Strong low-pass filter

Do you have a question? Post it now! No Registration Necessary

Translate This Thread From English to

Threaded View
I want to make a low-pass filter (-0.1 dB/20kHz, -150dB/22050Hz) for  
digital audio 44.1...192kHz (from S/PDIF).
What use for this ?
DSP processor OR FPGA OR something other?
What is the easiest way to do this?

Thanks in advance
Pawel

Re: Strong low-pass filter
On 2/14/2013 9:14 AM, pw wrote:
Quoted text here. Click to load it

   Could you tell the difference between -130db and -150db?

                                 Mikek


Re: Strong low-pass filter
Quoted text here. Click to load it

If you want -130, it might be the -130.

Re: Strong low-pass filter
wrote:

Quoted text here. Click to load it

Doesn't such an application require an analog filter?
        
                                        ...Jim Thompson
--  
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
We've slightly trimmed the long signature. Click to see the full one.
Re: Strong low-pass filter
On Thu, 14 Feb 2013 09:59:50 -0700, Jim Thompson wrote:

Quoted text here. Click to load it

Hmm.  I'm trying to decide if I'm suffering from unthinkitis today or not.

For the cost of digital hardware these days, upsampling internally,  
filtering, then converting to analog and having a relaxed requirement on  
the reconstruction filter would probably make it much easier to build  
quality equipment at a lower price.

(Or, if you're recording instead of playing back, turning everything  
around, but still sampling fast at the ADC).

((Which is part of the allure of sigma-delta converters))

But obviously if you're sampling at 44.1kHz then any reconstruction or  
anti-aliasing is going to happen in analog-land.

--  
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
We've slightly trimmed the long signature. Click to see the full one.
Re: Strong low-pass filter
On 02/14/2013 11:59 AM, Jim Thompson wrote:
Quoted text here. Click to load it

That's one tough filter spec.  It'll need a lot of taps and a lot of  
bits, too.  What on earth would actually need a spec like that?

Cheers

Phil Hobbs

--  
Dr Philip C D Hobbs
Principal Consultant
We've slightly trimmed the long signature. Click to see the full one.
Re: Strong low-pass filter
On Thu, 14 Feb 2013 12:26:58 -0500, Phil Hobbs

Quoted text here. Click to load it

An audiophool.  But you could probably approach -60dB with an analog
filter with stop-band zeroes.
        
                                        ...Jim Thompson
--  
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
We've slightly trimmed the long signature. Click to see the full one.
Re: Strong low-pass filter
On Thu, 14 Feb 2013 10:49:28 -0700, Jim Thompson wrote:

Quoted text here. Click to load it

Meeting his specifications with some honkin' big FPGA would be a snap;  
with a bit of shoe-horning one could probably make it work with one of  
the new fixed-point DSP chips.

The problem comes about when you want to reproduce this sound in the real  
world.  -150dB down from 1V p-p is about 32nV.  Where are you going to  
find a DAC, or even an analog audio chain, that's not going to bury all  
of your filtering efforts in noise and distortion that's much bigger than  
that?

--  
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
We've slightly trimmed the long signature. Click to see the full one.
Re: Strong low-pass filter
On 2/14/2013 10:07 AM, Tim Wescott wrote:
Quoted text here. Click to load it

If you did that, what would the phase plot look like in the passband?
Amplitude is only part of the problem with most applications???

Quoted text here. Click to load it


Re: Strong low-pass filter
On Thu, 14 Feb 2013 10:32:17 -0800, mike wrote:

Quoted text here. Click to load it

That depends on how you actually implement the filter.  To some extent  
it's the wrong question: for audio folks have pretty much decided that  
its better to use a minimum phase filter (which would cause huge phase  
wingwazzles close to the filter cutoff) rather than a linear phase filter  
(which is the "easy" way to do an FIR filter) because the linear phase  
filter causes "pre ringing".

But no matter what, you'd have latitude for playing around with the phase  
response as well as the amplitude response.

--  
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
We've slightly trimmed the long signature. Click to see the full one.
Re: Strong low-pass filter

Quoted text here. Click to load it

150 dB is equivalent to the headroom between the human threshold of
hearing, and a jet engine 100 feet away.

Some years ago, one poster on one of the audio newsgroups proposed
that digital audio (including the delivery-to-the-home media) should
be designed to use 32-bit numbers, "just to make sure".

I did a bit of back-of-the-envelope calculation, and concluded that if
you set the LSB to deliver energy some distance below the random noise
generated by the thermal collision of air molecules with the eardrum,
then a full-scale signal for a brief period would be distinctly
hazardous to the listener's health.  It would be louder than a stun
grenade and approaching the point at which you stop getting sound and
start getting shock waves (pressure peaks > 1 atmosphere, valleys all
the way down to vacuum).

The commercial utility of this much dynamic range in consumer-audio
applications hasn't really been established, IMO :-)


--  
Friends of Jade Warrior home page:  http://www.radagast.org/jade-warrior
We've slightly trimmed the long signature. Click to see the full one.
Re: Strong low-pass filter
On 2/14/2013 2:34 PM, Dave Platt wrote:
Quoted text here. Click to load it

The resolution of digital representations isn't all about the analog  
dynamic range.  Signal processing does calculations which often require  
rounding of results.  So noise is introduced which is proportional to  
the resolution of the numbers used, but not necessarily equal to that  
resolution.  If you perform N successive rounding operations which are  
uncorrelated, you can expect to see N/2 lsbs of noise to be added.  
Depending on the processing being done, this can add up to the loss of  
several equivalent bits.  So the idea of going beyond 24 bits is not  
obviously absurd.

However, the proof of the pudding is in the eating.  I remember years  
ago a friend showed me with a simple A/B comparison how poor cassette  
tapes were, even with Dolby noise reduction and all the other bells and  
whistles.  I've yet to hear the difference between a 16 bit CD recording  
and any other representation.

At some point you are much more limited by the rest of the system I  
think.  Can we make mics and speakers with under -120 dB of distortion  
and noise?  What about ears?  Sometimes I can't get past the accents on  
Downton Abbey.  I need the direct digital signal of closed captioning.

--  

Rick

Re: Strong low-pass filter


Quoted text here. Click to load it

Oh, I agree... signal-processing math ought to be done with oodles and
oodles of bit-width, all the way through the math pathway.  Then, do a
proper dithered-and-shaped truncation to your final delivery
resolution... that way you've only got one new set of significant
quantization noise to dither away.

It's the idea of trying to actually deliver content to the consumer at
32 bits, and pretend that the lower bits are meaningful, which I found
amusing.

Quoted text here. Click to load it

Well, CD can certainly be done badly... by making just the sorts of
mistakes you're alluding to!

Quoted text here. Click to load it

Agreed!

--  
Friends of Jade Warrior home page:  http://www.radagast.org/jade-warrior
We've slightly trimmed the long signature. Click to see the full one.
Re: Strong low-pass filter
On 2/15/2013 8:41 PM, Dave Platt wrote:
Quoted text here. Click to load it

I had to check to make sure this post was in sci.electronics.design!  
Someone discussed a topic and was courteous, friendly and agreeable!  It  
must have been a glitch in my newsreader and this message is from some  
other group!!!

--  

Rick

Re: Strong low-pass filter

Quoted text here. Click to load it

Early (classical music) digital recordings were made with ADCs
operating directly at 44.1 or 44.056 kHz, which were not necessary
monotonous and sometimes without proper dithering. They needed a very
complex analog LCR anti-alias filter, with horrible phase response.
any digital processing was done with 16 bits, truncating any
fractional bits at each stage.

Of course, these early recordings could sound really bad.

Quoted text here. Click to load it

Nominally 24 bits and up to 192 kHz sampling rate ADCs actually have
120 dB (20 bit) dynamic range _measured_ with 0-20 kHz bandwidth.

Of course it would be nice to plug any microphone (with unknown
sensitivity) into any input channel without having to use any noisy
and unreliable potentiometers (at least after longer usage) and do all
the processing in digital domain.

IMHO, 16 _active_ bits is more than sufficient, i.e. linear 16 bits
for low dynamic range popular music and for classical recordings, some
MSB compression (in 1 ms - 1 s range) would be sufficient.


Re: Strong low-pass filter


Dave Platt schrieb:

Quoted text here. Click to load it

Hello,

but 150 dB will damage the hering in a very short time. More than 134 dB  
will cause pain and 120 dB are enough for damages in a short time.

Bye


Re: Strong low-pass filter

Uwe Hercksen wrote:
Quoted text here. Click to load it



   Is that another red hering joke?

Re: Strong low-pass filter
On Mon, 25 Feb 2013 13:57:13 -0500, "Michael A. Terrell"

Quoted text here. Click to load it


maybe his herring has been impaired ?


Re: Strong low-pass filter
Quoted text here. Click to load it

Hmm I think it's 20dB :^)

George H.

Re: Strong low-pass filter
On 14/02/2013 17:51, George Herold wrote:
Quoted text here. Click to load it

That last 20dB might well make the difference here between an algorithm  
that could be implemented in 32bit arithmetic realistically and  
something that can't. Sounds decidedly like audiophool stuff to me.

--  
Regards,
Martin Brown

Site Timeline