I thought I'd post a belated follow-up to Gianni's question from last
summer (see below) regarding the Fostex FR-2 and aliasing. The short
answer is that the unit appears to perform correctly when presented
with an ultrasonic input. There is some visible aliasing as the signal
exceeds the nyquist in each of the tests (including 44.1, 96 and 192
kHz). The extent of the aliasing was very similar at all frequencies,
nothing like the odd behavior Gianni observed in the FA-101. If anyone
needs more detail about the test feel free to ask. FYI, I did these
tests as part of a detailed review of the FR-2 which will be published
soon in Electronic Musician magazine. While I won't re-cap my findings
here, I ended up buying the review unit!
I'm still contemplating the broader issues of ultra-sonic capable
recorders (and might even be wrong about the following--feel free to
correct me if you think I'm in error).
Most A/D designs have a fixed frequency analog input filter before the
A/D. If the A/D must operate at frequencies as high as 192, then this
fixed filter must be set high. The consequence of this (at least
theoretically) is that recordings made at lower sampling frequencies
(for example 44.1 on a device capable of 192) will have more aliasing
than would be the case if the input analog filter was set lower, to a
value appropriate for a 44.1 recorder, rather than a 192 model.
In our evaluation of the Fostex's aliasing, we did note that the Alesis
24-track hard disk recorder, which only records at up to 96 kHz, showed
slightly better aliasing performance than the FR-2 at 96 kHz--perhaps
because the FR-2 needed a higher fixed frequency filter to allow it to
optionally make 192kHz recordings.
The other issue about ultrasonic capture is the noise concerns
mentioned by Bernie a month or two ago when he reported on our
ultrasonic bat-call recording tests. I'm curious about a new Sanken mic
that I think has bandwidth up to 100kHz, but I have no idea what kind
of noise performance to expect. I think that the added acoustical gain
of a parabolic dish might be a wise addition to an ultrasonic rig if
our experience with the MKH-800 was an indication of the
design-tradeoffs mic manufacturers are taking with extended bandwidth
mics.
Cheers,
Rudy
> Message: 9
> Date: Tue, 10 Aug 2004 13:01:24 +0200
> From: Gianni Pavan <>
> Subject: freq response and aliasing of Edirol FA-101 at 192 kHz
>
> Hi all,
> I'm testing the new Edirol FA-101 to verify its performances at 96
> and 192
> kHz.
> Performances at 96 kHz are as expected, with a flat response up to 42
> kHz
> and sharp filtering close to Nyquist.
>
> On the contrary, performances at 192 kHz are really unsatisfactory. The
> unit behaves completely different, with a roll-off that begins at 30
> kHz
> and scarcely attenuates beyond the Nyquist frequency. Frequencies above
> Nyquist are poorly attenuated and thus, if they are present in the
> analog
> signal, they can fold back down to 40 kHz.
>
> Given a sinusoidal input signal of 8Vpp on the LINE INPUT (much less
> than
> the overload level, of course), by varying frequency I get the
> following
> response:
>
> 30 kHz, 0 dB (arbitrary)
> 36 kHz, -0.14
> 39 kHz, -0.22
> 42 kHz, -0.325
> 45 kHz, -0.46
> 48 kHz, -0.64
> 51 kHz, -0.885
> 54 kHz, -1.22
> 57 kHz, -1.66
> 60 kHz, -2.23
> 63 kHz, -2.955
> 66 kHz, -3.85
> 69 kHz, -4.95
> 72 kHz, -6.28
> 75 kHz, -7,87
> 78 kHz, -9.77
> 81 kHz, -12.065
> 84 kHz, -14.855
> 87 kHz, -18.355
> 90 kHz, -23.03
> 93 kHz, -30.375
> 94 kHz, -34.45
> 95 kHz, -41.0
> 96 kHz - difficult to measure
>
> and now, by increasing the input frequency beyond Nyquist, I get the
> following values for "aliased frequencies":
>
> (real input freq -> aliased freq, level of aliased freq vs input level)
> 97 -> 95 kHz, -42.0
> 98 -> 94 kHz, -36.5
> 99 -> 93 kHz, -33.485
> 102 -> 90 kHz, -29.25
> 105 -> 87 kHz, -27.7
> 108 -> 84 kHz, -27.35
> 111 -> 81 kHz, -27.75
> 114 -> 78 kHz, -28.7
> 117 -> 75 kHz, -30.07
> 120 -> 72 kHz, -31.82
> 123 -> 69 kHz, -33.92
> 126 -> 66 kHz, -36.34
> 129 -> 63 kHz, -39.08
> 132 -> 60 kHz, -42.18
> 135 -> 57 kHz, -45.58
> 138 -> 54 kHz, -49.3
> 141 -> 51 kHz, -53.5
> 144 -> 48 kHz, -58.3
> 147 -> 45 kHz, -63.8
>
> By the way, I tested the LynxTwo PCI boards that exhibits very sharp
> filters and flat bandwidth up to 90 kHz!
>
> Now I wonder about the performances of the relatively cheap hard-disc
> recorders now available, in particular the Fostex FR2: is there
> somebody
> able to test their performance with ultrasonic signals like I did with
> the
> FA101 ?
>
> Gianni
________________________________________________________________________
________________________________________________________________________
|