Build -- Active Twin-T notch filter for distortion analysis

Status
Not open for further replies.
Jump to Latest
#103
If that is the case, wouldn't it be worth while to replace the caps in the twin-T with teflon caps?
I know they are considered better in most regards. I just don't know about THD.

Oops, above I meant extended foil polystyrene, not polyester.
 
Last edited:
#104
myhrrhleine said:
If that is the case, wouldn't it be worth while to replace the caps in the twin-T with teflon caps?
I know they are considered better in most regards. I just don't know about THD.

Oops, above I meant extended foil polystyrene, not polyester.
Click to expand...

Polystyrene is also excellent and for this app, just as good as Teflon. For deep nulls and less drift - either is better. -Thx-RNMarsh
 
#105
I'd be careful with attributing fixed distortion rankings to generic capacitor types, particularly polymer film types. Distortion is not necessarily a very consistent parameter, i.e. we must expect considerable spread from specimen to specimen. I have seen Polystyrene parts with higher distortion than good Polypropylene types, and the very few Teflon capacitors I looked at weren't immaculate either.

Ceramic C0G seems to be much more consistent than polymer film types (and is also more stable, comes with lower tempco and is available with 1% precision at reasonable cost). I've observed very little spread amongst different specimen, and even between various types/brands (of course with the same voltage rating/size). I suspect this is because forming a ceramic capacitor involves much less ill-defined mechanical processing steps than winding a polymer film part. As far as I can measure this at the moment, C0G rated at 100 V or higher has consistently less than -140 dB harmonic distortion with +20 dBu across the capacitor. At lower AC levels, this quickly drops to vanishingly low levels.

The most effective and at the same time cheapest solution to improve a standard passive notch filter is to add a carefully designed input attenuator. This is to make sure the filter operates at modest voltage only (say 1-2 Vrms maximum) to keep distortion in the passives low. The input attenuator is probably best constructed from multiple series resistors (again to keep their voltage drop and thus distortion contribution low). Take e.g. ten 1k resistors in series--the last tap forms the output of a 10x divider.

Samuel
 
#107
Stable TC of caps are important consideration for high Q, deep notch/null circuits. Especially for non-auto null notched circuits. Polystyrene is excellent in this regard. However, for best performance for such a circuit, I think the Polyphenylene Sulphide (PPS) would be excellent. Drift away from a deep, sharp. High-Q notch will be minimised.

Thx-RNMarsh
 
Last edited:
#109
> At lower AC levels, this quickly drops to vanishingly low levels.

So then, would a 100 V device have lower distortion than a 50 V device at the same signal level?
Click to expand...

That's not what I'm saying here. I'm saying that any given capacitor has lower distortion at 0 dBu than it has at +20 dBu. Relative to the fundamental, you can expect the 2nd to be 20 dB lower, and the 3rd 40 dB (4th 60 dB etc.). 2nd falls proportional with level, and 3rd with the square. That's why it can be so extremely effective to add an input attenuator to a passive notch filter.

However, it is also my experience (which is backed up by Bruce Hofer's findings, see his "Building Analog in the 2010s" slides) that C0Gs rated for higher voltages tend to have lower distortion for a given AC level. This is perhaps a result of the higher thickness of the plates, and may not hold for other capacitor technologies.

BTW, I think I've remembered the distortion figures for C0Gs I quoted in post #105 wrong. Will need to check my measurements again, suspect it's a couple of dBs higher than -140 dB at +20 dBu. But surely mighty low at 0 dBu!

Samuel
 
Last edited:
#110
Hmmm. Just as I got the QA400 running and getting some data -- I am about 5dB different in one vs another test for harmonics... Then the QA400 FFt went nuts on me. stopped working and only works sporatically now. Waiting to see what can be done about it -- fix/replace or what?

But before it bit the dust, it appeared that the ShibaSoku 725D analyser had a fundemental notched null that was at -150dB (!!).

Thx-RNMarsh
 
Last edited:
#111
RNMarsh said:
Hmmm. Just as I got the QA400 running and getting some data -- I am about 5dB different in one vs another test for harmonics... Then the QA400 FFt went nuts on me. stopped working and only works sporatically now. Waiting to see what can be done about it -- fix/replace or what?

But before it bit the dust, it appeared that the ShibaSoku 725D analyser had a fundemental notched null that was at -150dB (!!).

Thx-RNMarsh
Click to expand...

What OS are you using?

I solved a problem by with the QA400 changing USB ports and reinstalling the drivers. I couldn't get 192ksps until I did this.
There is a timing issue with XP. What exactly is QA400 doing?
 
Last edited:
#115
RNMarsh said:
BTW -- whats the basic difference between a twin-t and Cordell's Distortion Magnifier?

-RNM
Click to expand...

Bob's DM is a null process. The source signal and it's distortions are subtracted from the output of the DUT leaving just the distortion products of the DUT. In the case of the DM it just removes enough of the fundamental to change the percent of distortion from the output of the DUT. The DM output is amplified to a level which an analyzer like the 339A can process. It's done this in a controlled amount. A bit of math has to be done with the reading on the analyzer to get your final result. It's the same math you would do taking a measurement from the monitor output of the 339A. Just scaling.

The difference is the DM changes the relative fundamental to distortion, It increases the THD with out adding too much to it by a controlled amount.

The TT filters the fundamental out without changing the relative FS.

The DM will not work for evaluating oscillators because it would remove the information we are trying to measure.
 
#116
I still have not returned to the usb issue... too many side-tracks. But I will get it running again.... somehow. probably with seller/company help. Because I dont want to spedn a long time on it fumbling around on my own.

-Thx RNMarsh
 
#120
One issue I want to look into further is the accuracy of harmonic levels measured by the various methods and test equipment. I have already noticed harmonics of 2H and 3H of the ADC at low distortion levels (below -100dB) are not accurately displayed. Today, I added to my arsenal, an Audio-Precision 2722. This will help as a cross check with the fabulous ShibaSoku AD725D... it is speced to maintain a specified accuracy/tolerance to -130dB. However, if the tolerance was loosened just a little, it is pretty much a straight line to the noise floor at -150dB.

These two instruments represent the SOTA and as such I can test the various ways being discussed here for oscillator distortion and ADC based FFT test data and check and report on their accuracy limits, tolerances and conditions especially at levels below -100dBr.

Thx-RNMarsh
 
Last edited:
Status
Not open for further replies.
Top Bottom