Let's design a portable audio test set

[NewYorkDave]

RE: transformers in the test set--definitely not a bad idea, and not just because of galvanic isolation. You can eliminate a few amplifiers (e.g., for input and output balancing) and a few mA saved here and there really add up when you're running off batteries. The Edcor PC-mount xfmrs are decent, cheap and lightweight. The WSM chassis-mount series are bigger and better but only slightly more expensive.

Here's an output amplifier idea I sketched this afternoon:

 

[NewYorkDave]

Quick breadboard test this afternoon... One half of a TL072 wrapped inside a two-transistor current booster can do +14dBM into 600, drawing 7-8mA from two 9V batteries. This was quickie and i didn't have time to attempt to "optimize" it.

 

[Rob Flinn]

[quote author="walter"] I suppose an O-scope display is beyond the scope of this project, but it would be cool to have a little sine wave on an LCD. Is that even possible? [/quote]

I have a small battery powered Vellemann LCD scope for field testing. It doesn't replace my 20Meg dual beam scope, but is invaluable in the field. I did some resoration work on a desk in a French Studio back in August where I would have been in trouble without it. This particular model has 4 meter functions displayed simultaneously which are programable. Very useful for 80 uk pounds

 

[sodderboy]

Is that Kev-erator based on a XR2206?

http://www.exar.com/product.php?ProdNumber=XR2206

That is a great chip and can be put in a very small package. But with a Fostex TT15 and a Fluke 8060A I have been busy enough to not have enough time to work on an all-in-one.

The TT15 is only a 4558, a PNP, and RC circuitry. I can dig-up a schemo if it would help with a test rig. It must be tweakable to generate pink as well.
Mike

 

[JohnRoberts]

[quote author="mediatechnology"]
all a bit messy inside as it was one of those quick projects that started and ended all in less than a day

I didn't post a front panel cos I was trying to save on forum noise
but here 'tis there is duplicate and other back panel connections as well
no picture for that ( don't know why ... didn't think to take one )

Nice job.. That looks like the same (Pactec?) box we put the Loftech TS-1 in.

Since this thread has turned into a kitchen sink wish list, a feature that I put in the old TS-1 might be considered here. Using a known resistive source impedance in series with the output of the sinewave generator and looking at the drop wrt frequency, you can perform crude impedance sweeps of speakers and such.

In the TS-1 I used a switching jack on the input to the meter with a feed from the sinewave output normalled to that switch contact. Due to limited range of your analog VU meter you might want to offer a range switch, which increases complexity quite a bit.

JR

 

[JohnRoberts]

John: How many of those TS-1s got sold? Must have been a bazillion.

Quite a few. AFAIK they are still being sold by Goldline. I don't get a penny so that's not exactly something I celebrate. :sad: I also know it's weak points and have the next generation improved version somewhere among the cobwebs of unfinished projects cluttering up my meat computer. Of course in light of modern technology, this too begs to be done with a small micro.

I used a 2206 in a console master section line up oscillator back in early '80s but it was too high distortion for the TS-1 so I rolled my own with a 13600 (OTA) in a state variable configuration (<.2%). The distortion was pretty much limited by settling time which with 30 KHz to less than 20 Hz on a single knob was a design concern.

In my current product I make sine waves by feeding a PWM output with sine wave data from a lookup table and then smoothing that with a simple filter. But sine wave purity for my app is not critical as long as distortion products are down low enough to not excite acoustic resonances by themselves. I can't even cite a proper THD figure but I expect it's better than the 2206. My old THD analyzer released it's smoke a while ago and is not worth repairing for this one measurement.

JR

 

[Samuel Groner]

A quick addition to this thread--I've been working on a similar test box for a friend and here's the current schematic: Anrauscher_r1.pdf

It'll get a built-in speaker based on a 600 ohm headphone, switchable white noise or 1 kHz sine and is powered by four 1.5 V cells. We breadboarded the important bits allready, but later on we decided to change from +/-12 V supplies to +/-9 V, so the values R301, R302, R304 and R305 might need to be changed. In addition to this, the oscillator (which I can give credit to W. Jung for the basic design) can't be set to low enough levels, so R202 will get a different value as well. I'll update the schematic as soon as we know more in case there is interest.

Samuel

 

[JohnRoberts]

[quote author="RogerFoote"]There are still CEM3340s around, quite a few actually.
I have a super low distortion sine wave add on driven by the triangle out if that would help the sine wave end of things...

Actually, aren't some of the function gen chips based on a triangle wave out? If so, you could do the sine wave converter there without the 3340.[/quote]

The triangle to sine wave convertors using a diode break amp or such were in my experience still audibly distorted when listing to loudspeakers for voice coil rubs and such. I guess it depends on how low is low distortion.

I found any one of several "BPF+ pos feedback" sine wave oscillators adequately clean but settling time is an issue when covering a wide frequency range and going down to sub 20 Hz frequencies. The distortion was typically dominated by the speed of the AGC loop (fast settling= higher distortion).

I have done some work with adaptive agc circuits that would use a fast and slow loop to quickly settle and then fade into lower distortion mode.

One approach that I never fully explored was using a trig identity to reduce ripple in AGC loop. SinX^2+CosX^2=1 so perhaps taking the BP and either HP or LP of a SVF gives you +/- 90' Sin and Cos representations. A couple of b-e junctions in series to square that term, and in theory you have a ripple free representation of sine wave level, should be fast and reduce distortion common to typical level detectors in AGC loop. I never pursued this as it was too complex for low cost test gear.

These days it's mostly of academic interest, or not. If I were to revisit my old TS-1 I'd surely do it all with a cheap microprocessor. Perhaps a little analog glue around it but mostly digital.

JR

 

[bcarso]

[quote author="JohnRoberts"]The triangle to sine wave convertors using a diode break amp or such were in my experience still audibly distorted when listing to loudspeakers for voice coil rubs and such. I guess it depends on how low is low distortion.

I found any one of several "BPF+ pos feedback" sine wave oscillators adequately clean but settling time is an issue when covering a wide frequency range and going down to sub 20 Hz frequencies. The distortion was typically dominated by the speed of the AGC loop (fast settling= higher distortion).

I have done some work with adaptive agc circuits that would use a fast and slow loop to quickly settle and then fade into lower distortion mode.

One approach that I never fully explored was using a trig identity to reduce ripple in AGC loop. SinX^2+CosX^2=1 so perhaps taking the BP and either HP or LP of a SVF gives you +/- 90' Sin and Cos representations. A couple of b-e junctions in series to square that term, and in theory you have a ripple free representation of sine wave level, should be fast and reduce distortion common to typical level detectors in AGC loop. I never pursued this as it was too complex for low cost test gear.


JR[/quote]

See the latest audioXpress (Dennis Colin, A Wide-Range Audio Sweep Oscillator, 2/07, pp. 26-30, 49) for an implementation of the trig thing. As I remarked elsewhere I suspect his distortion would plummet below the stated 0.1% with just a bit of filtering after the squared summation. Settling without filtering is in any case virtually instantaneous.

It's also possible, for improved distortion and noise, that better and lower-noise OTA parts need to be used for the oscillator integrators and squaring---the current mirrors in the CA3280 and their ilk are quite noisy for one thing.

 

[JohnRoberts]

[quote author="bcarso"]

See the latest audioXpress (Dennis Colin, A Wide-Range Audio Sweep Oscillator, 2/07, pp. 26-30, 49) for an implementation of the trig thing. As I remarked elsewhere I suspect his distortion would plummet below the stated 0.1% with just a bit of filtering after the squared summation. Settling without filtering is in any case virtually instantaneous.

It's also possible, for improved distortion and noise, that better and lower-noise OTA parts need to be used for the oscillator integrators and squaring---the current mirrors in the CA3280 and their ilk are quite noisy for one thing.[/quote]

cool.. I'd lost track of Ed Dell and Audio Amateur mag, I'm guessing that is the current incarnation?

I used even cheaper OTA in old TS-1 (13600) even used the cheezy built in darlington buffers, but noise wasn't an issue and ballpark .1% THD was dominated by AGC loop.

JR

PS: I may need to break down and try to find a copy of the schematic... It seems like it could be pretty fast (real time in theory) but any log errors could cause distortion so some filtering would be useful. One really nice side effect of squaring the terms is that higher rate of change portion of waveforms occur at lowest levels and will have reduced impact on combined result.

 

[bcarso]

[quote author="JohnRoberts"]
cool.. I'd lost track of Ed Dell and Audio Amateur mag, I'm guessing that is the current incarnation?

JR
[/quote]

Yes---aXp now incorporates Glass Audio, Speaker Builder, and Audio Amateur/Audio Electronics. Ed is still at the helm fighting the good fight.

Roger: No, I'm not talking about a sinewave shaper.

 

[PRR]

> I suppose an O-scope display is beyond the scope of this project, but it would be cool to have a little sine wave on an LCD. Is that even possible?

Got $200? 2MHz Panel Mount Oscilloscope, 6.5"W x 3.5"H x 1.4"D, eats 9VDC 300mA, 60-page Manual (well, 15 pages in each language...)

It's not the spiffiest 'scope in town, but it is 10 times faster than the 'scope I started on, and I learned a LOT of audio that way. And it never goes out of focus!!!!

Inside NYD's stated goal, general test-set, it's more than plenty. Leave the 50mHz on the heavy-research bench. An objection inside NYD's "portable" goal is that it could gulp batteries; might make the buzzer and squawker battery-power, and use a wall-wart if you find you need to see what you're doing. (But then you may as well just buy the pre-made LCD 'scope, which is chock-full of battery...)

> Using a known resistive source impedance in series with the output of the sinewave generator and looking at the drop wrt frequency, you can perform crude impedance sweeps of speakers and such.

My Heath spent half its life that way, just with the resistor outside the box.

The gold-plated kitchen-sink feature would be a Constant Current output, so that impedance magnitude could be read directly on a Voltmeter. Push-pull pair of current mirrors and two resistors to the voltage-mode output. Just have to remember that when it is unloaded, the output will be brutal big square-waves (unless your microprocessor can watch that too.)

> I'd lost track of Ed Dell and Audio Amateur mag

No wonder. This is the second time (in 30 years) that Ed let my subscription expire with only one notice. Try getting off any other mag that easy.

My impression is that Ed has stepped-back from the driver's seat. Anybody know how he is?

> aren't some of the function gen chips based on a triangle wave out?

That's about the only kind there is. (ARP had tracking ramp-and-triangle function generators, but that's a frill not a difference.)

> sine wave add on driven by the triangle out

As John says: I've not heard one that fools the ear near as good as it fools the THD meter.

Look at the shapes. The tri is "sharpest" where the sine is "flattest". You must move from nearly no change around zero to a very extreme change RIGHT AT the peak. The diodes and diff-amps make the triwave more mellow, but must be fine adjusted to reduce the high buzz, and it never goes-away completely. Post-filtering helps, but becomes as complicated as a decent sine-wave oscillator.

 

[JohnRoberts]

[quote author="RogerFoote"]This one fools this ear... Anyone else heard this shaper? Seems like everyone assumes this shaper just sounds like the diode type.[/quote]

I don't debate what other people can or can't hear, while I do hold opinions.

I found personally when dialing in sine wave oscillators, fairly small differences were audible when compared side by side. The nature of the distortion can make a difference in sundry applications. I found it interesting that I was still hearing fairly modest levels of distortion played through nominally much higher distortion loudspeakers. I accepted higher distortion than I wanted (mostly 2nd and 3rd), to realize acceptable settling time at LF.

A diode break function generator approach has no settling time issues so if acceptably low distorion for your application, enjoy. Many people have.

JR

 

[bcarso]

[quote author="RogerFoote"]This one fools this ear... Anyone else heard this shaper? Seems like everyone assumes this shaper just sounds like the diode type.[/quote]

The tri-driven diff amp produces an approximation to a hyperbolic sine*, which one can suppose from the name is a pretty decent approximation to a circular function. The devil is getting the gain to go just so as the triangle peaks.

I suspect it could work even better if driven differentially.

I've always liked the shaper concept as it is an example of a nonlinear transfer function that actually removes harmonic energy from an input signal. Of course it is a very specific input signal. It would be interesting if an example could be found of such a TF that was amplitude-insensitive.


*see for example http://en.wikipedia.org/wiki/Hyperbolic_function

 

[bcarso]

I played with the circuit a bit. It is interesting that minimum THD has a lot of 5th, as opposed to somewhat higher total numbers with less of it but more third. There is probably a way to combine the outputs of a couple of such circuits adjusted differently and get a still better result. But doing a few changes I got a sim down to about 0.3%, which is not too shabby.

 

[bcarso]

Might be time for a sine wave shaper/sine wave generator thread though.