xor mixer

Of course that assumes that the collector current is quiet, as it will be with a quiet tail source and hard switching, and the collector load resistance is small. Otherwise the collector shot noise and resistor Johnson noise will get coupled into the base and make the base current noise go up with frequency.

Cheers

Phil Hobbs

NIST and JILA pipe atomic clock signals in fibre back and forth for a few miles under Boulder traffic. Makes a dandy seismometer too. ;)

Cheers

Phil Hobbs

Yup. Parseval's theorem says that the for cross-correlation,

g star h => GH*

so you can use some convenient-length FFT, pick just the coefficients you care about, and multiply-accumulate all the cross terms into a single array. (Personally I'd probably keep the individual cross terms to allow sanity checking.)

Cheers

Phil Hobbs

Looks like a quad-core 3 GHz Xeon can do a 64K point single-precision FFT in vaguely under 1 second. So we'd get some serious noise floor reduction in an hour.

It is hard to tell how fast FFTs run on various CPUs. Execution times are given in Mflops or somethings per nanosecond and all sorts of units.

Yikes, that's only about 20 Mflops. It should be a good thousand times faster than that. What is it, a batch file? ;)

Cheers

Phil Hobbs

Rob says a 64K point 1D complex FFT, using the Python library, will be done before you take your finger off the key. He says a 10 megapoint FFT won't be especially boring to wait for.

Rob just did a 16M 1D complex FFT in 4 seconds. It's the Python library, which uses FFTW.

So 64K should take about 11 milliseconds. That's "vaguely under 1 second."

If Python is like Scilab, it'll be fast as long as you're doing a lot of work in library code and not much in the "glue code" that actually executes in the Python interpreter.

I tend to prototype that sort of thing in Scilab, then if it's not mostly running out of library code, put all of it or chunks of it into C++.

Here is the latest improvement in DMTD, from a Time Nuts posting. Get your copy of the Rev Sci Insts article now, before the one month expires.

============================================

Message: 11 Date: Wed, 25 May 2016 16:01:51 +0000 From: "Sherman, Jeffrey A. (Fed)" To: " snipped-for-privacy@febo.com" Subject: [time-nuts] Commercial software defined radio for clock metrology Message-ID: Content-Type: text/plain; charset="us-ascii"

Hello,

A recently published paper might be of interest to the time-nuts community. We studied how well an unmodified commercial software defined radio (SDR) device/firmware could serve in comparing high-performance oscillators and atomic clocks. Though we chose to study the USRP platform, the discussion easily generalizes to many other SDRs.

I understand that for one month, the journal allows for free electronic downloads of the manuscript at:

(Review of Scientific Instruments 87, 054711 (2016))

Afterwards, a preprint will remain available at:

There are commercial instruments available with SDR architecture under-the-hood, but they often cost many thousands of dollars per measurement channel. In contrast, commercial general-purpose SDRs scale horizontally and can cost

Nice. I've downloaded it.

Hey, we met John Miles, the Timepod guy, at the MTTS show. He made the pilgrimage by foot (about a mile) to Highland World Headquarters. He may give us some help on this.

The pod only works to 30 MHz, and I don't think it likes square waves. But it is a fact that a really good ADC can have a fast s/h in the front end with fs apearture jitter. But lots of LSBs of noise.

I need to read the Timepod manual. It's 163 pages.

True, but my concern was not with stochastic noise alone; there's also input from any interfering signal. The wideband nature of a square-wave-handling mixer opens it up to beating of any harmonic of the signal with any harmonic of any oscillator nearby. I've seen this happen, and it took a LOT of time and effort to track to the source. The system had dozens of DC/DC converters, and one of them interacted with a ground loop... and interference showed up in a sensitive signal path, a foot away.

I caught it by applying freeze-mist while watching the output spectrum.

Probably (because both the described signals ARE square waves) there's no way to improve matters for the original problem, as stated. I'd hope for better performance, though, with linear mixing and narrowband signals, just because the out-of-band pickup stung me that once... The band of interest is small, and out-of-band is big and scary.

The instrument literature is just like the rest of the peer-reviewed scientific literature - 90% of it is rubbish. Not necessarily wrong, but mostly unhelpful.