Building the Project
These instructions are somewhat brief but I will be happy to help with questions and problems, should you wish to build the project. The photo below provides an overview.
Click on the icon at right for a bill of materials in PDF format or click here for Excel 2002 format. Sorry that the total cost figure can’t be taken as a valid estimate. For reference, my cost estimate before starting was $533. The BOM figure isn’t valid because: (1) it doesn’t include items on hand, it DOES include extras for stock and vendor considerations and (3) certain items are extravagant. I wanted this reproduction to be “first class” and was willing to spend extra to avoid risks and blemishes. Examples include the $6 line cord (was more flexible and longer), $30 5Y3 tube (the backstory made it trusted NOS), $41 for a once-common, 10W output transformer (wanted a new one) and $55 for tube sockets (chose snap-ring NOS type, similar to ones used in the original). For those who aren’t very concerned about reproducing the original closely, the $61 L2 choke could be eliminated with a solid state regulator. The $13 bottom cover could easily be left out; I haven’t even attached mine (yet). I splurged on four of the pots, at $11-each. (Was concerned about stability.)
On the other hand, the $18 NOS, 25W rheostat at right was a bargain, as I was looking at $50 for a new one. Check out the foil-lined, sealed packaging! Also, the crystal is still listed as “on-hand,” even though Bill’s donated original didn’t work. What it will cost from eBay vendors depends on how many you need to buy to find the right one.
Perhaps the most difficult issue found when buying parts was finding a suitable crystal. Bill had supplied the original, vacuum-tube-packaged crystal he used (seen at left) but sadly, it was internally disconnected. I later saw that the solder attaching a lead to the crystal had slightly separated from the golden coating on the quartz. This also happened to another unit of this series which was 100kHz, instead of 120kHz.
I tried to order one (in any package) from many different crystal manufacturers, willing to pay dearly for a good unit. But most can no longer even make such large crystals. Generally, the lower frequency niche is now occupied by tiny, “tuning fork” types. Those have much higher series resistances than what this circuit was designed for. The only 120kHz crystal stocked by Digikey, for example, is specified at 12Kohms, whereas ones like Bill’s achieved 500ohms. The oscillator would no doubt be less stable with the higher resistance, if it worked at all.
Fortunately, eBay came through with a good crystal but I had to buy about ten units before finding the right one for this circuit. (Sellers don’t usually test each crystal.) The chosen unit is a James Knights (now CTS-Knights) H-9, packaged in a can about two inches long (at right) and dated July, 1953. I mounted it under the chassis, in the corner near the oscillator, wiring it in parallel with the existing octal socket. Bill’s original still proudly occupies the socket, giving the reproduction an element of authenticity.
Most units tried were too far off to be adjusted on-frequency. Part of the problem is that there are two principal ways to operate a crystal (series versus parallel resonance) and they’re ground to slightly different frequencies. This type of oscillator circuit requires a parallel resonant crystal. Those operate with a specified “load” capacitance, often in the 10-100pF range. Fortunately, cap values in the oscillator can be changed (within limits) to match the load cap the crystal requires.
If you try a crystal and the frequency is too high, increase the load capacitance (Cload). (Decrease it if the frequency is too low.) C33 and C3 must be changed together to vary Cload. As the circuit stands, Cload is 89pF. The table below gives you values for C33 and C3 to produce different load capacitances:
Cload C33 C3
16pF 0pF* 33pF *C1+C2 reduced by half
29 0 68
35 10 82
42 22 100
49 33 120
59 47 150
71 68 180
89 100 220 Current values
104 120 270
124 150 330
136 180 330
At the larger values, it might be better to use larger trimmers for C1, C2. Else, small values of cap may be paralleled to center C1 and C2 in their adjustment ranges. The only downside to larger trimmers is that adjustment becomes more sensitive. Caveat: Values in the table other than the current ones have not been tested; some experimentation may be needed.
Different Story for 50Hz Clocks
Of course, all this about 120kHz crystals assumes that you have a clock which operates on 60Hz power. If you have a synchronous motor clock designed to operate on 50Hz, the crystal frequency would be 100kHz, which is much easier to find. The divider frequencies would change but should be within the adjustment range of the pots. The real problem would be that (probably) the clock motor would also need 230V instead of the 120V which the PA can deliver. A voltage conversion transformer might not work, as the additional core losses could load the PA too heavily. That part of the circuit could be tested on the bench to see if it would work though.
Machining the Chassis
Templates (at right) simplified the chassis work. The single front panel hole is documented on the rear template. Original CorelDraw (X5) files are here: Top chassis, Rear chassis. The Top chassis file has a layer for the Template and another for the Wiring diagram.
Attaching the Templates
You can print the 15x9” template using a letter size inkjet printer (at right). It requires two prints, with the long direction of the paper vertical with respect to the images above. I scaled the images slightly on a temp copy to correct printer error in the feed direction. Using a small light table helps in aligning the two at the small overlap. Clear tape on the front and back held them together. The rear template was printed the same way and the results were quite accurate. Cut the templates exactly on the outside line.
In affixing them to the chassis, it’s a good to minimize residual adhesive to avoid a messy cleaning job after the template is removed. 3M Spray Mount worked well for this but I needed to let it dry for a lot longer than the one-minute specified for repositionable bonds. I guess it was about 10-minutes but you should experiment before trying the real thing. I still had a trace amount of residue but Goo Gone quickly took it off.
Apply a single, light pass of Spray Mount and let it dry for 10-minutes. To position accurately, it’s helpful to put a sheet of wax paper between the template and chassis. Slide the wax paper out to expose about 0.5” of adhesive. Align carefully and press down lightly with one finger, on one side of the exposed strip. You can then adjust the rotation at the other side and press down there. After bonding the strip, gradually slide the wax paper out while pressing down.
It’s best to bond and drill only one template at a time, else there will be conflict at the corner of the top and rear. Note that the outer black line indicates the full outside extent of each surface. A light blue line shows the inside wall of the enclosure. There is no template for the single hole in the front but placement details are noted on the rear template.
About the Rear Panel
The rear view of the Quartz Clock at right reveals that the AC outlet, output test points, phase indicator, power switch fuse and line cord are located on the rear panel. Why are the power switch and phase indicator put there? Those weren’t in the science fair project and needed to be kept out of sight for the sake of the reproduction. Since it’s a clock, the power switch on the back seems appropriate, anyway.
Photo Tour of the Construction
Next, we have a fast, visual overview of building the project:
1. The template is used to drill the top panel holes. Rear panel template is done separately.
2. Chassis drilling is done. Also, the rectangle for the AC outlet is completed using a nibbling tool. Next steps are polishing the chassis using a wheel and applying lacquer.
3. Holes are being punched for tube sockets. Note that the template shows 3/8” holes for the draw bolt. Your punch might have a different size draw bolt. The NOS Cinch 8R1 (9634) sockets require a 1-11/64” punch. Great sockets, by the way. The “wavy,” snap-ring mount saves drilling screw holes.
4. The completed chassis with a couple sockets. Notice the single hole at the left side of the front panel.
5. All chassis parts are mounted except the transformers and choke and power cord. Leaving those until the end makes the assembly much lighter and easier to deal with. Be careful not to hit the rheostat windings; they’re fragile. The greenish trimmer cap assembly is seen at the lower right. See image 10 for details.
6. Completed chassis wiring.
7. Topside of completed quartz clock.
8. The pictorial diagram made wiring easier, reducing errors. The top view below is more readable but the mirrored bottom view in Image-9 is what is used to guide the actual wiring. Clicking the image brings up a fairly high resolution GIF file but the PDF version here has better resolution, as it’s vector-based. (Note-2)
9. The mirrored, bottom-view, pictorial wiring diagram below is used to guide the actual wiring. (Compare to Image-6.) Clicking the image below brings up a fairly high-res GIF file but the PDF version here has better resolution, as it’s vector-based. (Note-2) Wire the 6.3V heater, ground and +150V lines first, keeping them close to the chassis. Then add the signal wires, keeping them away from heater lines. Transformers and choke are mounted and wired at the end.
Initial Checkout of the Completed Clock
After finishing construction, it’s time for the initial checkout. Needed test equipment includes a dual-trace oscilloscope, digital multimeter (DMM) and an accurate frequency counter. A variac is optional. We will first want to make sure that no damaging current or voltage is present. Preparation steps include:
- Install the 2-amp, slow-blow fuse.
- Insert the tubes.
- Set rheostat R30 to full CCW (max resistance).
- Plug-in a synchronous motor clock to J1.
- Set R13 and R18 to full CW (max resistance).
- Set all other pots to mid-position (Note-4)
- Set the two trimmer caps to mid-position.
- Connect the DMM between ground and +250V.
- Power the unit from a variac, if available.
Start with the variac at minimum; raise the line voltage gradually to 120V while observing AC line current. It normally runs about 0.56A but at initial turn-on may well-exceed 1A. With a gradual increase in voltage, it should not exceed 1A. If it does, shutdown and check for problems. Else, verify that the DMM reads roughly 270V, as mine did. If yours deviates more than (say) 20V from this in either direction, shutdown and check.
Next, check the voltage (relative to ground) at V7-8. It’s nominally about 12.2V and let’s us calculate cathode current for V7: Ic = 12.2/270 = 45.2mA. If you see more than about 50mA, shutdown and check.
Measure the voltage across R36, which runs from a lug on the rheostat to the terminal strip near V1. It shows current on the +150V rail, expressing amps as volts. As I recall, it was about 24mA (seen as 24mV). If you see more than 30mA, shutdown and check for problems. Note that the 150V rail voltage may be low at this point.
Connect the DMM across R34 to measure the V9 voltage regulator current. The half-watt resistor is on the terminal strip between V9 and L2. Adjust rheostat R30 for 20mV across R34, indicating 20mA current. Now measure the voltage from ground to the 150V rail. I found 155V there. If you see a deviation of more than 10V from that, something must be wrong.
Verify that the 6.3V heater voltage at V1 (pins 2,7) and the 5.0V heater voltage at V8 (pins 2,8) are each within 10%. This completes the build phase of the project.
Operational Tests and Adjustments
Now we’re ready to do operational checks and adjustments: For this, you will need the oscilloscope with a 10:1 probe attached. Initially set the scope for AC coupling and 50V/div sensitivity (5V/div and the probe). You should see roughly 13Vpp at 120kHz there and roughly 150Vpp at V2-2. Be careful about attaching a counter to the topside test points, as high voltage is present. Use a 10:1 scope probe on the counter, which should be AC-coupled. Beware of electric shock!
About a Frequency Counter
It’s best to use a frequency counter which can show fractional hertz, without having to count actual cycles. This is called a “reciprocal” counter, as opposed to a “direct” counter. For example, my HP5316B reads frequency with 0.01Hz resolution in just a few seconds. A cycle-counting (direct) counter would need 100 seconds to make such a measurement. At least 0.1Hz resolution is needed to properly adjust the oscillator. Also, the counter time base must be very accurate; 1ppm would be nice. If it has a 10MHz reference output, you can check it against WWV. Reciprocal counters from the 1980 HP catalog include: 5300A+5307A, 5315A/B, 5370A, 5335A and the 5345A. As of this writing, working examples can be found for around $75 on eBay. Incidentally, with the 5316B (below), NO beat with WWV could be found from its 10MHz ref output. (It has option 004, which is the high stability oven time base. See Note-3.)
Adjusting the Oscillator
We covered the issues related to finding the right crystal, above. If you have the right crystal, you should be able to tune the oscillator both higher and lower than the 120kHz center frequency. Two adjustments are provided, coarse (C2) and fine (C1). The coarse trimmer is on the right in the top-view wiring diagram. Connect an accurate frequency counter (see sidebar), with a 10:1 scope probe, to TPA. The displayed frequency should be close to 120kHz. If it’s more than (say) 15Hz off, something may be amiss; C3 and C33 may need to be changed, as discussed above.
The clock should be run at least an hour before adjustment (preferably for days). First center C1 in its range; then adjust C2 to bring the frequency as close to 120kHz as possible. Use C1 to correct any residual error. You may need to touchup the adjustment a day or two later.
To put the oscillator accuracy in perspective, here is a little table of the errors involved:
Hertz PPM Sec/Day Sec/Month
0.05 0.4 0.04 1.1
0.12 1.0 0.09 2.6
0.5 4.2 0.36 10.8
1.0 8.3 0.72 21.6
Adjusting the AMVs
Ideally, the divider test points, TPB, TPC, TPD and TPE would show signals at 12kHz, 1200Hz, 240Hz and 60Hz, but they can vary widely, unadjusted. We will adjust the AMVs, starting at V3. Notice that each AMV is associated with a tube, a test point and two pots. For the first, we have V3, TPB, R8 and R9. Ones which have equal pots (V3, V6) are adjusted by varying the two pots together, keeping them about equal value. The two asymmetric AMVs (V4, V5) are adjusted by varying the high-value pot only. The lower value pot is left at maximum resistance (full CW). Those pots were included as an insurance policy and to allow for possible future improvements to the adjustment procedure. Note that turning the pots CW decreases frequency (increases the time constant). AMV adjustments are summarized in the table:
Tube Adjust Test Point Frequency
V3 R8,9 TPB 12kHz
V4 R14 TPC 1200Hz
V5 R19 TPD 240Hz
V6 R22,23 TPE 60Hz
Bear in mind that the AMVs are rather squirrelly and adjusting them is something of an art. They’re affected by adjustment of the preceding stage and, to a small extent, the subsequent stage as well. With the counter’s probe attached to TPB, adjust R8 and R9 in concert to set the frequency to 12kHz. Be wary that the complex waveforms present sometimes cause counters to show gross errors in frequency. Use an oscilloscope at the test point to verify that the frequency is roughly accurate. You might need to adjust the counter’s sensitivity or threshold to get a good reading.
Carry on with the table, in given order. For each divider, we want to set the adjustment in the center of its sync range. That leaves some margin for drift to keep the AMV properly synchronized. To center it, find the settings at which sync is lost in each direction; then set it halfway between. If two pots are used, find the center with both of them kept at the same position. Then center each one individually. Since the dividers interact, it’s a good idea to repeat the adjustment process. After proper adjustment, my unit has maintained solid sync for months, without further attention.
Setting Output Level
Whereas the original design achieved an output of only 100Vrms, the modified version is able to produce up to 130V. With a synchronous motor clock plugged into outlet J1, adjust R28 for 120Vrms at the output. I found some hum from the power line in the output, causing the amplitude to vary with the relative phase of the line and the clock. Variation was 7Vrms, indicating a hum component of 3.5Vrms. After eliminating the power supply and the V7 grid signal as possible sources, I had to conclude that it must be magnetic coupling between choke L2 and T1. However, that isn’t confirmed and it needs further investigation. I set the output to swing between 115 and 122Vrms.
The phase indicator lamp (NL1) should slowly vary between darkness and full brightness. Depending on local power line frequency, it might spend some time, not appearing to change. If you suspect something wrong, you can check 60Hz output with a scope on TPF. Set for AC line sync, the TPF waveform should drift slowly in phase.
Clicking the thumbnail at right brings up a PDF file with:
- DC and AC voltage readings
- Measured values of adjusted multivibrator pots
- Waveforms of two multivibrators, the driver amp and power amp
As mentioned before, I will be happy to help if you have questions or problems.
The issues found in the original design all turned out to be correctable with only modest changes. Happily, this reproduction remains true to my brother Bill’s original science fair project, for the most part. Considering the effort that I had to put into it, I’m very impressed with how well that high school student did with it, back in 1961. I doubt I could have come up with a better layout of the chassis-mounted components.
The results of extended operation have been gratifying. As you can see from the chart at right, the oscillator has been quite stable. However, I have to admit that the room temperature has also been fairly stable during much of the time period. The deviation with rising temperatures after 4/24/16 suggests that some temperature control for the crystal or perhaps use of temperature compensating components might be needed. During the summer, I have to allow the house temperature to rise well into the 80s to acclimate for running. That’s quite a challenge for a time standard!
This has been an enjoyable and challenging project. It certainly has been a real time machine, taking me back to those days, so long ago, when I watched in awe as Bill traveled this path.
Many thanks go to my dear brother Bill, who helped with this project in so many ways. His encouragement, historical research, precious crystal, memories, photos, copies and scans have made a huge difference.
<<<< Back to “All About the Time Machine”
Note-2: Please note that the 100pF cap values shown in the wiring diagrams, have been changed.
Note-3: In comparing the 5316B 10MHz reference output to WWV, I was unable to detect a beat with a receiver —was too close. A spectrum analyzer set for it’s minimum 10Hz bandwidth and 20Hz/div span, couldn’t show any distinction between the reference and WWV. (The signals were adjusted to the same level.) All I can say from that is that the difference is considerably less than 10Hz (1ppm). It’s specified at 0.03ppm/mo drift. When introduced in 1988, the 5316B cost $2075 (about $4200 today), including the high stability oven option 004. Recent sales of working 5316B counters on eBay have been in the $100 range, including one with option 004 for $90.
Note-4: Instead of presetting the pots to center, you could set them according to the measurements given in the “Quartz Clock voltages and waveforms.pdf” file.
The Quartz Clock’s colorful glow at night.
(30-second exposure at f/5, ISO 1600)
Copyright © 2016 by Stephen H. Lafferty