Poor's man DIY oscilloscope calibration

[picofarad]

Hi,
as I said on another topic, I just got an HP/Agilent 1707A analog oscilloscope. The instrument works well but there are some small calibration issues. I would like to calibrate it by myself, but the other only gear I have is an analog DIY function generator, apart a digital multimeter.
I know that without specific gear it's hard to do a good work, but I'm a beginner and at the moment I don't need so much precision.

I've already adjusted trace rotation and astigmatism, but now I have to calibrate vertical and horizontal amplitude.
The scope has the typical square calibration output, 1Vpp +/- 1% and 1kHz +/- 10% . I was thinking to check the frequency with a software frequency counter and then adjust the scope using this signal. Can I do it?

Any advices, tips and tricks? Please don't tell me to buy a 50.000€ function generator plus a 20.000€ frequency counter: it's a poor man calibration. 

[Paul Moir]

You may be able to do the lower ranges with a multimeter without much difficulty, but it depends on the specifications of your multimeter.  What kind is it, so we can know the specifications?  Does it measure Hz?

(The fundamental accuracy of oscilloscopes is not all that great, so you can do a lot with the multimeter.)

[alm]

I'm not familiar with the exact procedure for this scope, but in general you can get away with a DC source (power supply, battery) for levels and a function generator plus counter for time base. You can ground the input to find the 0V point, and then adjust the amplitude calibration until the level on the screen matches your DMM. I'm not sure what you mean by software frequency counter, but an accuracy of about 1% should be enough, the scope is probably specced for only 2% accuracy or so to begin with. Just don't touch any adjustments involving 'high-frequency', 'transient response' or fast timebase settings.

[picofarad]

What kind is it, so we can know the specifications?  Does it measure Hz?

My digital multimeter is a cheap one like this: https://webshop.cashconverters.com.au/thumbnail/850x520/1432995-boxed-digitech-qm-1320-fequency-digital-multimeter-1.jpg
It can measure Hz, but it has only the 20kHz range, so very low frequencies will not be shown.

I'm not sure what you mean by software frequency counter...

I mean one of those software (for example Visual Analyzer) that take a signal from the computer audio input and measures frequency.
I hope you understand. 

So, using a battery should I be able to calibrate? That's a good idea.

[Paul Moir]

Ok thanks.  I was thinking of doing AC sine waves measured by your DMM, but alm is right, you can just use a DC source measured by your DMM.  That's good because with that DMM you'd have to pretty much stick to measuring 50 or 60Hz sine waves since there's no real stated accuracy beyond that.

For accurate frequency signals, hunt around old gear for some crystal oscillators that look like this:  http://media.digikey.com/photos/ECS%20Photos/ECS-TTL%20CLOCK%20OSCIL.jpg

Those will be way more accurate than your oscilloscope ever will be.  You can power them with 5v.  Alternatively you can order a couple; they're only a couple bucks each.  You can grab them off newer surface mount gear but they're far less common because everything is geared to work with a raw crystal these days.  You'd also have to figure out if they're 3.3v or 5v. 

Another good source for a signal is an electronic clock of some sort.  If you probe the watch crystal in it, you should get a nice accurate 32.768KHz signal.   

Finally the AC mains measured through a transformer can be an OK low frequency signal.  In theory they can be off a fair bit at any instant, but generally they're pretty close to dead on.

[Shock]

You could also check by taking your gear or improvised reference components to someone who has a recent calibration or reference you can measure off.  Check for local hackerspaces there is about a dozen in Italy or amateur radio/electronics groups.

[picofarad]

I know that probing a raw crystal oscillator could stop it. It's true?

Can I use a sine audio file generated by a computer? The frequency should be precise enough.

[picofarad]

So, today I opened the scope to look a little bit inside and try to calibrate. I downloaded the manual and I looked for calibration trimpots, but I have some doubts:
 - there's a trimmer named 'HORIZ GAIN', which I assume it's for the timebase calibration, right?
 - I think there are two vertical gain trimmers: one it's named 'X1 GAIN ADJ' and the other one 'X10 GAIN ADJ'. Are them the vertical gain trimpots (I guess that X1-X10 is for the MAG function)?

If you need to watch the manual, it's here: http://www.hparchive.com/Manuals/HP-1707B-Manual.pdf

Thanks.

[alm]

The adjustment section in the manual is not just filler material. See pages 5-22 to 5-26 of the manual you linked to, the adjustments you mention are all described in there. Order is also often important. For example, adjusting power supply voltages may affect all downstream adjustments. In general I wouldn't adjust power supply voltages unless out of spec. You will have to improvise or skip some steps (eg. high voltage power supply adjustments, high-frequency pulse adjust).

[madshaman]

Does function gen take an external reference clock?  Of so, the first improvement you can make is buy yourself a 44€ rubidium reference and hook that up.  I can help you if the idea seems daunting, really those boxes just need a power supply and some decent heatsinking (they get hot and will draw about 2 amps during startup at about 12-15V).

The other thing that might be useful is an RF power meter with a calibrated probe and attentuator, but that won't be as cheap (maybe €250 or so?).

A good frequency counter that goes up to a Ghz or so doesn't hurt.

That aside, driving your function gen from an external reference will get you further along for generating timing signals.

I also mentioned the Tek 184 time mark generator in another thread, not as a joke.  It's old, but it's really good and can probably be found cheap.

imho, calibrated and accurate timing signals are much easier and cheaper to generate than accurate level amplitude and/or power signals.

Scope calibration really does require some half-way decent gear imho, but you can still acquire it without breaking the bank.

[alm]

A rubidium oscillator to calibrate the RC-based timebase of an analog scope that might be 1% accurate on a good day? That is seriously overkill. Any frequency counter (even those built in a DMM) should be plenty accurate, though it may not have enough bandwidth. I don't see why you'd need signals up to 1 GHz for calibrating a 100 MHz scope, though.

I don't see the need for a power meter either. Amplitude is generally adjusted with a fairly slow chopped DC voltage (pulse train). A DC voltage and manually shorting the input can substitute for this if you're not in a hurry. HF adjustment is only for best pulse response, which requires a fast pulse with a known shape. In the absence of a known-good scope I'd suggest skipping this. The performance verification procedure contains a bandwidth test, but it's only a pass/fail test without any adjustment.

[madshaman]

Maybe for 100Mhz, but it's not uncommon to own a 400Mhz analog scope and some own 1Ghz analog scopes.

The bigger point is using any good reference to discipline the OP's function gen will have a good chance of ensuring it's output is correct with minimal phase noise, and where else could one find a cheaper frequency reference, even vs an ocxo?

I'd by hard pressed to remotely consider myself an oscilloscope expert, but even if you don't beed one, I'll bet you want a rubidium reference to plug into all your gear.

[picofarad]

The scope has a 35 MHz bandwidth. As I said in the first post, I don't want to buy extra stuff, specially if I'll use it once every 10 years to calibrate my scope.
However, I found that sine audio files generated by softwares like LTSPICE or even Audacity are precise enough, measured with a GW Instek frequency counter at school.

[victor]

I think you can use almost any digital electronic that works and probe the crystal while it is running and assume that the frequency is spot on the crystal spec.

Grab anything that may have a micro crontroller (webcam, mp3 player, modem/router) anything that have a lcd will probably have a crystal to run the clock, leave it running for like 5 minutes, and use it as a reference clock.

[picofarad]

If I probe a crystal, there is no risk to stop it? I read it on the web...

[AndrejaKo]

It depends a lot on the configuration of the crystal oscillator as a whole. What you will do with a scope probe is add capacitance to the circuit. If that capacitance is great enough, you could stop the oscillator or significantly change its frequency.

Even if you do not stop the crystal, you will definitely change the frequency of the crystal. The change of the frequency will probably be insignificant enough not to matter, if you only need rough calibration. The procedure is safe enough, so I don't see a reason why not to try it and report back.