[time-nuts] OCXO DC Current Question
ed_palmer at sasktel.net
Mon Feb 18 23:29:26 EST 2013
On 2/18/2013 6:19 PM, Bob Camp wrote:
> The -Z3811 is indeed a part out of a HP GPSDO. HP is a big enough customer that the part they get *may* have little relation to the standard part.
True, but it doesn't look like there were many Z3811s made. The order
wouldn't have been that large. Maybe there were enough other orders
from HP that they wanted to keep them happy.
> If they all started out at the same temperature, then the noisy ones finish warmup quicker than the quiet ones. Take a look at when the final "glitch" happens in the warmup process. It's definitely happening earlier on the noisy parts. If warmup finishes quicker - the part is running at a lower temperature. That of course is only the time between the cutback from full current to the glitch. It may not mean it's running cooler.
Between possible temperature differences due to crystal tuning and
component tolerances that affect the internal voltage regulators ( and
therefore affect the maximum current available to the oven ), I don't
know what, if any, conclusions can be drawn for such a small number of
samples. But when I rechecked the startup current, it looks like the
noisy oscillators draw a few milliamps ( up to 10 ) more than the quiet
one. That might allow them to warm up quicker.
> On Feb 18, 2013, at 5:32 PM, Ed Palmer <ed_palmer at sasktel.net> wrote:
>> Hi Bob,
>> On 2/18/2013 2:42 PM, Bob Camp wrote:
>>> Like any control loop, gain, bandwidth and noise are all related. In a DOCXO you have two control loops and they do interact. That said, there's nothing grossly wrong with the four OCXO's. The noisy parts have a bit more gain in the controller. The quiet parts have a bit less gain. The easy way to see this is the ringing after the current changes.
>> Yes, when I zoom in on the ringing after the first step change, they all show a period of ~9 seconds. The blue trace (the quiet one) has a nice smooth damped sine wave while the other ones have varying amounts of noise superimposed on the sine wave. The amplitude of the ringing on the noisy ones is greater than the amplitude of the quiet one. In one case, more than twice as big.
>>> Without ADEV numbers there's no way to know which one is good (or better) and which one is bad (or worse). The noisy parts may be responding to the ambient temperature rumble (thus correcting for it) and the quiet ones may be ignoring it (allowing it to hit the crystal).
>> I hadn't thought of that possibility. Thanks for that!
>>> It's also possible that the noisy ones have an electrical issue in the loop that generates the noise.
>> I was thinking of maybe a dead filter capacitor - hence the noise.
>>> If all the ovens started from the same temperature, there is a variance in the oven set points. Some take longer to warm up than others. You don't mention if they started the same, so that may or may not be significant.
>> They all started at the same temperature. Any apparent difference in warmup time is mostly due to the arbitrary offset of the traces. The offset was just so you could see all four traces clearly. However, isn't it also likely that each oscillator has had it's oven tweaked to match that particular crystal?
>>> To further complicate things, in a double oven, you can have a noisy outer oven that gets suppressed by the inner oven. Are your specific 260's double ovens? No way to be sure without tearing one open. The second "step" in the current plot could be an inner oven cutting back.
>> I wondered about that second step. The period of the ringing on the second step is about 1 sec. longer than the ringing on the first step so it's apparently a different circuit. The 260 series datasheet does NOT say anything about it being a double oven. They talk about customizing the unit to suit the customer, but that seems a bit extreme.
>> By the way, I purchased these oscillators from our favorite auction site. If anyone's interested, the 260 series includes both AT and SC crystals. Since these start out ~150 Hz low, they appear to be SC cut. The output is a 5 MHz sine wave @ ~+7 dBm into 50 ohms. They work fine with a 12V power supply. They have no mechanical frequency adjustment (unless it's under a suspicious spot of solder on the case) but my 4 were all adjustable to 5 MHz via EFC. They were apparently removed from Z3811 GPS receivers. They each have a sticker on the side that says "Z3811-80010". There doesn't seem to be much info available on the Z3811. I have no relationship to the vendor - just a customer.
>> So it sounds like the proper thing to do is file this information and carry on. After making some Allan Deviation measurements, review everything to see how, or if, oven 'noise' correlates to Allan Deviation results.
>> Thanks Bob,
>>> On Feb 18, 2013, at 2:40 PM, Ed Palmer <ed_palmer at sasktel.net> wrote:
>>>> I know that when making AC measurements on various OCXOs of the same type, you have to expect wide variations in the results. e.g. TVB's Allan Deviation measurements on a selection of 10811A oscillators at http://www.leapsecond.com/pages/z3801a-osc . But what about DC current measurements? How much variability should you expect?
>>>> I recently bought 4 MTI 260 oscillators with thoughts of doing some 3-cornered hat experiments. I thought I'd use the best 3 of 4. One test I always do on an OCXO is to measure the DC current drain as it warms up. Nothing radical - I have an HP 6622A GPIB-equipped linear power supply. I just do GPIB queries as fast as I can and log the results. I get about 6 readings per second. More than enough for my needs.
>>>> This time, I was surprised by the results of this test. The attached picture shows why. I've offset the traces horizontally and vertically for clarity so I deleted the axes. The horizontal lines are 200 ma apart, but the position of each trace is arbitrary. All four oscillators start at a current-limited value of ~1 Amp and have a steady-state current drain of ~230 mA. The length of the graph is ~20 minutes.
>>>> Although the family resemblance is obvious, I was surprised by the different noise levels. I let one of the noisy units run for a day to see if it would settle down, but there was no improvement. Are these results reasonable, or do I have one oscillator with a good oven (blue trace), one marginal (purple), and two rather poor ones (red and green)? I'm thinking that the noise on the oven could affect the Allan Deviation due to either or both of the thermal inconsistencies or varying load on the power supply.
>>>> Any thoughts?
More information about the time-nuts