I would think an access box with a port for the phone line and then phone and ethernet ports for the costomers equipment would be the obvious way to do it. Maybe add a panel with some punchdown blocks so it can be hardwired if desired.
This service sounds like it would need a custom modem anyway. I can't imagine building a voip analog adaptor into that box would be too expensive.
911 with VoiP? Assuming the VOIP soloution is part of the package and not a third party service I don't see why this should be a problem. The biggest problem with VOIP and 911 is when the voip is seperate from the line there is no way to tell if the user has moved the VOIP equipment without telling the provider.
You won't be able to utilize the full bandwidth of that frequency range until you get everyone switched over. For the most part phone lines should be seperate channels. There may be a little big of crosstalk but if there was significant crosstalk in the voice band people would have complained ages ago.
See here [vonl.com] for more explanation. They are a necessary evil on phone lines, unless you complerely re-engineer the line for higher frequencies. That may be though the impression I got from other sources is that the telcos often overdid it on the loading coils.
You would certainly want to get rid of them if moving the line to a modern all digital soloution. The extra bandwidth would be more usefull than a slight performance improvement for parts of the voice band
Modern 56k modems have trouble even connecting at 56k on rural phone lines, simply because the lines suck. 56K modems tend to fall a fair way short of 56K even on good phone lines which can easilly support several megabits per second of DSL.
56K modems suck because the device at the other end is not a modem it is a voice digitiser/undigitiser running at 8khz with 7 usable bits.
What that means is that in order to get even close to 56K the modems behaviour has to be similar to that of the voice digitiser/undigitiser. And that means it can't really adapt to line conditions except by dropping further below the 56K hard limit.
With modern equipment on BOTH ENDS you can do much better. You can use all the bandwidth of the line rather than being artificially limited to 4khz and you can split the line into many narrow channels which solves a couple of problems (though it does unfortunately increase latency). That is how DSL gets way way more out of phone lines than anyone could have anticipated before the DSP revoloution.
Splitting the line into many narrow channels achives the following 1: it means that narrowband noise is not so much of a problem. The hardware can simply avoid using noisy channels. That means that bandwidth that would previously have been regarded as unusable can become usable. 2: it means reflections are much less of a problem because symbol periods can be longer while still achiving a good overall data rate.
The download on 56k modems is already digital without overhead, and is already at the shannon limit. The signalling rate may be at the limit assuming the line really has a usable bandwidth of exactly 4khz (which I find highly unlikely) but signalling rate is only one part of the equation for data rate. Practically though you won't get more than that down a 4KHz channel because data rate is proportional to the logarithm of signal to noise ratio.
Of course the key word there is under ideal conditions. Afaict most of the gain from ADSL2 comes from extending to higher frequencies but those higher frequencies are lost even faster with distance. So on moderate length lines ADSL2+ is not much better than ADSL1.
Currently a fair chunk of the best (lowest) frequencies are used very inefficiantly to carry analog voice communication. That is bad if you are trying to optimise the utilty of a poor line.
The DSP revoloution has allowed us to extract performance from channels that would not previously have been thought possible. Probablly the best example of this is DSL extracting performance from phone lines way way beyond what they were originally designed for.
There are limits though. beyond a certain frequency the attenuation will reach a point where the line is unusable. You can increase the SNR by increasing the power but unfortunately achivable data rate is proportional to the logarithm of SNR so there are limited gains to be had there.
Improving the modulation may help a bit but there are fundamental limits on how much can be achived for a given bandwidth and SNR and afaict we are already pretty close to them.
For the purposes of this post bandwidth is defined in the traditional sense of the range of frequencies availible for your transmission.
ALL real world mediums are analog. Signals reflect off discontinuities. Noise gets added and higher frequencies get attenuated. Your channel may be all the usable bandwidth of a cable or it may be only a subset of it but it is still most certainly a meaningfull figure. A differential pair has a limited bandwidth just like any other cable (it has good noise immunitiy though especially if twisted, that along with being much cheaper than coax is why twisted is such a popular medium).
Even fiber has a limited bandwidth though unfortunately the limitations of current technology means most of it gets left unused.
We may choose to use the whole bandwidth of a communications medium as one channel. This is usually done for short distance links where we can relatively easilly install a high bandwidth low noise channel.
Often however it is to our advantage to divide the bandwidth we have availible into a number of narrower channels. It reduces sensitivity to reflections in cables or multipath distortion in radio. For point to point links it allows our equipment to negotiate different numbers of bits per symbol depending on how noisy that bit of spectrum is.
The problem with redefinitions like this is that they leave a lot of ambiguity. Especially as in the digital world you usually need to refer to both.
Data centers are a prime target for green work, and I bet with a little development work you could EASILY cut the power utilization by 30%. Upgrading and replacing four older machines with two newer machines will cut power usage. Replacing machines has a cost though both in terms of staff time and in terms of risk. I would imagine this is especially true of datacenters that perform a wide range of different functions, possiblly for a wide range of different customers.
Virtualisation is a possible soloution but the enterprise grade soloutions come at a rather high price and some customers may want the extra security/peice of mind of being on thier own dedicated machine.
Bulldoody, the cause of the Athlon cooking itself was a defective thermal diode on the motherboard. That is not what your source says.
Your source says that the issue was that the protection circuitry on the motherboard in question could only throttle the CU, not cut the power completely and that throttling was insufficiant to save the CPU in a complete loss of cooling scenario.
No I don't remember a followup from toms hardware, I do remember a lot of discussion about the reason for the XP burning up.
Was the retest done with the same make and model of motherboard or with a later one? I was under the impression that the proplem at the time was blamed on incorrect motherboard design.
I guess it is a sign that i'm an electronics/computer geek that I don't consider a soldering iron and a decent selection of screwdrivers to be special tools. Putty knives are a bit of an odd tool to be using for electronics work but they aren't exactly an unusual tool in general.
And because there are realtively few models of mac it is generally fairly easy to find information on what order things come apart in (which is generally the biggest challange when working on laptops and similar IMO)
The P4 did best, it just slowed down. The P3 crashed but worked fine afterwards The athlon XP cooked itself The thunderbird athlon did worst cooking itself and the motherboard too.
IIRC the XP was supposed to be the first AMD CPU with overheat protection. Unfortunately the motherboard manufacturers screwed up and at least earlier athlon XP systems still cooked themselves on heatsink loss (as demonstrated in the toms hardware video).
It seems the sempron 2600+ was a cut down athlon64 by which time they had thermal protection that actually worked properly.
These are prototypes. Back in the day prototypes were wirewrapped nightmares and they cost a lot more than $1500! Hmm, I dunno what volumes theese guys are planning to do but my guess is it is sufficiant that even back in the days when wirewrap was feasible the design would have been committed to PCB.
While in many ways progress is good most electronics hobbyists nowadays are left with the stark realisation that dealing with chips that are anywhere near performance competitive with PC hardware is a huge and expensive pain in the arse. In the through hole days with some determination and a wirewrap you could build almost anything yourself. Nowadays with all the high end stuff coming in BGA packages your only real option is to get both the board and the assembly done commercially.
Nah there are lots of other FPGA boards availible many of which are frankly a much better deal than this board.
Essentilly if you don't want the card for graphics what you get is a relatively small FPGA (one of the smaller members of the spartan 3 family which is xilinx's current low end family) on a PCI-X card. This board is way overpriced for that.
and this one isn't that big either, only 8K logic units (disclaimer: i've been working mostly with altera stuff and i'm not sure how spartan 3 logic units compare to cyclone 3 logic units but I would guess they are reasonablly similar).
Basically apple was actually doing comparatively well on the greenness front but they didn't announce what thier future plans for making thier products greener were (or at least not in as much detail as some vendors).
Greenpeace went on the assumption that no publically announced plans meant no plans, put them at the bottom of a ranking table and launched a campaign against them. I don't know if they asked apple for the info and were refused or if they just wen't straight ahead with the campaign.
In theory yes, in practice colinux beats vmware in some areas but loses out in others (the disk and network code seem to be particular bottlenecks). And colinux has no native GUI system so you have to use the network to run your GUI.
IMO the free (as in beer) versions of vmware have removed most of the reason to use colinux.
I think that dual booting does give people time to adjust to Linux, secure in the knowledge that if they need to do something quickly while they figure out Linux, they can boot Windows and just do what they need to. I think that dual booting is a poor choice for your main machine. At least if you are anything like me you will frequently have a lot of stuff open which you really don't want to have to close to reboot into another OS. So the linux install will probablly end up forgotten unless there are some important things you need the machine for that can only be done in linux.
quite some adaptation would be needed, while the core of colinux runs in kernel mode many of the management functions (such as actually making it run) and communication functions are controlled by user mode daemons which are as visible as any other process. CPU time used by colinux ends up attributed to the colinux-daemon process, not sure about memory.
Slashdot Mirror
I would think an access box with a port for the phone line and then phone and ethernet ports for the costomers equipment would be the obvious way to do it. Maybe add a panel with some punchdown blocks so it can be hardwired if desired.
This service sounds like it would need a custom modem anyway. I can't imagine building a voip analog adaptor into that box would be too expensive.
911 with VoiP?
Assuming the VOIP soloution is part of the package and not a third party service I don't see why this should be a problem. The biggest problem with VOIP and 911 is when the voip is seperate from the line there is no way to tell if the user has moved the VOIP equipment without telling the provider.
You won't be able to utilize the full bandwidth of that frequency range until you get everyone switched over.
For the most part phone lines should be seperate channels. There may be a little big of crosstalk but if there was significant crosstalk in the voice band people would have complained ages ago.
See here [vonl.com] for more explanation. They are a necessary evil on phone lines, unless you complerely re-engineer the line for higher frequencies.
That may be though the impression I got from other sources is that the telcos often overdid it on the loading coils.
You would certainly want to get rid of them if moving the line to a modern all digital soloution. The extra bandwidth would be more usefull than a slight performance improvement for parts of the voice band
Modern 56k modems have trouble even connecting at 56k on rural phone lines, simply because the lines suck.
56K modems tend to fall a fair way short of 56K even on good phone lines which can easilly support several megabits per second of DSL.
56K modems suck because the device at the other end is not a modem it is a voice digitiser/undigitiser running at 8khz with 7 usable bits.
What that means is that in order to get even close to 56K the modems behaviour has to be similar to that of the voice digitiser/undigitiser. And that means it can't really adapt to line conditions except by dropping further below the 56K hard limit.
With modern equipment on BOTH ENDS you can do much better. You can use all the bandwidth of the line rather than being artificially limited to 4khz and you can split the line into many narrow channels which solves a couple of problems (though it does unfortunately increase latency). That is how DSL gets way way more out of phone lines than anyone could have anticipated before the DSP revoloution.
Splitting the line into many narrow channels achives the following
1: it means that narrowband noise is not so much of a problem. The hardware can simply avoid using noisy channels. That means that bandwidth that would previously have been regarded as unusable can become usable.
2: it means reflections are much less of a problem because symbol periods can be longer while still achiving a good overall data rate.
The download on 56k modems is already digital without overhead, and is already at the shannon limit.
The signalling rate may be at the limit assuming the line really has a usable bandwidth of exactly 4khz (which I find highly unlikely) but signalling rate is only one part of the equation for data rate. Practically though you won't get more than that down a 4KHz channel because data rate is proportional to the logarithm of signal to noise ratio.
Of course the key word there is under ideal conditions. Afaict most of the gain from ADSL2 comes from extending to higher frequencies but those higher frequencies are lost even faster with distance. So on moderate length lines ADSL2+ is not much better than ADSL1.
Currently a fair chunk of the best (lowest) frequencies are used very inefficiantly to carry analog voice communication. That is bad if you are trying to optimise the utilty of a poor line.
The DSP revoloution has allowed us to extract performance from channels that would not previously have been thought possible. Probablly the best example of this is DSL extracting performance from phone lines way way beyond what they were originally designed for.
There are limits though. beyond a certain frequency the attenuation will reach a point where the line is unusable. You can increase the SNR by increasing the power but unfortunately achivable data rate is proportional to the logarithm of SNR so there are limited gains to be had there.
Improving the modulation may help a bit but there are fundamental limits on how much can be achived for a given bandwidth and SNR and afaict we are already pretty close to them.
Afaict if you are getting 128K you are already bonding two channels (a BRI interface carries two ISDN channels).
Some ISPs may well let you use more than two though.
For the purposes of this post bandwidth is defined in the traditional sense of the range of frequencies availible for your transmission.
ALL real world mediums are analog. Signals reflect off discontinuities. Noise gets added and higher frequencies get attenuated. Your channel may be all the usable bandwidth of a cable or it may be only a subset of it but it is still most certainly a meaningfull figure. A differential pair has a limited bandwidth just like any other cable (it has good noise immunitiy though especially if twisted, that along with being much cheaper than coax is why twisted is such a popular medium).
Even fiber has a limited bandwidth though unfortunately the limitations of current technology means most of it gets left unused.
We may choose to use the whole bandwidth of a communications medium as one channel. This is usually done for short distance links where we can relatively easilly install a high bandwidth low noise channel.
Often however it is to our advantage to divide the bandwidth we have availible into a number of narrower channels. It reduces sensitivity to reflections in cables or multipath distortion in radio. For point to point links it allows our equipment to negotiate different numbers of bits per symbol depending on how noisy that bit of spectrum is.
The problem with redefinitions like this is that they leave a lot of ambiguity. Especially as in the digital world you usually need to refer to both.
Data centers are a prime target for green work, and I bet with a little development work you could EASILY cut the power utilization by 30%. Upgrading and replacing four older machines with two newer machines will cut power usage.
Replacing machines has a cost though both in terms of staff time and in terms of risk. I would imagine this is especially true of datacenters that perform a wide range of different functions, possiblly for a wide range of different customers.
Virtualisation is a possible soloution but the enterprise grade soloutions come at a rather high price and some customers may want the extra security/peice of mind of being on thier own dedicated machine.
Bulldoody, the cause of the Athlon cooking itself was a defective thermal diode on the motherboard.
That is not what your source says.
Your source says that the issue was that the protection circuitry on the motherboard in question could only throttle the CU, not cut the power completely and that throttling was insufficiant to save the CPU in a complete loss of cooling scenario.
No I don't remember a followup from toms hardware, I do remember a lot of discussion about the reason for the XP burning up.
Was the retest done with the same make and model of motherboard or with a later one? I was under the impression that the proplem at the time was blamed on incorrect motherboard design.
I guess it is a sign that i'm an electronics/computer geek that I don't consider a soldering iron and a decent selection of screwdrivers to be special tools. Putty knives are a bit of an odd tool to be using for electronics work but they aren't exactly an unusual tool in general.
And because there are realtively few models of mac it is generally fairly easy to find information on what order things come apart in (which is generally the biggest challange when working on laptops and similar IMO)
It was a four way test
IIRC the results were:
The P4 did best, it just slowed down.
The P3 crashed but worked fine afterwards
The athlon XP cooked itself
The thunderbird athlon did worst cooking itself and the motherboard too.
IIRC the XP was supposed to be the first AMD CPU with overheat protection. Unfortunately the motherboard manufacturers screwed up and at least earlier athlon XP systems still cooked themselves on heatsink loss (as demonstrated in the toms hardware video).
It seems the sempron 2600+ was a cut down athlon64 by which time they had thermal protection that actually worked properly.
and pretty much all chips are powered by DC.
sorry I misread the comparision table (mixed up the 400 and the 4000).
sorry my mistake, I read the 3S400 entry rather than the 3S4000 entry by mistake.
These are prototypes. Back in the day prototypes were wirewrapped nightmares and they cost a lot more than $1500!
Hmm, I dunno what volumes theese guys are planning to do but my guess is it is sufficiant that even back in the days when wirewrap was feasible the design would have been committed to PCB.
While in many ways progress is good most electronics hobbyists nowadays are left with the stark realisation that dealing with chips that are anywhere near performance competitive with PC hardware is a huge and expensive pain in the arse. In the through hole days with some determination and a wirewrap you could build almost anything yourself. Nowadays with all the high end stuff coming in BGA packages your only real option is to get both the board and the assembly done commercially.
Nah there are lots of other FPGA boards availible many of which are frankly a much better deal than this board.
Essentilly if you don't want the card for graphics what you get is a relatively small FPGA (one of the smaller members of the spartan 3 family which is xilinx's current low end family) on a PCI-X card. This board is way overpriced for that.
and this one isn't that big either, only 8K logic units (disclaimer: i've been working mostly with altera stuff and i'm not sure how spartan 3 logic units compare to cyclone 3 logic units but I would guess they are reasonablly similar).
Basically apple was actually doing comparatively well on the greenness front but they didn't announce what thier future plans for making thier products greener were (or at least not in as much detail as some vendors).
Greenpeace went on the assumption that no publically announced plans meant no plans, put them at the bottom of a ranking table and launched a campaign against them. I don't know if they asked apple for the info and were refused or if they just wen't straight ahead with the campaign.
It isn't going to work in wine, wine only emulates win32, colinux uses a custom driver to do it's dirty work.
well it doesn't have built in GUI support so you would have to use networked X or XVNC or similar.
other than that I would imagine wine would run just fine.
In theory yes, in practice colinux beats vmware in some areas but loses out in others (the disk and network code seem to be particular bottlenecks). And colinux has no native GUI system so you have to use the network to run your GUI.
IMO the free (as in beer) versions of vmware have removed most of the reason to use colinux.
I think that dual booting does give people time to adjust to Linux, secure in the knowledge that if they need to do something quickly while they figure out Linux, they can boot Windows and just do what they need to.
I think that dual booting is a poor choice for your main machine. At least if you are anything like me you will frequently have a lot of stuff open which you really don't want to have to close to reboot into another OS. So the linux install will probablly end up forgotten unless there are some important things you need the machine for that can only be done in linux.
quite some adaptation would be needed, while the core of colinux runs in kernel mode many of the management functions (such as actually making it run) and communication functions are controlled by user mode daemons which are as visible as any other process. CPU time used by colinux ends up attributed to the colinux-daemon process, not sure about memory.