Hartley oscillator stability

[James Arthur]

I read in sci.electronics.design that dave vanhorn <[email protected]>


Isn't it the beat frequency between your two oscillators? If it's
amplitude is that big, I'd say that your mixer is seriously misbehaving.

I'd expect a beat note to be sinusoidal and of comparable
amplitude, especially if the oscillators are latching onto
each other (as has already been suggested and a virtual certainty).

To me these amplitude-modulating 'jumps' sound like squegging,
an undesired oscillation on top of the intended one. The feedback
network C11 * R9 = 470pF * 56K yields a time constant of 38KHz --
suspiciously about 1/20th of your oscillator's fundamental, just
like your bogey, and a prime suspect. Q2's base bypass network
has a time constant that's only about 220nS -- too short. It
oughtta be lots longer.

Try C5,C9 = 10nF, and consider temporarily disabling the other
oscillators (Q3, Q4) to minimize interactions. That done, the
remaining oscillators should then start locking solidly onto
one other near zero-beat, plus other annoying behaviors!

James Arthur
(posted this yesterday, apparently unsuccessfully. Apologies in advance if it
shows up twice.)

 

[dave vanhorn]

I'd expect a beat note to be sinusoidal and of comparable
amplitude, especially if the oscillators are latching onto
each other (as has already been suggested and a virtual certainty).

To me these amplitude-modulating 'jumps' sound like squegging,
an undesired oscillation on top of the intended one. The feedback
network C11 * R9 = 470pF * 56K yields a time constant of 38KHz --
suspiciously about 1/20th of your oscillator's fundamental,

I jumped C11 to 1000pF, seems to help, the wierd thing has not
recurred.

Try C5,C9 = 10nF,

I did that, but for other reasons, the 100pF was letting a lot of RF
onto the pot wiring, which caused tuning problems. It's better, at
least in that way.

James Arthur
(posted this yesterday, apparently unsuccessfully. Apologies in advance if it
shows up twice.)

I know the feeling, for the moment I'm stuck using google groups,
which has a tremendous time lag. :0P

 

[dave vanhorn]

Firstly, your oscillators must produce clean sinewaves. Any clipping or
distortion of the sinewave means the LC circuit is not in control of the
oscillation, and that makes your circuit *much* less stable against other
influences such as transistor variations and coupling from the other
oscillator.

Makes sense to me.

The secret of reducing clipping is to reduce bias current until
the amplitude drops well below clipping level - actually, you are looking
for a peak to peak swing of well under twice the supply rail less bias
drops, so that collector voltage never drops below emitter.

Hmm.. The original design is supposed to run "about 12Vpp".
I increased the drive (emitter resistor to 22k) which has definitely
helped on the mixer side, and I was actually able to play it a bit
last night, for the first time. I tried 10k, but the osc waveform was
somewhat distorted.
Scope says clean at 22k.

Once you have a clean waveform, you have to reduce coupling between the
oscillators. You can use separate LM78Lxx regulators or R and C decoupling
on the power rails of the two oscillators. A CMOS inverter makes a good
simple buffer.

Hmm I didn't think of taking it through a gate before, but that's an
idea.
Probably separate '14's for each osc. Separate regs may be where I
end up also.

Your project sounds just like a simple metal detector I built my kids, where
the search coil is part of an oscillator. A second oscillator can be tuned
to produce and audio beat.

It's very similar to a BFO type detector.
The original theremin was rather clever about the volume control
though, sweeping the osc through a tuned circuit and getting amplitude
modulation.

All this designed with nothing more than a crude meter to work with..
No scopes in the 30's AFAIK.

 

[James Arthur]

[email protected] (James Arthur) wrote in message



I jumped C11 to 1000pF, seems to help, the wierd thing has not
recurred.

Interesting -- I was going to suggest the opposite. First,
doubling that feedback cap isn't guaranteed to cure squegging,
only cut the squegg frequency in half.

Secondly, this massive amount of feedback contributes to the
oscillators' excessive willingness to oscillate over a wide range
of frequencies. Clearly a bonus in a theremin, but maybe a bit
too much of a good thing here given your descriptions -- your
beast sounds a little wild.

Increasing the tank capacitance, as has been suggested, would also
limit each oscillator's "pullability." Were it me, I'd be very
tempted to do exactly that to the BFO oscillators Q2 & Q3 -- they've
been designed to be as squirrely as motion-sensing oscillators Q1 & Q4,
but for no obvious advantage. You could increase tank caps C14/18
a bit, reduce feedback caps C15/19, keep the same operating frequency,
improve stability, and all without having to modify the coils.

I did that, but for other reasons, the 100pF was letting a lot of RF
onto the pot wiring, which caused tuning problems. It's better, at
least in that way.

Good. That likely *will* cure any squegging, plus kill the RF leakage.

Considering the stability of oscillators in ordinary AM
radios -- the local oscillator in my swapmeet-special handheld
stays true to 100Hz or so for days-to-weeks -- methinks
achieving decent stability oughtn't be difficult at'all.

Fun project.

James

 

[dave vanhorn]

I jumped C11 to 1000pF, seems to help, the wierd thing has not

Interesting -- I was going to suggest the opposite. First,
doubling that feedback cap isn't guaranteed to cure squegging,
only cut the squegg frequency in half.

Ok, I did that before dropping the emitter resistors, so maybe it will
be ok with the original value, or even less now. I'll give that a
shot.

Secondly, this massive amount of feedback contributes to the
oscillators' excessive willingness to oscillate over a wide range
of frequencies. Clearly a bonus in a theremin, but maybe a bit
too much of a good thing here given your descriptions -- your
beast sounds a little wild.

The maximum difference between the two, should be about 20-30kHz.
Granted, on this scale that's a little wide..
The original theremins ran at about 180kHz, and had to have the same
range of pull on the pitch VCO.

Increasing the tank capacitance, as has been suggested, would also
limit each oscillator's "pullability."

Tried that before lowering the emitter Rs, it wasn't a good thing.
Maybe I need to also try that again, with the "turbo" enabled

Good. That likely *will* cure any squegging, plus kill the RF leakage.

Considering the stability of oscillators in ordinary AM
radios -- the local oscillator in my swapmeet-special handheld
stays true to 100Hz or so for days-to-weeks -- methinks
achieving decent stability oughtn't be difficult at'all.

Good sanity check!

Fun project.

I'm learning rather more than I expected to..
Playing rather less though..

 

[dave vanhorn]

Boy this is dredging up some old neurons!

Does anyone know the formula for calculating the # of turns of an
inductor when wire diameter is known, and desired inductance, but
length is not?

Assuming air core, close wound.


I'm also looking for a working formula for capacitance and inductance
of a rod of a given diameter and length. I'll probably have to make
some guesstimate with regard to some sort of environmental factor
there, unless I can get there by knowing that a rod of X length and
diameter is resonant at F, and has an inductance of Lx.

I've found some formulas in older amateur radio handbooks but mathcad
vomits when I try to enter them, whining about mismatched units.. I
haven't figured out what exactly it's complaining about yet..

 

[Tim Shoppa]

I've found some formulas in older amateur radio handbooks but mathcad
vomits when I try to enter them, whining about mismatched units.. I
haven't figured out what exactly it's complaining about yet..

The numerical constants in those formulas are usually given without
units. The text accompanying the formulas will tell you things like
"d = coil diameter (in inches), s = turn spacing (in inches),
L = inductance (in microhenrires)", etc. From the formula and the text
you can deduce the units that go with the numerical constants.

For what you're doing, where you don't care much about the frequency
(within say a factor of two), all you really want is for the two
inductors to be fairly well matched (and the mismatch to be fudged out
by whatever tuning you use).

Tim.

 

[colin]

Boy this is dredging up some old neurons!

Does anyone know the formula for calculating the # of turns of an
inductor when wire diameter is known, and desired inductance, but
length is not?

Assuming air core, close wound.


I'm also looking for a working formula for capacitance and inductance
of a rod of a given diameter and length. I'll probably have to make
some guesstimate with regard to some sort of environmental factor
there, unless I can get there by knowing that a rod of X length and
diameter is resonant at F, and has an inductance of Lx.

I've found some formulas in older amateur radio handbooks but mathcad
vomits when I try to enter them, whining about mismatched units.. I
haven't figured out what exactly it's complaining about yet..

ive got an old book wich gives such infoormation, if u give me the length
inside diameter, outside diameter, and required inductance i can tell you
the turns you need, lenght isnt that important but mainly length/diameter
ratio afects the inducatance

(inches and microhenries prefered)

Colin =^.^=

 

[Tim Wescott]

Boy this is dredging up some old neurons!

Does anyone know the formula for calculating the # of turns of an
inductor when wire diameter is known, and desired inductance, but
length is not?

Assuming air core, close wound.


I'm also looking for a working formula for capacitance and inductance
of a rod of a given diameter and length. I'll probably have to make
some guesstimate with regard to some sort of environmental factor
there, unless I can get there by knowing that a rod of X length and
diameter is resonant at F, and has an inductance of Lx.

I've found some formulas in older amateur radio handbooks but mathcad
vomits when I try to enter them, whining about mismatched units.. I
haven't figured out what exactly it's complaining about yet..

If it's similar to the one below the problem is that they're giving you
conversion factors without dimensions. You have something like a^2
N^2/(9a + 10l), which will spit out inch-turns^2. You need to set the
dimensions of your "9" and "10" to inch/uH, then your dimensions will match.

 

[dave vanhorn]

If it's similar to the one below the problem is that they're giving you

conversion factors without dimensions. You have something like a^2
N^2/(9a + 10l), which will spit out inch-turns^2. You need to set the
dimensions of your "9" and "10" to inch/uH, then your dimensions will match.

Ok, that's what I was expecting.
I just picked this back up to look at, and your reply was just what I
needed.

I've set up an iterative approach, which is a pain, but you guess at a
len, and then it calculate turns. From that and the wire dia, I get
len, and if the difference is > 1 turn, you try agian.. :0P But, it
works.

Now, about that 30" antenna rod that I need to know the reactance
of...

 

[Vin Roth]

Waw, sweet thread.

I think I'll actually go with 'bittersweet" as I am in Dave's
shoes so many years later. Well not exactly, not yet.

I have completely built the Tmax but not yet powered it up.
Pretty excited about the project.

Now brushing up on my Hartley oscillator theory on some
random safari on the internet, I find this thread and all of you!

With many hours invested already, I pray that something
use-able will come out the other end. Curious about how
it all came out for Dave.

Changing C11 (and C15?) to 1000 pF gets a thumbs up.
Changing C5 and C9 to 10 nF also.

Thanks a lot.