Tech Note 1 Troubleshooting the ICF-6800W (WA) PLL
(last rev September 2009)


Note: the problems of the ICF-6800W PLL mainly apply to the W version with white model lettering on the front. The WA model (serial > 30.000) has orange model lettering. If you have an orange version it is most likely "just" an alignment issue. If you have a white one chances are high you have a real problem.  If you don't intend to dive in deep, stay away from the white W version.

1. PREAMBLE

So... you could not resist and picked up this nice (or scruffy) ICF-6800W... Chances are high it doesn't work on shortwave bands.... The 6800W is a synthesized unit, with a phase locked loop transferring frequency stability of a crystal to the LO (VCO1 in this case). Now these days PLLs are a piece of cake taking a few 10s of MHz reference and a VCO at frequencies up to GHZ directly. One programs the comparison frequency and dividing numbers and out comes the correction signal, you only have to supply the loop filter. Not so on the 6800W... this one use a nice discrete approach for most of the RF pieces of the PLL to get down to a low comparison frequency.  The 6800 incorporates aspects of the Wadley loop but, imho, it is not a traditonal application thereof. Yes, one selects a harmonic of a reference to mix down, but not in the signal path, only in the VCO - so the triple conversion is avoided.
Enough said... all you are interested in is to get it going again. So, let me ask whether:

You have a lot of time to waste.
You have decent mechanics skills and close-by eyesight.
You have good electronics skills, HF centric.
You like to go to bed with schematics and service manual (SM) instead of whiskey.
You have access to some test equipment.
You really really want to get it working again.


If the answer to all the above is yes, please read on. If not, please put the unit back on Ebay. There's always a fellow soul who has not read this yet who's willing to pick it up - with today's prices on 6800W's you may even make a decent profit - doesn't that look a LOT better suddenly? If you take a shot at it but can't get it to work, reassamble correctly and have the decency to tell potential buyers what you did.

2. WHAT'S GOING ON IN MY 6800 LOOP?

Here's a basic flowchart I use to fix (working under the assumption that all else is sort of fine) - spend a few hours on the schematic/explanation of the service manual first so that you understand all circuit principles and how the building blocks relate:

a. The counter doesn't work: not good, you will need it. Check the 10MHz reference, and the divider on the counter board.
b. MW doesnt work: there is no readout on MW and no reception. You need to fix this first. The VFO around Q22 used on MW also feeds directly into the phase comparator IC3. Look into the MW board and the counter buffers Q25/26.
c. SW: the counter shows a number like 89300 or so... VCO1 is not oscillating (most likely) or the counter buffer is gone. Check/clean switches and the capacitor switch. Maybe it is really out of alignment.
d. SW: in the most common problem scenario you will see a number that is a bit higher than the band you are on. For example when on 0/3MHz (=3-4 MHz) the number 4325 is on display, and it does not move a you spin the dial. The good news is VCO1 is working - the bad news is it is running unlocked. Many, but not an unlimited, number of  causes are possible:
- VCO2 does not work (usually misaligned)
- the FET mixer Q31 does not work (not very likely)
- the 10MHz amplifier Q32 does not work (mmm, sometimes is weak, but it should not be top problem, look into this if 10/ range works and 0/ 20/ don't)
- the output of the FET mixer and 10MHz are mixed together in Q33 (a likely source of problems). This one has to work good, or you will not have enough drive for the filter and the subsequent squarer/divider.
- one of the various buffer stages has degraded.

So all in all, not so easy to tell what's going on. It usually is one of the 2SC930 transistors failing. In fact, most of them are below spec after these years on the W version. But since we clean up everything in the squarer some loss of signal is not too bad - we just need "enough" at the end, and can tolerate quite a lot. That is why Q32/33/34/35 are critical. So I don't believe in ripping out all the 2SC930s... You just have to find the key one(s). Replace with, for example, BF241 or new 2SC930. Also investigate Q23 on the MW board if you have weak MW reception and/or consistent partial lock (e.g. XY200-XY500 kHz) regardless of X/Y setting.

3. DIAGNOSING THE PATIENT

Find the Q31 FET, next to it sits C113. A long leg is exposed, and you don't have to do further disassembly once the case is off. Here you can see a lot of interesting stuff. You will load the FET with the 50 ohms of your analyzer causing the voltage to collapse so the PLL will go out of lock - no problem... That is not relevant here.  Also the absolute power measurements are hard to relate to the scope Vpp's of the schematics because of different impedances etc.

This picture show the 1/0 =(10 MHz) setting. One can see all relevant signals: The 28MHz VCO2, the 30MHz VCO1, and the 2MHz difference. The 10MHz is visible, but remember 10MHz mixing stage is not used in the 1/ range.

This picture show the 0/3 =(3 MHz) setting. One can see all relevant signals: The 31MHz VCO2, the 23.5MHz VCO1, and the 7.5MHz difference. The 10MHz mixing stage is used. The 10-7.5=2.5MHz is hard to see in this measurement.

... but you can see it really well at the output of Q33. 10MHz, 7.5MHz and the difference. It is this difference signal that goes into the Q34/35 amp and then into the squarer.

So looking at the FET really gives you a lot of info. Do the VCOs work, is VCO2 locked (you ll see it jump if not because of the sweep circuit) and on the right harmonic, and do the mixers work. That's basically the whole PLL. The chance something is wrong with the squarer/divider/comparator chain is low. You may want to check the Cs in the loop filter, but in the 80ies, Sony components used were in general good quality.

Looking at the input pins of the squarer IC3 also gives a good estimate of signal quality - compare the usually clean VFO (if Q23 is ok) with the signal out of the low pass filter/buffer. Note the importance of Q35/36 for setting Vavg on the squarer/divider - if Vpp is weak the DC bias becomes important.

4. ALIGNMENT

Put all the compartment and PCB shields back in place. If not, you will not achieve lock on several frequencies.

a. First make a good VCO2 alignment
Follow the service manual procedure. A, B, C are around 3.25 2.32 4.25V. Note error regarding adjustment locations in SM. The outputs from 28 to 37MHz in 1MHz steps should be rock solid.  Confirm on FET Q31 as above.

b. IFT 10MHz L25 adjustment, peak for maximum mixer product in the 0/ or 2/ range.

c. Then VCO1 alignment.
This one is a whole lot tricker. The better approach would be to characterise the low and high frequency of VCO1 for each band setting (29 in total) and check for anomalies first. The yellow wire at the underside of the PCB has the VCO1 correction voltage. If you unsolder it you can swing the VCO between min (Vcorr), about 0.8V and max(Vcorr), about 4.7V. You can calculate the capture/lock range of the VCO1 and attempt to adjust the VCO so that it covers the band as intended whilst keeping the correction voltage away from 0.8V and 4.7V.

The procedure in the service manual is a bit more ad hoc. It suggests a setting of the max VCO frequency on each band at max(Vcorr), given factory VCO behaviour. Not a problem as such, but a few considerations need to be made:
i. the VCO behaviour changes because of aging.
ii. you cannot change the 3 decades independently as presented in the SM. CT8 is common to all decades. That's important: you can set up SW1-CT8 correctly and then take a shot at SW2-CT9/SW-3-CT10 never finding a satisfactory setting for the latter. So the alignment process should be iterative.
iii. the band switches and the switched capacitor bank need to be working fine or you will never get it completely right.
iv. the tuning dial position is entirely irrelevant during adjustment. Once you touch the CV1-11, you are basically destroying the VCO2 xMHz signal, and the mixer side of the comparator input goes away. The VFO input remains in place, and Vcorr will go max.
 
OK, let's proceed. YMMV, but this works for me.
Step 1. First check (iii), while disabling the PLL as indicated as one swicthes between the 3 decades the difference x/9 - x/0 should be of similar magnitude. If you have one that is 5 times the others, you are probably losing Cgang in the swicthed capacitor. 0/0 is a bit of an exception being a bit high than the others, say 1500 to 1900 kHz. Adjust incorrect Cgang by adding/removing C on the switched capacitor.
Step 2.  Set x/9-x/0 to be about 9 to 10 MHz. Again, because of 0/0 being a bit higher 0/0-0/9 will be more like 8.5-9Mhz. For this adjustment, use CT8.
Step 3. Align SW2, using L33/CT10. Try not to change CT8.
Step 4. Align SW3, using L32/CT9. Try not to change CT8.
Step 5. Adjust SW1 using L34.
Step 6. Recheck Step 2 if you find higher MHz of a decade not locking. To check this capture first at low end of MHz using dial, then move up.

You should end up after a while with a PLL remaining in lock everywhere, even though it may not capture everywhere immediately upon each switch. I think that's normal to some extent after 30 years and the discrete swicthing.

I like the loop to be tight, and use these numbers as guidelines. Service manual values between brackets. This gives an idea how the numbers can vary.
 
kHz 0/ 1/ 2/
/0 2000 (2145) 11200 (11745) 21200 (21745)
/1 2750 12275 22325
/5 6400 16750 26600
/9 10500 (11045) 21200 (20745) 30750 (30645)

By now, the PLL should be stable... and one can get to the rest of the alignment of the radio... a lot easier.

5. DOCUMENTATION

One can find decent scans of the service manual on the web (Mauritron original). It is good quality except for the PCB layouts... this could be better. Anyone have a better source or pdf?

back