It is true that the number of A/D bits is not always the critical
number. High-end audio gear sometimes uses 1-bit D/A coverters, for
example, clocked at a very high rate. That avoids the nonlinearity due
to uneven bit weighting that can be a problem with multi-bit D/As. IIUC
(if I understand correctly).
In any case, having enough bits in and out does not prevent problems due
to bad algorithms (codecs, AGC, etc.) that may allow overflows,
clipping, or loss of significant digits in internal calculations.
73 Martin AA6E
Rob Atkinson, K5UJ wrote:
> Hi,
>
> I'll try to explain this. think of a dsb circuit as a data processing
> circuit.
>
> If you have two circuits, one using 16 bit processing and the other 24 bit,
> but they both have the same clock speed, then the 24 bit circuit processes
> more data for a given amount of time. That means the digital simulation of
> an analog waveform for a fixed time period has greater precision beause it
> is sampling at the same rate, but processing it more rapidly, 1/3 more for
> each clock tick. Processing bigger chunks of data at a time means you can
> do things like employ more complex processing algorithms that can result in
> a more faithful analog result.
>
> Assuming a CD player is a 16 bit system, it could be extremely high
> fidelity, if it's clock speed is extremely high because then, it would be
> sampling and processing more data than a 24 bit circuit on a much slower
> speed for a given amount of time.
>
> So besides the word length, we also need to know the clock speed for the dsp
> circuits, which I for one, do not know off hand (I think the 870 dsp runs at
> 10 MHz but don't hold me to that).
>
> a few folks have mentioned that the Orion is (depending on who you talk to)
> a 24 or 32 bit dsp processor. That's great and I humbly apologize for
> implying anything else, but its clock speed is also important and in any
> event, there seem to be other components that limit its flat frequency
> response.
>
> 73,
>
> rob / k5uj
>
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|