Is this funtion generator junk?

[eswets]

I just purchased a Madell CA1640-20 function generator and right out the box, the frequency knob fell off.  then I go to test the thing and in all different frequency ranges, the frequency changes after I set it.  So if I start it at 2khz even, in about 45 second it will climb up to 2,010hz then down to 1990hz.  Is this normal or junk?  Here is the link:
http://www.madelltech.com/m1-8.html
So I don't need a top of the line generator, I just need one that I can play around with to help teach myself.  Let me know if I should return this and if so what would you suggest for a $250-500 budget?  Again I don't know much about these so I took a chance with this.  I should have asked here before I bought.
Thanks
Eric

[rossmoffett]

All of the function generators I've worked with in that price range (I've never seen anything more expensive because of educational budget constraints) have that problem.  Spot-on frequency stability isn't necessary for most things, +/- 10 Hz isn't too bad.

Just look at the FM radio bands, in the USA they skip every tenth of a MHz.  You'll see 106.9, then 107.1 and so on.  So when you consider that even these super high powered and highly regulated devices can't keep within 50 kHz of their band a +/- 10 Hz drift doesn't seem so bad.  Really, it's necessary to drift that much because of the nature of FM radio, but I still think it's a good example.

[septer012]

I don't think so ross.. I think they have that gap because the tuning  filter in your radio is junk.  There isnt a 16 pole filter in your clock radio.

Edit: the spacing is mandatory by the FCC here in the states for better co-channel and receive-ability.

[Andrew]

The specs say it has up to 1.5 % (in)stability at some frequencies. That would be +/- 30 Hz at your 2000 Hz. Your 1990 Hz to 2010 Hz are well within the specs. So you got what they promised. Not that I would have bought something with +/- 1.5 % stability, but they were at least honest upfront.

If this happened right after you turned the instrument on then there is hope that it gets better. Instruments like these need to be "warmed up" before they reach their most stable operating conditions. This typically means to have the instrument running for at least an hour. I know instruments where the manufacturer even requires four hours warm up to reach the guaranteed specs. Some people therefore never even turn of instruments.

However, the whole instrument as such looks like typical last decade Chinese design, copied from some western design from the '70 or '80 of the last millenium, re-badged for Madell. The OEM is likely Yang Zhong Ketai Electronic Instrument Co.,Ltd, alias Yang Zhong City Light and Electronic Instrument Factory, alias Micaltek, alias Caltek, alias Y-Caltek. As you see from the many names, we are deep down in the Chinese OEM naming game. And, of course, they have "tek" in some of their names in the hope that some of the reputation of Tektronix would rube off.

This is your generator in its natural habitat: http://www.y-caltek.com/cn/products/generator.htm

It is an outdated design, although designs like this are still sold by many, usual Chinese, manufacturers. Typically the stuff is robust but not very accurate. $250 is IMHO OK for such a generator, but not a bargain.  

BTW: If you want to learn, then one of the things to learn is that it is Hz and kHz, not hz and khz. Sloppy notation will bring you into trouble sooner or later.

[qno]

Maybe your sweep input is picking up some hum or noise.

Try to short circuit the sweep with a 50 ohm terminator or a short piece of wire.

Is there an FM modulation input?
If you short that maybe the stability improves.

[Zero999]

All of the function generators I've worked with in that price range (I've never seen anything more expensive because of educational budget constraints) have that problem.  Spot-on frequency stability isn't necessary for most things, +/- 10 Hz isn't too bad.

That's not good enough for FM, since the +/-10Hz, gets multiplied at higher frequencies,10Hz of drift at 3kHz is 10kHz at 3MHz and 300kHz at 90MHz.

[rossmoffett]

That's not good enough for FM, since the +/-10Hz, gets multiplied at higher frequencies,10Hz of drift at 3kHz is 10kHz at 3MHz and 300kHz at 90MHz.

Assuming a linear relationship, that the drift is not due to noise, etc etc.  It was just an example, not a technical statement.

[jahonen]

Reason for gaps on the FM is that there is finite information bandwidth in addition to the carrier, not due to frequency instability of the transmitter. Theoretically, the bandwidth of FM modulated signal is infinite (even with sinusoidal modulation), but in practice the sidebands fall off very quickly after certain bandwidth.

AM-modulated signal takes twice the space of modulating signal bandwidth. FM takes even more bandwidth, but gives advantage of better signal-to-noise ratio in presence of low carrier-to-noise ratio.

Regards,
Janne

[Zero999]

Assuming a linear relationship, that the drift is not due to noise, etc etc.  It was just an example, not a technical statement.

It won't be linear, if it were then when it's set to 10Hz, it'll vary between DC and 20Hz which would be horrible.

[jimmc]

Assuming a linear relationship, that the drift is not due to noise, etc etc.  It was just an example, not a technical statement.

It won't be linear, if it were then when it's set to 10Hz, it'll vary between DC and 20Hz which would be horrible.

Eh? Surely linear in this context means that drift is a constant percentage of frequency. To extend your previous example

...10Hz of drift at 3kHz is 10kHz at 3MHz and 300kHz at 90MHz.

or 1Hz at 300Hz or 0.1Hz at 30Hz or 0.03 Hz at 10Hz.

Jim

[jimmc]

I think this thread has gone a bit of track, function generators are not designed for RF testing.
Their forte is the variety of waveforms available and their wide sweep range.

In general there are two common types of function generator..

'Analogue'  based on a relaxation oscillator producing a triangle waveform which is then shaped to give sine and square waveforms.
See MAX038 for an example of the basic chip. http://datasheets.maxim-ic.com/en/ds/MAX038.pdf.
Frequency readout is directly from a dial or via a counter.
Poor stability but cheap and good enough for a lot purposes.

Arbitrary Waveform Generator this uses an accurate frequency (usually from a DDS) to clock data from memory through an D/A converter. By loading the appropriate data into the memory any waveform can be output.
In order to have a reasonable number of samples within each period of the output, the clock frequency must be many times the output frequency. Thus a high speed memory and a fast AD converter is required, the costs of these has dropped enormously in the last few years, making this approach more affordable.
For a example see our old friends Rigol http://www.rigolna.com/products_dg1000.aspx
Crystal controlled stability but still more costly.

To answer the OPs question, no it's not a heap of junk, it's performing as it should; just be aware of it's limitations.

Jim

[Zero999]

Assuming a linear relationship, that the drift is not due to noise, etc etc.  It was just an example, not a technical statement.

It won't be linear, if it were then when it's set to 10Hz, it'll vary between DC and 20Hz which would be horrible.

Eh? Surely linear in this context means that drift is a constant percentage of frequency. To extend your previous example

...10Hz of drift at 3kHz is 10kHz at 3MHz and 300kHz at 90MHz.

or 1Hz at 300Hz or 0.1Hz at 30Hz or 0.03 Hz at 10Hz.

Jim

Yes, that's what I meant by saying it's an unacceptable amount of drift for an FM oscillator - 300kHz @ 90MHz.

[Simon]

Assuming a linear relationship, that the drift is not due to noise, etc etc.  It was just an example, not a technical statement.

It won't be linear, if it were then when it's set to 10Hz, it'll vary between DC and 20Hz which would be horrible.

its a case of percentage, nothing to do with linaer ! at the frequency increases so will the drift as its expressed in percentage not a set figure

[rossmoffett]

Plot y = .02x

Is it a line?

[Simon]

that will not produce a line !

[rossmoffett]

that will not produce a line !

http://www.wolframalpha.com/input/?i=y%3D.02x