Not to thread-hijack -- but parenthetically, the Edison National Museum reopened last year after a huge refurbishment -- I know that when folks come to the US they have limited time to browse places like this, (and it's in West Orange NJ which takes about hour to get to from NYC) -- but it's worth the trip if you have interest in the history of engineering.
One thing that might help is to roll off the freqs below 10hz going to the DUT as they only contribute to unwanted jitter/freqs.
Thx-RNmarsh
Last edited:
@ Scott: could you check the spice models of the SSM2212 and SSM2220? are they correct when it comes to noise simulation? RB seems to be a bit optimistic. I prefer and use mostly MEXTRAM models for small signal BJT's, but I suppose it's too much to ask for.
(Scott, when I look at the data sheets for these transistors, I wonder if there are some among you who doesn't know the difference between NPN and PNP)
Stein
We use MEXTRAM but most/many? customers don't. I've been there on those datasheets, but can't decree a change.
EDIT - Yikes you made me look, the stupid noise graphs with the physically impossible bumps are still there. Yes the rbb listed would give .65nV at 1mA not .85 (the spec).
Last edited:
I don't know how exactly you define noise in an oscillator, but what we (I?) have discussed is reducing THD+N for a given bandwidth (e.g. 22 kHz). And that surely decreases with decreasing frequency (at least as a general statement, the exact behaviour is pretty complex), as the equivalent noise bandwidth of various sources is proportional to oscillation frequency.
Samuel
Both true.... noise increasing at lower end/freqs (10Hz) from active circuitry (opamps etc) to be lowered and thd+n (22KHz BW) to be lowered. Different mechanisms and issues but both need to be reduced.
Thx-RNMarsh
Amplitude variation of 0.05dB over 2 seconds from peak to trough and about 4 seconds for a complete cycle. Not referenced to any other frequency.
It's actually a less than 0.05dB but I'm going to call it that.
Last edited:
Yes we can set our own specs. The question is a comparative one.
I haven't looked that closely at other oscillator at this frequency.
And how do you measure that? Its the gray area where thermal converters don't work. Most DVM's with good accuracy need averaging for low frequencies. The averaging hides the fluctuations. KH used a 7A13 to monitor the peaks to test cycle to cycle stability.
I'm watching it on the FFT with no averaging at 96k and 8192 points.
I can see the variation on the scope using the 7A22 with DC offset to see the peak at 20mV/div.
I don't use a DVM for this kind of measurement.
Here are some amplitude noise measurements I've run several years back where I've started to look into oscillator design:
As I'm looking at the graph again, I'm pretty sure the +4 dBu and +14 dBu (which denotes the oscillator operating level) labels are interchanged.
This is for a SVF oscillator with sin^2 + cos^2 level detector (AD633), at 1 kHz. For the +4 dBu plot (red, mis-labeled as +14 dBu) the noise contribution of the squarers are clearly visible. At +14 dBu that's gone, but instead there was much more residual ripple from secondary effects (probably mostly nonlinearity) in the two multipliers.
As far as I recall I had to use the slowest integration time of the SYS-2722 RMS meter for these measurements, and average several readings for each data point. But that's long time ago.
Samuel
As I'm looking at the graph again, I'm pretty sure the +4 dBu and +14 dBu (which denotes the oscillator operating level) labels are interchanged.
This is for a SVF oscillator with sin^2 + cos^2 level detector (AD633), at 1 kHz. For the +4 dBu plot (red, mis-labeled as +14 dBu) the noise contribution of the squarers are clearly visible. At +14 dBu that's gone, but instead there was much more residual ripple from secondary effects (probably mostly nonlinearity) in the two multipliers.
As far as I recall I had to use the slowest integration time of the SYS-2722 RMS meter for these measurements, and average several readings for each data point. But that's long time ago.
Samuel
I guess I do need one. I would buy one if you know a source. I don't trust the EBay guys.
I dug my old 7A12 which has a much greater dc offset range. Managed to get a measurement before it crapped out.
I don't have the amplitude stability problem I thought.
The long term drift is 0.003dB and short term looked more like 0.00045dB.
I'll need to confirm this with better working equipment.
52 PPM? Even the best thermal converters are only barely that accurate under pretty ideal conditions. DVMs can do that for DC but I don't trust them for AC to that accuracy. The HP 3458 best is 100 PPM for AC.
I was using low res to speed up the software. Sound card stuff and software is not meant for doing this. Perhaps something intended for this type of measure.
Scope is the best thing if I can get it into the range needed.