Is it acceptable to calibrate my oscilloscopes with just a x-tal?

[BlownUpCapacitor]

Hi, the main purpose of this post is to settle a disagreement between my friend and I. When my friend saw how I calibrated my own oscilloscopes, he went on a rant saying how I shouldn't be using what I am using to calibrate my scopes.

For context, I am using essentially just a x-tal to calibrate my scopes. More specifically, a Global Specialties 4010 10MHz Crystal Controlled Pulse Generator. This is what I consider to be my lab frequency standard for calibrating all sorts of things, mainly oscilloscopes. This is only for the calibration of horizontal timing.

I don't calibrate modern DSOs, but rather old-school scopes such as the Tek 7904, Tek 2230, Tek SC502, 5110, 7613... etc.

Anyways, back to the point. You see, my friend insists that I use very accurate frequency standards like rubidium frequency standards. I personally think this is overkill, and just a simple X-tal controlled pulse generator is good enough.

What are your guys' thoughts?

[bdunham7]

It should be obvious to anyone that can read and understand the scopes manual and specs that even a generic AWG will be sufficient for almost all CRO calibration.  Some higher performance versions could also use a fast-edge pulser of some sort and an accurate RF generator.  The specs for sweep rate and linearity are 1% at best in most cases.  I was able to do the calibration procedure on my 2465B with just what I've listed plus normal bench stuff and there's no faking it in that case since the scope either accepts or rejects your inputs.

[TimFox]

I have a military-version Ballantine 6130A "time-mark generator" which is a good crystal oscillator and a switch-selectable frequency divider to get time marks in a 1-2-5 sequence over a typical analog CRO range of time-base settings.
Since one can resolve roughly 1% of the total time base (10 divisions, seeing 1/10 of a division) when comparing the pulse train to the total sweep, this is plenty adequate to calibrate an analog CRO.
The 22-position rotary switch allows checking the time base over a wide range of time/div.
Better oscillators, such as GPSDO or fancy Rb standards can be useful to calibrate a frequency counter, which has far better resolution (better than 1 in 10⁶).
I learned about significant figures in 11th-grade chemistry class back in high school.

[Whales]

The V and H gains on my CRO drift as it heats up, far more than the ppm of a standard quartz crystal.  Tweaking them to a rubidium standard would keep them within that level of cal for a few... seconds?  minutes?  I'd have to ovenise my CRO.

[langwadt]

The V and H gains on my CRO drift as it heats up, far more than the ppm of a standard quartz crystal.  Tweaking them to a rubidium standard would keep them within that level of cal for a few... seconds?  minutes?  I'd have to ovenise my CRO.

and how could you possibly see the difference on the screen?

[tggzzz]

Hi, the main purpose of this post is to settle a disagreement between my friend and I. When my friend saw how I calibrated my own oscilloscopes, he went on a rant saying how I shouldn't be using what I am using to calibrate my scopes.

For context, I am using essentially just a x-tal to calibrate my scopes. More specifically, a Global Specialties 4010 10MHz Crystal Controlled Pulse Generator. This is what I consider to be my lab frequency standard for calibrating all sorts of things, mainly oscilloscopes. This is only for the calibration of horizontal timing.

I don't calibrate modern DSOs, but rather old-school scopes such as the Tek 7904, Tek 2230, Tek SC502, 5110, 7613... etc.

Anyways, back to the point. You see, my friend insists that I use very accurate frequency standards like rubidium frequency standards. I personally think this is overkill, and just a simple X-tal controlled pulse generator is good enough.

What are your guys' thoughts?

You should have fun getting your friend to explain why a rubidium standard is more beneficial in this case than a simple crystal oscillator. Get him to put numbers on his contentions; no adjectives.

Unfortunately it will be fun at your friend's expense.

[tggzzz]

The V and H gains on my CRO drift as it heats up, far more than the ppm of a standard quartz crystal.  Tweaking them to a rubidium standard would keep them within that level of cal for a few... seconds?  minutes?  I'd have to ovenise my CRO.

Horizontal gain/linearity/accuracy is far better with modern digitising scopes than with old CRT scopes.

Vertical gain is little different; digitisers have different classes of errors. (See my .sig )

Fundamentally a scope's principal purpose is to faithfully represent the waveform's shape. Anything more than that is a bonus. If you want time/amplitude accuracy/precision, there are better tools.

[David Hess]

A crystal oscillator is more than accurate enough.  When I first calibrated oscilloscopes, I used a function generator, frequency counter, and AC voltmeter which is still more than good enough for horizontal and vertical calibration.

Transient response calibration however requires a reference level pulse generator.  Common pulse and frequency generators are not good enough.

[mendip_discovery]

Bradley 192 is a a dedicated scope calibrator.

That isn't super special.

Iirc the Fluke multi product calibrators also come with a scope option.

[Conrad Hoffman]

AFAIK, even the most modern scope "timebase" is controlled by a crystal. I put timebase in quotes because it's all just processors today anyway. Scope with frequency counter readouts usually don't have a ton of digits, so you only need accuracy good enough for that. You'll probably never see errors on the normal display. A decent crystal should be fine and probably zero advantage to a fancy disciplined oscillator.

[Andreas]

AFAIK, even the most modern scope "timebase" is controlled by a crystal.

Hello,

on better scopes its a bit more than a standard crystal with +/-50 .. 100 ppm drift.
Especially when using long FFTs or ETS sampling a better stability is essential.
Unfortunately not all oscilloscopes spec such things.

with best regards

Andreas

[bdunham7]

on better scopes its a bit more than a standard crystal with +/-50 .. 100 ppm drift.
Especially when using long FFTs or ETS sampling a better stability is essential.
Unfortunately not all oscilloscopes spec such things.

It didn't even occur to me that we might be talking about non-TCXO crystal oscillators, so I guess my answer was a bit incomplete.

IMHO, 50ppm would be joke quality by modern standards.  Of course, a scope could be perfectly useful for many things even with that low level of precision, but why?  Much better is very easy to come by.

For the mid-to-low end of my expectations (one notch above entry-level budget models), take the Siglent 2104X+

±1ppm initial; ±1ppm 1st year aging; ±3.5ppm 10-year aging

For the high end of my expectations, I'd want to see a 10MHz in connection for a reference clock and a reasonably good implementation of locking the scope clock to that reference.

[bdunham7]

Fundamentally a scope's principal purpose is to faithfully represent the waveform's shape. Anything more than that is a bonus. If you want time/amplitude accuracy/precision, there are better tools.

That may be true for the decades-old models that the OP is calibrating, but for any decent modern DSO it may in fact be the best tool on the bench for many uses that previously merited a specific instrument.  AC RMS volts for complex waveforms, pulse width, pulse amplitude, gated measurements, various timing applications, power analysis--the list is long and there are quite a few where the DSO is either the best or the only game in town.

[ejeffrey]

AFAIK, even the most modern scope "timebase" is controlled by a crystal.

Hello,

on better scopes its a bit more than a standard crystal with +/-50 .. 100 ppm drift.
Especially when using long FFTs or ETS sampling a better stability is essential.
Unfortunately not all oscilloscopes spec such things.

with best regards

Andreas

You want decent stability for ETS, but that is governed by the timebase in the scope not the calibration.  Absolute accuracy doesn't matter at all for ETS.  If you really want to measure absolute time / frequency at the ppm levels then it matters but for 99% of the measurements I have ever done, a 1000 ppm offset in the timebase would have no effect.  And 50 ppm is a joke, a cheap watch crystal is better than that.

With modern scopes with accurate hold off and deep memory you can actually do that sort of measurement, so it's nice to be able to calibrate but there is definitely no need to use an accurate standard for ordinary measurements.

[TimFox]

The OP's argument with his friend pertained to analog CROs.

[Wallace Gasiewicz]

If you look at the OP.s scope trace, it is right on the graticules at the beginning and end of the trace but in the middle is is slightly off, then comes back onto the graticle.  This is quite common with CRT scopes and due to non linearity maybe in the CRT itself. Most of the time no amount of adjustments can cure this entirely.  I think some of his scopes have on screen readings for freq and other values and these can be affected by good calibration. 
Unfortunately most scopes do not have a good freq standard,  even my newer Agilent digital scope has a poor freq standard.    And Agilent did not even put in a connector for an external standard.
I think his pulse generator is probably more than adequate for using as an adjustment standard. 
With most CRT scopes pretty close is as good as you can get. How close does the trace have to be to the graticule line? How thick is the line etc...

It might be more important to compare the pulse generator to a known good standard though.

[tggzzz]

If you look at the OP.s scope trace, it is right on the graticules at the beginning and end of the trace but in the middle is is slightly off, then comes back onto the graticle.  This is quite common with CRT scopes and due to non linearity maybe in the CRT itself. Most of the time no amount of adjustments can cure this entirely.

Any pulse generator, even a 555 timer, can be used to adjust the sweep linearity. All the scopes I've fiddled with have some capability in this respect, but I haven't bothered with low-end scopes.

The horizontal magnification does require an accurate frequency source. But remember that each sweep speed will be determined by selecting combinations of resistors and capacitors, and it may not be possible to adjust each range independently of the others.

Most importantly, see the scope's specification. 2% translates to one tick in 10divisions.

[David Hess]

CRTs are good enough that maintaining horizontal linearity requires the horizontal amplifier to use shunt feedback.  The tricky part about adjusting it, especially with the 10x magnification engaged, is that an accurate high frequency is required which will likely be outside the range of most signal generators.

Horizontal amplifiers are also surprisingly slow, with a bandwidth of 1 or 2 MHz being typical.

[Wallace Gasiewicz]

So the OPs trace looks OK? I think so.


Quote from: tggzzz on Yesterday at 10:51:07 pm

Most importantly, see the scope's specification. 2% translates to one tick in 10divisions.

[BlownUpCapacitor]

Horizontal timing-wise, if the x-tal is a good freq standard, the scope is well within cal. The manual specifies the 7904 horizontal accuracy to be 0.1div or better. Mine looks like 0.05 or 0.06, so it's good. I didn't need to adjust much of anything even though this 7904 spent 20 years on the shelf of an e-recycle center, got rained on several times (because the recycle center likes to leave some stuff outside), and the fact corrosion is everywhere inside this thing.

The power supply is in spec, horizontal is in spec. Horizontal is in spec, (timebase plug-in only needs to be adjusted), vertical is mostly is spec, z-axis is in spec. Calibrator output is a bit off, but that was adjusted.

These were checked with the Global Specialties 4010, Tek PG501, Tek FG501, Tek 191, Tek 2230, and Tek DC503. All these have been calibrated recently (2019) by my local college, because that's where I got them from.

I have since cleaned up the scope very nicely, removing most of the corrosion.
I would also like to provide an update that my friend has accepted defeat. But he says he'll still use his two rubidium frequency standards to calibrate his scopes, averaging out the two frequencies and finding "sweet" spots.

[Wallace Gasiewicz]

Yes, it certainly looks fine to me, I have done a few of these adjustments on my old scopes and your trace looks just fine. 

Your friend has too many frequency standards.  Like the man with two watches. He never is sure of the time of day.   
He needs to give one to you., because you were correct in your estimate of using a pulse generator.
On second thought, then you would have two things to compare so this is kinda like gifting a "white elephant".
HA!!!

[tggzzz]

So the OPs trace looks OK? I think so.


Quote from: tggzzz on Yesterday at 10:51:07 pm

Most importantly, see the scope's specification. 2% translates to one tick in 10divisions.

Well, a spec is a limit, not what's achievable with tweaking. I haven't checked that scope's spec, either.

My course of action would be to verify all power rails voltage and ripple, make sure backplane/switch contacts are stable, then get the service manual and tweak it until happy

I imagine the linearity could be improved a bit, but RTFM