Challenge Thread: The fastest breadboard oscillator on the mudball

[joeqsmith]

If you keep increasing your speeds and I will have to find some better parts..   Has your signal dropped from -10 to below -40dBm or are you attenuating the signal?   

Everything I have shown so far was the fundamental.   

Good luck and looking forward to watching your progress! 

[joeqsmith]

I'll match your 2GHz.... 

[trash]

I might call for some adjudication of joeqsmith's "Breadboard" oscillator.

Not to sound like I'm sour, quite the opposite. I giggle with an evil laugh at all the ways I can easily bend the rules and take the title with no effort. If I use minimal effort then I consider myself to be cheating. The very point of a contest is to get the maximum effort.

Lets start with is actually oscillating here?
Is his source as he says in the video the YIG, because they YIG is what I consider to be an integrated circuit.
If you remove the breadboard from this picture does the oscillator stop working?

Def: Integrated - with various parts or aspects linked or coordinated.

The YIG itself is an active device. You give it power, it gives you signal. All of the components that make it work are inside the YIG chassis.
For the same reason I have not just walked into this competition and pulled a 70GHz gunn diode module off the shelf with a varactor multiplier to take it to +200GHz. By that logic somebody will then bring an LED to the contest and claim 470THz before I retaliate with a Walton-Cockroft voltage multiplier and start cranking X-rays of at 50PHz.  And that would actually qualify as a valid signal. I can then take it one step further with a source of gamma rays. An Am-241 source stuck in one hole of the breadboard and we're done.


Ok, so lets come back and look at what the limiting factor is here. It's the breadboard. If the breadboard is not integral to the oscillator, then it's not a breadboard circuit, it's an external circuit plugged into the breadboard and defeats the purpose of the engineering competition.

Now the power requirements I have not paid a lot of attention to simply because when the definition was first set nobody thought we'd be playing in the GHz where 0dBm is a lot of signal.  There's nothing stopping me just putting an amplifier on the output and I'm back up to +3dBm and well within the rules.   It's a bit of a known quantity.  So I stick with keeping the signal clearly defined above the noise.
 That's easy for me on a spectrum analyser. 10dB above noise floor is almost no signal. If there was doubt, I'd just amplify the carrier.

I've also tried to stick with the generated frequencies of the oscillator so far. I'm not sure if I consider multiplying up fair play.
My dirtiest trick is just to use a waveguide varactor diode and plug the lead into the breadboard, dump the highest frequency signal I can into it with as much power as possible and claim the highest frequency on the comb as the winner.

 My problem is not so much with using the varactor diode as the multipler, but the waveguide is not part of the breadboard.
My dirty trick would be to claim the waveguide as the transition interface from the breadboard to the spectrum anaylser. 
Cheating logic says I can't connect the varactor to the breadboard and it needs the waveguide as a chassis to support it.

If I can get away with that, then I can change the varactor diode for a gunn diode in exactly the same chassis and claim a win of 70GHz without any real effort.

So you can see from my own posts the limitation of my output frequency is the breadboard, not the gunn diode module that I can attach to it.
To avoid the use of the module I have even though of ways of soldering leads to the gunn diode so I can plug it into the breadboard and then run it without the modular chassis.  You've seen the trickery I can do with a 1970's tunnel diode.
Make the breadboard work for you, don't fight it.

Anyhow, would some people like to consider if Joe's YIG oscillator is fair play and then I'll make my next move
I would also like comment on any of my own circuits. If you think something is cheating, please by all means speak up, because I will cheat any chance I can get.  (within the rules). And don't think I can't better the energy levels on Am-241 gamma rays

[joeqsmith]

I might call for some adjudication of joeqsmith's "Breadboard" oscillator.

Not to sound like I'm sour, quite the opposite. I giggle with an evil laugh at all the ways I can easily bend the rules and take the title with no effort. If I use minimal effort then I consider myself to be cheating. The very point of a contest is to get the maximum effort.

 

Bend the rules as much as you like.   It won't bother me one bit and I will gladly turn over the lead position.   Let see what you come up with. 

Lets start with is actually oscillating here?
Is his source as he says in the video the YIG, because they YIG is what I consider to be an integrated circuit.
If you remove the breadboard from this picture does the oscillator stop working?

Def: Integrated - with various parts or aspects linked or coordinated.

The YIG itself is an active device. You give it power, it gives you signal. All of the components that make it work are inside the YIG chassis.
For the same reason I have not just walked into this competition and pulled a 70GHz gunn diode module off the shelf with a varactor multiplier to take it to +200GHz. By that logic somebody will then bring an LED to the contest and claim 470THz before I retaliate with a Walton-Cockroft voltage multiplier and start cranking X-rays of at 50PHz.  And that would
actually qualify as a valid signal. I can then take it one step further with a source of gamma rays. An Am-241 source stuck in one hole of the breadboard and we're done.

Ok, so lets come back and look at what the limiting factor is here. It's the breadboard. If the breadboard is not integral to the oscillator, then it's not a breadboard circuit, it's an external circuit plugged into the breadboard and defeats the purpose of the engineering competition.

The YIG is only used as the local oscillator for the downconversion.   I was actually concerned that some viewers may not fully understand and this is why I showed the signal at the splitter with the YIG disconnected: 

https://youtu.be/AllzRi4uCy0?t=88 

The downconversion process was explained here:
https://youtu.be/D33lCZAYmMM?t=376

If you watch the last video, you will notice that was not using the downconverter to display the waveform on the scope.  I explained the reason why at the start of the video.  I had tried it and in 8GHzYIG1, you can see the our oscillator running at almost 12GHz.  The YIG was adjusted to 8GHz for this test, giving an IF of 4GHz.  It looks somewhat clean but all of the parts used in the downconversion take their toll on attenuating the signal.   In 8GHzYIG2, you can see the scope is reading roughly 4GHz from the IF. 

13p2GHz is showing pretty much the limit without getting a better part.   Notice again, there is no YIG.  Just a counter.  I would have liked to have shown what it looks like on the scope but the YIG and mixer I have just won't cut it at these speeds. 

Of course, the power supply, counter and sections of coax are not mounted on the breadboard.   From the OP's rule:  "7. Post an image of your breadboard and counter/scope(Rule 6!) and brief info on the type of oscillator used."  I understood these were not part of the oscillator.   I am at the point where I can no longer display the waveform on my scope.

Now the power requirements I have not paid a lot of attention to simply because when the definition was first set nobody thought we'd be playing in the GHz where 0dBm is a lot of signal.  There's nothing stopping me just putting an amplifier on the output and I'm back up to +3dBm and well within the rules.   It's a bit of a known quantity.  So I stick with keeping the signal clearly defined above the noise.
 That's easy for me on a spectrum analyser. 10dB above noise floor is almost no signal. If there was doubt, I'd just amplify the carrier.

I don't see a problem with it being low.   

I've also tried to stick with the generated frequencies of the oscillator so far. I'm not sure if I consider multiplying up fair play.

The OP had covered this "EDIT:  of course it is allowed to use common frequency doubling techniques, e.g. distorting and leeching harmonics. Knock yourself out."  Also see their comment below:

For a schematic with a diode that generates avalanche noise near reverse breakdown, followed by a filter and an amplifier, would that be cheating?

If you can pull that off with discrete components, go for it.
I would expect that you will have a hard time getting a narrow enough discrete filter to pick a single frequency from noise. But i'm always ready to be proven wrong.
Therefore, my advice would be to build the fastest oscillator you can, use a snap-off diode to clamp it, and filter one of the fundamentals that are created by the brutal distortion of the diode. Harmonics will have enough separation to have a chance of selecting one without an opamp.

Again, it's just for fun.   

My dirtiest trick is just to use a waveguide varactor diode and plug the lead into the breadboard, dump the highest frequency signal I can into it with as much power as possible and claim the highest frequency on the comb as the winner.

 My problem is not so much with using the varactor diode as the multipler, but the waveguide is not part of the breadboard.
My dirty trick would be to claim the waveguide as the transition interface from the breadboard to the spectrum anaylser. 
Cheating logic says I can't connect the varactor to the breadboard and it needs the waveguide as a chassis to support it.

If I can get away with that, then I can change the varactor diode for a gunn diode in exactly the same chassis and claim a win of 70GHz without any real effort.

So you can see from my own posts the limitation of my output frequency is the breadboard, not the gunn diode module that I can attach to it.
To avoid the use of the module I have even though of ways of soldering leads to the gunn diode so I can plug it into the breadboard and then run it without the modular chassis.  You've seen the trickery I can do with a 1970's tunnel diode.
Make the breadboard work for you, don't fight it.

Anyhow, would some people like to consider if Joe's YIG oscillator is fair play and then I'll make my next move
I would also like comment on any of my own circuits. If you think something is cheating, please by all means speak up, because I will cheat any chance I can get.  (within the rules). And don't think I can't better the energy levels on Am-241 gamma rays

Hope this helps explain the purpose of the YIG and downconversion.   

[trash]

Yep, that's all cool. Originally I figured the YIG was your down converter Local oscillator, just the last video looked like you were using it as the oscillator. 
 I'm not going to go off the reservation to try and win a title.  I'm staying on the breadboard
And my hope is that other people will realise that the circuits we're using are actually really simple ones.  I have about five or six different common base oscillator circuits or variations on them. They're simple and stable but I've never built them on breadboard before, which is the challenge.

I've still got some great legit dirty tricks up my sleeve and when I hit a limit I'll toy with those ideas.
But I'm a long way from component limits at the moment.
I'll also be away from my work bench for a few days so you won't see any more development on it until next weekend.




 

[trash]

For those interested in the circuit I'm using as the basis, it is a comm base oscillator similar to this circuit.
I change the design every 15 minutes to improve something but this is the basis for most of them.
The transistor I prefer to use is a BFR96, but I do use other transistors. You can even try running this circuit with a 2N2222A.
It works but you may find that you're having trouble breaking past 600MHz until you add some more bits

If you really wanted to make the competition fun, you'd restrict everybody to BC548's
You can even get 1GHz out of them if you're clever.

[0culus]

Interesting thread, this. I was looking through my components and I came across some 2SC5227A epitaxial planar silicon transistors I guess came with a kit I had bought. They are mounted on a tiny PCB with standard header "legs" for breadboarding. Wonder what one could accomplish with these? Info seems to be rather scarce though....not even any suggested applications in the datasheets. 

[joeqsmith]

.... I came across some 2SC5227A epitaxial planar silicon transistors I guess came with a kit I had bought. ...

I don't see why you couldn't use it.  If you have a way to test it, I would give it a try. 

[0culus]

.... I came across some 2SC5227A epitaxial planar silicon transistors I guess came with a kit I had bought. ...

I don't see why you couldn't use it.  If you have a way to test it, I would give it a try.

My fastest scope is 1 GHz, but I do have an 8566B spectrum analyzer available. I'll brainstorm a bit and see if I can't come up with something.

[joeqsmith]

Your scope and especially your SA would work fine.  Can you use your SA to downconvert the signal for the scope? 

[0culus]

Yeah, it has a standard 21.4 MHz IF out. I'm "warming up" by playing with Colpitts oscillators on the breadboard tonight. So far, with a 2N2222 and absolutely zero lead trimming I have achieved 19.5 MHz. Paltry, of course, compared to what you guys have been achieving but still faster than anything I've ever breadboarded. Is Manhattan style construction allowed?   

[tautech]

I'll match your 2GHz.... 

And raise you 10 !   

Still, from what I've read you might have a worthy challenger for the title in trash.
 

[T3sl4co1l]

Seeing as I can't compete in the unlimited class, how about an entry for the vacuum tube class?

B+ supply 106.7V, from a DC-DC module; heater 6.0V (the power resistors are just to take up the slack, the bench supply is running about 10V).  Tube is 5702, a subminiature version of 6AK5, an RF pentode.  Ideal for breadboarding!



Running a good 233MHz, with harmonics apparent:



(This is a plot from my 8590A spectrum analyzer, and I don't know why the text is so badly placed.  Axis is 100 to 800MHz, detector is peak, amplitude is 10dB/div, 0dBm ref (top axis).)

There's also a 20dB attenuator in line, between the pickup link and transmission line.  So that's about 9dBm real power.

The tube is wired thus (pins from right to left):
A: B+
G2: B+
H2: GND
H1: 6V
G3: GND
K: 4.7uH + 220R to GND
G1: 3.5 inch jumper ('U' shaped) to GND

A 22nF film cap bypasses B+ to GND.  A short jumper ties G2 to A.  A slightly taller jumper ties G3 to H2.

Another single turn jumper, in a BNC-binding post adapter, couples to the G1 link and goes off to the spectrum analyzer.

The circuit is completed by the ~4pF between adjacent positions, making this a common-anode Colpitts oscillator.

With over 9dBm of output power (and probably a lot more with a little tuning and stronger coupling), it seems to qualify for the purposes of this thread.

The self-resonant frequency of this tube is ca. 400MHz.  It's not a hard cutoff; some amplification can still be possible up there.  It's just very awkward to make use of, and might not even be usable as an oscillator.  More compact types are preferred -- Nuvistors and planar types in particular.

This is a good point to mention some vacuum tube physics -- which is not necessarily poorly known (and not just because tubes aren't well known today), or unexpected; more that it's the kind of thing that, if it never crosses your mind, you'd never know, but if you see it and think about it, it'll become obvious what's going on.

At these frequencies, the mass of the electron cloud itself is substantial; the grid input impedance will be around 200 ohms || 9pF I think.  (That may not seem like much compared to a BJT or MOSFET, but for a device that's normally comfortable with 10s of kohms around it, that's quite low!)

Mass by itself is conservative, so manifests as increased grid capacitance -- when cold, the capacitance of the grid to surrounding electrodes is only about 5pF, but when the cathode is hot, an additional ~3pF appears, and a little more when hot and biased on.  These are due to the space charge and electron beam!

The resistance?  Literally, work done on the electron beam.  This power modulates the beam's intensity and velocity.  And since the beam also carries space charge, the modulation is sensible on subsequent electrodes -- screen, suppressor and plate!

When the modulation reaches the plate, the beam intensity of course is received as plate (DC and signal) current.  The velocity modulation smears it out, effectively reducing AC (signal) plate current at very high frequencies.  Very fast tubes use incredibly close cathode-grid-plate spacing for this reason.

This works even when the electrode is not drawing beam current.  That is, there is a nonzero transadmittance y_g1g2, y_g1g3 and so forth.  This is a nonreciprocal effect, too -- that is, y_g1g3 >> y_g3g1.

Reciprocity is a big deal, thermodynamically speaking.  It generally means there's an energy input to the system.  There may not necessarily be actual power gain, but the resulting isolation is still very unusual otherwise.  In this case, of course, the heater is hot, and plate current is flowing, so there is ample input to the system.  (Indeed, the cathode itself is a heat engine: electron emission is spontaneous even at 0V bias, thus net electrical power is generated.  The efficiency is of course laughably pitiful, yielding microwatts or less for ~watts of heater power. )

In this circuit, g2 is bypassed to ground, and g3 has a useless cross section (it really is just a suppressor, or at best has a few uS of transconductance to the plate, and even less coming from g1), so there isn't any good illustration of these effects with this tube type.  It does however make things interesting for types with significant grid transconductance: dual-control pentodes, gated sheet beam tubes and pentagrid converters!

Tim

[BreakingOhmsLaw]

Updated record to 13,11 GHz.
@joe: sorry for late update. job kept me busy.

Kudos to T3sl4co1l for bringing a tube to the fight!

[madires]

Lovely! Tim, I'll give you 10 vintage bonus points! BTW, when you reach the microwave range you could add a traveling-wave tube for more output power.

[T3sl4co1l]

Alas, about all I have among the fancier types are a klystron or two, neither of which is very amenable to breadboarding! 

Aside from the de facto easy answer of the 2.45GHz tube in my (and everyone else's) kitchen...

Tim

[joeqsmith]

Wow, a tube!!  This is the first time I have ever seen one installed in a breadboard.  I own an old Millen 90651 grid-dip meter that can run up to 300MHz.   Looking at the schematic, they used a 9002.   

http://www.r-type.org/pdfs/9002.pdf

I watch glasslinger on YT where he's made a few tubes from scratch which I find very interesting.  This one is an X-ray tube:

[joeqsmith]

Yeah, it has a standard 21.4 MHz IF out. I'm "warming up" by playing with Colpitts oscillators on the breadboard tonight. So far, with a 2N2222 and absolutely zero lead trimming I have achieved 19.5 MHz. Paltry, of course, compared to what you guys have been achieving but still faster than anything I've ever breadboarded. Is Manhattan style construction allowed?   

I think most of mine move around more than 22MHz.  Maybe yours will be more stable.   If not,   I had a look at the manual for your 8566.  It appears you may have access to the first LO on the front of the instrument as well and it has +5dBm drive.   Appendix H talks about it.

https://www.eevblog.com/forum/index.php?action=post;quote=3095892;topic=243355.100

You may be able use this as the LO for an external mixer that could directly drive your scope.   Guessing it would be a lot more stable than the open looped YIG I show.   With your GHz scope, you may be able to see something over 7GHz. 

***
Some advice, make sure you show the signal on the SA as well or it may confuse Trash into thinking you are using it for your oscillator.. 

[tim_]

It was suggested to me that I join this competition with a ring-oscillator based contraption of mine, which I quickly reproduced on a breadboard.

So, let's have a look at my entry:



As you can maybe make out, it is a three stage ring oscillator based on 2N3904 (chinese off-brand) and 390 Ohm load resistors. As a specialty, the output of each inverter (collector) is connected to the input of the next inverter (base) with an LED. So what can it do?



When we apply 10V of bias to it, the LEDs will light up and it oscillates at a frequency of 35 MHz. As you can see, this design also follows the rule of a full 5V swing. The frequency suggests a tpd of 4.8 ns for each inverter, which is already "interesting", if you look up the storage time of a 2N3904 in the datasheet. Btw, with red LEDs this goes up to 55MHz, corresponding to 3ns. The same design is much faster on a PCB, so the parasitics of the breadboard take their toll.

Ok, but maybe that is not the most interesting part of this oscillator. Let's have a look at some X-Y plots where X is the input of one inverter and Y is at the output:



At 10V we see a somewhat circular output, which suggests that the phase delay of this inverter is constant - as expected for a proper oscillator. However at 11.8 V we can suddenly see a convoluted path that almost looks random, but is clearly repeating - a limit cycle? Changing voltages further can transition from noise to different structured trajectory.

It looks like this simple circuit is acting as a chaotic oscillator. I though that was quite neat. You can find a longer write up here: https://hackaday.io/project/170697-evaluating-transistors-for-bipolar-logic-rtl/log/179154-using-a-led-as-base-resistor-chaotic-ring-oscillator.

[SilverSolder]

It was suggested to me that I join this competition with a ring-oscillator based contraption of mine, which I quickly reproduced on a breadboard.

So, let's have a look at my entry:

(Attachment Link)

As you can maybe make out, it is a three stage ring oscillator based on 2N3904 (chinese off-brand) and 390 Ohm load resistors. As a specialty, the output of each inverter (collector) is connected to the input of the next inverter (base) with an LED. So what can it do?

(Attachment Link)

When we apply 10V of bias to it, the LEDs will light up and it oscillates at a frequency of 35 MHz. As you can see, this design also follows the rule of a full 5V swing. The frequency suggests a tpd of 4.8 ns for each inverter, which is already "interesting", if you look up the storage time of a 2N3904 in the datasheet. Btw, with red LEDs this goes up to 55MHz, corresponding to 3ns. The same design is much faster on a PCB, so the parasitics of the breadboard take their toll.

Ok, but maybe that is not the most interesting part of this oscillator. Let's have a look at some X-Y plots where X is the input of one inverter and Y is at the output:

(Attachment Link)

At 10V we see a somewhat circular output, which suggests that the phase delay of this inverter is constant - as expected for a proper oscillator. However at 11.8 V we can suddenly see a convoluted path that almost looks random, but is clearly repeating - a limit cycle? Changing voltages further can transition from noise to different structured trajectory.

It looks like this simple circuit is acting as a chaotic oscillator. I though that was quite neat. You can find a longer write up here: https://hackaday.io/project/170697-evaluating-transistors-for-bipolar-logic-rtl/log/179154-using-a-led-as-base-resistor-chaotic-ring-oscillator.

Interesting contraption!  -  is it light sensitive?  -  it might behave differently depending on if it's in the light or dark,  or if the different LEDs can "see" each other or not....

[0culus]

Yeah, it has a standard 21.4 MHz IF out. I'm "warming up" by playing with Colpitts oscillators on the breadboard tonight. So far, with a 2N2222 and absolutely zero lead trimming I have achieved 19.5 MHz. Paltry, of course, compared to what you guys have been achieving but still faster than anything I've ever breadboarded. Is Manhattan style construction allowed?   

I think most of mine move around more than 22MHz.  Maybe yours will be more stable.   If not,   I had a look at the manual for your 8566.  It appears you may have access to the first LO on the front of the instrument as well and it has +5dBm drive.   Appendix H talks about it.

https://www.eevblog.com/forum/index.php?action=post;quote=3095892;topic=243355.100

You may be able use this as the LO for an external mixer that could directly drive your scope.   Guessing it would be a lot more stable than the open looped YIG I show.   With your GHz scope, you may be able to see something over 7GHz. 

***
Some advice, make sure you show the signal on the SA as well or it may confuse Trash into thinking you are using it for your oscillator.. 

My first attempt with the 5GHz transistor substituting for a 2N2222 was an abject fail. Absolutely nothing came out. No magic smoke though, so hopefully I didn't blow it up. I'm going to have to go back to the drawing board for this one. Attached is a plot from the 8566B with my original 2N2222 Colpitts oscillator circuit (taken with a near field probe, so the level is lower than it would be otherwise; there's a healthy DC voltage on the output which I didn't want to do anything foolish with even though I keep an 18GHz bandpass dc block on the input of the 8566B). However, it's build does not meet the rules (I had to cram a couple of leads in one hole to make it the loop size smaller) so I am *not* submitting this. I'm still quite pleased with it as a first effort though! I didn't trim any leads, either.

[tim_]

Interesting contraption!  -  is it light sensitive?  -  it might behave differently depending on if it's in the light or dark,  or if the different LEDs can "see" each other or not....

I checked the light sensitivity with the red-LEDed version. There was a very small impact on frequency with strong illumination, but not sufficient to change trajectory.
In any case, that is what I would expect, as the circuit uses currents in the tens of mA while the photocurrent induced in a LED is in the range on nA or even lower.
Maybe if you take a laser...

[0culus]

OK, stepping back from that failure, I instead redid my 2N2222 colpitts oscillator in a way that (I believe) is much closer to conforming with the challenge rules within the confines of what I have on hand. I was able to get more output and slightly better performance after snipping leads and playing with the layout. I can experiment with different inductors now to change the oscillation frequency.

Measurement conditions: again, spectrum analyzer measurements are done with a near field probe (pictured). I do not wish to take unnecessary risk feeding a big DC signal into my 8566B, as stated before. Oscilloscope measurement was done using a P6201 FET probe with 10x attenuator (bare tip, no hook) and a very short ground wire connected to my 7104 with 2x7A29 + 7B15 and 7B10 plugins. I found that passive probes detuned the circuit considerably.

Lessons learned: this is a lot of fun, and I look forward to continuing to make faster breadboard designs.  Also, passive probes are actually fairly poor performing. The beauty of breadboards is it's super easy to make changes, so attempting to exploit the well known problems of breadboards is a neat challenge. Additionally, bypass caps on the power supply line are required. Without them, it doesn't work very well at all I found.

[T3sl4co1l]

As you can maybe make out, it is a three stage ring oscillator based on 2N3904 (chinese off-brand) and 390 Ohm load resistors. As a specialty, the output of each inverter (collector) is connected to the input of the next inverter (base) with an LED. So what can it do?

Try a B-E resistor, say 470 or so.  Maybe a little ESR for the LEDs as well.

Chaos is interesting; likely it's a combination of rectification by the LEDs, and variable time constant due to the lack of B-E discharge.  Lots of nonlinearity around.

Tim

[T3sl4co1l]

Updating the screenshot -- much clearer.  Circuit changed slightly, with a shorter jumper, hence the higher frequency (and coupling kinda sucks, hence the lower power).



Tim