falkTX
/
Carla
mirror of https://github.com/falkTX/Carla


			
				
					
						
						
							
							<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/strict.dtd">
<html><head>
<meta http-equiv="Content-Type" content="text/html"; charset="utf-8">
<meta http-equiv="Content-Language" content="en">
<link rel="stylesheet" type="text/css" href="styles.css">
<title>BLS1 - Quick guide</title>
</head>

<body>

<table border="0" cellpadding="0" cellspacing="10" width="950">
<tbody>

<tr height="50">
<td width="150">
<img src="redzita.png" border=0>
</td>
<td width="850" align="center">
<h2>BLS1 - Quick guide</h2>
</td>
</tr>

<tr>
<td></td>
<td class="content">
<p>BLS1 is a digital realisation of the 'Blumlein Shuffler', invented by Alan
Blumlen in the early 1930s and analysed in detail by Michael Gerzon in a 
paper presented at the 1993 AES Convention in San Francisco.</p> 
<p>The Blumlein Shuffler is used to convert binaural stereo signals into a
form suitable for reproduction using a convential stereo speaker pair.
Binaural signals are provided by e.g. a dummy head, or a pair of closely
spaced omnis with a baffle (e.g. a Jecklin disk) in between them. Such signals
have no significant level differences for off-centre signals at low and low-mid
frequencies, only a phase difference which depends on the frequency, the distance
between the mics and the source direction. The shuffler turns these phase
differences into amplitude differences by applying a specific filter to the
(L-R) signal. This can be done only as long as the phase difference is not
ambiguous, i.e. less than half a cycle, so the shuffler is normally set up
to not affect higher frequencies. For these, sufficient channel separation must
be provided by the input signals themselves, typically by using some sort of
baffle between the mics.</p>
<p> The ideal difference channel filter would be of the form 1 + a/s, i.e. unity
plus an integrator, or the inverse of a first order highpass. But since this has
unbounded gain as frequency goes down it can't be used. An analog implementation
would use an LF shelf filter instead. The problem with this is that limiting the
gain at LF also means that the phase response returns to zero, while it should
ideally stay at 90 degrees. This could be compensated for by using all-pass filters
in both the sum and difference channels, but this complicates the circuit (in
particular if the shuffler parameters are variable) and of course also affects
the overall phase response.</p>
<p>Zita-bls1 uses a FIR filter (1024 taps at 48 kHz) that has the amplitude
response of the shelf filter but the phase response of the ideal filter, and
the sum channel is just a matching delay. The finite length of the FIR means
there will be some roll-off at very low frequencies, in this case -3dB at 40 Hz,
which is probably not a bad thing. It could be avoided by using a longer FIR.</p>
<p>Apart from the basic shuffler algorithm some extras are provided: an input
gain balance control, an 18 dB/oct highpass filter, and a variable frequency
LF shelf filter.</p>
</td>
</tr>

<tr>
<td></td>
<td align="center">
<img src="zita-bls1.png" border=0>
</td>
</tr>

<tr>
<td></td>
<td class="content">
<p> The rotary knobs can be used in two ways:</p>
<p> * Click on the knob with the <b>left</b> mouse button, keep it pressed
and move either left..right or up..down.</p>
<p> * Using the <b>mouse wheel</b>. This provides 'sensible' steps, e.g.
1 or 1/2 dB for gains and 1/3 or 1/6 octave for frequencies. Press <b>Shift
</b>for smaller steps.</p>
<p>From Left to right we have:
</p>
<p><b>* Input balance.</b>
Not all mic preamps provide perfect matching of channel gains, and the mics
themselves may have slightly different sensitivities as well Since the
shuffler amplifies the difference between the L and R inputs, even small
gain errors could have a significant impact. This control is provided to
allow exact matching of the two input signal gains.</p>
<p><b>* Highpass filter.</b>
As the shuffler boosts LF signals it's advisable to remove any low
frequency rumble. Ideally this should be done by the mic preamps, but
not all of them all provide such a filter. Filter slope is 18 dB/oct.
When set to 10 Hz it has no significant effect in the audio band.</p>
<p><b>* Shuffler gain.</b>
This the maximum gain the difference channel filter will have as 
frequency goes down. Depending on the combined gain and frequency
settings It may not reach this value since there is also a fixed
rolloff (-3 dB at 40 Hz).</p>
<p><b>* Shuffler frequency.</b>
This is the frequency at which the difference channel filter will
have 3dB gain.</p> 
<p><b>* LF shelf filter frequency and gain.</b>
The LF shelf filter is provided to correct the tonal balance in
the low frequency range which may be affected by the action of the
shuffler. It is a second order design and has a somewhat steeper
slope than most shelf filters.
</p>
</td>
</tr>

</tbody>
</table>

</body>