DIY distortion measurements better than -120dB

[Mr Evil]

Improving on the test equipment described in a previous thread, I have constructed some circuitry dedicated to distortion measurements. It consists of one bandpass filter to purify a sine wave and then one notch filter to magnify the relative amplitude of distortion components.

Tests suggest that it can measure distortion down lower than -120dB. Much of the time that's lost below the noise floor. It's fixed at 1kHz only, but that's usually enough to get a good idea of how well a circuit performs.

You can read details on this page on my website.

Attached are some FFT results from the test equipment itself (red), an OPA134 (blue) and an OPA604 (yellow). Since the 1kHz fundamental is attenuated by 30dB by the notch filter, the magnitude of harmonics etc can be read directly off the scale. Note that the OPA604 is rubbish. You can also see a lot of noise picked up below 1kHz. I'm still working on trying to reduce that, although it doesn't affect this particular type of measurement.

[Sch3mat1c]

Nice. What's the filter made of, passive I presume? Hope you used air core inductors, I hear any kind of core will introduce distortion - for sure, no matter how little, it will show up at -120dB!

Tim

[Mr Evil]

Both are active. I tried a passive RLC bandpass filter at first, but the distortion was significant. The required inductor size was too large for air-core, so I had to go active.

Still, the performance is very good, as you can see. I used an OPA227, which is a new one to me. It's quite possibly the best op-amp I've seen in terms of THD+noise.

[jackinnj]

to get the 60Hz "spurs" down make sure that the test setup is completely shielded (I use a cookie tin), don't use a scope probe -- a piece of RG58 with a BNC on one end and a "directly soldered" connection to the DUT on the other, you can thread the RG58 through an E-Core which will also knock down some of the RFI. You don't want things like a soldering laying about either. These are all suggestions from Linear's Low Noise Measurement application guides and they work.

I use an SSM2019 as a 1000X noise amplifier -- the OPA227 (and its quad family member) would appear to make a dandy replacement in a differential configuration.

[ghg]

@ Mr. Evil

Please, tell us about the FFT sw, you use.


CU

Gary G.

[Mr Evil]

Originally posted by jackinnj
to get the 60Hz "spurs" down...

Thatnks for the advice. I am assaulted by EMI on all sides here.

Originally posted by ghg
...Please, tell us about the FFT sw, you use...

I use Cool Edit. It's a very useful program for all sorts of audio things.

[capslock]

The spurs are 50 and 100 Hz, the AC frequency being 50 Hz in Europe. However, I wonder what the 65 Hz spur might be (it's almost as strong as the 50 Hz).

[moamps]

Originally posted by capslock
..I wonder what the 65 Hz spur might be ...

CRT monitor vertical frequency?

[Mr Evil]

Originally posted by moamps
CRT monitor vertical frequency?

There are no CRTs in the room, and that spike is still there even when none of the ones in the house are on. The only equipment in close proximity is my PC + two LCD monitors (both of which refresh at 60Hz, not 65).

[mwmkravchenko]

Mr. Evil

Can the refresh rate be altered? Gremlins in the mix seldom make any sense. Sometimes the only way to find them is to start a process of elimination.

Mark