What's a good signal generator?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Kingston

Well-known member
Joined
Nov 1, 2005
Messages
3,716
Location
Helsinki, Finland
So I got my first proper oscilloscope. I'm enjoying my new-found window to signal chain internals, poking around all over the place.

But I bumped into a problem: I need a good signal generator. I've been using RME HDSP, but its limited to sinewaves (2.55VAC peak to peak, 50ohm output, roughly 10hz to 40khz). Of course I can output any waveform imaginable from a soundcard, but I found this method is severely bandlimited when viewed with a decent oscilloscope. Square wave output is not really squarewave, as expected, and the bandlimited "square" is useless trying to find how gear on the bench is handling transients. Sines of course not affected.

So what's a good signal general to be paired with a decent 50-100mhz DSO?

The equivalent (in value) signal generator to my oscilloscope would be Rigol DG1022 - often sold as a kit with DS1052E oscilloscope - but the feature set seems excessive for my needs. Perhaps not, please advise. I don't really know what features I should look for in a signal generator, except more waveforms than just sine, and wider frequency, impedance and voltage range than allowed by RME HDSP.

Also not looking to spend as much as I did for the DSO, but perhaps I have no choice. And it doesn't have to be digital either, or maybe it does. Again, please advise. Maybe there are even DIY kits to make good signal generators?


Thanks,
Mike
 
£50 looks like a decent buy. If you want one then I could forward to Finland for you, if the sellar won't.

http://www.ebay.co.uk/itm/Farnell-sine-square-audio-oscillator-LF1-10Hz-1MHz-/200687545900

I have a similar old Farnell analog generator that has been great over the years. It is not quite this model, but very close. I was going to get one of these as a backup anyway - let me know if you want me to grab one for you.

Stewart
 
Here is a slight upmarket model

http://www.ebay.co.uk/itm/FARNELL-FREQUENCY-SINE-SQUARE-OSCILLATOR-LFM3-50-60Hz-/270869493538?pt=UK_BOI_Electrical_Test_Measurement_Equipment_ET&hash=item3f11147722

What I like about this is, if the meter is connected directly across the output as a voltmeter, will show the true level. So if there is a short on the circuit you are injecting the signal into, the meter will show the change in level. My old sig gen used to do this but my new HP does not......

Peter
 
I see. those are good starting points. Now I know what to look for feature wise in basic models. But I think I'd be more happy with a something brand new from the factory for critical tools like these. These aren't like weller soldering iron transformers that you can drop on the floor as you please and they'll be like nothing happened.

Perhaps something like Instek SFG-1003? It's still on the expensive side, and I don't really need the frequency counter feature, but getting closer. I also don't need any metering on the signal generator, I think the oscilloscope and multimeter will take care of all those needs.

The search continues.
 
Have you searched at EEVblog.com?  They have some good threads on test equipment.  They LOVE the Rigol DSO's (I have one) but they are very critical of the Generator.  I recently decided to get an older one on Ebay for about $50 after researching the Rigol.
 
FWIW I use one of these not too shabby for the $  No metering

I believe it was ~$160 USD New

Instek GFG 8215a

Kingston I'm blown away that you've been doing what you do with no generator/scope  :eek: 
I find myself shadowing your moves on several builds now, if I can only grasp the theory as you have so well.
 

Attachments

  • gfg8215a.jpg
    gfg8215a.jpg
    164.4 KB · Views: 47
Echo North said:
Have you searched at EEVblog.com?  They have some good threads on test equipment.

Yeah, that's how I chose my oscilloscope as well. Awesome stuff there. "first thing you do when opening the box of some new gear, take in that new gear smell"

and next

"don't turn it on, take it apart!"

MicDaddy said:
FWIW I use one of these not too shabby for the $  No metering

I believe it was ~$160 USD New

Instek GFG 8215a

Kingston I'm blown away that you've been doing what you do with no generator/scope  :eek:

I see. Looks like the $160 area is giving me some solid choices.

By the way it's not like I've gone blind this far, a good AD/DA converter and a high quality FFT tool like voxengo SPAN is a solid testing companion. View into distortion, noise floor and spectrum in general has been very detailed already.

In fact I'm having some difficulties finding efficient uses for my new found oscilloscope poking freedom. I'm just that used to FFT viewpoint.
 
Echo North said:
Have you searched at EEVblog.com?  They have some good threads on test equipment.  They LOVE the Rigol DSO's (I have one) but they are very critical of the Generator.

I'm on Tequipment.nets mailing list and they recently sent this offer

http://www.tequipment.net/RigolDS1102E.html?utm_content=ruairioflaherty%40eircom.net&utm_source=VerticalResponse&utm_medium=Email&utm_term=&utm_campaign=100MHz%20DSO%20%24399%2C%20Back%20in%20Stockcontent

$399 for the scope that is normally $795.  I thought the cheaper digital scopes were looked down upon compared to the old Tektronix etc?  I need to pick up a scope stateside, my Tek 2250 is in storage in Ireland.

Cheers,
Ruairi
 
That's exactly the one I also got. It's the de facto cheap but good DSO today, even after several years apparently. For someone new to scopes like me, it was completely intuitive from the first time I turned it on (after I found the setting to turn the language from chinese to english).

Well, I actually have the DS1052E scope, but I already flashed the DS1102E firmware. Was a bit tricky, but since the hardware is identical why not. I always despised this kind of artificial handicapping. Screw the man.
 
Kingston,

It is handy to have a counter somewhere in the line as the sig gens rarely have accurate scales.

The square wave is very useful for detecting ringing on transformer inputs and optimizing a damping circuit.  Funny thing, most square waves much over 10kHz turn into sine waves through tube amps.  They are really useful for showing up the flaws in an amp.  Your scope will help find all sorts of nasty little oscillations that you had no idea were there.

best of luck

DaveP
 
I designed and sold a modest test piece a few decades ago, and I would love to revisit that using modern (digital) technology but one of my first rules of product management, is not to sell a hardware product that can be replaced by an I-phone and software.

Digital sound files of sine waves can handle most test needs cheaply and easily.

JR

 
DaveP said:
It is handy to have a counter somewhere in the line as the sig gens rarely have accurate scales.

The square wave is very useful for detecting ringing on transformer inputs and optimizing a damping circuit.  Funny thing, most square waves much over 10kHz turn into sine waves through tube amps.  They are really useful for showing up the flaws in an amp.  Your scope will help find all sorts of nasty little oscillations that you had no idea were there.

I also read that the scales might not be accurate.

The reason I thought the freq counter would be overkill is that it's probably not useful to know whether something is ringing at some exact frequency. If there's something happening at 2khz sharp then I'd expect it'll be there at 1.5k-2.5k as well. That kind of inconsistency is fine with me as long as the scale starts with stable 10hz at least.

Also, my scope shows the frequency.

The most important factor I've understood this far is stability of the generated signal. I saw in some of those EVblog reviews strange issues of analog generator stability, and that those issues would be completely gone in reasonably modern DDS generators.

I also now saw some compelling arguments on arbitrary waveform generators, like simulating noisy or somehow flawed DC filtering, rectified signals etc error proofing. But they're still too expensive.

JohnRoberts said:
Digital sound files of sine waves can handle most test needs cheaply and easily.

Yes, as I outlined in the first post. But I need more amplitude and real square waves.
 
JohnRoberts said:
I designed and sold a modest test piece a few decades ago, and I would love to revisit that using modern (digital) technology but one of my first rules of product management, is not to sell a hardware product that can be replaced by an I-phone and software.

Digital sound files of sine waves can handle most test needs cheaply and easily.

JR

It's 1AM here, I'm a little tired, so forgive me if I haven't thought this through...

Given that the DAC's in most cellphones and soundcards are noise shaping delta sigma based, I would be concerned that unless you had severe low pass filtering, that the high frequency content may cause misreadings.

Once you add significant filtering, then you have to start worrying about phase shift in the audio content?

If your dealing with audio frequencies, maybe this isn't a major concern... i'm not sure.

That's not to say that an iPhone and a 3.5mm Jack to RCA won't solve problems. We had one of our apps guys at a customer in deepest darkest China use one to debug a problem in 45 minutes.

My next test equipment will be real simple. stupidly simple infact - USB Codec, plus a small speaker with a probe. CheepCheerfulnEase.
But then - it won't be for nice measurements... It'll be for dirty, simple audio continuity type tests.
 
Kingston said:
So what's a good signal general to be paired with a decent 50-100mhz DSO?
It depends very much on what you intend to do with it. If you want to do strictly audio, I would say that 10Hz-100k covers 99.9% of your needs. Square-wave is not strictly necessary, but a good plus. This you could do with a good stereo soundcard capable of 192k operation.
Experience told me that the difference between toys and tools is in other aspects, in particular balanced outputs, ultra-low distortion, controlled output impedance, residual noise at low levels (and in particular when muted), how accurate and easy is the level control. But again it depends what you want to do and how you value your time.
Just to give you an example: I had to set up a bench for performance assessment of mixers. Noise performance assessment was impossible with a Neutrik (now NTI) A1 because it had too much noise when muted; an AP System 1 was brought in and performed flawlessly. With the A1, the operator would have had to connect the gen, then disconnect and replace with a 200R dummy, then... and this 24 or 32 times in a row. With the AP, no more manipulation.
I suspect this is not relevant to your case.
In my home lab, I still use a 1960's LEA generator because it has a wonderful decade output attenuator and transformer-isolated balanced output. I could buy an used AP S1, but in the end, including the computer, it would be even more bulky, and I'm concerned about the unavailability of replacement parts if anything goes wrong.
 
Rochey said:
JohnRoberts said:
I designed and sold a modest test piece a few decades ago, and I would love to revisit that using modern (digital) technology but one of my first rules of product management, is not to sell a hardware product that can be replaced by an I-phone and software.

Digital sound files of sine waves can handle most test needs cheaply and easily.

JR

It's 1AM here, I'm a little tired, so forgive me if I haven't thought this through...

Given that the DAC's in most cellphones and soundcards are noise shaping delta sigma based, I would be concerned that unless you had severe low pass filtering, that the high frequency content may cause misreadings.

Once you add significant filtering, then you have to start worrying about phase shift in the audio content?

If your dealing with audio frequencies, maybe this isn't a major concern... i'm not sure.

It is a major concern. One of the main reasons I got an oscilloscope was to get acquainted with the spectrum above 20khz. Square wave output is nearly useless, apart from a very rough view. And this is not some cheapo mobile DAC output solution, but very well spec'ed RME card.

Looks like this, except the ringing caused by the band-limiting is actually much sharper than this example:

bl_square.gif


It's too difficult to study ringing with those artifacts. Very high frequency squares are a complete mess already. You can actually see the sampled points in that. I should post some captures from the scope when I feel less lazy.

Sine waves of course are perfect, and not too noisy either, but 2.55Vpp is too limiting for what I have in mind. And in fact roll off starts somewhere around 20khz even for 192khz output so I can't get even that amplitude for high frequencies.

abbey road d enfer said:
Just to give you an example: I had to set up a bench for performance assessment of mixers. Noise performance assessment was impossible with a Neutrik (now NTI) A1 because it had too much noise when muted

I can see how annoying that would be on a streamlined test bench, but not that relevant for my meager set up. But I'll keep that in mind when choosing the generator.
 
While admittedly my test equipment design experience was 30+ years ago, my product (TS-1) is still around and used by lots of people. The sine wave generator a simple OTA based oscillator covered 30kHz to 15Hz, adequately (<.5% thd+N). For troubleshooting audio systems simple frequency response is 99% of what people hear.... My test set had a dB meter with .1dB resolution and frequency counter (for audio frequencies.)

Yes, I would probably use a cell phone codec, the outputs are 16B+ which could make a higher purity sine wave than my old piece, and the input level meter could be scaled. Only if I decided to make it a distortion analyzer too, would I need a full resolution A/D. But I repeat, this is not a viable product IMO... too easy to do it with a computer sound card.

JR
 
For actual audio (not exploring the envelope of high-NFB loops; nor super-low THD), I was always happy with a Heath IG-18. 10V out, sin square, <1Hz to 111KHz, decade and fine attenuator, solid 600 ohms except when maxed.

There is a "low THD mod" around. IMHO, this is a downgrade unless you really-really need to use the IM-18 as your only THD source. Like the H-P 200AB, there is a tradeoff between distortion and stability, and a jazzed-up lamp-Wien can be awkwardly unstable.

The H-P 200AB, though limited to 40KHz, has BIG output, 25V in 600 and over 10 in 150, and fully floating. Readily available used. "untested" (hah!) is fine, it's all big enough to fix.

> limited to ...2.55VAC peak to peak

Yeah, and the Heath won't dump more than 60mA peak which is not enuff for speaker testing. My answer was to build a current-buffer (car-sound chipamp) inside which would put 2.8V RMS into 3 ohms with dead-nuts-zero source impedance. You could wire four 9V batts and an opamp to boost to 28V p-p in 2K or 600. Modern DSL drivers will loaf pumping big squares at 100KHz. I have also used a yard-sale Sansui hi-fi to boost sig-gen output.

Square wave from sine is fairly easy. Devil is in the details, but for 7V peak you can just use CMOS Schmitt.

> sig gens rarely have accurate scales.

Good ones are not bad.

Old ones have drifted.

The IG-18 is mostly switched and unlikely to drift in our lifetime.

> expect it'll be there at 1.5k-2.5k

Yes, mostly, depending what you do. If you want smooth broadband response then 10%-20% errors are unimportant. If you do steep filtering you want some accuracy. If you get very steep you also want some resolution: the 200AB is infinite (with uncertainty in reading), the IG-18 is switched (certain) plus a trim. OTOH there are handy-oscs with just 9 or 27 frequencies.

IAC, if you have a 'scope, you Lissajous against your 50/60Hz hum to check the lower half of the audio band. Find another freq-stable osc, Liss it to 600Hz, then reference that to check to 6KHz. (After that, who cares?)

> my scope shows the frequency

Or you can cheat with your snazzy toy.

> I would say that 10Hz-100k covers 99.9% of your needs.

Agree 90%. However having the IG-18 (which will go below 1Hz) helped me find sub-sonic troubles which the old EICO 20Hz never got near. Depends what you are doing. Some phono amps tend to have a bump very near 0.55Hz (1/33RPM). Slight imbalance in a Williamson may cause 1Hz bumping which eventually motorboats.
 
PRR said:
I have also used a yard-sale Sansui hi-fi to boost sig-gen output.

I also thought about using this one clean tube preamp of mine that has an oversized line driver for the task. But that means I would first have to intimately learn the artifacts of this amp in question and "substract" its behaviour from a particular test case. Unnecessary complication in an already convoluted set up.

Anyway, thank you for the input everyone. I have all the info I need.
 
A strategy used in the not so good old days to clean up a less than pure sine wave source, for spot distortion measurements, was to run it through one or more stages of filter. A LPF could knock down higher harmonic distortion components several dB. A very narrow band-pass filter (opposite of a notch filter) could clean it up even more. This filtering usually involves giving up level accuracy, so it was not unusual to use a relatively dirty function generator for simple frequency response sweeps and then do what you needed to do, to get a clean enough source for low distortion measurements.

JR

Note: when I did my old TS-1, listening for speaker rubs and otherwise broken speakers was a popular application. My low order THD of a couple tenths of a percent was clean enough for that. While I could hear the difference in a close A/B comparison to a cleaner sine wave source, a broken speaker would be much higher distortion than that (so were many not broken speakers.
 
Back
Top