Low-distortion Audio-range Oscillator

[Samuel Groner]

I don't see the need for certainty just a max limit.

I was referring to the (un)certainity of the end result, not the (un)certainity of the notch filter distortion contribution. Yes, an upper limit for this is surely sufficient.

Samuel

[coluke]

Those who are looking for a simple high-performance sinusoidal source should give a try to the venerable HP239A/339A Bridged-T generator - despite the simplicity of the design its performances are quite amazing: my (slightly modified) implementation only employs a couple of OPA2134s and a TL082, and is capable of ~0.00004% THD @ 1kHz, 20Vpp out. I think that better opamps and an ALC loop optimized for fixed frequency generation (instead of 10 Hz - 100 kHz continuous coverage) could further improve THD and THD+N figures.

Ciao,

L.

[DDB]

Those who are looking for a simple high-performance sinusoidal source should give a try to the venerable HP239A/339A Bridged-T generator - despite the simplicity of the design its performances are quite amazing: my (slightly modified) implementation only employs a couple of OPA2134s and a TL082, and is capable of ~0.00004% THD @ 1kHz, 20Vpp out. I think that better opamps and an ALC loop optimized for fixed frequency generation (instead of 10 Hz - 100 kHz continuous coverage) could further improve THD and THD+N figures.

Ciao,

L.

I have also built a HP339A version similar to Modifications: Heathkit IG-18 Audio Generator using a pc board.

Can you post your schematic so we can see your mods?

[dirkwright]

Has anyone tried this technique for measuring distortion on differential input amplifiers? I don't know if it could be used for other kinds of amps. Substitute our low distortion oscillator for the source and a super nice ADC/computer for the analyzer input.

[coluke]

Originally Posted by DDB

Can you post your schematic so we can see your mods?

Nothing really special - I did add a full-wave peak detector with diode-drop compensation, reworked a bit the ALC loop time constant, and employed a not-so-hard to find FET as gain controlling element. The clone seems to work just a little bit better than the original one - if I remember well the operating AC Vds of the J111 is slightly less than 80 mVpp @ 1kHz, and the level of the higher harmonic (II) is about -135 dBV, more than 100 dB below the fundamental, which seems to account for the very low output THD I see with my notches.

Ciao,
L.

[scott wurcer]

Originally Posted by dirkwright

Has anyone tried this technique for measuring distortion on differential input amplifiers? I don't know if it could be used for other kinds of amps. Substitute our low distortion oscillator for the source and a super nice ADC/computer for the analyzer input.

Pease wrote a related app note too, but this technique is pretty old. You can also obviate somewhat the need for a super oscillator.

[richiem]

Steve Lafferty has modded a Heath IG-18 to the HP 339A circuit with the HP cap ratio of 100:1. I've modded an IG-18 to the HP 339A circuit, too, then remodded using the stock Heath caps and range switch. The results are very good. See my page IG-18 #4, the IG-339B. My page has a link to Steve's page as well.

[DDB]

Quote:
I can post some pix later his weekend if you want to see it.


Yes please!
Terry

@Terry

Here is a couple of pix

[zigzagflux]

So I received Victors’ 1 kHz oscillator yesterday, and completed a first round of testing. I also built a twin-t notch filter in a separate box, to test out the real distortion of this and other sources. It is a passive design on the front end followed by a fixed gain amplifier of 100 (+40dB).

I have performed no real calibrations or loopbacks; this is just to see if things are functional or not. I have good confidence in my M-Audio Profire 610 (other than Win7 frustration) so cursory results are very encouraging.

As far as the oscillator, I built my own 35V non-regulated outboard power supply, which was actually putting out closer to 37V at 35mA. This caused the oscillator’s local power supply to run a little too warm for my comfort; you could really see drift of circuit performance as the entire pcb warmed up. The TL431’s were running around 55-60C absolute. So a 100 ohm resistor in series was sufficient to bring voltage down to 34.7V, where the shunt power supply was much more happy and cooler.

Frequency of oscillation was 1001.2 Hz, well within specifications. Ideally let the assembly run for 15-20 minutes before doing any serious measurements. Gain control appears very decent- I had the Fluke 87 bouncing +/-1mV, but that could just be the meter itself. It doesn’t appear to have a significant effect on distortion results, anyway.

Now that I knew the oscillator’s frequency, I could custom tailor my notch filter specifically to that frequency, getting maximum performance. Happy to report, that worked out really well- I was able to get a -82dB notch, which followed by the +40dB fixed gain amplifier resulted in total fundamental attenuation of -42dB. Now I had a low fundamental, which would not (probably) add any distortion products on the soundcard’s FFT results. Any distortion I see would now be from the oscillator and notch filter. Although the notch tends to drift a little with time, all I was interested in was getting it low enough to not be an issue with the soundcard; I believe I have accomplished that.

So by the numbers, my low-Q notch filter performs as follows:
2nd: +30.9 dB
3rd: +34.9
4th: +36.7
5th: +37.7
6th: +38.3
7th: +38.7
8th: +39.0
9th: +39.2

I set the oscillator for around 2V rms output (-9dBFS) and began measurements. Keep in mind, this is a mess of wiring at present; I have attached a picture showing the spaghetti setup, so you understand that a little noise here and there is expected. Pre-notch into the soundcard and post-notch into the soundcard attached. My observation:

2nd harmonic (which has been boosted by 30.9 dB by the notch filter) is actually at -137 dB below the fundamental.
3rd unmeasureable.
4th at -147 below fundamental.
5th at -inf.
6th at -145.
7th at -inf.
8th at -141.
THD therefore 0.000018%

The pre-notch capture indicates to me that in the presence of a strong fundamental, the soundcard does in fact distort, resulting in erroneous measurements of harmonics. So a notch filter is absolutely required in order to test the distortion of Victor’s oscillator, and for the price, that’s a darn nice oscillator.

Your thoughts?

[zigzagflux]

For kicks, I measured the output of my DCX2496 (1.65V rms) , which is modded by using a Jensen JT-11-DMPC directly connected to the D/A pins. Significantly higher distortion than Victor’s (though still not too bad):
2 = -99 dB
3 = -107
4 = -123
5 = -109
6 = 0
7 = -121
THD = 0.0013%