Share:
Notifications
Clear all

Vintage crystal.

Page 1 / 2
 
Doz
 Doz
(@doz)
Noble V-Ratter Registered

Hi All,

In a recent, possibly daft (no, most certainly daft) project. I'm using a GEC crystal in a B7G envelope.

 

DSC 3142

 I've experimented with various capacitance loading, but the thing seems to drift badly with temperature. It's marked 600KC/s, I expect it to have aged, but it just not stable... even though I'm dividing by 16384 to get a ~28Hz output. 

It does seem more stable with 100pF capacitance load than my initial experiments using 40pF. Where should I look next? Or is it just , well, "sub-optimal"...

oscillatopr

 

ReplyQuote
Topic starter Posted : 21/08/2020 12:34 pm
Nuvistor
(@nuvistor)
Famed V-Ratter Registered

Can it be isolated with a buffer in case the IC is changing load with temperature?

Another thought is the seal has gone and the Crystal is contaminated.

ReplyQuote
Posted : 21/08/2020 1:56 pm
Doz
 Doz
(@doz)
Noble V-Ratter Registered

@nuvistor .. from the datasheet " All inputs and outputs are fully
buffered. " ... unlikely I would have thought. The thought of the seal being broken however is a very real possibility. 

 

 

ReplyQuote
Topic starter Posted : 21/08/2020 3:33 pm
Doz
 Doz
(@doz)
Noble V-Ratter Registered

Incidentally, there's a typo in my original post...it's not  600KHz, nope... 460!

 

ReplyQuote
Topic starter Posted : 21/08/2020 3:35 pm
Till Eulenspiegel
(@till)
Famed V-Ratter Registered

Two 405 line colour TV sub-carrier crystals. 2657,8125Kc/s. = 10125 X 262.5.

Crystals X2

Till Eulenspiegel.

ReplyQuote
Posted : 21/08/2020 5:03 pm
ntscuser liked
Katie Bush
(@katie-bush)
Illustrious V-Ratter Moderator

Probably a daft question, and being something I've never even seen, I just have to ask, if these crystals are sealed in an evacuated glass envelope, are they also gettered?

ReplyQuote
Posted : 21/08/2020 8:26 pm
Till Eulenspiegel
(@till)
Famed V-Ratter Registered

Hi Marion,

                 No getters in the TV crystals.

TV crystal 3

 

ReplyQuote
Posted : 21/08/2020 10:47 pm
sideband
(@sideband)
Famed V-Ratter Moderator
Posted by: @till

Hi Marion,

                 No getters in the TV crystals.

TV crystal 3

 

No indeed. They don't need the high quality of vacuum that a valve needs.....no electrons to bump into gas molecules. 

ReplyQuote
Posted : 21/08/2020 11:27 pm
Katie Bush
(@katie-bush)
Illustrious V-Ratter Moderator
Posted by: @sideband
Posted by: @till

Hi Marion,

No getters in the TV crystals.

No indeed. They don't need the high quality of vacuum that a valve needs.....no electrons to bump into gas molecules. 

I thought that might be the case, but how handy it would have been if they were gettered - you would be able see at a glance if the vacuum had been destroyed.

 

ReplyQuote
Posted : 22/08/2020 2:49 pm
Till Eulenspiegel
(@till)
Famed V-Ratter Registered

Perhaps the glass bulb is filled with an inert gas?

ReplyQuote
Posted : 22/08/2020 3:27 pm
Doz
 Doz
(@doz)
Noble V-Ratter Registered

I think I've solved the problem...

I checked my ~28Hz output and it drifts a miniscule amount, not enough to give problems. I warmed it up, and cooled it , and it's pretty darn good. More than good enough to give project a good clock signal... but still it drifted. This ~28Hz signal drives a micro with a bit of code in it to output seconds over a serial interface. There's plenty of code space in the micro so I wrote a cal routine, which allows me to show output seconds against a time-elapsed counter which is referenced to an internet clock. Yes, it drifts, but again, not the huge errors ( a gain of 3-4 minutes in a 24h period) .... This micro feeds a second micro which decodes the seconds, and generates a video signal clock. The issue was, under certain situations, once every now and again, the receiver would receive the 00 seconds minute marker signal twice, thus incrementing minutes twice ... I've now implemented a bit of code to prevent this from happening. Now, on a count of 59, a flag is set, telling the code to expect 00 seconds. When this happens, the minute is incremented, and the flag reset. Let's see how it performs now...

ReplyQuote
Topic starter Posted : 22/08/2020 3:44 pm
Doz
 Doz
(@doz)
Noble V-Ratter Registered
Posted by: @till

Perhaps the glass bulb is filled with an inert gas?

Most likely.

ReplyQuote
Topic starter Posted : 22/08/2020 3:44 pm
Boater Sam
(@boater-sam)
Reputable V-Ratter Registered

@katie-bush  Getter is metallic magnesium, I would imagine it would upset the crystal somewhat. There is not much control over where it goes when it is fired.

And, firing is done with an RF induction heater usually, don't think crystals would like that.

ReplyQuote
Posted : 23/08/2020 12:21 pm
Doz liked
Boater Sam
(@boater-sam)
Reputable V-Ratter Registered

Are the crystals suspended in the "valve" or mounted in the base near the pins?

Just thinking about capacitance with unconnected pins.

ReplyQuote
Posted : 23/08/2020 12:59 pm
Doz
 Doz
(@doz)
Noble V-Ratter Registered
Posted by: @boater-sam

@katie-bush  Getter is metallic magnesium, I would imagine it would upset the crystal somewhat. There is not much control over where it goes when it is fired.

And, firing is done with an RF induction heater usually, don't think crystals would like that.

... or barium. Either way there's none there...

ReplyQuote
Topic starter Posted : 24/08/2020 10:54 am
Doz
 Doz
(@doz)
Noble V-Ratter Registered

@boater-sam. Very much suspended. My software fix proved useless. It's still drifting. Gaining about 6 mins every 24 hours...

Didn't want to oscillate with 22pF in each leg. Started up reliably with 100pF ... maybe I should try somewhere in between. 

DSC 3143
ReplyQuote
Topic starter Posted : 24/08/2020 10:58 am
Doz
 Doz
(@doz)
Noble V-Ratter Registered

Actually, why would you have two in there? They're isolated from one another. Perhaps there supposed to be in parallel or something?? I'm not sure what happens with series or parallel connected crystals. Time for google.

ReplyQuote
Topic starter Posted : 24/08/2020 11:00 am
crustytv
(@crustytv)
Vrat Founder Admin
Posted by: @doz

Very much suspended...... 

DSC 3143

Wow cool photo, I've never seen a B7G enveloped crystal before, let alone inside. ? I'm just used to those little oval cans on decoders. As Spock would say, 'Fascinating'.

ReplyQuote
Posted : 24/08/2020 11:05 am
Katie Bush and Doz liked
Doz
 Doz
(@doz)
Noble V-Ratter Registered

I've also had another thought. When I said "My 28Hz looks stable enough" , that is, it's hooked up to the PC in my workshop, and left to run for 24Hrs, and any difference worked out. 

I originally thought that therefore my oscillator was OK, and the fault's in the software... but maybe not. The PC is chucking out a bit of hot air into a small room... perhaps that is stabilising any temperature drift... or maybe not. 

I think I may run a more brutal test with a hot air gun and bag full of ice cubes whilst it's connected to the counter/timer. I may not trust my counter/timer's absolute accuracy, but it would show a change if the thing's drifting. 

ReplyQuote
Topic starter Posted : 24/08/2020 11:37 am
Doz
 Doz
(@doz)
Noble V-Ratter Registered

Nope, the oscillator is innocent. Heating up with a hair dryer to ouch levels, and cooling with a bag of ice (not straight away, didn't want to crack the bottle) does vary it's output, but not by any significant amount.

It's got to be in the software. There are 3 bits of software running, one receives the ~28Hz from the oscillator via an interrupt, and (in normal use) outputs seconds over the serial interface at 9,600 bits per second, in the "calibration mode", it outputs minutes and seconds elapsed over the same interface. The second bit receives the serial stream, generates a video waveform for display on a CRT monitor. The third bit runs on a PC, and receives the data from the first chip in calibration mode. It compares this after a period of time to an NTP time server.... any difference is calculated as a number bigger or smaller than 1. 

I ran calibration for an hour, and sure enough it gained 28 seconds. Not good. The PC software spat out the difference as a number. That number is multiplied by the original value stored in the flash of the first 1Hz generator, and thus *should* correct the error. I dutifully did this and ran the software again for an hour. It gained 54 seconds .... hang on ... Sure enough the software again spat out a number greater than 1.... that's not right... that will increase the error again... Yep... my maths was dividing number of elapsed seconds form the oscillator by the true NTP elapsed seconds. Ah .... 1/x 

I thought this time I had it, and reversed the original changes to the cal value, and tried again... OK, now it's better, but still a couple of seconds gained over an hour.

I ran cal again, and had a very, very small error, certainly nowhere near enough to introduce introduce a 2 second gain over an hour.

I stared at the code for ages. Then I spotted it.

It all concerns the amount of data transmitted, and how the serial interface is implemented. When the clock is running normally, one byte of data is sent, the number of seconds. When the calibration mode is running, each character is sent over the interface as a byte , so , for example 60 minutes in is sent as 60:00, that's 5 bytes including the colon. Now I looked at my serial interface, and realised when it's transmitting, the interrupt from the oscillator is being missed occasionally. A change of interrupt priority was implemented, and bingo. It's been running for about 90 minutes now, and is still "to the second". I suppose there's a slight change in the data rate of the interface, but it's not dropping data. 

The moral? "To err is human, to make a real cock-up requires a computer"

ReplyQuote
Topic starter Posted : 24/08/2020 9:07 pm
Page 1 / 2