Calculable Amplitude Colpitts Oscillator (?) - Elecraft XG2

[quantumvolt]

This 50 (and attenuated to 1) microvolt signal generator is probably intended for calibration of S-meters. On page 8 in:

www.elecraft.com/.../E740084%20XG2%20Manual%20Rev%20E.pdf it reads:

"Output accuracy is determined by the precision voltage regulator and 1% resistors."

To me the circuit seems to be just an (emitter feedback stabilized operating point) Colpitts Crystal Oscillator with a precision supply voltage and a very low current and output amplitude (collector tapped) attenuated and presented with 50 ohm output impedance.

My question is - keeping in mind the formulas for Emitter Resistor DC-Biasing - will not the actual transistor's Vbe and Hfe influence the collector operating point current and also (?) the RF-signal amplitude?

EDIT: Can't get the link over to work. A description and the manual is linked in http://www.elecraft.com/mini_module_kits/mini_modules.htm . It also says (new to me - quite a claim ...):

"In addition to receiver testing, the XG2 can be used as a reference to calibrate other lab instruments."

Whaddayathink?

[T3sl4co1l]

If it's running with a high enough feedback ratio and load impedance, the amplitude will be limited by Vce(sat), which has little tempco.

If it's running with a lower load impedance, it might be limited by Vbe instead, which would be a bit of a problem.

It doesn't strike me as a precision device, not one that can be expected to hold constant with respect to temperature.

Tim

[quantumvolt]

Thanks. I've been thinking about that: If the oscillation occurs from transistor cut-off to saturation, then the math works out so so ignoring Vce-sat.

But can one expect 1-3 (or 5) % accuracy? Does anyone know of other HF-frequency oscillators that has a predictable calculable amplitude?

[uncle_bob]

Thanks. I've been thinking about that: If the oscillation occurs from transistor cut-off to saturation, then the math works out so so ignoring Vce-sat.

But can one expect 1-3 (or 5) % accuracy? Does anyone know of other HF-frequency oscillators that has a predictable calculable amplitude?

Hi

A standard generic cheap CMOS output oscillator will give you a specific output frequency. If you put it into a divide by 2 (D flip flop) the duty cycle will be pretty close to 50/50. It also should be fairly predictable from the data sheets on the divider. Run that into a buffer ('125) that runs off of a precision supply. The logic high and low should be quite close to the supply and ground. It's not (yet) 50 ohms, but you have a very predictable peak to peak output.

Next step is to run an L pad to get it to 50 ohms. Exactly what you run depends a bit on the logic you are running and the frequency. Something like a 470 ohm combined with a 56 ohm is s good start. You now have a pretty good 50 ohm source and a known p-p signal. A check of the relative amplitudes at the fundamental, second, and third harmonics will give you a pretty good idea how well the divider is doing. Since it's a square wave, you have levels that should be useful out to at least the 5th and possibly the 7th harmonic.

Bob

[quantumvolt]

Thank you. Yes, a rail to rail square signal has crossed my thought. But my primary goal is to see what I can do with analog RF (only HF) circuits first.

I rigged up an LC oscillator at around 300kHz. I managed to get a decent correlation over a wide voltage range (LT 10mV to GT 700mV) between (the visual scope and) the AC Voltage function of an HP 34401A and a cheap eBay AD8307 based RF power meter that also displays mV - like the one here http://www.eham.net/ehamforum/smf/index.php?topic=103001.0.

So I will go to 1Mhz / 10MHz crystal oscillators first and check the power meter against the scope. Then I will try to build a precision rectifier front end for the 34401.

If I can get a precision oscillator working at around 0dBm / 223 mV RMS in 50 ohm, I should be able to use the instruments I have. Then I can step down towards microvolts by standard dB attenators.

RF millivoltmeters are few and IMO overpriced on eBay. So I will try this way first.

But I might go for the square wave also. I might learn more about filters and other stuff.

[uncle_bob]

Thank you. Yes, a rail to rail square signal has crossed my thought. But my primary goal is to see what I can do with analog RF (only HF) circuits first.

I rigged up an LC oscillator at around 300kHz. I managed to get a decent correlation over a wide voltage range (LT 10mV to GT 700mV) between (the visual scope and) the AC Voltage function of an HP 34401A and a cheap eBay AD8307 based RF power meter that also displays mV - like the one here http://www.eham.net/ehamforum/smf/index.php?topic=103001.0.

So I will go to 1Mhz / 10MHz crystal oscillators first and check the power meter against the scope. Then I will try to build a precision rectifier front end for the 34401.

If I can get a precision oscillator working at around 0dBm / 223 mV RMS in 50 ohm, I should be able to use the instruments I have. Then I can step down towards microvolts by standard dB attenators.

RF millivoltmeters are few and IMO overpriced on eBay. So I will try this way first.

But I might go for the square wave also. I might learn more about filters and other stuff.

Hi

A conventional oscillator will have a voltage dependance and a temperature dependence. It's output level (without an AGC) will also be a function of the crystal resistance and tuning. None of that is a deal breaker in terms of an accurate source. Simply don't tune it, run it on a voltage regulator, and don't subject it to large temperature changes. As with any other circuit, you will need to isolate it well from the load and do a good broadband match. A 20 db pad is a pretty common solution to most of those issues.

Bob

[quantumvolt]

Yes, good practical advice. My main attempt will be the analog way. But it turns out I should make a CMOS Clock Oscillator for DC mV accuray level calibration of the AD8307 [url]http://www.cqham.ru/forum/attachment.php?attachmentid=52007&d=1269630856[/url ]. And then maybe a filter to get the fundamental ... Should keep me occupied for a while.

Thanks again for all answers.

[uncle_bob]

Yes, good practical advice. My main attempt will be the analog way. But it turns out I should make a CMOS Clock Oscillator for DC mV accuray level calibration of the AD8307 [url]http://www.cqham.ru/forum/attachment.php?attachmentid=52007&d=1269630856[/url ]. And then maybe a filter to get the fundamental ... Should keep me occupied for a while.

Thanks again for all answers.

Hi

Filters add a lot of complexity to something like this. They are always going to be sensitive to source and load impedances. Essentially the impedances "tune" the filter. That can / will / does change the level you are looking at in a different way than a 50 ohm source. The net result is an added error term.

The main use for precision level sources is in calibrating gear like spectrum analyzers or in testing receivers. In both cases the device you are running it into will select out the fundamental for you. A broadband power meter will not select the fundamental. In that case you use the square wave power as your calibration reference.

If you want to protect the filter from load variations, the quick and dirty way is to put a 10 or 20 db pad in front of and after the filter.

Bob

[quantumvolt]

Yes, I know. The calibrator in the link over does include the harmonics in the output power / DC, so a filter must not be used.

However - I need training in making IF filters by software, cookbook and experimentation. Most filter schematics out there need impedance matching networks either for going down to 50 ohm or up to a few kohm (MC1350P). So when I am at it I will see how well I can do in isolating the 10MHz sinusoid from the square with a specified termination impedance filter. No further reason needed.

[uncle_bob]

Yes, I know. The calibrator in the link over does include the harmonics in the output power / DC, so a filter must not be used.

However - I need training in making IF filters by software, cookbook and experimentation. Most filter schematics out there need impedance matching networks either for going down to 50 ohm or up to a few kohm (MC1350P). So when I am at it I will see how well I can do in isolating the 10MHz sinusoid from the square with a specified termination impedance filter. No further reason needed.

Hi

If the objective is to play with filters, then by all means, play with filters. Simply be aware that anything was designed as a filter is likely to effects that significantly degrade the level accuracy.

Cascading matching networks with filters is equally a dangerous thing to do. The right way to do it is to synthesize the entire circuit as a properly terminated filter. Unequal termination filters are simply one of any filter categories. Sensitivity analysis is one way to dig into the problem, it often is less than fully helpful. Monte Carlo simulation is another way to go, it also has its limits.

Bob