>ISDN and GSM systems have no DC response at the communications channel >level. However, they carry digital data, and what you convey in that >digital data is completely open. If you carried audio in a modem signal >across the analogue PSTN, that could also have response down to DC.It depends how you define the end points of the channel. If you terminate an ISDN line with an ISDN card in a PC and use the CAPI interface library, then the stream of audio data which the CAPI library passes on to the application program is definitely dc coupled. There is no special coding involved - the telephone network simply passes A-law or u-law coded samples from end to end. The network does not normally do ANY filtering (unless the link is so long that echo cancellers are introduced by the telco). For example, one can connect an answerphone application such as Capitel http://www.2tec.com/capitel/ to an ISDN line and call it from an analogue phone. Capitel will make a good quality recording with a frequency response defined by the originating end. If the originating end is another digital application then that frequency response can go from dc to Nyquist. An ISDN telephone could have a non-standard codec with a high-pass response at 80Hz for example. That frequency response would be propagated across the telephone network until it reached a codec at the far end. If that codec had a similar response, perhaps in another modified ISDN phone, it really would be propagated end-to-end. A pc with a sound card and an ISDN card would easily do this using readily available software. With analogue pstn phone lines the codecs are at the exchange (central office) where you can't get at them - and of course there are other good reasons already mentioned why there needs to be aggressive hp filtering. John
Why the low frequency limitation on POTS?
Started by ●September 9, 2005
Reply by ●September 11, 20052005-09-11
Reply by ●September 11, 20052005-09-11
On Fri, 09 Sep 2005 23:40:32 -0400, Jerry Avins <jya@ieee.org> wrote:>Jon Harris wrote: >> "Tim Wescott" <tim@seemywebsite.com> wrote in message >> news:542dnZEbj-t_gr_eRVn-1w@web-ster.com... >> >>>glen herrmannsfeldt wrote: >>> >>> >>>>jharris wrote: >>>> >>>> >>>>>I'm curious to know why the 300 Hz low frequency limitation exists on POTS. >>>>> Since fundamentals of the male human voice can easily extend below this, >>>>>it would seem beneficial to accomodate low frequencies, at least down to >>>>>say 150 Hz. Is there some historical reason that the low frequencies are >>>>>limited? >>>> >>>>My guess would be to avoid 60Hz pickup problems. >>>> >>>>Among others, as I understand it early telegraphs systems used >>>>ground (real earth) return. Maybe not for telephones, though. >>>> >>>>Still, with open wire phone lines, probably near power lines, >>>>the 60Hz pickup could have been a problem. >>>> >>>>Otherwise, how hard was it to make amplifiers work down to >>>>lower frequencies? It requires larger coupling capacitors, >>>>for example. >>>> >>> >>>Originally all that was transformer coupled, and low frequency, high fidelity >>>transformers are expensive. A POTS line still has to supply 48VDC along with >>>audio, and it has to deliver 96V at 20Hz to make the ringer go -- those are >>>both things that will drive the lowest audio frequency higher in a >>>cost-effective system. >> >> >> I figured it had something to do with the analog properties of the phone line. >> Someone I spoke with recently claimed that it was a psychoacoustic issue: that >> since the high frequency end was limited, the low frequencies had to be >> similarly limited in order to "sound right". I had never heard of anything like >> that, but thought I would investigate his claim (he was a stranger, but in the >> audio industry in some form). Has anyone else heard of that? > >Yes. It is generally accepted -- and listening tests indicate -- that >poor low- and good high-frequency response (or vice-versa) sounds less >"natural" than the two being balanced. For low-frequency cutoffs between >20 and 100 Hz, the recommended product is -- surprise! -- 400,000 Hz^2. >Personally, I think that calls for restricting the highs too much above >about 60 Hz, but you might not agree. Telephones run from 300 to 4000, >giving a product of 1,200,000 Hz^2. (Score one for me.) 100 to 4,000 >would meet the "hifi" norm, but (when I could hear that high) I >preferred 100 to about 8,000.I read something similar about audio design that dated to way back when (I probably read this in the '70's and it was a decades-old book then), but calculated in a slightly different (but apparently equivalent) way. If either end is bandimited to less than the human hearing limit, the other end should also be limited so that their geometric mean (square root of their product) is 1000 Hz. That would lead to 1M (Hz^2) in your calculation system, a higher point than what you gave, but closer to the telephone standard. I don't know when this "standard" was first proposed, though I wonder if the original POTS design predates it.>Jerry
Reply by ●September 12, 20052005-09-12
"Ben Bradley" <ben_nospam_bradley@frontiernet.net> wrote in message news:ico9i1d7vq3hjmah40keu3vhnkb04u1dl5@4ax.com...> On Fri, 09 Sep 2005 23:40:32 -0400, Jerry Avins <jya@ieee.org> wrote: > >>>>> >>>>>>I'm curious to know why the 300 Hz low frequency limitation exists on >>>>>>POTS. >>>>>> Since fundamentals of the male human voice can easily extend below this, >>>>>>it would seem beneficial to accomodate low frequencies, at least down to >>>>>>say 150 Hz. Is there some historical reason that the low frequencies are >>>>>>limited? >>>>> >>>>>My guess would be to avoid 60Hz pickup problems. >>>>> >>>>>Among others, as I understand it early telegraphs systems used >>>>>ground (real earth) return. Maybe not for telephones, though. >>>>> >>>>>Still, with open wire phone lines, probably near power lines, >>>>>the 60Hz pickup could have been a problem. >>>>> >>>>>Otherwise, how hard was it to make amplifiers work down to >>>>>lower frequencies? It requires larger coupling capacitors, >>>>>for example. >>>>> >>>> >>>>Originally all that was transformer coupled, and low frequency, high >>>>fidelity >>>>transformers are expensive. A POTS line still has to supply 48VDC along >>>>with >>>>audio, and it has to deliver 96V at 20Hz to make the ringer go -- those are >>>>both things that will drive the lowest audio frequency higher in a >>>>cost-effective system. >>> >>> >>> I figured it had something to do with the analog properties of the phone >>> line. >>> Someone I spoke with recently claimed that it was a psychoacoustic issue: >>> that >>> since the high frequency end was limited, the low frequencies had to be >>> similarly limited in order to "sound right". I had never heard of anything >>> like >>> that, but thought I would investigate his claim (he was a stranger, but in >>> the >>> audio industry in some form). Has anyone else heard of that? >> >>Yes. It is generally accepted -- and listening tests indicate -- that >>poor low- and good high-frequency response (or vice-versa) sounds less >>"natural" than the two being balanced. For low-frequency cutoffs between >>20 and 100 Hz, the recommended product is -- surprise! -- 400,000 Hz^2. >>Personally, I think that calls for restricting the highs too much above >>about 60 Hz, but you might not agree. Telephones run from 300 to 4000, >>giving a product of 1,200,000 Hz^2. (Score one for me.) 100 to 4,000 >>would meet the "hifi" norm, but (when I could hear that high) I >>preferred 100 to about 8,000. > > I read something similar about audio design that dated to way back > when (I probably read this in the '70's and it was a decades-old book > then), but calculated in a slightly different (but apparently > equivalent) way. If either end is bandimited to less than the human > hearing limit, the other end should also be limited so that their > geometric mean (square root of their product) is 1000 Hz. That would > lead to 1M (Hz^2) in your calculation system, a higher point than what > you gave, but closer to the telephone standard. > I don't know when this "standard" was first proposed, though I > wonder if the original POTS design predates it.I know it was a long time ago, but any idea where that came from? I would be interested in seeing a reference.
Reply by ●September 12, 20052005-09-12
Ben Bradley wrote: ...> I read something similar about audio design that dated to way back > when (I probably read this in the '70's and it was a decades-old book > then), but calculated in a slightly different (but apparently > equivalent) way. If either end is bandimited to less than the human > hearing limit, the other end should also be limited so that their > geometric mean (square root of their product) is 1000 Hz. That would > lead to 1M (Hz^2) in your calculation system, a higher point than what > you gave, but closer to the telephone standard.So that a response from 20 Hz to 20 KHz would fall short of the ideal? Did the author advocate cutting off bass below 50 Hz?> I don't know when this "standard" was first proposed, though I > wonder if the original POTS design predates it.Probably. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●September 12, 20052005-09-12
"Jon Harris" <jon99_harris7@hotmail.com> wrote in message news:%k6Ve.4860$Hs6.3094@trnddc07...> "Ben Bradley" <ben_nospam_bradley@frontiernet.net> wrote in message > news:ico9i1d7vq3hjmah40keu3vhnkb04u1dl5@4ax.com... >> On Fri, 09 Sep 2005 23:40:32 -0400, Jerry Avins <jya@ieee.org> wrote: >> >> I read something similar about audio design that dated to way back >> when (I probably read this in the '70's and it was a decades-old book >> then), but calculated in a slightly different (but apparently >> equivalent) way. If either end is bandimited to less than the human >> hearing limit, the other end should also be limited so that their >> geometric mean (square root of their product) is 1000 Hz. That would >> lead to 1M (Hz^2) in your calculation system, a higher point than what >> you gave, but closer to the telephone standard. >> I don't know when this "standard" was first proposed, though I >> wonder if the original POTS design predates it. > > I know it was a long time ago, but any idea where that came from? I would be > interested in seeing a reference.I just found this with Google: http://www.smeter.net/daily-facts/10/fact18.php It claims an 800 Hz geometric mean is ideal.
Reply by ●September 12, 20052005-09-12
jrwalliker@gmail.com wrote: (snip)> It depends how you define the end points of the channel. If you > terminate an ISDN line with an ISDN card in a PC and use the CAPI > interface library, then the stream of audio data which the CAPI library > passes on to the application program is definitely dc coupled. There > is no special coding involved - the telephone network simply passes > A-law or u-law coded samples from end to end. The network does not > normally do ANY filtering (unless the link is so long that echo > cancellers are introduced by the telco).Also, note that the current 56k modem standards require that the answer mode end be an ISDN line. It depends on the ability of the modem to recognize the outputs from the D/A converter and decode the digital value input to the DAC. That won't work if there are two DACs in the process. -- glen
Reply by ●September 12, 20052005-09-12
glen herrmannsfeldt wrote:> jrwalliker@gmail.com wrote: > > (snip) > >> It depends how you define the end points of the channel. If you >> terminate an ISDN line with an ISDN card in a PC and use the CAPI >> interface library, then the stream of audio data which the CAPI library >> passes on to the application program is definitely dc coupled. There >> is no special coding involved - the telephone network simply passes >> A-law or u-law coded samples from end to end. The network does not >> normally do ANY filtering (unless the link is so long that echo >> cancellers are introduced by the telco). > > > Also, note that the current 56k modem standards require that the > answer mode end be an ISDN line. It depends on the ability of > the modem to recognize the outputs from the D/A converter and > decode the digital value input to the DAC. > > That won't work if there are two DACs in the process. > > -- glen >I'll admit much ignorance ;) Glen's statement that "56k modem standards require that the answer mode end be an ISDN line" confused me. I _had_ understood that 56k modems were designed around limitations of totally analog telephone service ( POTS? ). I Googled "ISDN". Chose "definition" and eventually ended at http://www.answers.com/main/ntquery;jsessionid=1effwr8fd197q?tname=basic-rate-interface&method=6&sbid=lc03a ( a *very* long line ) The article apparently implys that my existing telephone service is really ISDN simulating/emulating/??? POTS. Should I be able to request ISDN service at a rate that would *LESS THAN* sum of ( [my current rate] plus [DSL service] ) ? I don't expect DSL speed, but something > "56k". Would there be hassles accessing my current DSP who is dial up oriented and provides *MUCH _BETTER_ customer service* than local 'baby bell' would ever consider providing. I suspect a reasonably short answer would be an educational URL. Thank you.
Reply by ●September 12, 20052005-09-12
Richard Owlett wrote:> glen herrmannsfeldt wrote: >-snip->> -- glen >> > > I'll admit much ignorance ;) > > Glen's statement that "56k modem standards require that the answer mode > end be an ISDN line" confused me. I _had_ understood that 56k modems > were designed around limitations of totally analog telephone service ( > POTS? ). > > I Googled "ISDN". Chose "definition" and eventually ended at > http://www.answers.com/main/ntquery;jsessionid=1effwr8fd197q?tname=basic-rate-interface&method=6&sbid=lc03a > > > ( a *very* long line ) > > The article apparently implys that my existing telephone service is > really ISDN simulating/emulating/??? POTS. Should I be able to request > ISDN service at a rate that would *LESS THAN* sum of ( [my current rate] > plus [DSL service] ) ? > > I don't expect DSL speed, but something > "56k". > > Would there be hassles accessing my current DSP who is dial up oriented > and provides *MUCH _BETTER_ customer service* than local 'baby bell' > would ever consider providing. > > I suspect a reasonably short answer would be an educational URL. > > Thank you.The 56kbit/sec modems are designed to have characteristics that match the ADC and DAC in the SLIC (oops - acronym overload). From what I know about them they'd have to be designed for a certain kind of filtering from the SLIC, and if you try to send them through an older system with more analog hardware between you and the SLIC they won't work at full speed. Modern phone systems are almost entirely digital. For instance, I live out in the boonies so my twisted pair goes a couple of miles down the road and terminates at a SLIC (subscriber line interface card) in a little green box by the side of the road. Inside the box it's converted to 64kbits/sec digital and sent on its way. Any conversation stays digital until it gets to the SLIC at the other end (which may be in the phone there if it's ISDN or a business with a digital PBX). So from that point of view it's kinda ISDN like up to the SLIC. Assuming that things haven't changed in the 15 years since I last worked on them "Real" ISDN lines such as you might get installed at your house are known as BRI (Basic Rate Iforgottherest). BRI consists of not one but two 64kbit/sec channels for data and/or voice, and a 9600 bit/sec channel for signaling. A BRI line would probably only get you 64kbit/sec, but it should be solid. ADSL gives, if I'm not mistaken, much more bandwidth, at least on the "downstream" end because the designers know a dirty picture has a lot more bits than the little packet that requests it. Around here (Oregon) an ISDN line costs a bit more than a regular business line, but I don't think you can get one at a "residential" rate -- so expect around $50 - $60 a month. To find out about ISP issues (and whether you can get ISDN at all) you should check with your local phone company _and_ your ISP. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●September 12, 20052005-09-12
Tim Wescott wrote:> Richard Owlett wrote: > >> glen herrmannsfeldt wrote: >> > -snip- > >>> -- glen >>> >> >> I'll admit much ignorance ;) >> >> Glen's statement that "56k modem standards require that the answer mode >> end be an ISDN line" confused me. I _had_ understood that 56k modems >> were designed around limitations of totally analog telephone service ( >> POTS? ). >> >> I Googled "ISDN". Chose "definition" and eventually ended at >> http://www.answers.com/main/ntquery;jsessionid=1effwr8fd197q?tname=basic-rate-interface&method=6&sbid=lc03a >> >> >> ( a *very* long line ) >> >> The article apparently implys that my existing telephone service is >> really ISDN simulating/emulating/??? POTS. Should I be able to request >> ISDN service at a rate that would *LESS THAN* sum of ( [my current >> rate] plus [DSL service] ) ? >> >> I don't expect DSL speed, but something > "56k". >> >> Would there be hassles accessing my current DSP who is dial up >> oriented and provides *MUCH _BETTER_ customer service* than local >> 'baby bell' would ever consider providing. >> >> I suspect a reasonably short answer would be an educational URL. >> >> Thank you. > > > The 56kbit/sec modems are designed to have characteristics that match > the ADC and DAC in the SLIC (oops - acronym overload). From what I know > about them they'd have to be designed for a certain kind of filtering > from the SLIC, and if you try to send them through an older system with > more analog hardware between you and the SLIC they won't work at full > speed. > > Modern phone systems are almost entirely digital. For instance, I live > out in the boonies so my twisted pair goes a couple of miles down the > road and terminates at a SLIC (subscriber line interface card) in a > little green box by the side of the road. Inside the box it's converted > to 64kbits/sec digital and sent on its way. Any conversation stays > digital until it gets to the SLIC at the other end (which may be in the > phone there if it's ISDN or a business with a digital PBX). > > So from that point of view it's kinda ISDN like up to the SLIC. > > Assuming that things haven't changed in the 15 years since I last worked > on them "Real" ISDN lines such as you might get installed at your house > are known as BRI (Basic Rate Iforgottherest). BRI consists of not one > but two 64kbit/sec channels for data and/or voice, and a 9600 bit/sec > channel for signaling. A BRI line would probably only get you > 64kbit/sec, but it should be solid. ADSL gives, if I'm not mistaken, > much more bandwidth, at least on the "downstream" end because the > designers know a dirty picture has a lot more bits than the little > packet that requests it. > > Around here (Oregon) an ISDN line costs a bit more than a regular > business line, but I don't think you can get one at a "residential" rate > -- so expect around $50 - $60 a month. To find out about ISP issues > (and whether you can get ISDN at all) you should check with your local > phone company _and_ your ISP. >I suspect your last 2 paragraphs give me ALL the *bad* news ;[ I rarely download pictures. SBC, my telco, forces you towards YAHOO which provides mediocre newsgroup access if you want DSL [ and *NO* mention that you _ONLY_ get ADSL ]. My dialup ISP provides access to supernews.com as part of their basic service. I'll check with my ISP. Doubt I'll really like answer ;} HOWEVER, I *can* drive *~mile* and _personally_ speak with the owner of my ISP ;}
Reply by ●September 13, 20052005-09-13
"Jerry Avins" <jya@ieee.org> wrote in message news:S8qdnbIEluSQbrneRVn-vA@rcn.net...> Ben Bradley wrote: > > ... > >> I read something similar about audio design that dated to way back >> when (I probably read this in the '70's and it was a decades-old book >> then), but calculated in a slightly different (but apparently >> equivalent) way. If either end is bandimited to less than the human >> hearing limit, the other end should also be limited so that their >> geometric mean (square root of their product) is 1000 Hz. That would >> lead to 1M (Hz^2) in your calculation system, a higher point than what >> you gave, but closer to the telephone standard. > > So that a response from 20 Hz to 20 KHz would fall short of the ideal? Did the > author advocate cutting off bass below 50 Hz?I've also seen 800 Hz listed as the "ideal" geometric mean (http://www.smeter.net/daily-facts/10/fact18.php). This is pretty close to what you would get for the 20-20k case (geometric mean = ~632 Hz), and would dictate 32 Hz - 20 kHz. As a matter of practicality, there isn't much in music below 32 Hz, few speakers can reproduce it anyway, and the ear rolls off pretty fast down there. As a point of reference, the low "E" string, which is the lowest note of both the orchestral double bass and most electric basses used in contemporary music, has a frequency 41.2 Hz. So I wouldn't lose too much sleep over the last half-octave or so!
