output measured 176mV, the same as before. Checking GL14, this measured about 13 volts at its cathode and 170mV at its anode meaning that the MC1569 output pin was either being held down or it was not reacting to the voltage produced at R1/R18 and R2. R1/R18 is calculated, with R2 being the recommended 6.8K from the formula R=2x5.2-7K = 3.4K.

Is the MC1569R faulty? It's not too easy to test but the fact is that the output pin is only 0.17 volts which seems odd. The straightorward option is to fit the working one, B2 in its place and see if we get 5.2 volts output. But if it is faulty.. why did it fail as I don't wish to damage a replacement? One possibility is that the power transistor lost most of its gain forcing B1 to pass more than its maximum rated current. That might cause it to overheat and fail.

I decided to compare B1 and B2 in the same way I used to test complex TV chips. That's using a diode test at each pin with the meter black or red lead grounded in turn (to the chip ground connection.. in this example the outer case is ground). 2.4 indicates the max value for my meter and is often unimportant.

See the manufacturers spec for the MC1569R.

PIN	1	2	3	4	5	6	7	8	9
BR (B1)	-	1.4	2.4	-	-	-	-	-	-
RB (B1)	1.0	0.7	0.6	0.7	0.75	-	0.7	-	0.6
BR (B2)		1.4	2.4	-	-	2.4	-	-	-
RB (B2)	0.7	0.7	0.6	0.7	0.75	-	0.7	-	0.6

The results appear to show Pin 1 has an anomaly and from the drawing of the MC1569R below my guess is the output transistor circuit is suspect which tends to support the hypothesis of the loss of gain of the 2N3772. I tested a 2N3055 as a guide and it measured 19 and then the 2N3772 and it was 22 so that hypothesis is not valid.

I then decided to fit the "bad" B1 into location B2 and much to my surprise it worked, giving me the correct +21 volts. That means B1 hadn't been damaged in its proper location, and B2 must be working because +21 volts had previously been present.. so nothing to lose fitting the working B2 into location B1. I tried this and it worked so I'm really confused. Presumably B1 works with a high voltage output but not 5.2 volts. My only explanations are (A) the circuit given below is not "complete" as indicated but must have additional biasing resistors which are not shown and that one of these is out of spec thus preventing the device from working at low voltages or (B) there's a bad ground connection within the case making one of the connections high resistance to the extent low voltages are out of range.

Anyway, as you can see below, after plugging in the 12-way power lead the equipment powered up.

I snapped a picture and pressed a button or two but then the display went out. It looks like there's an intermittent fault and one way of tracking tis down might be to unplug the 12-way PSU lead and use an external 5-volt supply to discover what's going on, but first a block diagram of the thing.

For the moment the diagram has labelling in German, but you might be able to see, in conjunction with the front panel, that not only is the SMDU a signal generator, but it also works as a frequency counter (or frequency meter). Fault finding, specifically looking for a short circuit on a voltage rail is going to be very difficult, however the single oint of power distribution (ie. the 12-way PSU connector) will help as an external power supply can be used to test the circuitry. A clue might be the fact that the 5 volt supply has already given trouble and the display may well be entirely dependent on 5 volts. Below is a wiring diagram which is pretty useless until you click to see full size..

The PSU outputs (ST92-BU92) are cabled to a "Voltage Distribution Board" (BU93-ST93) together with other discrete locations as indicated in the table below.

Then BU94 to BU97 outputs are wired to the various modules.

BU92 PIN	VOLTAGE	TO	TO	TO
1	+21V	BU27 Pin 11	ST53/BU53 Pin 2	Overload Protector
2	+15V	BU93 Pin 6	ST95/BU95 Pins 1-7
3	+15V	BU93 Pin 4	ST95/BU95 Pins 8-13
4	+65V	BU112 Pin 4	Range Switch
5	+28V	BU303 Pin 3	Frequency Doubler Option
6	-24V	BU112 Pin 2	Range Switch
7	0V	BU93 Pin 13	ST97/BU97 Pins 1-16
8	0V	BU93 Pin 14	ST97/BU97 Pins 1-16
9	-15V	BU93 Pin 10	ST96/BU96 Pins 1-14
10	+5.2V	BU93 Pin 1	ST94/BU94 Pins 1-11	ST404/7, ST203/3, ST112/1, ST80/17&18, ST83/18, ST63/2, ST303/6, ST53/4, BU16/40
11	+5.2V	BU93 Pin 2	ST94/BU94 Pins 1-11
12	+12V	BU93 Pin 8	ST96/BU96 Pins 15-18

I used an external test power supply connected to each voltage rail in turn and gradually adjusted the voltage from zero to the appropriate pin at ST92. I initially set the current limit low in order to trap any short circuit or intermittent fault then increased these to the rated values.

Ground connection was made to a copper coax cable feeding the main assembly making some test voltages lower than stated.

No rail exhibited any signs of variation except when some of the front panel controls were adjusted. Some "nominal" figures may be dependent on functions selected on the front panel.

** this figures represents the total 5.2V current drawn by Pin10 plus Pin11.

ST92 PIN	VOLTAGE	TEST	NOMINAL	NOTES
1	+21V	136mA	150mA
2	+15V	831mA	1000mA	Isolated from Pin 3
3	+15V	380mA	1000mA	Isolated from Pin 2
4	+65V	0mA	25mA
5	+28V	0mA	300mA
6	-24V	2mA	-30mA
7	0V	Gnd	Gnd
8	0V	Gnd	Gnd
9	-15V	324mA	-750mA
10	+5.2V	294mA	3500mA**	Isolated from Pin 11
11	+5.2V	1070mA	3500mA**	Isolated from Pin 10
12	+12V	0mA	600mA

Having tested each rail independently the equipment was powered up normally. The display lit for several seconds before the 5.2V rail dropped to 170mV as it had done previously, indicating a PSU problem rather than a fault in the main equipment.

Studying the 5 volt circuit diagram I tried to figure which component must be responsible for a sudden loss of voltage, and once lost, pretty permanent. What about the power transistor? This type of device can lose gain over a long period so I checked it once again on two different meters. The LCR-T7 said my transistor was "two diodes" but my UK-made Peak tester declated it was an NPN transistor with a gain of 22 at 2.5mA. I checked a variety of TO3 devices and most were low gain. The 2N3772 spec says that 22 is normal, so this is a dead end, particularly because the trip happens with only a low nominal load anyway.

Could the small transistor be faulty? One way perhaps to check this part of the circuit is to remove the TO3 transistor and check the output voltage. Sure enough... no change. The output read about 0.5 volt. What about the input voltage? The MC1569R uses a potentiometer as previously discussed, but I didn't test Pins 6, 8 and 9. I put my voltmeter on the end of the 2.2K pot R18 and switched on with the DC power connetor unplugged. It read about 1.5 volts. That's wrong so I checked contunuities and resistor values by unplugging the regulator and found everything was exactly as the circuit shows. I tried again and oddly the voltage this time measured 5.2 volts... but after a few seconds dropped back to half a volt. So, that's a little more information. The regulator is shutting down, but the shutdown pin (Pin 2) is correctly grounded so that's not the reason. The fault isn't the regulator either because my first step had been to swap B1 and B2 and the other regulated voltage was fine with both devices.

I decided to look at the spec for the MC1569R yet again. The notes refer to a couple of relevant figures. One is the minimum input voltage of the device stated as 8.5 volts. The reason for this relatively high figure for a 5 volt regulator is that it uses a 7 volt zener diode. A second figure is a differential voltage of 2.7 volts being the headroom over the output voltage. So, the input rail must be at least 8.5 + 2.7 or 11.2 volts. The drawings show this figure to be 12 volts so there's only a margin of 0.8 volts. In fact off-load I recorded 16.25 volts, but that was with only 0.17 volts at the output of the 5 volt regulator. If you study the way the two voltages are produced you'll notice they are a bit odd. This is because of a clever technique to save power. The series pass 2N3772 transistor is fed from a separate DC supply (GL1/GL2) to the MC1569R. The latter has its own supply which comes from GL3/GL4. What is happening at switch-on though? Is it possible that due to ageing the minimum voltages have drifted upwards and the MC1569R is now underpowered?

One way to test this hypothesis is to run the transformer at the next lower tapping or to connect an external DC power supply in place of the existing arrangements. I tried the latter, unplugging the 2N3772 and applying a voltage to the cathode of GL14 but the output stuck at 0.51 volts with the input voltage set to anything between 10 and 23 volts.

At this stage I decided to explore an alternative solution. One is to drive the 2N3772 with a 5 volt regulator and another is to use a regulator directly to provide 5 volts. To this end I actually have a number of uA78H05SC devices each capable of delivering up to 5A of current. I rigged up a test circut and found this method was satisfactory. In fact, as there are two unconnected 5 volt supplies to the body of the equipment (ie. the two 5 volt rails are fed separately), a good solution would be to provide a pair of uA78H05 devices. This would decrease individual device dissipation and improve reliabilty especially if ballast resistors were incorporated to set the device inputs to a minimum of say 9 volts under maximum load. Below is the test rig. If the resulting voltage is too low (ie the original runs at 5.2V to cater for resistive losses) there is a method of increasing the regulated output by using a fixed drop in the regulator ground feed but this means isolating the heatsink from the equipment chassis. The fixed drop could be achieved with Schottky diode having a suitable power rating. However, in practice this might not be necessary.

I added a second regulator on the heatsink and fed it from the wire going to the 2N3772 case after unplugging both that device and the MC1569R. The pair of uA78H05SCs will now feed the pair of 5 volt wires connecting to BU93 Pins 1 & 2. I need to measure the worst case current at each connection and if necessary insert a suitable ballast resistor in the input lead. This will reduce the heatsink temperature.

During testing I allowed the uA78H05SC to get extremely hot and sure enough, as stated in the manufacturer's spec, the chip shut down and restarted once it had cooled down. To reduce dissipation I can insert a ballast resistor in series with its input pin to reduce the input voltage to its minimum working level of around 8 volts. The original design of the power supply needed at least 11.2 volts for the MC1569R.

Below are the results of the power supply modification; fitting a pair of uA78H05SC 5-volt regulators in place of the MC1569R circuit.

As you can see the display, driven by 5-volts, is now working whereas previously this would appear for only a few seconds before the MC1569R failed. Thinking about this I believe the MC1569R devices must degrade over time. One possibility is the growth of an internal whisker from the area of the circuit dealing with the shut-down facility. This circuit is supposed to be disabled by grounding Pin 2. Many years ago I had the misfortune of using 82S06 devices which were prone to a similar fault. And.. I should also mention those infamous Mullard germanium transistors found in things like old Roberts radios.

ST92 PIN

VOLTAGE

TEST

@5.2V

UA78H05SC

+5.00V

294mA

300mA

+4.99V

1070mA

740mA

And for reference some published documentation.

SMDU MANUALS

DESCRIPTION of SMDU PART 1 in German

PART 2 circuits German/English

PART 3 circuits German/English

PART 4 circuits German/English

SMDU BROCHURE

SMDU_249.3011_Vol.1_Manual.pdf

Now that I can switch on the SMDU and twiddle the knobs and see things happening I can investigate its features. One thing of note was sticky push-buttons. These are generally inter-linked like the types on old portable radios. Pressing one button should release one already pressed, but that doesn'y always happen. Often dried grease is the culprit and this should respond to switch cleaner. I also noticed the analogue meter is prone to a lot of twitching... and as yet I haven't proved there's an output at the N-type connector, or that it responds to inputs at the BNC sockets for frequency or voltage readings..

The first test other than checking basic operation of the signal generator operation was to see if the SMDU correctly generated and read audio.

AUDIO OUTPUT

Select NF-AF push switch

Select Mod-Gen push switch

Select frequency range.. rotary control knob

Select amplitude.. rotary control knob.. set to 0dBm

This worked perfectly but with frequency indication one division low.

Loudspeaker (=8ohm) connected to MOD.GEN (=50ohm) provides audio output but loading output level to roughly half.

AUDIO INPUT- FREQUENCY

Plug audio generator into BNC 15Hz-30MHz

Select INT NF-AF

Counter reads frequency

Select various frequencies/amplitudes

Works perfectly

AUDIO INPUT-AMPLITUDE

Plug in audio generator into BNC NF-AF voltmeter

Select NF-AF VOLTM.

Select AUTO

Select various frequencies/amplitudes

Works perfectly with meter auto-ranging with meter scale indicated in steps of full scale 3mV/30mV etc to indicate level of applied voltages

Next, a test of the RF Signal Generator

I found the operation of the SMDU very strange compared with "normal" signal generators. First of all, although there are twin tuning knobs, one coarse and the other fine tuning, neither has end-stops. The method of selecting a specific frequency is to initially select the appropriate range, then rotate the coarse dial to the matching calibration line and the approximate setting commensurate with the desired frequency, then twiddle the fine tune knob to accurately select the desired frequency (it's not too easy to tune the final digits but of course you're then reading down to less than two or three hundred Hz.)

Most of the ranges worked fine but one or two seemed to be a bit reluctant to lock onto the desired frequency. The fine tune was tricky to adjust and I found that once locked to the correct range, if you tuned to the end frequencies (lowest or highest), then a little more for example tuning to 525MHz, you could actually get up to say 543MHz or down to 390MHz before the frequency lost lock. Losing lock made the counter read "506MHz" or result in an "out of range" message.

On the lowest frequency range, although marked "0.14 MHz" (= 140KHz), tuning down to zero frequency was possible although probably not within spec. In this respect it's similar to the Marconi TF2008 and the Racal RA17 receiver. In fact the TF2008 is very similar in the its features except one needs an external frequency counter to verify the dial readings, and of course the latter doesn't have those extra test facilities.

Later I discovered one could continue tuning downwards in Range 9 because of a bad microswitch.

The frequency coverage is given below, with figures over 525MHz handled by an optional piece of hardware. In fact the dial is calibrated for this option and a red button fitted for its operation, meaning (given the parts) it could be retrofitted easily.

RANGE	CCT BOARD	MIN MHz	MAX MHz	NOTES
I or 9	Y61/Y6/Y63	0.14	50	TUNES DOWN TO ZERO
II or 8	Y11	49	64.5	-
III or 7	Y12	63.5	88	-
IV or 6	Y13	85	119	FM BAND II
V or 5	Y14	118	198	506MHz displayed
VI or 4	Y15	196 (510)	290 (580)	(510) ARE IN RED
VII or 3	Y16	286 (572)	395 (790)	(572) ARE IN RED
VIII or 2	Y15/Y26	392 (784)	525 (1050)	(784) ARE IN RED
IX or 1 RED BUTTON	NOT FITTED	FIGURES IN BRACKETS	FIGURES IN BRACKETS	HARDWARE OPTION

Everything seems to working other than there's reluctance to lock frequencies on some ranges (Range VI above isn't locking at all) so I need to investigate further..

As yet I haven't checked the attenuator operation (pictured below).

Well, I checked the attenuator fairly easily (except the knob is very heavy probably due to dried grease).

Remember that the frequency coverage of the built-in voltmeter is 15Hz to 150KHz so it's possible to connect the RF output socket to the meter input and see a true reading if the RF output is less than 150KHz. I did this and the meter showed the voltage output at about the correct level and changing as the attenuator setting was altered. It's reasonable therefore to assume the RF output is likely to be OK. Of course I can check this with my Rigol spectrum analyser later. I say later because the equipment is 50 feet away from my workshop whilst I'm initially checking it out and it's no easy matter moving it those 50 feet.

Next I need to figure out why Range V is producing an output across the tuning range of 506MHz. I'll first try to analyse circuitwise to see what's involved.

The area handling these buttons is the "Range Switch", circuit board 250.1019 where switch S101 selects various functions. Note this switch carries numbers 1 to 9 which is opposite to other parts of the documentation using IX to I in Roman numerals with I (9) being the red button.For Range 5 or V to be bad, one possibility is a bad range switch. This uses no less than six banks of nine changeover switches meaning that up to 54 switches must be in perfect working order for Range 5 to work properly in all its configurations. The solution to finding the fault may be relatively easy in that, although there are 6 switch banks, only bank 4 is responsible for powering the board generating 118-198MHz. Bank 4 provides a 5 volt supply to Y14 so that must be the first check. If the 5 volt supply is turned on and no RF output is present there may be a faulty component on Y14. If Y14 isn't getting its 5 volts (via ST114 Pin 2), then bank 4 of the switch will need to be checked.Another possibility is that, although Range 5 oscillator is working, the digital display might not be receiving correct inputs and S101 bank 1 is responsible for that. Switch banks 2, 3, 5 and 6 are used for handling the intricasies of FM (banks 2 &3) and the selection of the correct output filter (bank 6) plus the handling of the VOR/ILS option which I don't have (bank 5).

I decided to detach the metal plate carrying the left handle so I could better access the Range Switch board and spotted this label. The arrow is pointing at a microswitch which measured about an ohm across all three of its terminals. The microswitch is activated by a large toothed gearwheel connected to the tuning knob and at one point the microswitch lever is pressed but fails to change the reading across the switch. But what's its function? Is it important?

I swapped the microswitch and found the tolerances between its lever and the slight projection on the toothed wheel were too critical and it didn't switch properly. The old microswitch was OK except its lever was stiff and it locked on. But what's it for? (see later)

Turning on the power resulted in nothing working. The display lit with all the digits reading zero and able to indicate either MHz or KHz but no frequency reading. I tried an external audio generator (10Hz to 110KHz) and this failed to indicate a reading when plugged into either the low frequency (EXT) or voltmeter (EXT) sockets. My guess is perhaps I'm missing one of the power rails?

I found PSU Pin 2 was about half a volt instead of 15 volts so checked everything. Nothing seemed amiss and when I refitted the 15 volt regulators after comparing them the rail read 15 volts and the frequency still read zeroes so I may have just misread the voltage at Pin 2.

Later I came to the conclusion that the 7815 devices used in the SMDU are suffering from a similar fault to the 5 volt regulator.

Still no frequency readings so I decided to check the 5 volt rail at the ICs because, if this has dropped to under 4.75 volts, it might explain the problem (In fact I vaguely remember that the military spec 54 series TTL chips performed better than basic 74 series and work down to a guaranteed 4.5 volts so in respect of those 54 series chips 5 volts is fine). The supply read 4.8 volts worst case so to test the theory I fitted a diode under the grounds of the two 5 volt regulators. This boosted the outputs to 5.6 volts but still no change to operation.. the frequency readings were still missing so I removed the diode and reconnected the ground direct to the heatsink. As I'd expected, still no change so I wondered whether the RF outputs were OK but with no display. I connected up my oscilloscope and tried again. Oddly the ranges now worked normally with both frequency readings and RF outputs present. This time however two ranges were bad rather than one as before. I selected Range 8 and measured the output with the attenuator set to 1 volt and set the frequency to 1MHz. Both values were OK on the scope, so before switching off I measured all the voltages at the PSU output so if the frequency readings disappeared again at least I could compare the voltages.

Could the intermittent operation be due to a tiny increase in mains voltage coupled with a dodgy circuit?

After carrying my oscilloscope from my workshop to the conservatory (because it's warmer there) the signal generator suddenly started working for no real reason that I can figure out after failing to produce an output. Is it something to do with the contacts on one of the push buttons?

Above, my scope shows 1 volt RMS at 1.002MHz with the R & S settings as shown. It's mighty difficult to set the output to precisely any specific figure especially using the lowst range of 0.14 to 50MHz. The scope input impedance is very high as I'm using a probe so the output voltage isn't being loaded and should be as indicated on the scale, left. I checked AM and FM and both seem to work. Oddly Range 3 and 5 are both failing with no display and no RF. This suggests that the display circuitry is OK and the fault is in the range oscillator area. That's one step nearer finding the fault.

I recorded the voltages at the PSU connector as below.

Pin	1	2	3	4	5	6	9	10	11	12
Volts	21.21	15.02	14.71	73.1	29.35	-22.88	-14.99	4.95	4.95	13.05
Nominal	21	15	15	65	28	-24	-15	5.2	5.2	12

I've now reached the stage where I need to understand the detailed operation of the SMDU in order to fault-find. The most important thing to understand is how the various ranges (using Roman numerals) are handled. For example, although each range from II to VII (8 to 3) uses a discrete oscillator, Range I (=9 or 0.14 to 50MHz) is more complicated with frequencies being derived from a 240MHz oscillator and Range VIII (=2 or 392-525MHz) using a doubler connected to the Range VI (=4 or 196-290MHz) oscillator. The fault-finding chart below reflects this design.

Currently (and note there's some intermittency at play) ranges V and VII are bad with no RF output and no indicated freqncy on the display. This would seem to point to oscillators Y14 and Y16. The decade circuit board (under the top cover) shown below includes a set of pins which reflect the switch settings.

These pins (11, 12 & 13) should reflect the states of the push buttons. I switched on and the previous frequency display had gone again so I checked the states of Pins 11-13. Maybe one (AF/NF INT was bad. Thinking back to voltage measurements I rechecked these and found PSU Pin read half a volt instead of 15 volts. I removed the two 7815 regulators and switchem them around after finding Pin 3 had 15 volts but this proved he regulators were both good and there was a drain on Pin 2s circuit. Switching off and powering Pin 2 from an external supply revealed a current draw of over 700mA. I checked with my thermal camera and found on the Range Selection board a slightly abnormal temperature of maybe 30C with about 24C for the remainder. Bearing in mind this was 30C viewed from the track side the true temperature was likely to be say 35C.

Removing the Range Selection board was not straightforward.

Later I discovered you have to slide the metal tray holding the other circuit boards backwards after unscrewing a couple of M5 bolts

After unscrewing the RF cable and a ground wire the circuit board was held in place by four screws. Two (at the connector end) are easy to get at but the other pair needed the removal of a front mounted metal bracket and slackening of the end of the front panel. After detaching the metal plate on the front, the end switch bank needed persuading past the end of the metal box holding the display boards, but eventually I detached it and unplugged its four connectors.

The source of the heat was probably a 47uFx40 volt capacitor which measured open circuit, as did an adjacent capacitor marked 470uFx40 volt. Two other maroon coloured capacitors (both 100uFx40 volt) were poorish and also a longish thin capacitor at the switch end. This is marked 68uFx20 volts and measured open circuit. Other capacitors in metal cases were fine.

Something odd though.. I retested the bad capacitors a few hours later and they looked completely different so is that the explanation for the intermittency? Maybe the contents are expanding when voltage is applied and a leak develops? The manufacturer was Roederstein and they're supposed to be reliable. I retested the capacitors again and all were fine.. I suspect it was slight tarnishing that gave me false results. No, it was a broken lead on my ESR meter. Anyway I powered the 15 volt (Pin 2) rail from my lab PSU and plugged in the board for soak testing. The PSU read 15.00 volts at 713mA of which the Range Selection board accounted for just 13mA of current.

Also, while the board is out I'll test those push switches to ensure the latching/unlatching and contacts are good. The operation of that microswitch is now clearly visible so I can fix it in place and bend its lever to match the toothed wheel. I sprayed the latching springs and all are now good. I then tested all their make/break contact sets and these were all good as well.

I bent the lever on the top microswitch and refitted it. It now works perfectly but I noted several other microswitches dotted around the periphery of the toothed wheel. The are not easy to get at so testing them if there are glitches will have to wait.

Click for larger view

Left, the replacement microswitch, centre... near the front and right... mounted on a circuit board at the rear of the toothed wheel. There are several others also.

These switches are designed to register the dial limits, so the faulty switch that I'd replaced allowed frequencies below 0.14MHz to be generated. Now that a new switch is in place that range displays zeroes if the dial is set below 0.14MHz.

Maybe I'm getting somewhere because the 15 volt test began with a drain of 713mA but after 20 minutes this had dropped by 95mA to 618mA (there's a 15% leak). In other words there's an intermittent and whatever it is the internal supply is being cut off to half a volt when the leak is present... very strange. Later, with the SMDU turned on with the 15 Pin 2 volt rail at half a volt, I coupled it to my external supply via a diode, set the current limit to 1.2A and increased the voltage. At about 10 volts the current dropped to zero after the SMDU suddenly kicked in and the 15 volt rail established. At that point the display showed Range I/9 started up.

I'd connected my scope to the output to monitor the signal out and oddly though, the displayed frequency read wrongly and a second decimal point had appeared. The table below shows the figures. I managed with some fiddling to get a second range working and this also showed a discrepancy. Either a chip in the display counter has failed or the 5 volt supply has dipped too low at a point on one of the circuit boards. The second decimal point that's appeared may also be indicative of this.

TRUE MHz	50	45	40	35	30	25	20	15	10	5
DISPLAY MHz	34	31	27	24	21	17.5	14	10.4	6.9	3.5

Now switch over to Page 2 of the commissioning.

Rohde & Schwarz Test Set SMDU

Model 249.3011.04

Selected drawings for the SMDU

Go to second page of commissioning

Click above for larger picture

SMDU selected set of drawings

Return to Reception