Ultra wide band oscillator solution needed

[neazoi]

Hello,
I need a circuit for an ultra wideband oscillator.
The simplest the best.
The highest the bandwidth, the best (khz to microwave)
discrete or IC solution is ok.
Is that possible?

 

[BigBoss]

kHz to Microwave is not possible.Max. practical bandwidth of an oscillator is 2-2.5 octaves.

 

[neazoi]

kHz to Microwave is not possible.Max. practical bandwidth of an oscillator is 2-2.5 octaves.

I do not mind switching 1-2 components to achieve max bandwidth.
Please let me know about possible solutions for max available bandwidth.

 

[BigBoss]

I do not mind switching 1-2 components to achieve max bandwidth.
Please let me know about possible solutions for max available bandwidth.

Passing from kHz range to Microwave zone is not possible by switching only few components.
Professional wideband signal generators use many switchable sub-oscillators to achieve this bandwidth.

 

[neazoi]

Passing from kHz range to Microwave zone is not possible by switching only few components.
Professional wideband signal generators use many switchable sub-oscillators to achieve this bandwidth.

An idea I had was to use a comb generator driven by a dc-100MHz modern DDS. At DC-100MHz the DDS will not be followed by the comb (no reason). Then, to produce 100-200MHz, the dds must operate at 50-100MHz and the second harmonic out of the comb must be taken. To produce 200-300MHz, the dds must operate at 66-100MHz and the third harmonic out of the comb must be taken. And so on...
Of course the comb denerator will produce a whole lot of other harmonics if it's output is not filtered, but there will always be a signal (harmonic) that will be able to be tuned, yelding in an always-present tuneable signal without gaps throughout a wide spectrum.
The DDS has sub-Hz tuning capability, so the tuning step of higher order harmonics will be good.

As an example, consider the 10th harmonic of the 0-100MHz DDS, that will cover the 900-1000MHz and will require 90-100MHz on the DDS.
Of course, if the DDS is tuned from DC to the 100MHz instead, then the 10th harmonic will cover the entire spectrum from almost DC to 1000MHz with 10x the dds tuning step accuracy.

How does this idea sound?

 

[D.A.(Tony)Stewart]

Simplest solution is PLL with programmable GaAs prescaler and CMOS divider and VCO.
The higher MHz/V sensitivity on VCO, the more phase noise.

 

[volker@muehlhaus]

Simplest solution is PLL with programmable GaAs prescaler and CMOS divider and VCO.

It's difficult to design PLL with proper loop gain across such a wide tuning range. One approach that we used long ago is to use two oscillators and a mixer, with output signal fo = f1-f2. For example, f1=2.0-3.0GHz and f2=2.0GHz fixed, for output frequency from kHz to 1.0GHz. This avoids a couple of problems with VCO bandwith and PLL loop stability. The only problem when you go down to a few kHz is injection locking of the VCOs, so you need good shielding and good backward isolation between them.

 

[neazoi]

It's difficult to design PLL with proper loop gain across such a wide tuning range. One approach that we used long ago is to use two oscillators and a mixer, with output signal fo = f1-f2. For example, f1=2.0-3.0GHz and f2=2.0GHz fixed, for output frequency from kHz to 1.0GHz. This avoids a couple of problems with VCO bandwith and PLL loop stability. The only problem when you go down to a few kHz is injection locking of the VCOs, so you need good shielding and good backward isolation between them.

I was just ready to post this as an alternative solution, but in a slightly different approach, to increase even more the range.
Two identical VCOs mixed and combined with a few filters (1-2). The VCO stability, phase noise and spurs should be as good as possible.
This vackar VCO https://www.qsl.net/va3iul/High_Fre..._VCO_Design_and_Schematics_files/image010.gif seems to have good characteristics.
I am thinking of make two such identical VCOs and mix them.

1. For dc-600MHz, the first VCO is tuned to 600MHz and the second to 600-1200MHz and the difference is taken.
A 600MHz LPF may be employed at the mixer output if desired.

2. For 600-1200MHz, only one of the two VCOs will be used and tuned to 600-1200MHz.
Theoretically no LPF will be needed at the output.

3. For 1200-2400MHz, the first VCO is tuned to 600-1200MHz and the second to 600-1200MHz and the sum is taken.
A 1200MHz HPF may be employed at the mixer output if desired.

The above, will give a range of (theoretically) DC-2400MHz from a single module.

The other solution is the Si570, but it is way too complex to program and the output is square wave, so lots of harmonics should appear. Not to mention about it's cost and availability problems...

How does it sound?

 

[BigBoss]

One possible solution is to use programmable divider who divides the signal coming from a constant frequency oscillator.
For instance, let the max. frequency is 6GHz that is also osc. frequency, lowest frequency 10kHz, so max divider ratio will be 600000.This number will be programmable 1 to 600000.(20bit divider)
https://www.hittite.com/products/view.html/view/HMC983LP5E

 

[volker@muehlhaus]

One possible solution is to use programmable divider who divides the signal coming from a constant frequency oscillator.
For instance, let the max. frequency is 6GHz that is also osc. frequency, lowest frequency 10kHz, so max divider ratio will be 600000.This number will be programmable 1 to 600000.(20bit divider)

I think there is a flaw in your logic. The divider solution can't cover the entire band with 10kHz steps.

6GHz : 1 = 6GHz
6GHz : 2 = 3GHz
6GHz : 3 = 2GHz
6GHz : 4 = 1.5GHz

 

[FvM]

How does it sound?

You didn't consider the necessary guard band for feasible filters. The design concept sketched by volker@muehlhaus is in contrast a straigthforward approach.

It's also instructive to study the block diagrams of commercial RF generators.

My Agilent ESG generator (250 kHz - 4 GHz) has e.g. a 0.5 - 1 GHz VCO, a switchable 2:1 frequency divider, optional 1:2 and 1:4 frequency multipliers, a switchable filter bank and a downmixer with 1 GHz oscillator and filters for the frequency range up to 250 MHz.

 

[neazoi]

You didn't consider the necessary guard band for feasible filters. The design concept sketched by volker@muehlhaus is in contrast a straigthforward approach.

It's also instructive to study the block diagrams of commercial RF generators.

My Agilent ESG generator (250 kHz - 4 GHz) has e.g. a 0.5 - 1 GHz VCO, a switchable 2:1 frequency divider, optional 1:2 and 1:4 frequency multipliers, a switchable filter bank and a downmixer with 1 GHz oscillator and filters for the frequency range up to 250 MHz.

Sure, however it is too complex to build such a thing. I was thinking more of a single small enclosure module that can easily produce a broadband signal. To be honest I do not know if this mixer approach will work at all below 1MHz or so, as the two oscillator signals are too close and PLL-unlocked VCOs are not that stable.

I think that by saying "guard band for feasible filters" you mean to consider the cut-off slope of the real filters right?

On the other side, the hittie dividers require require a GHz LO to work and I think (not sure) that they produce a square wave output, don't they? I do not think this is an easier solution. It requires the GHz LO which is usually achieved by multiplier and filtering stages and I do not know if the output is clean enough. The output frequency step could also be a problem..

 

[vfone]

If you want a processional example (but not hard to build) of how to mix a 1 GHz LO with a 1GHz-to-2GHz VCO, take a look to the schematic below of the IFR-1500 (RF test monitor)
Go to page 7-14 (Dual VCO Module Schematic).

https://www.repeater-builder.com/test-equipment/aeroflex/ifr1500/ifr-1500-maint-man-rev.pdf

They mix there a 1.21 GHz fix oscillator to a 1.3 GHz to 2.3 GHz VCO, resulting after mixing, a 90 MHz to 1090 MHz signal. You can make the fix oscillator to be 1.3 GHz and you get a signal starting from almost DC.

Special care how to build the two oscillators, because they should be very well thermal coupled (to get the same frequency drift over temperature), but well RF isolated (to avoid any spurious mixing).

 

[BigBoss]

I think there is a flaw in your logic. The divider solution can't cover the entire band with 10kHz steps.

6GHz : 1 = 6GHz
6GHz : 2 = 3GHz
6GHz : 3 = 2GHz
6GHz : 4 = 1.5GHz

It's a fractional divider buddy..

 

[rfmw]

Your BFP420 VCO idea might (or let me say will) give you a very noisy signal in the low sub-1MHz range due to its high phase noise. From your first post I was thinking that mixing two YIG oscillators might be the only "good" solution. But you never explained the final application and the noise levels...

 

[Borber]

Here is project to build spectrum analyzer and tracking generator containing wideband VCO, microwave mixer and microwave oscillator from which you can build 0-1.7GHz generator.
http://lea.hamradio.si/~s53mv/spectana/sa.html
http://lea.hamradio.si/~s53mv/spectana/tg.html

 

[biff44]

some general comments to the above discussion.

I would not use a comb generator (like a diode or fet type) so much as I would use a gigabit digital gate. If done properly, you can create a .5 nS wide single pulse for each input sine wave cycle, and it would act like a pretty good comb generator in the 1 to 1000 MHz range.

mixing two VCOs in a mixer will technically work, but gawd you will have a bunch of harmonics and intermodulation products...so many that I fail to see how it would be useable.

You might be able to instead make a tunable bandpass filter that covers that range, and have a standard DC to 2 GHz amplifier to connect the input and output of the filter, and make it oscillate. You would need to vary the loop phase shift to that you have positive feedback at each desired frequency (some sort of vector phase shifter). Variations of this might include a fiberoptic delay line to make the dPhase/dFrequency big enough to get good phase noise. This would be one frequency...you would only have to filter the output 2nd, 3rd, etc harmonics. No intermodulation products.

 

[D.A.(Tony)Stewart]

Wideband sweep is used in SA's and VNA's. a narrow band microwave VCO from F1 to F2 is generate and mixed with a fixed LO with a tuneable IF that results in a wide multi-decade range Fout.

 

[volker@muehlhaus]

It's a fractional divider buddy..

I see ... but that generates the desired average frequency by switching back and forth between two division ratios. No problem if we use that as a PLL reference. But used directly as a source and looking at the spectrum, wouldn't we end up with two output signals at f/n and f/n+1, instead of one single tone?

- - - Updated - - -

mixing two VCOs in a mixer will technically work, but gawd you will have a bunch of harmonics and intermodulation products...so many that I fail to see how it would be useable.

Agreed, you need to look at the frequency plan and think about all the mixer products. In our example, for 100KHz to 200MHz output frequency, we had used f1=1-1.2GHz and f2=1.0GHz. This was used as the signal source for EMI test instrument.

 

[neazoi]

Your BFP420 VCO idea might (or let me say will) give you a very noisy signal in the low sub-1MHz range due to its high phase noise. From your first post I was thinking that mixing two YIG oscillators might be the only "good" solution. But you never explained the final application and the noise levels...

Although this oscillator is described as a low phase noise oscillator by the author, I believe it can never be too good at these frequencies as a low frequency HF Vackar type. So yes I believe you are right about this. Although I intend to use this for short term measurements, not for permanent installation.

- - - Updated - - -

Here is project to build spectrum analyzer and tracking generator containing wideband VCO, microwave mixer and microwave oscillator from which you can build 0-1.7GHz generator.
http://lea.hamradio.si/~s53mv/spectana/sa.html
http://lea.hamradio.si/~s53mv/spectana/tg.html

I have seen this VCO with great interest. The best version has almost 2GHz range. I would build this if I do not find anything smaller in size. I need two of these a mixer and a filter to acomplish the task.

- - - Updated - - -

You might be able to instead make a tunable bandpass filter that covers that range, and have a standard DC to 2 GHz amplifier to connect the input and output of the filter, and make it oscillate. You would need to vary the loop phase shift to that you have positive feedback at each desired frequency (some sort of vector phase shifter). Variations of this might include a fiberoptic delay line to make the dPhase/dFrequency big enough to get good phase noise. This would be one frequency...you would only have to filter the output 2nd, 3rd, etc harmonics. No intermodulation products.

I think a variation of this tuneable amplifier is the s53mv oscillatoir, which uses a tuneable transformer for feedback, thus it achieves wide range.
I believe that intermodulation will occur only of the mixer is driven hard. Like the old SA612, if it is driven hard then it is almost useless.
And if it is common to all diode mixers to intermodulate even in low drives, then a chip approach may exist?

- - - Updated - - -

Wideband sweep is used in SA's and VNA's. a narrow band microwave VCO from F1 to F2 is generate and mixed with a fixed LO with a tuneable IF that results in a wide multi-decade range Fout.

I am not so sure what you are talking about. How a narrowband VCO and a fixed OSC can lead to a wideband IF?