Re: Twin T Notch Anomaly
- From: Ian Bell <ruffrecords@xxxxxxxxx>
- Date: Tue, 27 Jan 2009 10:13:01 +0000
Patrick Turner wrote:
Ian Bell wrote:Patrick Turner wrote:Ian Bell wrote:LOLI have built a 2KHz twin T notch filter to help measure low levelNo.
distortion harmonics by removing the fundamental (2KHz). However a
strange thing has happened. I have measured the notch response so I know
it is 12dB down at 4KHz and 8dB down at 6KHz etc so I can correct the
readings. However I was getting different distortion readings with and
without the notch and at first I thought I had the calibration wrong but
I rechecked it and is seems OK.
Now I have discovered that attaching the Twin T itself alters the
distortion products and it makes no difference to the fundamental. I
just connected the Twin T but measured across its input rather than its
output. I have tried this on both the oscillator (low impedance) and the
output of one of my 6SN7 mu followers with similar effects namely the
2nd harmonic is reduced by about 10dB and the third is raised by about
the same amount.
I could understand it the happened with perhaps one particular circuit
but not with two entirely different ones.
Any ideas what it is?
CheersNone of us have a clue what your problems with your twin T filter
Ian
actually are.
Twin T R&C notch filters should be driven with low Z and have very highIt feed my HP Wavemeter which has a 1Meg input impedance. Both the
Z loads on their output such as a grid circuit of a tube.
oscillator and the mu follower have low output impedances.
They do not give a sharp notch, and positive FB is required to sharpenOr you calibrate it as I did.
the notch to avoid the attenuation of the 2H
You don't have a full detailed schematic that you used for us to lookI am sure you know the circuit of a twin T - all you need are the
at.
component values - the caps are 510pF 1% and the resistors 150K 0.1%.
With the RC twin T filter set for maximum attenuation of maybe 80dB atI get about 60dB at the fundamental and 2H is at -1dB and 3H at -7dB.
Fo, ie, 2kHz, you should see -3dB at 4 kHz and 1kHz, and almost no
attenuation at all below 500Hz and above 6 kHz.
I don't follow you.
Don't you mean that for 2kHz, the attenuation is -60dB, so if the input
was 1vrms, you'd have 1mV at the output at 2kHz.
Yes, soryy Patrick, I forgot the minus sign.
BTW, with careful trimming, you should get much more attenuation than
-60dB.
60dB is plenty since the wavemeter can resolve 80dB below that.
But two such filters cascaded with a very slightly staggered Fo, say
1,980Hz and 2,020Hz you get a broader null and overall much more
attenuation at 2kHz.
Now at 1/2 Fo, or 2Fo, the normal T will be -3dB, and maybe -1dB at
1/3Fo and 3Fo. Using such a "natural" T filter is useless because you
cannot measure harmonics of 2H easily without performing calculations,
hence the need to use the T in FB network and have positive FB maybe as
well and then all F further away from about 3/4 Fo and 1.5Fo are
unattenuated, which is exactly what you want to measure noise and
distortions. But you should also incorporate a LF filter with -3dB at
700Hz so that rectifier hum and mains hum artifacts are largely
excluded.
A much better type of notch filter is the Bridged T LC notch filter.
See http://www.crbond.com/papers/ent22.pdf
Thanks for that link Patrick. Someone else mentioned it too. It has the advantage of being easily adjustable too.
Searching more on Google brings up piles more stuff.
I got a pile of useful info at uni archive books and from old Wireless
World articles about 1994-5 when I was puzzled deeply about this basic
facet of amplifier phenomenae. Then I began designing and building my
own THD and IMD measuring gear, all analog of course.
Now you'd use a sound card and a program and PC, and let the PC count
out what's there.
Basically, the LC resonance is formed by the C and L chosen.
The L can have a CT with one C, but its easier to have a single L and
two exactly equal C.
The Adjustable R from the join of the two C to 0V should be in fact two
good pots, one high value for broadly adjusting the null,
and one for fine adjustment, and you can get -90dB Fo attenuation
easily. And the -3dB attenuation points either side of Fo are +/- 10% of
the Fo, so you don't need any active circuit of NFB or PFB. Its the best
way to make a passive nulling circuit for testing an amplifier output
N&D.
Such bridged T notch filters can be made up and used cascaded to work on
the harmonics of the test signal at the input so that the 2H, 3H and
maybe 4H of the test signal is further reduced to make your test signal
have THD < 0.002%.
I use an LC bridged T filter as the initial adjustable filter for 1kHz,
then follow this with an active op-amp based bandstop filter with a very
steep roll off below 1.6kHz and after 10kHz, so all that gets through
are harmonics between 2khz and 10kHz.
To make low level harmonics visible on the CRO, I have an opamp with FB
and 20dB gain, so that 0.1mV becomes 1mV, and just visible on my CROs.
This means I can see 0.01% thd, or thd = -80dB in a 1Vrms signal.
Maybe you should invest in an HP wavemeter. I got mine for 100 USD. Mine is model 3581A - you can still get the manual for it from the Agiolent web site.
Cheers
Ian
.
- References:
- Twin T Notch Anomaly
- From: Ian Bell
- Re: Twin T Notch Anomaly
- From: Patrick Turner
- Re: Twin T Notch Anomaly
- From: Ian Bell
- Re: Twin T Notch Anomaly
- From: Patrick Turner
- Twin T Notch Anomaly
- Prev by Date: Re: 6SN7 Mu Follower Distortion Paper
- Next by Date: Re: Favorite Tube Pre-amp
- Previous by thread: Re: Twin T Notch Anomaly
- Next by thread: Favorite Tube Pre-amp
- Index(es):
Relevant Pages
|
