phemt colpitts: beads

OOPs, it's erratic at 301 ohms in the gate. It really likes 499.

I got a Rigol DP821 precision supply for the +-5, and $5K Keysight SMU for the varicap voltage, which helps too.

An LC oscillator can be really good if the supplies are super-stable.

I think you are doing something wrong. I have never heard of a colpitts that needed 499 ohms in the gate.

Another thing that is questionable: a proper colpitts should be relatively insensitive to the supply voltage. I would have to see the circuit to tell what is going wrong, but the oscillators I have designed and built simply don't work that way. Sure, you can get second or third-order effects due to minor capacitance changes with voltage, but not to the point of needing $5k power supplies to run the oscillator.

You have never mentioned having this problem before. What are you doing different now compared to your previous designs?

HP 4145B, man, I keep telling you. ;) (What's a few boot floppies among friends?)

No huge surprise that the 25-ohm and higher beads didn't work--they crap out at lower frequencies than the 5- and 10-ohm ones.

A 301 ohm resistor is a pretty nasty noise source to add to a 300 pV/sqrt(Hz) pHEMT--how about reducing the drain current by 2-3x and sticking with a 5- or 10-ohm BA-series bead?

It will surprise no one that I'm a fan of the HP 611xA series power supplies. They have 4-digit thumbwheel switches, temperature-stabilized

Cheers

Phil Hobbs

OK and I was thinking by 'super-stable' John meant a supply with low output impedance. stable but not necessary accurate. (I guess that means low noise and ripple too.)

George H.

They handle BFC loads pretty well.

Cheers

Phil Hobbs

OK but on some side of the peak an LC looks inductive. George H. My only LC oscillator is a ~70MHz Rb lamp drive... BFL (or is it BFI ?) 1.5->7 turn autotransformer wrapped around the bulb. and a tuning cap across the whole thing.

Floppies? I've heard rumors about them. They apparently stored nano-terabytes of data.

I want very constant supply and VCO voltages, to keep the phase noise and drift down during my measurements so I can see what the oscillator itself does. I'm measuring/tweaking temperature coefficients too, which is a seriously boring PITA.

I want a 4-channel SMU!

Am 02.02.21 um 02:02 schrieb John Larkin:

Use 2 or more 18650 Lithium batteries IN the cookie box. If the voltage does not fit --> LT3042, 2nV/rtHz You won't get it better without great effort, let alone from a power supply outside the box. Look at the area between the power supply leads.

If the square wave is generated directly in the oscillator, it will mix up all the nasty GaAs 1/f noise onto the carrier. Plus the effects of modulating the transistor capacitance to the max, and the excessive loop gain also amplifies the noise.

Gerhard

John Larkin wrote in news: snipped-for-privacy@4ax.com:

Try a toroid directly on the leg, right up next to the device body.

I spent considerable time researching common-collector and common-base colpitts oscillators. Not one of them required a gate resistor, let alone

Gerhard was right in noticing the output is a square wave. If that is coming directly from the oscillator, there is something seriously wrong.

On a sunny day (Wed, 3 Feb 2021 06:23:02 -0000 (UTC)) it happened Steve Wilson wrote in :

Indeed, hard to tell without a circuit diagram that is 'top secret' I am sure.. but for an LC it sucks, and 160MHz is not really that high and why even use phemts and trying to imagine a 'triggered Colpitts'??? But that picture of the board made me think: No No No No Dunno how many LC oscillators I have build, never a gate resistor and never parasitic oscillation, and always close to a sine wave. Sure looks like a flipped out multivibrator to me? That would trigger nicely! John is it a multivibrator with some LC in it?

John and Phil have discussed these before. Basically he wants a trigger signal to start an oscillator to full power and zero spuria from the first half-cycle of oscillation, and with minimum lag or jitter from the trigger.

How would you do that Jan? I'm genuinely curious.

I mean, with a perfect LC, you just need to inject exactly the right amount of charge in an instantaneous impulse.

Quite a few spherical cows there...

CH

On a sunny day (Wed, 3 Feb 2021 20:44:57 +1100) it happened Clifford Heath wrote in :

OK, did not see that discussion. So what's the circuit then???

Indeed it is self-contradictory sort of requirement. You want (he ??) LC for stability, so some Q must be there. That means stopping a pendulum in motion and putting it at zero deflection at _any_ moment in next to zero time? Logic then says do not use the LC, use a stable multivibrator that you can reset. The frequency of multivibrators is very sensitive to supply voltage, as is his whatever it is circuit. And you always have 2 stages to bother about. For lower frequencies you could use an UJT, I have, it is very easy to sync to a higher frequency / reset by pulling one electrode down a bit, used that for TV H and V sync. You could create something similar with semiconductors for higher frequencies, tunnel diode comes to mind too. Basically I would not use an LC, 160 MHz or so is low frequency, even a decent audiophile can hear that as you undoubtedly know.

Pendulum... do not use it. Even if you managed to set it to zero it will then need time to speed up again.

Nah its easy :-)

just get a predator, that will fix your cows problem,

Why why why?

It's actually a 4-layer quickie proto board with several experiments.

It's an instant-start triggered Colpitts sinewave oscillator, feeding an ECL comparator to make a clock. My FPGA kids are complaining that I'm pushing them too hard (they actually believe the Xilinx tools worst-case timing reports) so it's down to 112 MHz.

The goal is zero propagation delay from trigger to clock start, zero TC, and zero time jitter (phase noise to the RF boys.)

Phil suggesed a ferrite bead in the gate to kill microwave oscillations, but I can't find one that works. The 500 ohm resistor works, which is a serious argument for including it. I'm estimating that it contributes femtoseconds of jitter to my clock.

It only needs to be really good for a few us, after which it's phase locked to an OCXO. That's the secret part.

How can people declare that I'm doing something wrong when they don't know what I'm doing, and it mostly works?

I'm tuning tempcos and Kvco's now, which is impressively tedious. FR4 makes a rotten capacitor.

There are not a lot of choices available in varicaps. Most are hyperabrupt parts, intended for wide tuning ranges, so I have to use them at high voltages to get a narrow, mostly linear tune range. And I'm not an RF guy. Some of the 15 volt max rated parts work past 40 volts, some don't.

First year EE we learned about passive circuits and differential equations and initial conditions. All that part works.

What's the voltage tempco of a lithium battery? I'm tweaking tempcos, and need the supplies to not confuse things.

Probably nanovolts induced by h-fields. In our place, the big problem is the 22 megawatts of RF from Sutro Tower. Why is RF even broadcast any more? 99.999% is wasted to space. Everybody has cable or fiber or dishes. Well, I guess we need FM for ancient car radios.

I think there is, basically, FM capture effect. The zero crossings of the oscillator sine wave are pretty fast, ballpark 1v/ns. First-order, sinewave amplitude doesn't affect clock timing.

Oscillators need loop gain! The phemt follower gain is less than 1.

On 2021-02-03 11:46, Jan Panteltje wrote:

I think I would use a diode switch to 'open' the resonator and to set up a DC current in the coil while the oscillator is stopped. Something like below.

Jeroen Belleman

================================= Version 4 SHEET 1 880 680 WIRE 224 -16 64 -16 WIRE 64 16 64 -16 WIRE 64 112 64 96 WIRE 224 144 224 -16 WIRE -320 192 -368 192 WIRE -192 192 -240 192 WIRE -64 192 -128 192 WIRE 64 192 -64 192 WIRE 112 192 64 192 WIRE 160 192 112 192 WIRE 64 208 64 192 WIRE -368 240 -368 192 WIRE 64 288 64 272 WIRE 224 288 224 240 WIRE 224 288 64 288 WIRE -64 304 -64 192 WIRE 224 304 224 288 WIRE 64 336 64 288 WIRE -368 352 -368 320 WIRE -64 416 -64 384 WIRE 64 416 64 400 WIRE 224 432 224 384 WIRE 224 528 224 512 FLAG 64 112 0 FLAG 224 528 0 FLAG -64 416 0 FLAG 64 416 0 FLAG 224 288 e FLAG -368 352 0 FLAG 112 192 b SYMBOL npn 160 144 R0 SYMATTR InstName Q1 SYMATTR Value bfr92a SYMBOL res 208 288 R0 SYMATTR InstName R1 SYMATTR Value 6k8 SYMBOL voltage 64 0 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value 3 SYMBOL voltage 224 416 R0 SYMATTR InstName V2 SYMATTR Value -3V SYMBOL ind -80 288 R0 SYMATTR InstName L1 SYMATTR Value 100n SYMBOL cap 48 208 R0 SYMATTR InstName C1 SYMATTR Value 20p SYMBOL cap 48 336 R0 SYMATTR InstName C2 SYMATTR Value 20p SYMBOL schottky -128 176 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName D1 SYMATTR Value HSMS-2820 SYMATTR Description Diode SYMATTR Type diode SYMBOL voltage -368 224 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V3 SYMATTR Value PULSE(-3 3 10n 1n) SYMBOL res -224 176 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R2 SYMATTR Value 220 TEXT -8 488 Left 2 !.tran 100n TEXT -384 536 Left 2 ;Switched Colpitts. Jeroen Belleman - 20210203

On a sunny day (Wed, 3 Feb 2021 11:16:05 -0000 (UTC)) it happened John Doe wrote in :

Interesting thing, recent experiment here:

I thought I would sleep better in all the noise here (from the neighbors) when I had some soft music playing. So I took my xiaomi smartphone and had it play some good music in a loop and put it under my pillow, lowest volume setting. After a day or 2 I noticed some what can best be described rumbling noise in my head during daytime. I suspected it came from the GHz (means Giggle Hertz as you undoubtedly know) processor in that phone interfering with the brains neural net. It was in airplane mode, and WiFi off, so should not transmit anything. So I replaced it with my old low frequency mp3 player and earbuds on were under my pillow, end of problem. This is interesting in many ways, people have been burning down cellphone towers here because of fear for 5G. Those higher frequencies .. From this experience I would not want to live next to a 5G tower either. Of course radiation goes down exponential with distance. but I have been in very high radiation levels often in my life partly because of my own transmitters, at lower frequencies, never noticed anything (beep beep be..~~~ ) So our perception goes way over what we just transfer to electric, chemical signals via the ear and eye to the neurons in our brain.

An argument against 5G!!!

And China is already moving to 6G!

We know very little about the brain and it probably uses QM effects to communicate with other life forms, things that can easily be jammed (for lack of a better word, but 'jammed' is a very good word actually) by close giggle Hertz radiation.

grin

Wilson

and zero time jitter (phase noise to the RF boys.)

Why not get seriously ambitious and try to get it started before the trigge r signal?

A zero propaganation delay is just as physically impossible

It has always struck me as a silly idea. Record the phase of a continuously running clock at the point when the trigger arrived, and digitise that.

When we did that back in 1989, we could do it to about 10psec. It took betw een one and two clock cycles (1.25 to 2.50 nsec) before we had a stable vol tage to digitiise, and it took some 20nsec before we had an eight-bit numbe r to pass on, but that was then.

w what I'm doing, and it mostly works?

You post here a lot.

es a rotten capacitor.

That's one of the reasons why we didn't use it back in 1988 at least not fo r the outer layers of the six layer board.