[time-nuts] OCXO and fluctuations after EFC adjustment (Ben Bradley)

[Bob kb8tq]

Hi

Electrolytic caps have a lot of leakage. That’s the bad news. The good news is 
that if you maintain a stable voltage on a tantalum part, the leakage decays 
( = drops off) with time. In a system with a fixed EFC, you probably can get away
with a tantalum part. You will have to wait a bit for it to “settle down”. 

Of course, if the Vref is quiet enough and the EFC narrow enough, there is no 
need for a big bypass on the EFC in the first place. 

If you dig into film caps, they have the same sort of decaying leakage …. 

All these caps are rated for leakage at some voltage after some time on
power. To keep the test easy, the time often is pretty short. Thus the data 
sheet numbers *should* give you a pretty good idea what the max leakage
will be. 

Bob

> On Apr 12, 2020, at 2:01 PM, Leon Pavlovic <leon.pavlovic at gmail.com> wrote:
> 
> Speaking of high-performance crystal oscillators, EFC inputs and Tantalum
> capacitors, how to select a good decoupling capacitor for its EFC input?
> 
> I'm attaching two PN measurements made yesterday on a HP3048A. There are
> two 100MHz crystal oscillators in the setup, (1) a XCO made by myself,
> which I know from previous measurements on a E5052 that has a stable! PN of
> -130dBc/Hz at 100Hz and below (Vtune from HP3048A is connected to this
> XCO). The other one (2) is a commercial Connor-Winfield OCXO, specified as
> -125dBc/Hz at 100Hz. I'm not sure if this one has a stable PN. The EFC
> input is decoupled by a 1uF ceramic in parallel to 33uF/16V Tantal and 10kR
> to ground, and connected by another 10kR to its Vref output.
> 
> So where's the problem? The two measurements attached were taken only some
> minutes apart. The first one is expected, the second one is a disaster
> below 1kHz offset. Is it possible that the heated Connor-Winfield OCXO
> affected the Tantal capacitor on its EFC node after some minutes (both are
> in close proximity on a PCB), increasing the leakage or whatever phenomenon
> that corrupted its PN (modulation of the EFC voltage)?
> 
> Possible reasons for a fluctuating/unstable PN to consider:
> - one of the oscillators has unstable PN
> - the Tantalum capacitor should be replaced by a film or a ceramic 10uF?
> - broadcast FM stations influenced the second measurement
> - ?
> 
> Thanks for any advice where to look at the problem ;)
> Leon
> 
> 
> ---------- Forwarded message ----------
> From: Ben Bradley <ben.pi.bradley at gmail.com>
> To: Discussion of precise time and frequency measurement <
> time-nuts at lists.febo.com>
> Cc:
> Bcc:
> Date: Sat, 11 Apr 2020 23:22:29 -0400
> Subject: Re: [time-nuts] OCXO and fluctuations after EFC adjustment
> Despite my interest in the precision high-end of electronic design
> (and thus being a subscriber to this list), I have very little
> experience with such high-fallutin' designs. Still, I've done and seen
> some "interesting" things in my career. It's amazing how the minutiae
> of even jelly bean components can cause product failures.
> 
> A few decades ago (many of my jobs were engineering positions at
> companies that made equipment for the POTS phone line), the company I
> was with had some rather large surface-mount ceramic capacitors that
> were failing short in a new product that used them across the phone
> line (they were RATED for this application right there on the data
> sheet).  The concern was our product would some day short out the
> phone line and someone with another phone on the line would have an
> emergency and wouldn't be able to call 911 to get help. The possible
> legal liability to the company was obvious. In a discussion with other
> engineers and managers, someone wondered if it was the way the board
> was manufactured, maybe the solder process caused some caps to crack,
> or whether the failures were strictly the fault of the capacitor
> manufacturer. I volunteered to test the caps in an environmental
> chamber that wasn't being used at the time, putting them near the edge
> of, but still within, specs. I made a jig that put a bias voltage on
> many in parallel (using a mechanical spring connection, not soldered)
> and left them for a few days/weeks. Out of 100 to 150 devices, about 5
> or 10 became unacceptably leaky. I wrote up my report and emailed it
> to the other engineers and managers involved, and didn't hear back
> anything more about it. The good news (!) is the product wasn't in
> production for very long for several reasons, more recently of course
> that the POTS phone line is no longer the most common form of
> telecommunications. The bad news is it's no longer helpful to have
> POTS product design on my resume.
> 
> More recently, I saw this Kemet presentation on Digikey about tantalum
> capacitors. Certainly for aluminum electrolytic capacitors, the rated
> voltage is "the rated voltage" and as long as the capacitor never goes
> ABOVE that voltage (and has no overcurrent that would heat it up,
> etc.), the cap is good for its combination of temperature and lifetime
> rating. I (and as far as I know, everone I've known) assumed this was
> the same for tantalums, but it appears that's not the case (this
> presentation mentions several failure causes and shows how they are
> multiplicative). As you go from 1/2 rated voltage to full rated
> voltage, the chances of a tantalum failing goes up substantially. The
> implied rule seems to be for maximum reliability, don't operate a
> tantalum above HALF the rated voltage. I'd heard a lot of anecdotal
> things about tantalums suddenly shorting out for this or that reason,
> but hadn't heard of this, and here it is straight from the
> manufacturer.
> https://www.digikey.com/en/ptm/k/kemet/derating-guidelines-for-surface-mount-tantalum-capacitors/tutorial
> <XCO_vs_OCXO_bad.BMP><XCO_vs_OCXO_good.BMP>_______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.