Phil mentioned the drive mech for the variable tuning cap in the wien bridge oscilator he has....
Post by Phil AllisonThe shaft is driven by a backlash free, 60:1 worm reduction drive and a
200 point scale.
Verrry silky.
** Ambiguity fix.
The 200 point scale is engraved on a 4.4 inch dia steel disk that rotates
with the worm drive, making 6000 setting points per range. 30 turns x 200 = 6000.
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.
And so often, a maker produces a dial in a prototype and then the cap used for production is slightly different to the prototype and dial is the same, and maybe a bit wrong.
I like to see no more more or less than 3mm between easy increments in frequency, and so the total number of markings around a fully used dial or 351mm would be 117, not 6,000, but then maybe I have a metal picture quite different from what a photo of your oscillator might show. Unfortunately, r.a.t does not allow images, probably because if we were allowed, the porn pedlers would saturate this web service with zillions of terrabites of cunts, arsoles, and megalitres of spoof sprayed on ladies' faces.
So, to escape the excessive and banal sexual ways of the world at large we all need our own website so that pictures of oscillators can be displayed with minimal interference.
I also like a frequency dial to have equal distance between octaves of F.
In my last effort with a 1H-1MHz WB oscilator, the F is controlled with a 12 pos switch for 12 F per decade, based on these numbers, 1, 1.25, 1.5, 2.0, 2.5, 3.2, 3.9, 4.7, 5.6, 6.8, 8.2, 10.0.
Now this seems awkward, because there are hardly any whole number F stops. But why do ppl insist we have convenient whole numbers? Well, its because much in electronics varies according to a logarithmic rate, and once anyone adopts the thinking behind Standard Resistance Values, then one begins to understand you don't need to depend solely of linear scales at all. Minds of our recent ancestors gave values some thought.
Standard R values are 1.0, 1.2, 1.5, 1.8, 2.2, 2.7, 3.3, 3.9, 5.6, 6.8, 8.2, 10.0. That's 12 values and sure you could use this number sequence instead of my sequence. But I wanted at least 4 nearly close octaves, 1.0, 2.0, 3.9, 8.2 per decade, and if you read a 6dB drop between 1kHz and 2kHz, then you immediately can suspect a first order LPF.
So I have slightly adjusted my resistance values to get there.
Having a 24 position make-before-break switch would be better! Maybe DACT make something suitable, but to get R values you start with a number just over 1.0 and then when that is multiplied by itself 23 times consecutively, the total of 24 numbers are used as factors for the R value that gives the lowest F and the 24 number should give 10 times the lowest R value. Then you have to carefully series or parallel R to get within 1% of the R values. Its all a lot of work.
But consider a well made dial with 12 marked stops on the F I have suggested.
Its not difficult to place additional dial marks at say 0.1 or 0.2 or 0.5 increments and highlight Whole Numbers with a longer marking, and this takes the guess work out of reading the dial. But with ONLY 12 F per decade range, you don't need to know any other number than I've suggested to plot a response, and you just have a sheet of paper with all F marked along a LOGARITHMIC sacle where magically, all my numbers appear about EQUALLY spaced, and much less than an octave apart so a reponse graph for 1Hz to 1MHz is a sequence of up to 72 dots, although some repeat at 1 & 10 so actually 66 dots are used, and when the dots are joined you get a smooth enough curve. You know what F you have by just reading the number on the dial.
But a variable cap does not have an exactly equal dial spacing as I have, and a circular dial or sliding horizontal scale like a radio set must be calibrated according to the F achieved when the unit is properly adjusted and the dial will only suit that tuning cap and no other. Every time you use such a thing you never repeat exactly the same F, and just what F you do get strains your eyes to interpolate what you have. An F meter is a godsend for those who don't trust the dials or F stops with a switch. My F meter is a 1Hz to 50Mhz kit I built about 15 years ago, I got fed up with guesswork, especially when using a function gene with all chips inside and with a crummy dial F control where markings could be 20% in error.
But today, I tested the tube stages I have created around a phase change network with 3 x R and 3 gang tuning cap each 30pF to 400pF. Well, I was most dissapointed. The darn thing was far more critical to adjust than anyone online has described, and the F stability was approaching a theremin. At all F above 50kHz, making voltage readings stopped oscillations. The NFB and PFB had to be ever so carefully adjusted, and a full decade F variation was impossible.
It was similar when I changed R from 3 x 1k5 to 3 x 150k.
So, the more I experiment with phase shift oscillator, the more difficulties arrise to be solved, and I think that explains why the Wien Bridge oscillator reigns supreme because it can be made to work so much more easily and with less parts.
The fundemental problem with a phase shift oscillator is that the voltage obtained after the 3 x RC H-pass sections is one that rises in amplitude id F goes a little lower, and falls if F goes a little higher. Now this is a far more difficult starting premise than if you have a wien bridge because at F0, the network output either side becomes lower. and at Fo, the is 0 degrees of phase shift, and at all F the ratio of Vo to Vin = 1/3. Then you can have a flat response more easily, and with a NFB network using a j-fet or LDR as a variable R to ensure the oscillator Vo remains at a constant level.
The LC oscillator for all F above 50kHz is a good option because L does not vary much and nor does the F value of a tuning cap. The LC has a higher Q than any RC arrangement, unless you have a twin T RC network or bridged T RC network
and some sort of active device and a NFB arrangement. LC is at least very simple, especially if you consider a j-fet with a tapped single coil and tuning cap. See the Hartley at http://qooljaq.com/LC_Osc.htm
But a C change of say 40pF to 400pF gives an F change of 3.16 times, ie, say from 50kHz to 158kHz. The L value must be 25.34mH. Then if the L value is changed to 2.534mH, you get 158kHz to 500kHz, and so on. The well made HP 606 I have has a beautifully made dial but there is a scale for each F range, because the coils used all have varying Q and must cope with skin effect and so on and not even HP could make an instrument with just 2 scales, 50 to 158 and 158 to 500, and 500 to 1580, and so on. The coil properties determine the dial marks.
My dear old 1993 kitset F reader is somewhat slow to read any F, but it does appear to be accurate enough. A later kit that was more expensive went to 1GHz, and I thought 50MHz was good enough.
But the WB oscillator can have just ONE set of dial numbers 1 to 10 for all 6 ranges with accuracy that is +/- 3% and good enough for 99% of what i do with the unit - test amps and stuff.
I have plans for a WB oscillator with multiple 3 gang tuning C. I have even more 2 gang types and they may be all fitted with the same size of plywood disc made by using a 70mm hole cutter and then hooked up with a dial cord that is backlash free. A fair sized box needs to contain the caps with electonics on top so heat from electronics does not heat the caps. But even with max C = 0.003uF, the R for WB must be 26.5Mohms for 2Hz.This R value is what the ancient AWA unit I had used, except lowest F was 20Hz with just 40 - 400pF 2 gang cap.
It seems criminal to send so many tuning gangs to the recycler to make modern junk the world wants. But the world isn't much happier.
Patrick Turner.
