bjt phase noise

Hi Ulrich,

Allow me also to welcome you to the forum, hopefully you can contribute on a regular basis. I have always appreciated your combination of theoretical foundations and a practical approach. It's a privilege having you here with us!

Just to pick up an unanswered question from the thread above: The FM noise of the crystal is the dominant noise contributor at 10/100Hz offset. How would you try to improve this performance?
* having crystal with better specs (better mechanical design which reduces chance of having acoustic flicker noise
* using noise feedback as you suggested in numerous publications?

Thanks already for your advice!

The drive level depending noise is really NOT an acoustic noise but closer to a flicker frequency noise. This can be best modeled by replacing the dynamic capacitance with a tuning diode of appropriate value. The voltage is loss resistance x rf current. This dynamic effect depends on errors in the crystal structures and could be found with an electron scanning microscope, which of course does not cure it.

Later in the day I will send a dc to 10 KHz patented feed back circuit, that largely but not fully reduces the effect.


Ulrich

http://www.tu-cottbus.de/fakultaet3/...r-patente.html

Hi Ulrich,

Thank you for your prompt response. Allow me to explain my experiments in a little more detail.

We have a product, which synthesizes an RF LO from a XO reference. We do not have a phase noise analyzer which is suitable for crystal oscillators, but as the reference is multiplied to microwave frequencies, so is the phase noise. As long as the loop bandwidth is a few tenths of kHz that's fine.

What we did was take a batch of 40 crystals and put them one by one in the oscillator circuit. Some resulted in good phase noise at 10/100 Hz, others were way off spec. that is 30 dB higher than the other. Hence my conclusion that the phenomenon was caused by the crystal and not the oscillator and the origin is therefore acoustic flicker noise, likely caused by the design and performance of the transducers.

Another clue: The manufacturing was done by a mainstream Chinese mass-producing site. Low cost but likewise quality ...

Do you have another explanation for this?

Thanks

This is a well known problem. Selection is a choice. Yes, to multiply it up several GHz is good if the analyzer is low noise .I use the R&S FSUP 8 with correlation and new enhanced dynamic range. If you are it that business such a phase noise system is a must. The Agilent is not so good below 10KHz, correction factor seems to be off, but better above 1MHz offset. Nice to have both

What is your requirement in phase noise ? Wenzel is the world leader (many SSSS) and I am trying to become less dependent of the crystal phenomena as you experience it . Please wait until my feedback circuit will be uploaded later.

Ulrich,

Welcome to this forum!

It is a real privilege and honor to have you on this forum.
I have many of your books and have read many of your publications. All excellent stuff.

Yes, Wenzel seems to be leading in low noise XTAL osc design at the moment. Their oscillators are very expensive and with very long lead times. One company I know have outstanding orders for their super low noise series for more than 8 months. So it appears that they are struggling to get a good yield for these.

Hope to see you around more often here.

E

Her is a circuit with about 10KHz feedback to reduce the close in noise. This approch is patended

Ulrich

Hi,

most of the crystals intrinsic noise ist generated by imperfections of the surface grinding and interactions between crystal surface and electrode material. Also "dirt" on the surface before deposition of the electrode causes noise from the crystal.

Unfortunately the BVA type resonators are not made anymore. Their construction avoided many surface-to-electrode associated problems.

Oscillator noise can be minimized by clever circuit design but close to the carrier the parameters of the crystal (Q and intrinsic crystal noise) are most dominant.
If you have a really bad crystal your oscillator will definately not deliver premium noise performance.


Ulrich,

can you tell me where the signal is fed back into the input of the oscillator amplifier?

Excuse me for not seeing it in the schematic.

Regards,

Schmocki

The feedback is part of the bias circuit. For this reason, have to be careful tuning the circuit because is possible to get some bumps in the SSB phase noise characteristic, which instead to improve, actually will degrade the phase noise at particularly offset.

Hi,

I see the 270nH and the 33pF, but isn't the BC857 too slow to get a reasonable feedback signal out of the collector? And there are the two capacitors of 1nF and 4.7pF that effectively ground RF at this node.
Also the 10k resistor does not let much of any RF signal through it.

Or does the feedback rely on the CCB of the BC857 and the parasitic capacitance of the 10k resistor?

Regards,

Schmocki

There is a guy in the UK that claims to have designed XTAL oscillators with better performance than Wenzel lowest phase noise units. Below are results of his claims for a 10MHz osc. He also claim to use a common available crystal, nothing really special.
One company I know wants to purchase one of these, but this guy wants more than 20,000 GBP for these so they are still considering if they should order a unit. The only thing it will be very difficult to verify the claimed specs because very few in the world can measure that low accurately.

Interesting reading
http://www.febo.com/pages/oscillators/wenzel_uln/

Twenty thousand quid for an idea like this is not too much for performances shown in the graphs. Would be interesting to see at least the block diagram.

Nice simulation results

I'd be curious how the measurement results really look like...

Schmocki

As I understand the company wants some agreement from him not to target their own client base which includes Gov, NASA and the like, and in return promise to keep him busy with many orders. He is still unwilling to do that. The units are not in production yet, they have been shown lab samples only. They also offered to pour cash into help setting up production facilities, but he is still non-cooperative. Sounds like he wants the whole pie himself.

Well, nice simulation which also violates the laws of physic. There has to be a flicker contribution, 30dB/decade, Here is the measured phase noise of what I consider the best possible solution . Also here is a simulaton plot for a 5 MHz Osc. The better noise is with a low fT transistor. To measure this you need at least the Wenzel multiplier .

I am not (yet) an expert in this technology but we have a Government contract to investigate this and I belief this can be solved at low cost with my few patented methods. The goal is to develop an affordable oscillator with the best performance and normal crystals. Well we will see.

Ulrich

Yeah, I can believe that using audio transistors give lower noise. I wonder if it is possible to get low drift performance from a standard AT cut crystal. Most high stab circuits seems to use SC type at an elevated temperature which may be higher than 70 degC.

Ulrich, how much noise does electronic fine tuning contribute to the total noise in comparison to a fixed or mech tuning?

Ulrich,

that's why I wrote nice simulation, too. Of course, if this designer found a transistor with a flicker corner frequency below 1hz... .

Can you explain to me how the feedback works in the circuit you posted?

And do you still consider the bft66 the best oscillator transistor?

Regards,
Schmocki

Added after 6 minutes:

You need SC type as AT cut crystals have a much higher dynamic F/T dependancy which introduces much worse short time stability due to small temperature fluctuations.
A much more sophisticated oven design you would need to compensate this for would kill all of the cost savings an AT cut could give you.

Schmocki

Phase noise and frequency accuracy are two different issues. You are referring to an oven controlled oscillator, the SC cut has higher Q, less temperature dependence . I don't do oven controlled oscillators I am just interested in the basic noise theory and improvements but I am not in the business of making these, this is pure research ..

For these oscillators you MUST have a coarse mechanical tuning and a less then 1 Hz electronic tuning

Below 30 MHz I would use 2N4416 and above a modern bipolar transistor with Ft of about 5 Ghz, and Ic max of 60 to 100 mA.

In the DC circuit, the PNP transistor samples the DC current up to some offset, and within the loop gain and bandwidth compensates this. The third transistor act as a temperature stabilizing diode. It inverts the voltage drop at the output, including the noise by -180 deg and feeds it back so canceling these contributions to some high degree

Ulrich

I believe that to get low phase noise you need to run the osc transistor at fairly high currents. How does this affect crystal dissipation and long term aging? I have read somewhere that it is a balance between acceptable phase noise and keeping crystal current low. Any views on this?