[patrick-turner]

I replied to Phil about his the oscilator dial mech.....

"""Well Phil, your dial certainly seems nice by what you say, but how accurate
are the 6,000 points on the dial? It would seem that the dial length around
the 4.4 inches is roughly 13.8 inches long, 351mm, so distance between each
"point" is 351 / 6,000 = 0.058mm and too small to be useful."""

Phil larfs 'is 'ed orf....

** ROTFLMAO !!!!!!!!!!!!!!

351 / 200 = 1.755mm

There a 30 turns of the dial for 180 degrees on the gang.

.... Phil

OK,OK, whatever.

Meanwhile, after I got back from a 115km bike ride I had a look at the ancient ex PMG Wien Bridge oscillator I must have re-wired for the 3rd time maybe 8 years ago or more.

I last made it with 5 ranges, 1Hz to 22Hz, 10-220, 100-2,200, 1kHz-22kHz, 10k-220k. I figured some overlap of F in the ranges was easy to do and useful, but not absolutely necessary.
Dial is 125mm dia and has around 300 degrees of turn, and no reduction drive was used - not deemed necessary, but there could be one.
I did once try two twin gang tuning caps for the first time I re-wired the unit before 2000.
The tunings caps have been replaced with an Alps "Black" 27mm mm square body normal direction 49.2k dual log pot, which means lowest F is with maximum clockwise turn of the F control knob. You get used to that in 2 seconds.

The pot tracks have a series R = 2k2, so the R range is actually 2k2 up to 51k2 giving C values of 3u3 for 1Hz up to 330pF for 220kHz. The dial was calibrated in pencil and later in blank ink on cream cardboard, then glued to front fact and then varnished. Its accurate within +/- 3%, mostly better..

The Alps pot gives "silky" F adjustment with very little Vo jitter, and no bounce occurs. I have ZDs back to back to clamp the voltage across the fixed R of the NFB network. No clamp between Vo and 0V was necessary. The operation is actually better than subsequent oscilators i made with opamps and pots or switched R, although the latest switched R type are quite nice to use.

The NFB network has 12 seriesed 12V x 50mA lamps.

Vo at output of tube stages = 16.4Vrms, and 5.46Vrms is acoss the lamps, and the current = 9.1mA, so the lamp R = 600r, so each lamp R average = 50r, and cold they measure about 27r, so they are operating at twice their cold R with average of only 0.45Vrms across each lamp, enough to cause a doubling of the cold R, and into the non linear region of lamp R increase..

The fixed R ( with parallel trimming Vr = 1,200r, and so the NFB network has total R = 1,800 r at normal operation.
The PFB network minimum Zin occurs when R = 2k2, and is a total of 4.67k and I in = 3.51mArms. The output attenuators are 10 log pot in series with 1k0 log pot so I get 0V to 1.4V, or 1.4V to 14.0Vrms. Ouput attenuator current = 1.26mA, rather feeble, but OK for most high Zin amps. No output buffer is used after the output level pots.
So the current max = 9.1 + 3.51 + 1.26 = 13.87mArms, or 19.5mA peak.

The tube line up = 12AU7 V1, V2 differential pair with one anode bypassed to 0V, gain approx 14 with other anode feeding V3 gain tube. PFB feeds V1 g, and NFB feeds V2 g, and commoned cathodes taken to current 8mA sink with MJE340. V3 = 6AC7, old metal jacket tube made maybe 1940. it was a high gain+gm octal pentode used commonly on instruments and for HF. Still very usable. Gain approx 200. Total amp gain = 14 x 200 = 2,800.
V3 feeds a White Cathode Follower with V4, V5 = 2 x 6CM5 with +250V anode supply and Ia = 25mA. This gives a surprisingly low Z class A output to drive the loads.

THD is not obviously noticeable on the CRO, and there no sign of any clipping, so I guess THD 0.5%, mainly 2H.

I do recall TRYING to get F higher to 1MHz, and only getting to 800kHz with a 6DJ8 input pair, and C would have needed to be 72pF in theory where R minimum = 2k2.
But trouble was that oscillations and Vo become unstable and slew rate limited at the high Vo level and the dial accuracy wasn't good above 300kHz. Maybe more Ia in white follower and and less Vo would have been more sensible, and an additional output cathode follower after the attenuators, but room in the box is a bit limited. The 5 pos range switch has one more position blocked, could be 100kHz to 2.2MHz maybe.

So this unit does the 1Hz to 220Khz very well, but I could have a tuning gang and a dial for a separate high range 100kHz up to whatever I might get. For this range, if the R ( fixed ) = 1k2 then C minimum for 1Mhz C mimumum = 132pF, and then oscillations should not become unstable and trim caps can adjust it for optimum because there is stray C lurking, plus input C of the tube grid. But then to get a decade of F the max C of a single gang would need to be 1,320pF, and that would require 3 gangs at say 445pF each, so there's a job for 2 x 3gang caps.

Depending how well the two triple gang caps worked, I could use the tuning caps for top 5 ranges and the largest fixed R = 12Megohms. The 50k pot can work to do the 1Hz to 10Hz. My experience is that the tuning caps can give surprising accuracy with 1 dial for 5 ranges because its easier to get accurate R than fiddle around with paralleling C to just the right C value.

The beauty of halving the Vo to 7Vrms means that if loads remain the same R, then the amp current output is halved. But loads can be reduced, for better HF, and same Ia in white follower should work OK, although maybe 35mA would be better.
The NFB network for 7Vrms would use 600r in series with 6 x 12V lamps in series, so that NFB network is a load = 900r. The attenuator can be lower value, and mimimum R value in PFB network.

I'll prepare a schematic of what I have now for my website because its not a bad oscillator even without the extension to 1MHz, and anyone could build it without any hard to get parts. The 6CM5 could be replaced with say 2 x EL84 in triode, and 6AC7 replaced with EF80/6BX6 or EF184/6EJ7. Plenty of these around, and the 12V x 50mA lamps are around, or better still, the type 327 lamp, 28V x 40mA are also easily getable. in a box fulla toobes, a j-fet or opto coupler would seem out of place.
The oscilator circuit would fit on an old AM radio chassis. The wooden box or bakelite box would be removed, and a new metal box made to fit. RF coild, tuning cap, and IF trannies and audio amp and speaker are all removed. There are plenty of such radio chassis around which are not at all worth restoring. Perhaps the dial mech parts could be saved to make a horizontal sliding dial, but to get 300mm of movement you need a wheel just the right size fixed on the pot shaft. Much thought needs to be applied. But I leave such imaginative craftwork to the thousands of old guys who like nothing better than to farnarkle around in their sheds with old gear.
Patrick Turner