Key ADSL contributor Joseph W Lechleider dies at 82

Re: 128kbps (ISDN) may be considered unsatisfactory now...

Agreed. From 1999 to 2002 I worked from home in Lincoln, Nebraska (the state capital and second-largest city in the region), and BRI ISDN was the only reasonable choice for Internet access - including VPN to the corporate network - there at the time. And it was perfectly adequate for many purposes, including telnet to our UNIX systems in the UK.

Wasn't much good for remote-GUI work (whether X11, Citrix, VNC, etc), but even that was possible in an emergency.

'course in the mid-90s I was using a 56kb/s digital line, and before that Telebit Trailblazers running either SLIP or that Telebit proprietary file-transfer protocol (or once in a while zmodem or Kermit). And while doing my undergrad degrees it was V.42bis and earlier 1200b/s async dialup.¹ Or, years before that, the occasional BBS use with a Hayes 300 baud modem, though that was only ever for larks, not real work.

So 128kb/s seemed pretty decent for remote access. And the connection time was magical, after years of +++ATDT DTMF tones ring train train train maybe connect get a cup of coffee...

¹Plus a smattering of SDLC for SNA. Say what you will about SNA and SDLC, but man, I did not miss async when I was working with those connections. App tries to connect to host and the line comes up with no fuss, and you get the full symbol rate for actual data. Async didn't do that until, what, V.42?

I think that you misunderstand. A voiceband modern uses echo cancelling with the upstream and downstream using the same frequency bands. ADSL uses non-overlapping frequency bands for the upstream and downstream.

It's a different issue. Voice band modems share the available resource through time, versus sharing the resource through frequency for xDSL. Time division versus frequency division.

The crosstalk problem doesn't occur in the voice band because the modulated waveform doesn't have the characteristics that cause it to impose itself on other nearby circuits. The signal must comply with the PAM/PCM rules being applied at the telephone exchange or it won't go anywhere - so 300hz to 3.4Khz, with an 8Khz sample rate. The PSTN was designed around that standard and a waveform that complies will, in a healthy network, not be of concern to any other metallic circuits.

xDSL is different though - multiple channels are set up at different frequencies on the pair of wires and some of these can impose themselves on adjacent circuits. A strong signal that has just left a transmitting device can easily swamp the low signal being received on an adjacent circuit from a transmitting device at the far end of that circuit. It's very hard to hear what someone at the end of your garden is saying, even if they're shouting, if your next door neighbour is standing the other side of your fence right next to you and also shouting. That's crosstalk.

They key to solving this problem is a system of rules about how all these devices must co-operate in terms of signal frequency and power to achieve the highest mean throughput for all the users in a given plan. That's what Lechleider solved.