Calibration.

[eclectic]

Calibration.
Just musings at the moment.

Professionals need calibrated instruments.
Most amateurs/hobbyists do not and, indeed, cannot afford them.

However, what methods are available to us, even if only for interest?

For example, I own a cheap oscilloscope, but how can I test its accuracy?
I've been thinking along the lines of a scanner-radio c/w a Picaxe pwm transmission.
Still using radio, hacking a cheap “radio controlled” clock. (Ex Rugby, now Anthorn)

Any other thought, perhaps using Picaxe timer circuits, or anything else?

Mmm, this might make an interesting student project?

 

[BeanieBots]

For frequency calibration I always used to use Radio 4 which used to broadcast on exactly 200kHz LW.
It's changed now but I think there is still a station on that frequency.
Maybe some other local station could be used.
A 'quality' quartz clock chip is good for a few ppm.

Voltage wise, a 'scope will never be very good so any old DVM will be good enough.

More importantly, make sure your probes are tuned properly or the waveform will be distorted. Use the built-in "calibration" terminal for that.
If it doesn't have one, (chuck it away) I mean use a PICAXE PWM signal.
DON'T use a PICAXE for frequency calibration.

 

[Dippy]

Obviously 'true' calibration has to be paid for.
Many products from places like RS and Farnell can be purchased with a calibration option, but it costs.
I can't imagine any cheapie stuff from Ebay vendors (and cetera) being calibrated.

I can only suggest, as an easy option for a limited range, connecting your scope to the OSC1 or OSC2 terminals of a PIC+Crystal.

If your 'scope leads are set up you should see a nice wave.

I've done this with a Tektronix scope and the results are good.
...unless both the 'scope and crystal were out by the same amount....
(I should add that this particular 'scope had been calibrated by Tektronix a year previously).

And as BB says use the PICAXE PWM (but IMHO only if PICAXE is using a quality crystal).
But, be careful, PWM frequencies are NOT infinitely adjustable on PICs. They are in 'steps' and the step (generally) gets bigger as your F goes up. (Similar for duty % btw). In many cases they can be bang-on.

You can calculate this from the PIC Data Sheets.
You will see that you may need to juggle crystal frequencies too in order to get accuracy.
e.g. You want 17kHz
With a 10MHz crystal you can get (according to calcs) 17.0 kHz
But, with a 20MHz crystal the best you can get will be 16.9kHz or 17.3kHz depending on your register choices.
... assuming my calcs are correct.
Anyway, bottom line, - think, read, calculate - don't assume.
Far easier is to download PIC MultiCalc then you can pretend you know the Data Sheets off-by-heart!

 

[manuka]

The 2009 leap second didn't bring this on did it?!

The science of calibation is indeed a costly field, but (akin perhaps to the original metre) secondary standards can often be sufficient. Hence wave a few ale over the electronic boffins at your local uni/poly & they may let you have access to their 2° standards. Be warned however that some of these can be long over due for checking against more professional standards, or they may be inherently prone to wandering deviations as a result of their yester-year design & service life. I've had past problems with abused Weston Standard cells (1.018659 V @ 20°C), which if used carefully hold this value indefinately.

Replies above tend to assume it's just the scope you want to calibrate- is this so?

 

[manuka]

As a guide to calibration costs, a NZ govt. standards facility I'm associated with charges ~US$500 to calibrate even a thermometer. However the supplied paperwork is legally bullet proof- a important issue.

 

[Dippy]

'tws only cos eclectic mentioned his scope.

And the trouble with cheap scopes is that todays calibration may be next months cockup.

I haven't got any Unis near me so I can't scrounge. And I forgot to steal some of the standards when I left UKAEA.

Within our budgets maybe get hold of some precision voltage references such as LM4030 (and there are many other devices that are even better) and maybe some quality precision resistors. Yes, you will have to trust the manufacturers, but as you don't want to spend any money then you're snookered.

Just stay away from anonymous rubbish and rejects from cheapo suppliers. (You know who I mean).

 

[BeanieBots]

Besides the few years I spent working for an instumentation company, I've never found absolute accuracy to be very important. It's consistancy, drift, ageing and temperature effects that are the killers. In most cases you need to be more than 10% out before you are likely to meet any real problems but a 5% drift caused by today's temperature as apposed to yesterday's can catch you out by surprise.

A good example of this might be powering a PICAXE via a few series diodes.
A few percent out due to a 'bad' DVM, followed by an extra 0.3v when the central heating comes on.

 

[Dippy]

I reckon on most occasions you're right Beanie re: accuracy in general.
And I agree with the other bits re drift etc.

But a bit different when you're designing products for others when voltages/ADCs are being used for example. Who could trust your product if you haven't tested accurately?
And, of course, for testing many components you need accuracy too.
You need a standard somewhere.
Horses for courses. Ditto the degree of accuracy/precision ie. 10% in 70% of cases, 1% in 25% of cases and 0.1% in 5% or something like that (whatever).

I can just imagine Jezzer C saying "and the new Ferazza does a 0-60 in 4 seconds +/- 50%"

Confucius once said: "If a 0.1% reference costs £2 and a 1% reference costs £1.80 then the wise man spends the extra 20p and the very wise man buys three of them to take an average".

 

[Andrew Cowan]

Confucius once said: "If a 0.1% reference costs £2 and a 1% reference costs £1.80 then the wise man spends the extra 20p and the very wise man buys three of them to take an average".

Do you mean three of the 0.1% ones?

 

[Dippy]

I didn't say it, Confucious did.
That is Miss Peter Confucious, Ward 3 (Gender re-assignment Dept), Broadmoor.

Well, I'd buy more than one - wouldn't you?
And the 2 out of 3 agreement method is a common practice in many fields.
What's a few quid for peace of mind?
Three pints of lager versus a near-calibrated DMM ... no contest.hic.

Anyway, any decent mathematician knows how to add uncertainties and drive an uncerainty for that.

 

[BeanieBots]

"If a 0.1% reference costs £2 and a 1% reference costs £1.80 then the wise man spends the extra 20p and the very wise man buys three of them to take an average".[/I]

And the China man bought up all those that were out of spec and sold them on Ebay and made his fortune.

 

[Dippy]

Haha, absolutely.

And you only make fortune if someone buys...

PS. I noticed I said "drive" in my previous post. Due to poor typing and lack of proof reading I didn't notice it at first. However, thanks to the rapid DippyFastView response update systems, it should say "derive".

 

[BeanieBots]

And you only make fortune if someone buys...

I think the bulk of issues on this forum is proof enough that there are plenty who will buy if it's cheaper.
Not supplying technical data and/or source so as not to confuse the unwary also seems to work!

Then again, "Found in manufacturers out-of-spec bin" might not help sales.

 

[eclectic]

Thanks for the interesting replies, so far, folks.
Excellent discussion.

@Stan.
“Replies above tend to assume it's just the scope you want to calibrate- is this so?“
Any ideas about SI base /derived units welcome.

@Dippy
“Within our budgets maybe get hold of some precision voltage references such as LM4030
(and there are many other devices that are even better) and maybe some quality precision resistors.
Yes, you will have to trust the manufacturers,
but as you don't want to spend any money then you're snookered.“

Actually, I'm quite happy to spend money on a good cause.
However, a secondary motive behind my question was to provide ideas
for the younger / student members of the Forum, who can't spend any money,
'cos they haven't got any.

I've just dug out a half-hacked MSF clock, and looked at the signal.
More things to play with.

e

I've just noticed the "hole" in the 'scope trace.
One day, I'll read the manual.
Or is the space the Cumberland Gap? ;-)

 

[papaof2]

Having good calbration and a device which can hold that calibration is always good, but sometimes "close enough" is good enough. I don't get the Fluke DVM out to check charge voltage to a gel cell battery - the $8US DVM is close enough when you don't have a temperature-compensated charger. I'm much more likely to reach for the Fluke if checking a lithium battery.

For timebase checking, I once brought the output of the color timebase crystal of a TV (NTSC) to the rear panel to have a signal that was of traceable calibration and strong enough to drive a frequency counter for calibration at 3.579545MHz. If you square that signal up, it also works for checking scope calibration.

The AT&T network design organization that I worked with in Atlanta had a secondary atomic frequency standard, but they didn't let me bring it home when I retired (that would have been a great retirement "clock").

For some portion of the forum, consistency is probably more important than absolute accuracy. The garden pump controller I'm building needs to deliver the specified amount of water each time it runs. Whether 50 gallons is delivered as 45 gallons or 55 gallons isn't nearly as important as having it always deliver the same amount.

The flowmeter used to measure the pump output provides 820 pulses/gallon; if the interrupt circuit *consistenly* missed 2% of the pulses, then the amount of water delivered would be the same each time - just a little more than expected. If the interrupt circuit missed 20% of the pulses, the extra water would be obvious and there would be a problem with my design (either wiring layout or clock too slow). So far there's been consistent delivery during testing (breadboard with the PICAXE running at 8MHz).

The "quantity to deliver" selection is yet to be added, but there are several examples of menu systems already on the forum.

John

 

[Dippy]

Well this could turn into 10 pages of A4 if we're not careful.

Yes, I agree, 'close enough' is perfect in many cases. BUT, I bet you checked your $8 DVM against your Fluke...... no? So, how do you know it's 'close enough' for good quality battery charging? What's 'close enough'?

You have to trust your instrumentation don't you. Would you trust an $8 DVM in isolation?
I wouldn't.

If your circuit consistently missed 2% that's fine. But without a true measurement how do you know? In that example you can calibrate it against a gallon bucket of water. But suppose you had to measure the flow in a closed circuit. Say you wanted to measure energy transfer from a Solar Panel (water) to your hot water tank. Maybe thats a bad example.

Suppose you are doing some ADC work to measure a 12volt battery. Unless you can guarantee the supply/reference to x% you can't guarantee your results to x% (plus the factor introduced by a potential divider).
It's almost chicken and egg.

Will the $8 DVM be good enough for that? I dunno, but if it agreed with a $300 calibrated Fluke then I'd be a helluva lot more confident.

People blindly use sensors - and the results are always bang-on aren't they. That DS18B says 20 degrees so its 20 degrees innit.


I would have thought that Papa's scope cal. method from the crystal timebase was perfectly good, as is testing from PIC crystals. If these broke people can afford a PICAXE they can afford a 40p crystal, surely?
Perhaps the kiddiewinkies could hack into their Playstations?
I assume there was enough money kicking around for those!

(wonders: Will a cheap Ebay scope actually handle >1MHz?)

But hacking Grandad's Radiogramme and assuming that the 1963 voltage regulator (covered in dust and hair) is perfect is not too clever.

I think the only way to be confident and at no charge is along Stan's lines ... borrow, beg or bribe with offers of mowing the lawn.

I'm just wondering if there is a way to do volts where uncertainties mathemtically cancel (you can do it with some things)... but lifes too short so I've stopped.

I guess (cost aside) that expectations, requirement and what the device can tolerate are the criteria.

Or you just explain to your parents that you'd like a few quid to further your education.... oh, but wait.. that would eat into the Wii tennis racket you need..

 

[BeanieBots]

Brand new (dry alkaline) battery within 2 years of it's expiry = 1.55v +/- 3% @ 20C.

 

[premelec]

[Slightly edited] "An egineer and a mathematician were one meter from a delicious picece of chocolate cake and allowed to halve the distance between them and the cake every second - the mathematician never got to the cake but the engineer got close enough for all practical purposes."

It usually boils down to how close do you really need to get... I saw a surveyor setting up a local GPS station the other day and asked him how accurate he could get with it and he said .1 inches per 7 miles. That sounds excessive furthermore it isn't really that accurate in terms of frequency measurements [.1/(7x5280x12]... Of course errors multiply and so it's good to stay a bit more accurate than you need...

We mustn't forget the difference between precision and accuracy - to be precisely accurate about this would be.... [old man's mind drifts into the fog... smile appears...]

 

[papaof2]

Close enough

Yes, I agree, 'close enough' is perfect in many cases. BUT, I bet you checked your $8 DVM against your Fluke...... no? So, how do you know it's 'close enough' for good quality battery charging? What's 'close enough'?

Of course - also checked the two $4US ultra-cheapies from Harbor Freight ;-)

All were within a few hundredths of a volt compared to the Fluke at 13.8v - well less than 1% difference, so better than I anticipated. The Harbor Freight leads have very limited life - the first one went open internally in about a year. I would never use the original leads for current measurement (wire too small for more than a few mA without serious voltage drop). However, these meters do stand up to more punishment than an analog meter would (a Simpson 260 would not survive a 3 foot drop to the concrete floor).

John

 

[manuka]

Any ideas about SI base /derived units welcome.

Eclectic: (Warning- non metric US readers may glaze over here.) What are you after & for whom?? For educational skinflints you can't beat using water & kitchen digital scales- & even a tape measure. (1 litre has mass 1kg etc). How about the original metre definition itself- a 10 millioneth of the distance between the Nth. Pole & equator thru' Paris.

And speaking of the globe- the accuracy of clocks behind the GPS navigation system is perhaps THE most crucial everyday calibration example. I should imagine even a microsecond drift may be unacceptable. Mmm-let's see- "c"= 3 x10^8 m/sec, so in 1 µsec that's 3x10^2 m = 300 metres. Yikes- try GPS driving around London with that sort of uncertainty!

 

[eclectic]

Eclectic: (Warning- non metric US readers may glaze over here.) What are you after & for whom?? For educational skinflints you can't beat using water & kitchen digital scales- & even a tape measure. (1 litre has mass 1kg etc). How about the original metre definition itself- a 10 millioneth of the distance between the Nth. Pole & equator thru' Paris.

!

What am I after?
Absolutely nothing, except looking at interesting ideas.
My original post was based on random musings,
on a freezing cold "hiding indoors" Sunday morning.

I think I've achieved my aim(s).
1. Interesting reading, and
2. Possibly some ideas for others.

e

 

[Dippy]

Aha, Papa, so you did 'calbrate' it/them (well, check really) against a standard then.

"...a Simpson 260 would not survive a 3 foot drop to the concrete floor)".
- Doh!

Glad you are sorted ec.