> What would be ideal/available for these?
There is no universal answer. Anyway, we often don't need "ideal" performance (whatever that would be).
> ~50uH is good for single inductors. 50mH is OK for a common mode choke.
Note the huge difference in values. To make sense of this, you have to remember that a common-mode choke (which isn't really used as a transformer, though the construction and symbol are the same) has two values. And the second one is rarely stated.
For unbalanced current, or when driven the same on both sides, it has a high impedance, the H given in the catalog.
For current the same in both sides (and assuming the two sides are connected out of phase), the inductance cancels. In theory to zero, in practice to some small value.
If you just stick 50mH in front of a mike input, it will suck. 50mH*20KHz*6.28= 6K ohms at the top of the audio band. Since many mike inputs are less than 5K, often 2K, the top octave or two will droop bad. (Also the treble noise will rise in BJT inputs.)
50uH is 6 ohms at the top of the audio band, which does nothing to the sound. It is only 190Ω at 600KHz, the bottom of the AM broadcast band, and by itself (with a resistive amp input) will do little about radio signals well up into the short-wave band.
But it is never by itself, and amp inputs are rarely resistive up to MHz. If we throw in a capacitor, we can swamp some of the amp input impedance with a known stable impedance, and get a 2-pole filter that will cut hard above a certain frequency. But AT that frequency, the L and C resonate, which can make things worse. Can even suck enough power through the feedback loop to make the whole amp oscillate at MHz.
50uH and 470pFd will resonate at about 1MHz. Will it peak or slump? Well, 50uH is about 300Ω at 1MHz, and so is 470uFd, but the 1-side input of the sample amp is 1KΩ. The L-C network is very lightly loaded. It will be flat up to maybe 500KHz, rise to about a 1:3 or 10dB peak at 1MHz, then fall fast, ultmately asymtotic to -12dB around 1MHz.
If you work next to KRAP-AM on 990KHz, it will BOOST the station's signal. Yes, it will severely knock-down most police, taxi, CB, TV, and other sources (though in the upper TV band, waves are small enough to flow over the surface of these physically large MHz-band parts).
You really need to grok RF circuits well enough to rough-analyze tanks at a glance to get anywhere with such filters. And you should have RF-band test gear (and know how to use it) to verify that the actual parts and layout do what you think they do.
In most DIY, ultimate RF rejection is not needed. Simple tricks may suffice.
One thing is: don't be ashamed to load your input with carbon-film resistors. R9 2K7 above serves no audio purpose, actually degrades audio a fraction-dB, but gives RF something to overcome before it gets to the input devices. It also gives some loading on the L-C network. We'd like more loading to kill that 10dB 1MHz peak, but that would degrade audio performance.
Don't be too afraid to blindly throw caps on the XLR pins. Any good mike can drive 100 feet of cable, but rarely does. 100 feet of cable is 3,000pFd, so 470pFd or 1,000pFd is "nothing". The caps will drain-off considerable RF.
Don't be over-obsessed with low series resistance. Yes, 100Ω in series with the amp will degrade noise figure. When was the last time you really pushed your mike and room noise floor? If you use the big hot condensers, series resistance is almost a non-issue: their internal amp's self-noise is like the random noise of a 5KΩ resistor, so a 470Ω + 1,000pFd R-C low-pass in each input will not raise working noise floor.
We need to consider common-mode and differential-mode separately, both the filter and the amp input susceptability. Many LTP BPJ input can eat volts of common-mode even up at 1MHz, but are sensitive to small amounts of diff-mode MHz signal.
And while I hate the weight and cost, the "best" mike transformer (which comes closest to passing radio waves) is still a better radio-rejector than most things you can put in front of a transformerless input.
FETs tend to accept radio signals more gracefully. They also have less gain and higher noise, for some of the same reasons they shrug-off radio. The hottest FETs do approach BJT gain and noise. Even at the same gain, the curvature of a FET is more benign and won't convert radio to audio byproducts as well (bad) as a BJT.
April 01, 2009, 02:00:24 AM
