AGC oscillator chips like Elantec EL4451 or the Motorola MC12061?

[Tim Shoppa]

I've been playing around with AGC amplitude-stabilized crystal oscillators
lately. Usually Meacham bridge type, with either semiconductors
or photocells or the good old #317 lamp for stabilizaion, although
there are some other things I've been tinkering around with. It's
been fun, especially tinkering with CD4007UB's and CA3046's. No real
application, just trying for low distortion sine waves with simple
circuits, but the "simple" circuits that don't use a #317 lamp often
end up with more than a dozen passive components and split power supplies
etc.

As a slight aside, the Silonex Audiohm LED+photocell parts are really cool
for anyone doing AGC/compression of any kind. They work well at least into
the low MHz, despite the "Audio" part of their name.

The Motorola MC12061 and the Elantec EL4451 have been mentioned
here in the past as a single-chip-low-discrete-part-count solution.
But neither of those are made or easily bought anymore.
I saw in the latest Digi-Key catalog that there's a line of 5-pin surface
mount parts from Panasonic, the AN8955SSMTXL being an example, that is some
sort of integrated oscillator chip. I've been unable to find any specs
(or even a pinout) for this chip, though, and it's just a guess of mine
that it might have built-in AGC. Panasonic and Digi-Key haven't
been much help, does anyone here know more details or a URL for a datasheet?
Or maybe recommendations for some other available AGC-stabilized oscillator
IC?

Tim.

 

[dd]

me.

The Motorola MC12061 and the Elantec EL4451 have been mentioned
here in the past as a single-chip-low-discrete-part-count solution.
But neither of those are made or easily bought anymore.
I saw in the latest Digi-Key catalog that there's a line of 5-pin surface
mount parts from Panasonic, the AN8955SSMTXL being an example, that is some
sort of integrated oscillator chip. I've been unable to find any specs
(or even a pinout) for this chip, though, and it's just a guess of mine
that it might have built-in AGC. Panasonic and Digi-Key haven't
been much help, does anyone here know more details or a URL for a datasheet?
Or maybe recommendations for some other available AGC-stabilized oscillator
IC?

Tim.

A maxim uhf oscillator chip could be used.
A chip that gives good LC oscillator results should give crystal
performance that is better by the ratio of the Qs.

 

[Jim Thompson]

A maxim uhf oscillator chip could be used.
A chip that gives good LC oscillator results should give crystal
performance that is better by the ratio of the Qs.

MC12061 is carried by Lansdale Semiconductor, http://www.lansdale.com/

Lansdale carries many Motorola legacy parts, including most of the
chips I designed in the 1960s ;-)

...Jim Thompson

 

[Joerg]

Hi Tim,

It's been 25+ years since I built an AGC oscillator and I don't have a
schematic anymore. Well, I could start an archeological excavation in
the basement if needed. What I remember is that there was no chip, no
bulb and no budget for anything beyond regular passives and transistors.
I used a FET as a controlled resistor and the AGC loop consisted of a
plain old diode circuit. Worked really nice, our RF institute tested the
heck out of that thing on their spectrum analyzer and had a hard time
finding harmonics. Of course, bets had been in place before this test.

Nowadays the FET path is much easier since you can get SD5004 and other
nice arrays for little money. You could probably even build it all
around an AD603. Now that would be the Rolls Royce edition.

Regards, Joerg

 

[Tim Shoppa]

MC12061 is carried by Lansdale Semiconductor, http://www.lansdale.com/

I'm not so interested in buying any particular chip, I'd rather learn how it
did what it did. In AoE there's a circuit shown using it, which seems to
say that there's a sine wave oscillator section and then another section
which squares up the output of the sine wave for TTL or ECL out levels,
but other than that I don't know how it internally did its AGC stuff.

Lansdale carries many Motorola legacy parts, including most of the
chips I designed in the 1960s ;-)

If you happen to have any MC12061 datasheets in a book or pile of papers,
I'd love to see it scanned in .

Tim.

 

[dd]

I'm not so interested in buying any particular chip, I'd rather learn how it
did what it did. In AoE there's a circuit shown using it, which seems to
say that there's a sine wave oscillator section and then another section
which squares up the output of the sine wave for TTL or ECL out levels,
but other than that I don't know how it internally did its AGC stuff.

A technique I first employed in 1968:
The activity of a colpitts type bipolar amplifier is dependent on the
gain of the device this can be reduced by changing the biassing to
decrease the collector current.The gm of colpitts / pierce oscillator
with collector to base feedback resistor to autobias can be reduced by
reducing the entire supply voltage by servo control .

 

[Tim Shoppa]

A technique I first employed in 1968:
The activity of a colpitts type bipolar amplifier is dependent on the
gain of the device this can be reduced by changing the biassing to
decrease the collector current.The gm of colpitts / pierce oscillator
with collector to base feedback resistor to autobias can be reduced by
reducing the entire supply voltage by servo control .

One thing that I'm learning (or I knew but never realized) is that many
ways of implementing AGC will inherently reduce large and even medium-signal
handling ability as bias points shift around and (in recommendations like
yours) you explicitly reduce the headroom available for output swing.

This isn't a problem if you keep the amplitudes low...
For example, the AGC loop in the Sulzer oscillator keeps the amplitude
of the 2.5MHz oscillator at 0.02V P-P. I always thought this was really
miniscule, but I'm beginning to realize that with classic AGC implementations
this may be a sweet spot.

The "new way", using gold-plated variable-gain amps like the AD603
don't seem quite so limited. (I'm sure there's a sweet spot for minimum
THD and it's not having the output go from rail to the other, though!)

Tim.

 

[dd]

One thing that I'm learning (or I knew but never realized) is that many
ways of implementing AGC will inherently reduce large and even medium-signal
handling ability as bias points shift around and (in recommendations like
yours) you explicitly reduce the headroom available for output swing.

This isn't a problem if you keep the amplitudes low...
For example, the AGC loop in the Sulzer oscillator keeps the amplitude
of the 2.5MHz oscillator at 0.02V P-P. I always thought this was really
miniscule, but I'm beginning to realize that with classic AGC implementations
this may be a sweet spot.

The "new way", using gold-plated variable-gain amps like the AD603
don't seem quite so limited. (I'm sure there's a sweet spot for minimum
THD and it's not having the output go from rail to the other, though!)

Tim.

The circuit I described was also for a 2.5MHz 5th overtone crystal in
competition to Sulzer!. the state of the art for crystal standards in
the 60s

The particular crystal design shows very little change of frequency
until the crystal dissipation exceeds 2.5uW, note that because of the
2.5M Q the displacement causes the quartz to approach the stress strain
non-linearity region above this dissipation and slight frequency changes
occur.

I design crystal maintaining circuits by determining the negative
resistance to commence oscillation, say -250 ohms for the 2.5MHZ , then
designing the bias / supply reduction to be able to reduce the negative
resistance to -50 ohms thus encompassing the range necessary to meet
start up and manufacturing tolerance.

You will find that this limited range of activity enables a linear swing
of 3V pp or so with a resistance in the collector.
If an inductor (parallel LC) is employed the swing can be greater but
stability may be compromised does not limit the linear swing. Again
note that as the servo reduces the collector current the start up swing
may be limited but as the activity reduces to a loop gain of 1 the
amplitude can me maintained in the linear swing region.

Note email no longer interrogated due to spam.




can equal a significant

 

[Tim Shoppa]

The circuit I described was also for a 2.5MHz 5th overtone crystal in
competition to Sulzer!. the state of the art for crystal standards in
the 60s

The particular crystal design shows very little change of frequency
until the crystal dissipation exceeds 2.5uW, note that because of the
2.5M Q the displacement causes the quartz to approach the stress strain
non-linearity region above this dissipation and slight frequency changes
occur.

Sure, in my primitive lab (well, I do have a Z3801A)
I can easily see FM of the crystal frequency
at the few ppm level as I crank Vcc up and down on a CMOS oscillator.
When you're aiming for stabilities 1000 times or more better (like in
a Sulzer or competitor) that must be enormous.

I design crystal maintaining circuits by determining the negative
resistance to commence oscillation, say -250 ohms for the 2.5MHZ , then
designing the bias / supply reduction to be able to reduce the negative
resistance to -50 ohms thus encompassing the range necessary to meet
start up and manufacturing tolerance.

You will find that this limited range of activity enables a linear swing
of 3V pp or so with a resistance in the collector.
If an inductor (parallel LC) is employed the swing can be greater but
stability may be compromised does not limit the linear swing. Again
note that as the servo reduces the collector current the start up swing
may be limited but as the activity reduces to a loop gain of 1 the
amplitude can me maintained in the linear swing region.

How much "extra" gain do you end up designing in? A few tens of percent,
a factor of two, more? My interpretation of your negative resistance
numbers would lead me to believe that 50 out of 250 ohms is twenty percent.
That sounds reasonable.

Tim.