Forums

A Basic Question or two on CD Emphasis/Deemphasis

Started by Unknown September 16, 2004
"Jon Harris" <goldentully@hotmail.com> writes:

> "glen herrmannsfeldt" <gah@ugcs.caltech.edu> wrote in message > news:ciclbp$uv$1@gnus01.u.washington.edu... > > > > > > > Note that (US) FM radio (1940s?) uses preemphasis with a 75us time > > > constant. Many other parts of the world (which got FM radio much > > > later than the US) use 50us. CD (1970s) is switchable to 15us or > > > none. > > > > Not to mention that Dolby encoding with 25us preephasis is also > > used. Dolby, on average, increases the high frequencies, which > > is somewhat corrected by going through 75us deemphasis for those > > without Dolby decoders. > > What's the deal with specifying filters in terms of time? Wouldn't a frequency > be more intuitive? > A naive question, what is the cut-off frequency (3dB-point I presume) of a 25us > filter? Is it 1/25us = 40kHz or 1/(2*pi*25us) = ~6.37kHz or something else? > I'm guessing it's the 6.37k.
Good guess. :) |H(w)| = 1/sqrt(w^2*C^2*R^2 + 1). -- Randy Yates Sony Ericsson Mobile Communications Research Triangle Park, NC, USA randy.yates@sonyericsson.com, 919-472-1124

Allan Herriman wrote:

(snip)

> The audio was 16 bit and the DAC was 14 bit. > The extra bits were made up with PWM. This generates high frequency > noise.
> (NB. This seemed like a good idea at the time, but that was before > people believed in noise shaping or could make true 16 bit DACs.)
When was it that 4X oversampling with digital filters into 14bit DAC's were claimed to be better than 1X at 16 bits? It seems reasonable to me that an analog anti-aliasing filter is easier to build with the higher frequency. -- glen
Allan Herriman wrote:

> On Thu, 16 Sep 2004 11:51:51 -0400, Jerry Avins <jya@ieee.org> wrote: > > >>Randy Yates wrote: >> >> >>>Not to look a gift whore in the mouse, but how do you know all this? >> >>Hey! it's Allan Herriman you're asking! That's as close to a dumb >>question as I've seen here in a long time. He's our resident >>encyclopedia for audio and broadcast specifications. ;^) > > > I got it wrong though. The emphasis curve has a single continuous > time zero with a time constant of 50us and a single continuous time > pole with a time constant of 15us. The pole was added to limit the > amount of high frequency boost (to about 10dB). > > Regards, > Allan
I didn't write that you were correct. I wrote that I believed you. :-) Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
Jon Harris wrote:

> "glen herrmannsfeldt" <gah@ugcs.caltech.edu> wrote in message > news:ciclbp$uv$1@gnus01.u.washington.edu... > >> >>>Note that (US) FM radio (1940s?) uses preemphasis with a 75us time >>>constant. Many other parts of the world (which got FM radio much >>>later than the US) use 50us. CD (1970s) is switchable to 15us or >>>none. >> >>Not to mention that Dolby encoding with 25us preephasis is also >>used. Dolby, on average, increases the high frequencies, which >>is somewhat corrected by going through 75us deemphasis for those >>without Dolby decoders. > > > What's the deal with specifying filters in terms of time? Wouldn't a frequency > be more intuitive? > A naive question, what is the cut-off frequency (3dB-point I presume) of a 25us > filter? Is it 1/25us = 40kHz or 1/(2*pi*25us) = ~6.37kHz or something else? > I'm guessing it's the 6.37k.
The time is the time constant tau; R*C. When picking an R and a C to match a spec, it's easier to get it right, but that's not the real point. 750 pF and 100 K give me a 75 ?sec time constant, easy to achieve with components known to be available. What C should I use with a 100 K resistor to get 2100 Hz? Where can I get one? (750 pF is close enough, but we like our specs to be round numbers exactly.) Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
On Thu, 16 Sep 2004 12:11:26 -0700, glen herrmannsfeldt
<gah@ugcs.caltech.edu> wrote:

> > >Allan Herriman wrote: > >(snip) > >> The audio was 16 bit and the DAC was 14 bit. >> The extra bits were made up with PWM. This generates high frequency >> noise. > >> (NB. This seemed like a good idea at the time, but that was before >> people believed in noise shaping or could make true 16 bit DACs.) > >When was it that 4X oversampling with digital filters into >14bit DAC's were claimed to be better than 1X at 16 bits? >It seems reasonable to me that an analog anti-aliasing filter >is easier to build with the higher frequency.
Using current technology, the 16 bit DAC gives better performance. Back then, it wasn't possible to make a cost effective monotonic 16 bit DAC. 14 bits with PWM gave better linearity, but generated more noise. Regards, Allan

Allan Herriman wrote:

(I wrote)
>>When was it that 4X oversampling with digital filters into >>14bit DAC's were claimed to be better than 1X at 16 bits? >>It seems reasonable to me that an analog anti-aliasing filter >>is easier to build with the higher frequency.
> Using current technology, the 16 bit DAC gives better performance.
> Back then, it wasn't possible to make a cost effective monotonic 16 > bit DAC. 14 bits with PWM gave better linearity, but generated more > noise.
I don't remember the PWM description. I thought there was a system that used a digital filter to generate a 4x data stream low pass filtered somewhere between 20kHz and 22kHz. As doubling the conversion rate is equivalent to an extra bit of resolution, D/A could be done at 14 bits. Is that what you mean by PWM? Many CD players I know of advertize one bit D/A conversion. I can imagine extending the double sample rate per bit, but I don't imagine them running >1GHz, 32768*44.1kHz. -- glen
glen herrmannsfeldt <gah@ugcs.caltech.edu> writes:

> Allan Herriman wrote: > > (I wrote) > >>When was it that 4X oversampling with digital filters into > >>14bit DAC's were claimed to be better than 1X at 16 bits? > >>It seems reasonable to me that an analog anti-aliasing filter > >>is easier to build with the higher frequency. > > > Using current technology, the 16 bit DAC gives better performance. > > > Back then, it wasn't possible to make a cost effective monotonic 16 > > bit DAC. 14 bits with PWM gave better linearity, but generated more > > noise. > > I don't remember the PWM description. I thought there was a > system that used a digital filter to generate a 4x data stream > low pass filtered somewhere between 20kHz and 22kHz. > As doubling the conversion rate is equivalent to an extra bit > of resolution,
The technique you're describing Glen is plain old oversampling. Each doubling buys you 3 dB which is only a "half" bit of resolution. Thus 16X oversampling would have been required to get from 14 bits to 16. This technique (oversampling followed by a lowpass digital filter) also has the advantage of reducing the order/complexity of the reconstruction filter. But I'm with you - I remember no such conversion mechanism as the one Allan described, but I do distinctly remember the oversampling systems - in fact I owned one. -- Randy Yates Sony Ericsson Mobile Communications Research Triangle Park, NC, USA randy.yates@sonyericsson.com, 919-472-1124
Randy Yates <randy.yates@sonyericsson.com> writes:
> [...] > Each doubling buys you 3 dB which is only a "half" bit of > resolution.
3 dB less quantization noise, that is. -- Randy Yates Sony Ericsson Mobile Communications Research Triangle Park, NC, USA randy.yates@sonyericsson.com, 919-472-1124
On 16 Sep 2004 12:17:18 -0400, Randy Yates
<randy.yates@sonyericsson.com> wrote:

>> The audio signal being recorded tends to have less power at high >> frequencies. > >Why? >
This thread is hard to keep up with! Randy, I think this is the same reason that FM radio systems use preemphasis/deemphasis. To keep the PLLs operating efficiently and keep the SNR from getting really crappy at the high frequencies, preemphasis is applied. It sounds like this was the sort of thinking that went into this as well. I think others have pointed out that the contemporary thinking when standards are made are key to understanding them. Another example of this sort of thing that comes to mind is the Grand Alliance choosing 8VSB as the modulation standard for DTV in the US. I know few comm people who think that was a good move, but it has some features that would make it comfortable to people familiar with the NTSC analog standard. The fact that there is zero need for them to be similar or compatible doesn't seem to have mattered much. Standards bodies seldom pick the best technologies, they pick the technologies on which the participants can agree. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
On 17 Sep 2004 14:19:11 -0400, Randy Yates
<randy.yates@sonyericsson.com> wrote:

>glen herrmannsfeldt <gah@ugcs.caltech.edu> writes: > >> Allan Herriman wrote: >> >> (I wrote) >> >>When was it that 4X oversampling with digital filters into >> >>14bit DAC's were claimed to be better than 1X at 16 bits? >> >>It seems reasonable to me that an analog anti-aliasing filter >> >>is easier to build with the higher frequency. >> >> > Using current technology, the 16 bit DAC gives better performance. >> >> > Back then, it wasn't possible to make a cost effective monotonic 16 >> > bit DAC. 14 bits with PWM gave better linearity, but generated more >> > noise. >> >> I don't remember the PWM description. I thought there was a >> system that used a digital filter to generate a 4x data stream >> low pass filtered somewhere between 20kHz and 22kHz. >> As doubling the conversion rate is equivalent to an extra bit >> of resolution, > >The technique you're describing Glen is plain old oversampling. >Each doubling buys you 3 dB which is only a "half" bit of >resolution. Thus 16X oversampling would have been required to >get from 14 bits to 16. > >This technique (oversampling followed by a lowpass digital filter) >also has the advantage of reducing the order/complexity of the >reconstruction filter. > >But I'm with you - I remember no such conversion mechanism as >the one Allan described, but I do distinctly remember the >oversampling systems - in fact I owned one.
This was prior to oversampling. IIRC, it was a Philips player from about '83 or '84 (i.e. one of the very first ones). I am unable to find a reference, however. Regards, Allan