Home |
Search |
Today's Posts |
#1
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Digitally, high quality audio is 24 bits at 96kHz.
The question is, does the bit resolution apply to a woofer eg at frequencies under 100Hz. Does the 24 bits of precision make any difference? At what point would signal degradation be hearable compared to mid and high frequencies? 12 bits? 10 bits?... Dirk |
#2
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
"Dirk Bruere at NeoPax" wrote ...
Digitally, high quality audio is 24 bits at 96kHz. That is debated in both directions. The question is, does the bit resolution apply to a woofer eg at frequencies under 100Hz. Why would it not? Does the 24 bits of precision make any difference? If you had hung about this neighborhood for any length of time, you would have understood that this is a perennial topic of debate. I commonly use 24-bit depth for recording live concerts to take advantage of the greater dynamic range, to avoid getting into trouble at either end of the scale. At what point would signal degradation be hearable compared to mid and high frequencies? 12 bits? 10 bits?... Why do you think it is frequency-dependent? |
#3
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
On 2 Jan, 19:31, "Richard Crowley" wrote:
"Dirk Bruere at NeoPax" wrote ... Digitally, high quality audio is 24 bits at 96kHz. That is debated in both directions. The question is, does the bit resolution apply to a woofer eg at frequencies under 100Hz. Why would it not? Does the 24 bits of precision make any difference? If you had hung about this neighborhood for any length of time, you would have understood that this is a perennial topic of debate. I commonly use 24-bit depth for recording live concerts to take advantage of the greater dynamic range, to avoid getting into trouble at either end of the scale. At what point would signal degradation be hearable compared to mid and high frequencies? 12 bits? 10 bits?... Why do you think it is frequency-dependent? I don't know - that's what I'm asking. Esp since the mass and stroke of the cone are vastly different in a sub compared to a tweeter. I suppose I could argue either way. For example, because the stroke is longer more bits are required to get the same precision of absolute movement. OTOH the extra mass might 'blur' the precision - but then again the extra power usually being fed in might cancel that effect. Then there is the sensitivity of the ear to very low frequencies compared to mid from which one might argue that extra bits of precision don't matter as much low down. etc etc What's the answer? Dirk |
#4
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Dirk Bruere at NeoPax wrote: Digitally, high quality audio is 24 bits at 96kHz. Why not 192 kHz ? The question is, does the bit resolution apply to a woofer eg at frequencies under 100Hz. What does the frequency or a woofer have to do with it ? Does the 24 bits of precision make any difference? 24 bit audio converters don't actually offer true 24 bit precision, so the number is largely a marketing thing. At what point would signal degradation be hearable compared to mid and high frequencies? 12 bits? 10 bits?... Why do you think higher frequencies are 'degraded' ? Your questions are normally quite same but this one has puzzled me. Graham |
#5
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Dirk Bruere at NeoPax wrote: "Richard Crowley" wrote: Why do you think it is frequency-dependent? I don't know - that's what I'm asking. Esp since the mass and stroke of the cone are vastly different in a sub compared to a tweeter. What's that got to do with ADCs or DACs ? You sound very confused. Graham |
#6
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Dirk Bruere at NeoPax writes:
Digitally, high quality audio is 24 bits at 96kHz. The question is, does the bit resolution apply to a woofer eg at frequencies under 100Hz. Does the 24 bits of precision make any difference? At what point would signal degradation be hearable compared to mid and high frequencies? 12 bits? 10 bits?... Dirk Hi Dirk, You ask a very good question, and the answer requires a surprising amount of knowledge of digital signals. In a nutshell, and simplistically speaking, the bit resolution is inversely proportional to the amount of wideband noise the digital signal carries. This noise is spread evenly from 0 Hz to Fs / 2 Hz, where Fs is the sample rate. From this you can see that your question depends on the sample rate. If you're reproducing a bass note at, say, 40 Hz but at a sample rate of 44.1 kHz, then you're going to hear the noise all the way up to 22.05 kHz (assuming you haven't listened to too much loud music when you were a kid). So even though the bass note is low, the noise from the digital signal will continue to be high and in fact will be more audible with a bass-only signal content due to the effect of masking (or lack thereof). Now if you KNOW you're only ever going to be producing frequencies under 100 Hz, you can utilize decimation to gain several bits of resolution due to the fact that you're grossly oversampling the signal. The number of extra bits gained N_b is given by the equation N_b = log_4(Fsi / Fso), where "log_4" denotes the logarithm base 4, Fsi is the input sample rate, and Fso is the output sample rate. So for your hypothetical situation, Fsi = 96000, Fso = 200, and N_b = 4.45 bits. So you would gain an extra 4.5 bits of resolution by decimating to 200 Hz. That means that you could use an input resolution of 20 bits and the output would still be better than 24 bits. You can read some of the basics of oversampling in the comp.dsp presentation I made a few years back on delta sigma data conversion: http://www.digitalsignallabs.com/presentation.pdf --Randy -- % Randy Yates % "Maybe one day I'll feel her cold embrace, %% Fuquay-Varina, NC % and kiss her interface, %%% 919-577-9882 % til then, I'll leave her alone." %%%% % 'Yours Truly, 2095', *Time*, ELO http://www.digitalsignallabs.com |
#7
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
On 2008-01-02, Dirk Bruere at NeoPax wrote:
Digitally, high quality audio is 24 bits at 96kHz. The question is, does the bit resolution apply to a woofer eg at frequencies under 100Hz. Does the 24 bits of precision make any difference? ... If you are talking of signal degradation due to quantization noise (which appears, from the question, to be the case) then signal to quantization noise ratio is *theoretically* independent of signal frequency (from just below 1/2 the sampling rate right down to "DC"). For N bits, sampling rate Fs and considering wideband noise (over frequency 0 to Fs/2): SNR = (6.02N + 1.76) dB Which does not depend on the signal frequency. See, for example, http://www.analog.com/en/content/0,2...F88014,00.html ... At what point would signal degradation be hearable compared to mid and high frequencies? 12 bits? 10 bits?... I am not sure what is the threshold of human audibility of quantization noise in the presence of a signal. Nor am I sure whether that threshold varies with frequency. However S/N is, as said above, nominally independent of signal frequency. -- John Phillips |
#8
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
"Dirk Bruere at NeoPax" wrote in
message Digitally, high quality audio is 24 bits at 96kHz. No, it only takes about 13 bits at 36 KHz to provide high quality audio. Everything past that is overkill, headroom, etc. The question is, does the bit resolution apply to a woofer eg at frequencies under 100Hz. Of course. Does the 24 bits of precision make any difference? No, not for just playing music. At what point would signal degradation be hearable compared to mid and high frequencies? 12 bits? 10 bits?... A full-utilized 13 bit digital path running at 36 KHz is sonically transparent for all audio frequencies. Going up to 14 bits cuts some slack for headroom. |
#9
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
On Thu, 3 Jan 2008 07:59:30 -0500, "Arny Krueger"
wrote: "Dirk Bruere at NeoPax" wrote in message Digitally, high quality audio is 24 bits at 96kHz. No, it only takes about 13 bits at 36 KHz to provide high quality audio. Everything past that is overkill, headroom, etc. NICAM sounds pretty clean, and that just uses the lowest possible 10 bits from 14. The question is, does the bit resolution apply to a woofer eg at frequencies under 100Hz. Of course. It is easier for a woofer, because the audio bandwidth involved is lower. By constraining the bandwidth to 100Hz rather than 20kHz, we see an improvement in SNR of 23dB. That means we can lose 4 bits of information with no added noise penalty. It just gets easier and easier for a woofer. Does the 24 bits of precision make any difference? No, not for just playing music. At what point would signal degradation be hearable compared to mid and high frequencies? 12 bits? 10 bits?... A full-utilized 13 bit digital path running at 36 KHz is sonically transparent for all audio frequencies. Going up to 14 bits cuts some slack for headroom. But for playback you don't need headroom, since the signal is already constrained by the number of available bits. d -- Pearce Consulting http://www.pearce.uk.com |
#10
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Eeyore wrote:
Dirk Bruere at NeoPax wrote: "Richard Crowley" wrote: Why do you think it is frequency-dependent? I don't know - that's what I'm asking. Esp since the mass and stroke of the cone are vastly different in a sub compared to a tweeter. What's that got to do with ADCs or DACs ? You sound very confused. It's got to do with the number of bits describing the absolute position of the cone, and its velocity, placement etc at a given moment. -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#11
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
On 2008-01-06, Dirk Bruere at NeoPax wrote:
Eeyore wrote: Dirk Bruere at NeoPax wrote: "Richard Crowley" wrote: Why do you think it is frequency-dependent? I don't know - that's what I'm asking. Esp since the mass and stroke of the cone are vastly different in a sub compared to a tweeter. What's that got to do with ADCs or DACs ? You sound very confused. It's got to do with the number of bits describing the absolute position of the cone, and its velocity, placement etc at a given moment. I think of it like this: 1. The signal to quantization ratio of a normal digitally-derived input voltage signal to the driver is constant (see the reference I provided earlier). 2. Inside its designed frequency of operation the audio driver has a nominally flat voltage versus audio SPL response which is nominally linear. These are basic design goals for a normal audio driver (cf a driver spec of typically about 89 dB SPL for 2.83 V rms input). From these two points we can deduce that the audio (SPL) signal to audio quantization noise ratio is nominally the same as the input's electrical (voltage) signal to quantization noise ratio. That is, it's not dependent on whether the signal is in the bass, mid or treble frequency range. We don't need to go into "the number of bits describing the absolute position of the cone, and its velocity, placement etc at a given moment". In reality the driver's own non-linearity will produce distortion at the 0.1% level or greater at usual listening SPLs (actually more in the bass region - 0.3% is not untypical and subwoofers produce even more). Thus the driver's own nonlinearity will exceed and so mask the audibility of an electrical signal to quantization noise ratio in the 50 - 60 dB range (8.5 - 10 bits). This question has the hallmarks of a Radium/Green Xenon question with just a little bit more subtlety, I do hope I'm wrong. -- John Phillips |
#12
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
John Phillips wrote:
On 2008-01-06, Dirk Bruere at NeoPax wrote: Eeyore wrote: Dirk Bruere at NeoPax wrote: "Richard Crowley" wrote: Why do you think it is frequency-dependent? I don't know - that's what I'm asking. Esp since the mass and stroke of the cone are vastly different in a sub compared to a tweeter. What's that got to do with ADCs or DACs ? You sound very confused. It's got to do with the number of bits describing the absolute position of the cone, and its velocity, placement etc at a given moment. I think of it like this: 1. The signal to quantization ratio of a normal digitally-derived input voltage signal to the driver is constant (see the reference I provided earlier). 2. Inside its designed frequency of operation the audio driver has a nominally flat voltage versus audio SPL response which is nominally linear. These are basic design goals for a normal audio driver (cf a driver spec of typically about 89 dB SPL for 2.83 V rms input). From these two points we can deduce that the audio (SPL) signal to audio quantization noise ratio is nominally the same as the input's electrical (voltage) signal to quantization noise ratio. That is, it's not dependent on whether the signal is in the bass, mid or treble frequency range. We don't need to go into "the number of bits describing the absolute position of the cone, and its velocity, placement etc at a given moment". In reality the driver's own non-linearity will produce distortion at the 0.1% level or greater at usual listening SPLs (actually more in the bass region - 0.3% is not untypical and subwoofers produce even more). Thus the driver's own nonlinearity will exceed and so mask the audibility of an electrical signal to quantization noise ratio in the 50 - 60 dB range (8.5 - 10 bits). This question has the hallmarks of a Radium/Green Xenon question with just a little bit more subtlety, I do hope I'm wrong. Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#13
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
On 2008-01-06, Dirk Bruere at NeoPax wrote:
John Phillips wrote: On 2008-01-06, Dirk Bruere at NeoPax wrote: It's got to do with the number of bits describing the absolute position of the cone, and its velocity, placement etc at a given moment. I think of it like this: ... From these two points we can deduce that the audio (SPL) signal to audio quantization noise ratio is nominally the same as the input's electrical (voltage) signal to quantization noise ratio. That is, it's not dependent on whether the signal is in the bass, mid or treble frequency range. We don't need to go into "the number of bits describing the absolute position of the cone, and its velocity, placement etc at a given moment". Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. To which the above is applicable, just as stated. -- John Phillips |
#14
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Dirk Bruere at NeoPax writes:
[...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). -- % Randy Yates % "So now it's getting late, %% Fuquay-Varina, NC % and those who hesitate %%% 919-577-9882 % got no one..." %%%% % 'Waterfall', *Face The Music*, ELO http://www.digitalsignallabs.com |
#15
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Randy Yates wrote:
Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#16
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Dirk Bruere at NeoPax wrote:
Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. I though that twice the Nyquist frequency would be adequate? eg for a sub designed to operate under 100Hz I'll sample/decimate to 400Hz. Any benefits going to a higher frequency? -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#17
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
On Sun, 06 Jan 2008 16:20:19 +0000, Dirk Bruere at NeoPax
wrote: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. I though that twice the Nyquist frequency would be adequate? eg for a sub designed to operate under 100Hz I'll sample/decimate to 400Hz. Any benefits going to a higher frequency? Just about every audio DAC/ADC system in use these days goes for sampling at perhaps 128 times the Nyquist rate. The hard bit is always the analogue anti-alias filter and at this rate it is easy. I can't think of any audio sampled systems that work at just twice Nyquist. d -- Pearce Consulting http://www.pearce.uk.com -- Posted via a free Usenet account from http://www.teranews.com |
#18
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Don Pearce wrote:
On Sun, 06 Jan 2008 16:20:19 +0000, Dirk Bruere at NeoPax wrote: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. I though that twice the Nyquist frequency would be adequate? eg for a sub designed to operate under 100Hz I'll sample/decimate to 400Hz. Any benefits going to a higher frequency? Just about every audio DAC/ADC system in use these days goes for sampling at perhaps 128 times the Nyquist rate. The hard bit is always the analogue anti-alias filter and at this rate it is easy. I can't think of any audio sampled systems that work at just twice Nyquist. Well, at least with subs I've got plenty of room for tweaking frequencies and seeing what happens. I want to keep the number of samples the DSP works with fairly low so I can do other stuff eg room eq, graphics eq etc. without pushing the DSP anywhere near its limits. -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#20
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
On Sun, 06 Jan 2008 16:54:14 +0000, Dirk Bruere at NeoPax
wrote: Don Pearce wrote: On Sun, 06 Jan 2008 16:20:19 +0000, Dirk Bruere at NeoPax wrote: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. I though that twice the Nyquist frequency would be adequate? eg for a sub designed to operate under 100Hz I'll sample/decimate to 400Hz. Any benefits going to a higher frequency? Just about every audio DAC/ADC system in use these days goes for sampling at perhaps 128 times the Nyquist rate. The hard bit is always the analogue anti-alias filter and at this rate it is easy. I can't think of any audio sampled systems that work at just twice Nyquist. Well, at least with subs I've got plenty of room for tweaking frequencies and seeing what happens. I want to keep the number of samples the DSP works with fairly low so I can do other stuff eg room eq, graphics eq etc. without pushing the DSP anywhere near its limits. What most DSP systems - CD and the like - do is this. Having captured the data at high speed you apply a steep digital lowpass filter at 100Hz. You can then decimate the data and do all your work at 200Hz to minimize processor load. You will need to upsample back to high speed again to drive the DAC. d -- Pearce Consulting http://www.pearce.uk.com -- Posted via a free Usenet account from http://www.teranews.com |
#21
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Dirk Bruere at NeoPax writes:
Don Pearce wrote: On Sun, 06 Jan 2008 16:20:19 +0000, Dirk Bruere at NeoPax wrote: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. I though that twice the Nyquist frequency would be adequate? eg for a sub designed to operate under 100Hz I'll sample/decimate to 400Hz. Any benefits going to a higher frequency? Just about every audio DAC/ADC system in use these days goes for sampling at perhaps 128 times the Nyquist rate. The hard bit is always the analogue anti-alias filter and at this rate it is easy. I can't think of any audio sampled systems that work at just twice Nyquist. Well, at least with subs I've got plenty of room for tweaking frequencies and seeing what happens. I want to keep the number of samples the DSP works with fairly low so I can do other stuff eg room eq, graphics eq etc. without pushing the DSP anywhere near its limits. Any modern DSP (TI TMS, ADI SHARC, etc.) should be well-within its limits when working with audio-frequency sample rates. The lowly and dated TI TMS32C54x, e.g., has a minimum of 100 MIPs, so at, e.g., Fs = 48 kHz you have 2000 cycles per sample to burn. What DSP are you using, and what high-level system architecture are you employing? -- % Randy Yates % "How's life on earth? %% Fuquay-Varina, NC % ... What is it worth?" %%% 919-577-9882 % 'Mission (A World Record)', %%%% % *A New World Record*, ELO http://www.digitalsignallabs.com |
#22
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Dirk Bruere at NeoPax writes:
Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. Then you aren't concerned with "subwoofer construction and performance." It appears you are attempting to design a digital crossover system that will be used to drive an existing subwoofer. In other words, you appear to be designing a wideband speaker system through the selection and use of various speaker driver units and digital crossovers. It would help if you told us your plan at the high level. -- % Randy Yates % "Maybe one day I'll feel her cold embrace, %% Fuquay-Varina, NC % and kiss her interface, %%% 919-577-9882 % til then, I'll leave her alone." %%%% % 'Yours Truly, 2095', *Time*, ELO http://www.digitalsignallabs.com |
#23
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Don Pearce wrote:
On Sun, 06 Jan 2008 16:54:14 +0000, Dirk Bruere at NeoPax wrote: Don Pearce wrote: On Sun, 06 Jan 2008 16:20:19 +0000, Dirk Bruere at NeoPax wrote: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. I though that twice the Nyquist frequency would be adequate? eg for a sub designed to operate under 100Hz I'll sample/decimate to 400Hz. Any benefits going to a higher frequency? Just about every audio DAC/ADC system in use these days goes for sampling at perhaps 128 times the Nyquist rate. The hard bit is always the analogue anti-alias filter and at this rate it is easy. I can't think of any audio sampled systems that work at just twice Nyquist. Well, at least with subs I've got plenty of room for tweaking frequencies and seeing what happens. I want to keep the number of samples the DSP works with fairly low so I can do other stuff eg room eq, graphics eq etc. without pushing the DSP anywhere near its limits. What most DSP systems - CD and the like - do is this. Having captured the data at high speed you apply a steep digital lowpass filter at 100Hz. You can then decimate the data and do all your work at 200Hz to minimize processor load. You will need to upsample back to high speed again to drive the DAC. As opposed to setting the sampling rate ie the number of samples per second taken from the ADC, at a few hundred Hz? -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#24
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Randy Yates wrote:
Dirk Bruere at NeoPax writes: Don Pearce wrote: On Sun, 06 Jan 2008 16:20:19 +0000, Dirk Bruere at NeoPax wrote: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. I though that twice the Nyquist frequency would be adequate? eg for a sub designed to operate under 100Hz I'll sample/decimate to 400Hz. Any benefits going to a higher frequency? Just about every audio DAC/ADC system in use these days goes for sampling at perhaps 128 times the Nyquist rate. The hard bit is always the analogue anti-alias filter and at this rate it is easy. I can't think of any audio sampled systems that work at just twice Nyquist. Well, at least with subs I've got plenty of room for tweaking frequencies and seeing what happens. I want to keep the number of samples the DSP works with fairly low so I can do other stuff eg room eq, graphics eq etc. without pushing the DSP anywhere near its limits. Any modern DSP (TI TMS, ADI SHARC, etc.) should be well-within its limits when working with audio-frequency sample rates. The lowly and dated TI TMS32C54x, e.g., has a minimum of 100 MIPs, so at, e.g., Fs = 48 kHz you have 2000 cycles per sample to burn. What DSP are you using, and what high-level system architecture are you employing? Not entirely decided yet, hence the questions. The other problem with (say) Fs = 48kHz is low frequency instability and rounding errors. I would like to implement a sixth octave graphic EQ below 100Hz -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#25
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Randy Yates wrote:
Dirk Bruere at NeoPax writes: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. Then you aren't concerned with "subwoofer construction and performance." I am to the extent that I want to know what features influence reproduction the most. It appears you are attempting to design a digital crossover system that will be used to drive an existing subwoofer. In other words, you appear to be designing a wideband speaker system through the selection and use of various speaker driver units and digital crossovers. It would help if you told us your plan at the high level. Just done that in another post. -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#26
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
On Sun, 06 Jan 2008 17:36:25 +0000, Dirk Bruere at NeoPax
wrote: Don Pearce wrote: On Sun, 06 Jan 2008 16:54:14 +0000, Dirk Bruere at NeoPax wrote: Don Pearce wrote: On Sun, 06 Jan 2008 16:20:19 +0000, Dirk Bruere at NeoPax wrote: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. I though that twice the Nyquist frequency would be adequate? eg for a sub designed to operate under 100Hz I'll sample/decimate to 400Hz. Any benefits going to a higher frequency? Just about every audio DAC/ADC system in use these days goes for sampling at perhaps 128 times the Nyquist rate. The hard bit is always the analogue anti-alias filter and at this rate it is easy. I can't think of any audio sampled systems that work at just twice Nyquist. Well, at least with subs I've got plenty of room for tweaking frequencies and seeing what happens. I want to keep the number of samples the DSP works with fairly low so I can do other stuff eg room eq, graphics eq etc. without pushing the DSP anywhere near its limits. What most DSP systems - CD and the like - do is this. Having captured the data at high speed you apply a steep digital lowpass filter at 100Hz. You can then decimate the data and do all your work at 200Hz to minimize processor load. You will need to upsample back to high speed again to drive the DAC. As opposed to setting the sampling rate ie the number of samples per second taken from the ADC, at a few hundred Hz? All that stuff has to be done. If the ADC will do it for you without your having to get involved, then great. I'm not familiar enough with the current SOTA to know whether you can take oversampling and decimation for granted with these things. d -- Pearce Consulting http://www.pearce.uk.com -- Posted via a free Usenet account from http://www.teranews.com |
#27
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Dirk Bruere at NeoPax writes:
[...] Not entirely decided yet, hence the questions. You missed my point. No matter what you use (within reason), you're going to have plenty of MIPS. The other problem with (say) Fs = 48kHz is low frequency instability and rounding errors. I would like to implement a sixth octave graphic EQ below 100Hz Oh, well why didn't you say so in the first place? You have opposing requirements (welcome to engineering). One is that you do your processing with the least number of bits possible in order to save CPU operations (e.g., to be able to avoid doing double-precision arithmetic, assuming you're implementing in fixed-point). The other is to lower the sample rate so that your filter designs are more easily achieved. These two desirables contradict one another, because once you go to a lower sample rate, you must use a HIGHER bit width in order to maintain the equivalent SNR that you would have at the higher sample rate. In this particular application at this juncture in history, however, reducing CPU cycles should be low on your list of concerns, as I've already pointed out, since most DSPs are likely to have PLENTY of horsepower. As for a good architecture for implementing the sixth-octave filter, that's a perfect question for comp.dsp - suggest you repost there. -- % Randy Yates % "Midnight, on the water... %% Fuquay-Varina, NC % I saw... the ocean's daughter." %%% 919-577-9882 % 'Can't Get It Out Of My Head' %%%% % *El Dorado*, Electric Light Orchestra http://www.digitalsignallabs.com |
#28
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Randy Yates writes:
[...] These two desirables contradict one another, because once you go to a lower sample rate, you must use a HIGHER bit width in order to maintain the equivalent SNR that you would have at the higher sample rate. I should also have stated that, in another sense, and one that has very real gains, going to a lower sample rate also allows you to save CPU cycles since a) you're running the filters at lower sample rates, and b) the filters themselves are likely to be more benign. So, via a good bit of hand-waving (or appealing to experience, whichever justification you prefer), using a wider bitwidth at a lower sample rate is almost certainly going to be a big net-win MIPS-wise. -- % Randy Yates % "And all that I can do %% Fuquay-Varina, NC % is say I'm sorry, %%% 919-577-9882 % that's the way it goes..." %%%% % Getting To The Point', *Balance of Power*, ELO http://www.digitalsignallabs.com |
#29
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Don Pearce wrote:
On Sun, 06 Jan 2008 17:36:25 +0000, Dirk Bruere at NeoPax wrote: Don Pearce wrote: On Sun, 06 Jan 2008 16:54:14 +0000, Dirk Bruere at NeoPax wrote: Don Pearce wrote: On Sun, 06 Jan 2008 16:20:19 +0000, Dirk Bruere at NeoPax wrote: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. I though that twice the Nyquist frequency would be adequate? eg for a sub designed to operate under 100Hz I'll sample/decimate to 400Hz. Any benefits going to a higher frequency? Just about every audio DAC/ADC system in use these days goes for sampling at perhaps 128 times the Nyquist rate. The hard bit is always the analogue anti-alias filter and at this rate it is easy. I can't think of any audio sampled systems that work at just twice Nyquist. Well, at least with subs I've got plenty of room for tweaking frequencies and seeing what happens. I want to keep the number of samples the DSP works with fairly low so I can do other stuff eg room eq, graphics eq etc. without pushing the DSP anywhere near its limits. What most DSP systems - CD and the like - do is this. Having captured the data at high speed you apply a steep digital lowpass filter at 100Hz. You can then decimate the data and do all your work at 200Hz to minimize processor load. You will need to upsample back to high speed again to drive the DAC. As opposed to setting the sampling rate ie the number of samples per second taken from the ADC, at a few hundred Hz? All that stuff has to be done. If the ADC will do it for you without your having to get involved, then great. I'm not familiar enough with the current SOTA to know whether you can take oversampling and decimation for granted with these things. IIRC some AKM ones do 128x oversampling. -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#30
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Randy Yates wrote:
Dirk Bruere at NeoPax writes: [...] Not entirely decided yet, hence the questions. You missed my point. No matter what you use (within reason), you're going to have plenty of MIPS. The other problem with (say) Fs = 48kHz is low frequency instability and rounding errors. I would like to implement a sixth octave graphic EQ below 100Hz Oh, well why didn't you say so in the first place? You have opposing requirements (welcome to engineering). One is that you do your processing with the least number of bits possible in order to save CPU operations (e.g., to be able to avoid doing double-precision arithmetic, assuming you're implementing in fixed-point). The other is to lower the sample rate so that your filter designs are more easily achieved. These two desirables contradict one another, because once you go to a lower sample rate, you must use a HIGHER bit width in order to maintain the equivalent SNR that you would have at the higher sample rate. In this particular application at this juncture in history, however, reducing CPU cycles should be low on your list of concerns, as I've already pointed out, since most DSPs are likely to have PLENTY of horsepower. As for a good architecture for implementing the sixth-octave filter, that's a perfect question for comp.dsp - suggest you repost there. Yes - been round the houses the-) I'm just going to have to try a few things and hear how they sound in real life. -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
#31
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Dirk Bruere at NeoPax writes:
Randy Yates wrote: Dirk Bruere at NeoPax writes: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. Then you aren't concerned with "subwoofer construction and performance." I am to the extent that I want to know what features influence reproduction the most. If the subwoofer input is analog, you aren't at all concerned about anything digital. The "construction and performance" of such a subwoofer is independent of any signal processing that may or may not be ahead of it in the system. Either you haven't adequately described the subwoofer system you have in mind to us (e.g., maybe you're including the equalizer as part of the subwoofer), or you have a problem seeing where the lines are in system design. You still haven't described from a system level what your "box" is going to do. What are the inputs and what are the outputs? What is analog and what is digital? -- % Randy Yates % "And all that I can do %% Fuquay-Varina, NC % is say I'm sorry, %%% 919-577-9882 % that's the way it goes..." %%%% % Getting To The Point', *Balance of Power*, ELO http://www.digitalsignallabs.com |
#32
Posted to rec.audio.tech
|
|||
|
|||
Bits and Bass
Randy Yates wrote:
Dirk Bruere at NeoPax writes: Randy Yates wrote: Dirk Bruere at NeoPax writes: Dirk Bruere at NeoPax wrote: Randy Yates wrote: Dirk Bruere at NeoPax writes: [...] Actually, I'm referring to the realworld construction and performance of subwoofers versus higher frequency units. Which has absolutely nothing to do with digital signal processing. My earlier response to you was apparently a gross misinterpretation of your question, such as it was. The sooner you forget about "bits" in relation to subwoofer construction and performance, the sooner you can get on to the things that matter, like B/L, mechanical resistance, enclosure volume, or whatever else it is that really matters (I'm not a speaker designer). Actually, what you wrote was very interesting and informative. Thanks. And having said that, I am going to use decimation (or at least much lower frequency sampling) of the signal used to drive the subs. Then you aren't concerned with "subwoofer construction and performance." I am to the extent that I want to know what features influence reproduction the most. If the subwoofer input is analog, you aren't at all concerned about anything digital. The "construction and performance" of such a subwoofer is independent of any signal processing that may or may not be ahead of it in the system. Either you haven't adequately described the subwoofer system you have in mind to us (e.g., maybe you're including the equalizer as part of the subwoofer), or you have a problem seeing where the lines are in system design. You still haven't described from a system level what your "box" is going to do. What are the inputs and what are the outputs? What is analog and what is digital? Inputs either analogue or digital, then DSP with graphics eq, then DAC, power amp, speaker. In one box. -- Dirk http://www.transcendence.me.uk/ - Transcendence UK Remote Viewing classes in London |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
Studer A-80 looking for bits... | Pro Audio | |||
Hot bits | Audio Opinions | |||
How many bits of dither? | Tech | |||
Bits n Bobs | Marketplace | |||
Edirol UA-20 and 24 bits capture | Pro Audio |