Forums

it oscillates!

Started by Unknown April 15, 2020
During this shutdown, I did a little pcb layout for my triggered
common-collector Colpitts oscillator, and had some boards fabbed. A
manufacturing person came in today and built a couple for me.

It's a triggered 125 MHz oscillator that will be used to time delays
in a laser system. We want minimal time jitter so I tried to keep the
Q up. The inductor is a Coilcraft Midi-Spring.

I used a BFT25 super-fast transistor, but it oscillates at more
frequencies than I intended. Tons of jitter. The choice was to add a
base resistor or go with another transistor. A BFS17 seems to work
fine. That's a great little npn, fast but not too fast.

Here's the board

https://www.dropbox.com/s/kalhm9aiq9hal2j/Colpitts_Bench.jpg?raw=1

and here's the roughly 250th rising edge after it's triggered to run

https://www.dropbox.com/s/vjzvh6v8616dt2e/Colpitts_4-15_2us.jpg?raw=1

at 100 ps/div. I haven't figured out how to get that scope to measure
the RMS jitter on that edge; it's obviously smarter than I am, and
lets me know. If I eyeball the p-p jitter and divide by 5 for RMS, I'm
estimating 4 ps RMS jitter at 2 us out from start, which is a ratio of
500K:1. That's unheard of, so I may be doing something wrong. 

We plan to phase-lock this to a good OCXO, but it will take a while to
lock, so the better the open-loop Colpitts behavior, the less frantic
we need to be about the DPLL. 2 microseconds is plenty of time to do
the math.

It's out on the bench, so I'll put it in a metal chocloate box with
some feed-thrus for better EMI shielding. Gotta empty the box first.

I need to temperature compensate it and play with the active guard
idea.


Phil H helped me think about this. Thanks.










-- 

John Larkin         Highland Technology, Inc

Science teaches us to doubt.

  Claude Bernard
  
jlarkin@highlandsniptechnology.com writes:

> https://www.dropbox.com/s/vjzvh6v8616dt2e/Colpitts_4-15_2us.jpg?raw=1 > > at 100 ps/div. I haven't figured out how to get that scope to measure > the RMS jitter on that edge; it's obviously smarter than I am, and > lets me know. If I eyeball the p-p jitter and divide by 5 for RMS, I'm > estimating 4 ps RMS jitter at 2 us out from start, which is a ratio of > 500K:1. That's unheard of, so I may be doing something wrong.
Too good a result feels good for a while always. Then the professional side starts wondering. What does 'Trigger on positive edge with Holdoff by time' exactly do, is it arming on first edge, then triggering on the second edge after holdoff time ? I think that would mean that you're triggering on the roughly 250th edge and thus seeing scope trigger jitter ? How about triggering on the first edge and scrolling to trigger+2us ? I think the scope timebase should be quite good compared to oscillator. -- mikko
On Thursday, April 16, 2020 at 2:36:22 AM UTC-4, Mikko OH2HVJ wrote:
> jlarkin@highlandsniptechnology.com writes: > > > https://www.dropbox.com/s/vjzvh6v8616dt2e/Colpitts_4-15_2us.jpg?raw=1 > > > > at 100 ps/div. I haven't figured out how to get that scope to measure > > the RMS jitter on that edge; it's obviously smarter than I am, and > > lets me know. If I eyeball the p-p jitter and divide by 5 for RMS, I'm > > estimating 4 ps RMS jitter at 2 us out from start, which is a ratio of > > 500K:1. That's unheard of, so I may be doing something wrong. > > Too good a result feels good for a while always. Then the professional > side starts wondering.
Going off on that point.... Many years ago I was working (peripherally) with a rather intimidating antenna designer. He had just built a radiating element using a 3dB 90 degree hybrid and was launching into the air from each output. He brought me over to the network analyzer and showed me his fantastic return loss. It turns out I had just been doing reading on hybrid couplers and so I pointed out to him that any reflected power at the two launching ends would reflect back into the 4th port with a 50 ohm resistor. He paused for a moment and you could see his countenance drop like a rock. He was nice to me for the rest of the time I worked there.
> > What does 'Trigger on positive edge with Holdoff by time' exactly do, is > it arming on first edge, then triggering on the second edge after > holdoff time ? > > I think that would mean that you're triggering on the roughly 250th edge > and thus seeing scope trigger jitter ? > > How about triggering on the first edge and scrolling to trigger+2us ? I think > the scope timebase should be quite good compared to oscillator. > > -- > mikko
jlarkin@highlandsniptechnology.com wrote...
> > During this shutdown, I did a little pcb layout for my triggered > common-collector Colpitts oscillator, and had some boards fabbed. > A manufacturing person came in today and built a couple for me. > > It's a triggered 125 MHz oscillator that will be used to time delays > in a laser system. We want minimal time jitter so I tried to keep the > Q up. The inductor is a Coilcraft Midi-Spring. > > I used a BFT25 super-fast transistor, but it oscillates at more > frequencies than I intended. Tons of jitter. The choice was to add a > base resistor or go with another transistor. A BFS17 seems to work > fine. That's a great little npn, fast but not too fast. > > Here's the board ... snip > > Phil H helped me think about this. Thanks.
Very interesting, show us the circuit! -- Thanks, - Win
On 2020-04-15 22:30, jlarkin@highlandsniptechnology.com wrote:
> > During this shutdown, I did a little pcb layout for my triggered > common-collector Colpitts oscillator, and had some boards fabbed. A > manufacturing person came in today and built a couple for me. > > It's a triggered 125 MHz oscillator that will be used to time delays > in a laser system. We want minimal time jitter so I tried to keep the > Q up. The inductor is a Coilcraft Midi-Spring. > > I used a BFT25 super-fast transistor, but it oscillates at more > frequencies than I intended. Tons of jitter. The choice was to add a > base resistor or go with another transistor. A BFS17 seems to work > fine. That's a great little npn, fast but not too fast. > > Here's the board > > https://www.dropbox.com/s/kalhm9aiq9hal2j/Colpitts_Bench.jpg?raw=1 > > and here's the roughly 250th rising edge after it's triggered to run > > https://www.dropbox.com/s/vjzvh6v8616dt2e/Colpitts_4-15_2us.jpg?raw=1 > > at 100 ps/div. I haven't figured out how to get that scope to measure > the RMS jitter on that edge; it's obviously smarter than I am, and > lets me know. If I eyeball the p-p jitter and divide by 5 for RMS, I'm > estimating 4 ps RMS jitter at 2 us out from start, which is a ratio of > 500K:1. That's unheard of, so I may be doing something wrong.
Very nice if true. (Nice scope, too.)
> > We plan to phase-lock this to a good OCXO, but it will take a while to > lock, so the better the open-loop Colpitts behavior, the less frantic > we need to be about the DPLL. 2 microseconds is plenty of time to do > the math. > > It's out on the bench, so I'll put it in a metal chocloate box with > some feed-thrus for better EMI shielding. Gotta empty the box first. > > I need to temperature compensate it and play with the active guard > idea. > > > Phil H helped me think about this. Thanks.
Glad to help. It's an interesting problem, for sure. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
On Wednesday, April 15, 2020 at 10:30:51 PM UTC-4, jla...@highlandsniptechnology.com wrote:
> During this shutdown, I did a little pcb layout for my triggered > common-collector Colpitts oscillator, and had some boards fabbed. A > manufacturing person came in today and built a couple for me. > > It's a triggered 125 MHz oscillator that will be used to time delays > in a laser system. We want minimal time jitter so I tried to keep the > Q up. The inductor is a Coilcraft Midi-Spring. > > I used a BFT25 super-fast transistor, but it oscillates at more > frequencies than I intended. Tons of jitter. The choice was to add a > base resistor or go with another transistor. A BFS17 seems to work > fine. That's a great little npn, fast but not too fast. > > Here's the board > > https://www.dropbox.com/s/kalhm9aiq9hal2j/Colpitts_Bench.jpg?raw=1 > > and here's the roughly 250th rising edge after it's triggered to run > > https://www.dropbox.com/s/vjzvh6v8616dt2e/Colpitts_4-15_2us.jpg?raw=1 > > at 100 ps/div. I haven't figured out how to get that scope to measure > the RMS jitter on that edge; it's obviously smarter than I am, and > lets me know. If I eyeball the p-p jitter and divide by 5 for RMS, I'm > estimating 4 ps RMS jitter at 2 us out from start, which is a ratio of > 500K:1. That's unheard of, so I may be doing something wrong. > > We plan to phase-lock this to a good OCXO, but it will take a while to > lock, so the better the open-loop Colpitts behavior, the less frantic > we need to be about the DPLL. 2 microseconds is plenty of time to do > the math. > > It's out on the bench, so I'll put it in a metal chocloate box with > some feed-thrus for better EMI shielding. Gotta empty the box first. > > I need to temperature compensate it and play with the active guard > idea. > > > Phil H helped me think about this. Thanks. > > > > > > > > > > > -- > > John Larkin Highland Technology, Inc > > Science teaches us to doubt. > > Claude Bernard
What do you mean the first circuit oscillated at more frequencies than you intended? Isn't that circuit tuned?
On Thu, 16 Apr 2020 09:36:18 +0300, Mikko OH2HVJ
<mikko.syrjalahti@nospam.fi> wrote:

>jlarkin@highlandsniptechnology.com writes: > >> https://www.dropbox.com/s/vjzvh6v8616dt2e/Colpitts_4-15_2us.jpg?raw=1 >> >> at 100 ps/div. I haven't figured out how to get that scope to measure >> the RMS jitter on that edge; it's obviously smarter than I am, and >> lets me know. If I eyeball the p-p jitter and divide by 5 for RMS, I'm >> estimating 4 ps RMS jitter at 2 us out from start, which is a ratio of >> 500K:1. That's unheard of, so I may be doing something wrong. > >Too good a result feels good for a while always. Then the professional >side starts wondering. > >What does 'Trigger on positive edge with Holdoff by time' exactly do, is >it arming on first edge, then triggering on the second edge after >holdoff time ?
LeCroy is always confusing. The oscillator makes a burst, a few us wide now, and my intent is that the scope triggers on the first rising edge and captures one burst. Seems to. If there's no holdoff, the display is a mess. Holdoff usually means that once a trigger is accepted, triggers are inhibited for some time. If I set the holdoff greater than the burst width, I should trigger on the first edge of each burst. I guess I should trigger the scope from the same pulse that gates the burst. I got the built board late in the day, and was pleased to see it oscillate. I need to spend a couple days carefully tweaking it.
> >I think that would mean that you're triggering on the roughly 250th edge >and thus seeing scope trigger jitter ?
LeCroy claims 1 ps RMS jitter. I checked it with our SRS clock generator as the input signal, and that has a lot less jitter than my oscillator, so the scope is good enough. My old Tek 11802 uses an LC burst oscillator as its timebase, but the jitter is something like 1/20000 of the delay, which isn't good enough. The LeCroy just digitizes the input at a constant ADC rate, pokes it into RAM, and does all the rest in software, including finding the trigger. I think.
> >How about triggering on the first edge and scrolling to trigger+2us ? I think >the scope timebase should be quite good compared to oscillator.
That's basically what's happening, probably. But I'll trigger on the oscillator gate, which is why I included the SMA loop-thru connectors on the gate input. -- John Larkin Highland Technology, Inc Science teaches us to doubt. Claude Bernard
On Thu, 16 Apr 2020 03:07:53 -0700 (PDT), blocher@columbus.rr.com
wrote:

>On Thursday, April 16, 2020 at 2:36:22 AM UTC-4, Mikko OH2HVJ wrote: >> jlarkin@highlandsniptechnology.com writes: >> >> > https://www.dropbox.com/s/vjzvh6v8616dt2e/Colpitts_4-15_2us.jpg?raw=1 >> > >> > at 100 ps/div. I haven't figured out how to get that scope to measure >> > the RMS jitter on that edge; it's obviously smarter than I am, and >> > lets me know. If I eyeball the p-p jitter and divide by 5 for RMS, I'm >> > estimating 4 ps RMS jitter at 2 us out from start, which is a ratio of >> > 500K:1. That's unheard of, so I may be doing something wrong. >> >> Too good a result feels good for a while always. Then the professional >> side starts wondering. > > >Going off on that point.... Many years ago I was working (peripherally) with a rather intimidating antenna designer. He had just built a radiating element using a 3dB 90 degree hybrid and was launching into the air from each output. He brought me over to the network analyzer and showed me his fantastic return loss. It turns out I had just been doing reading on hybrid couplers and so I pointed out to him that any reflected power at the two launching ends would reflect back into the 4th port with a 50 ohm resistor. He paused for a moment and you could see his countenance drop like a rock. He was nice to me for the rest of the time I worked there. > > >
I almost never trust measurements, especially good ones. Especially using an oscilloscope that nobody understands, even the LeCroy support people. -- John Larkin Highland Technology, Inc Science teaches us to doubt. Claude Bernard
On 16 Apr 2020 03:35:25 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

>jlarkin@highlandsniptechnology.com wrote... >> >> During this shutdown, I did a little pcb layout for my triggered >> common-collector Colpitts oscillator, and had some boards fabbed. >> A manufacturing person came in today and built a couple for me. >> >> It's a triggered 125 MHz oscillator that will be used to time delays >> in a laser system. We want minimal time jitter so I tried to keep the >> Q up. The inductor is a Coilcraft Midi-Spring. >> >> I used a BFT25 super-fast transistor, but it oscillates at more >> frequencies than I intended. Tons of jitter. The choice was to add a >> base resistor or go with another transistor. A BFS17 seems to work >> fine. That's a great little npn, fast but not too fast. >> >> Here's the board ... snip >> >> Phil H helped me think about this. Thanks. > > Very interesting, show us the circuit!
Here's the non-secret oscillator part. https://www.dropbox.com/s/ct32117dmho0cbr/Z384_Colpitts_3.jpg?raw=1 One key thing to test and tweak is tempco. I cut an island out of the layer 2 ground plane as an active guard that can float, be grounded, or be driven from the Q1 emitter. That might mitigate some of the horrible FR4 capacitance. C3 can tweak tempco too. The real board will be 10 layers with just 8 mils from the top to the L2 ground plane, which means it will be different from this proto board. So I'm just learning principles now. Varicaps have bad tempcos too, so things will get tangled. I'd like to make the board thicker and make the dielectrics thicker, but there are mechanical limits, including the Vbite connectors. -- John Larkin Highland Technology, Inc Science teaches us to doubt. Claude Bernard
On Thu, 16 Apr 2020 07:57:06 -0700 (PDT),
bloggs.fredbloggs.fred@gmail.com wrote:

>On Wednesday, April 15, 2020 at 10:30:51 PM UTC-4, jla...@highlandsniptechnology.com wrote: >> During this shutdown, I did a little pcb layout for my triggered >> common-collector Colpitts oscillator, and had some boards fabbed. A >> manufacturing person came in today and built a couple for me. >> >> It's a triggered 125 MHz oscillator that will be used to time delays >> in a laser system. We want minimal time jitter so I tried to keep the >> Q up. The inductor is a Coilcraft Midi-Spring. >> >> I used a BFT25 super-fast transistor, but it oscillates at more >> frequencies than I intended. Tons of jitter. The choice was to add a >> base resistor or go with another transistor. A BFS17 seems to work >> fine. That's a great little npn, fast but not too fast. >> >> Here's the board >> >> https://www.dropbox.com/s/kalhm9aiq9hal2j/Colpitts_Bench.jpg?raw=1 >> >> and here's the roughly 250th rising edge after it's triggered to run >> >> https://www.dropbox.com/s/vjzvh6v8616dt2e/Colpitts_4-15_2us.jpg?raw=1 >> >> at 100 ps/div. I haven't figured out how to get that scope to measure >> the RMS jitter on that edge; it's obviously smarter than I am, and >> lets me know. If I eyeball the p-p jitter and divide by 5 for RMS, I'm >> estimating 4 ps RMS jitter at 2 us out from start, which is a ratio of >> 500K:1. That's unheard of, so I may be doing something wrong. >> >> We plan to phase-lock this to a good OCXO, but it will take a while to >> lock, so the better the open-loop Colpitts behavior, the less frantic >> we need to be about the DPLL. 2 microseconds is plenty of time to do >> the math. >> >> It's out on the bench, so I'll put it in a metal chocloate box with >> some feed-thrus for better EMI shielding. Gotta empty the box first. >> >> I need to temperature compensate it and play with the active guard >> idea. >> >> >> Phil H helped me think about this. Thanks. >> >> >> >> >> >> >> >> >> >> >> -- >> >> John Larkin Highland Technology, Inc >> >> Science teaches us to doubt. >> >> Claude Bernard > >What do you mean the first circuit oscillated at more frequencies than you intended? Isn't that circuit tuned?
I think the NPN was also oscillating at some microwave frequencies, all on its own. Fast emitter followers do that. It was probably still oscillating when the 120 MHz tank was quenched and the oscillator theoretically stopped. Touching various counter-intuitive nodes with a tiny screwdriver changed things a lot. I even stopped the extra oscillations by touching something that made no sense. A series base resistor is the usual fix for an oscillating emitter follower, but that would need a hack and might reduce Q. Going to a slower transistor seems like the best fix. It's only 120 MHz. -- John Larkin Highland Technology, Inc Science teaches us to doubt. Claude Bernard