Hey, it’s me, I want to give you some good frequencies. (The part I’m referring to is the very end, and the bandpass filtered beat you hear in the background is the beat to Eple, which follows it on the album. Eple will sound familiar to anyone who’s ever fired up a fresh install of Mac OS 10.3…)
But all that aside, this is about metrology and frequency standards and things my cat likes to loaf herself on top of because they’re warm.
We’re preparing for the installation of a new GatesAir Maxiva DTV transmitter at work. I was gonna say it’s an ATSC transmitter, but… I’d at least like to hope… it’s ATSC3 ready, whenever that rolls out. Sitting in the space it was going to reside in was a weird old Axcera transmitter that never worked right and was yanked out in pieces to be e-waste’d. Sitting on one of the pallets of refuse left over was the reference oscillator for the exciter, which, interestingly, was just a standalone thing without GPS synchronization. The tub in the middle is an insulated chamber containing an oven controlled crystal oscillator. Basically, this is an oscillator in a thermostatically controlled heated chamber that keeps it stable. It MUST be allowed to warm up to full operating temperature before use, or, well… it just ain’t gonna be in spec!
(insert commentary here on how silly it is that I’ve seen OCXOs in battery powered equipment that has a shorter battery life than the warmup time)
Most modern stuff uses GPS sync because it’s a good inexpensive way of obtaining a stable reference frequency and timecode. The usual arrangement is to have a voltage controlled oscillator that’s PLL locked to a 10khz timing signal output from a GPS receiver head. Aside from a little bit of phase noise possible in the system, it’s always spot on. This is why you’ll see funky little cone shaped GPS receiver antennas all over the place at broadcast facilities.
Here’s the Evertz system we have that takes GPS time and frequency references and generates our facility master clocks, black burst, and trilevel video sync. I’ve never really gotten that good a look at the way it operates but I think the black burst is generated inside the automatic changeover unit which also has some distribution amplifiers in the back as well. One of the outputs is a 10.000.00000 (I’m not sure how many significant figures) reference which can be used by a wide range of equipment. After having an, uh, experience, with one of these changeover units (see link above) I wisely do not even look at it hard while we’re in anything but 4:00 AM Sunday morning backwash programming. A frame of Grass Valley distribution amplifiers near it is used to distribute its black burst, LTC timecode, and 10mhz signals to where they’re needed throughout the facility.
This will come into play later.
The toroidal power transformer has two primary windings which were series wired for operation on 240vac. That’s why it says 240 on the AC terminal block shield. I swapped them to paralleled for 120.
More pictures and calibration process — onward
Out of curiosity, I decided first to try checking the output of this reference unit using a frequency counter, as there was a note on the front regarding the pilot frequency of the exciter it had been connected to, and it was a very, uh, uneven weird lookin’ number.
I found a frequency counter.
Boy did I find a frequency counter, WOW this thing… is…. CHONKY. It has an oven controlled crystal oscillator in its for its timebase. This was what I got after letting both units warm up for a couple hours then counting the output of this OCXO reference unit. It’s a bit off. Now, why on earth was the counter SO DAMN HEAVY???
WELL THAT’D DO IT!!!
The rectangle at the bottom of the last picture is the oven controlled crystal oscillator. The thing that looks like an old Ford starter solenoid is a YIG filter, and the cards are the counters, dividers, and control logic.
The way this counter works for microwave frequencies is apparently pretty clever; I didn’t try using it up in those ranges, but basically, it’s generating a reference signal with a comb oscillator that has a lot of harmonics to its output, using the YIG filter to lock onto the input signal and filter that and the local comb oscillator, then counting the heterodyne frequency generated by mixing them and presenting that to the user.
And heavy. But I digress.
I wonder if it works? These nice microwave counters are notorious for being found with the input and mixers blown to bits by someone having connected a few watts of input to the input jack (OUCH!). Personally I never run into that issue as I’m a paranoid son of a booger and usually put a 30dB pad on the input of everything… removing that, begrudgingly, only if the input signal from the system under test is too low level to make it through that pad at a usable strength.
An external reference jack is provided (which seems to be pretty much standard equipment on every counter I’ve used) so if you have a high quality 10mhz reference source available, you can use that. I threw a cable from the house 10mhz distribution amp to the counter, flicked the switch on the back, and the EXT REF light came on happily.
(Note that if you ever find a frequency counter seemingly dead, stuck on zero, and not operating correctly, the first thing you should check is the ext ref switch on the back – that 10mhz clock is used for pretty much everything including the gate time, so the counter will just sit there waiting for pulses that don’t arrive…)
The resulting display checking the OCXO reference unit was a little different, but only by maybe 10 hz or so… the OCXO inside the counter was also pretty dang good.
that sure is a lot of significant figures of…….. whaaaaat?
Next step: With the EIP counter on house reference, I pulled the plug from the frequency adjust hole in the top and dialed it in.
On a side note – the recommended calibration procedure for a TCXO also involves testing and setting the operating temperature to manufacturer specifications. Unfortunately, no manufacturer specifications could be found on this unit, so I’ve skipped that. If I ever do find specs on the oscillator used I’ll revisit this.
Now, I began to realize there’s a “you’re heating the whole neighborhood!” problem; the room I was in has a lot of air movement to flush heat away from racks of equipment, and once the plug is pulled out of the frequency adjust hole, the 1hz digit starts to walk around. This high-tech problem was fixed with a low-tech piece of masking tape over the hole with the adjustment tool stabbed through it to seal out the drafts.
Now, I’ve run into the resolution limit of that frequency counter. The maximum gate time on it is 1 second; I’m used to using some lab grade counters where they have a 10 second gate time and can thus give you resolution down to 0.1 hz. Fortunately, it’s easy to use a neat little trick with an oscilloscope… I disconnected the inputs to the counter at this point and connected them to the X and Y axes of a scope and adjusted until the resulting, uh, splat, on the screen stopped moving.
The output waveform isn’t really sinusoidal. It’s weird looking but every piece of equipment I’ve thrown at it works fine with it. If I probe the circuit at the input of the distribution amp cards, it IS sinusoidal, so I’m guessing either these just have some weird distortion inherently or they want to be terminated into something other than a piece of 75 ohm coax jammed on the 1M/22pF input jacks of an oscilloscope…
With the reference oscillator unit calibrated to, at my best estimate, better than 0.1hz accuracy, I took it back to my own lab and threw it at a couple other pieces of equipment to see how they’re doing:
Rohde & Schwarz FSH3 spectrum analyzer – not too bad (I don’t trust its exact frequency count anyway)
And my old Fluke counter…. not so great. This unit uses a temperature compensated crystal oscillator (TCXO) system for its local reference. I followed its calibration instructions but found that the oscillator trim coil bottoms out before it reaches 10.0000000.
That being said…… it does have an external reference input, so I just have a cable running to it from now on. Problem solved.
This counter’s ok but I’ll probably replace it when I find a better one. It’s got A, B, and C inputs, with the C input being good up through UHF frequencies. The C input feeds a divider full of unobtainable custom silicon. Normally on counters like this, when you set it to FREQ C, the counter gate cycle will not begin until it sees input above a usable threshold. On this one, it just starts counting the background noise, and will cheerfully give you almost-but-not-at-all-right readings when the signal is just below the usable level. Also it doesn’t have a Nixie tube display, which is an inexcusable fault.