Saltar al contenido ForoHackers

Why 56k modems never delivered 56k

Historia de la computaciónProtocolos de redProcesamiento de señales
#957 · carrier_lost_1200 · · Denunciar

Short version: the number on the box was a theoretical ceiling, and the physics was more stubborn than the marketing.

A normal voice line is digitised at the phone exchange: 8000 samples per second, 8 bits each (u-law in North America, A-law in Europe). That is 64 kbit/s. In the V.90 design, the ISP side connects to the network digitally, so its modem can feed the exchange the 8-bit PCM values directly, with no analog-to-digital step on that end. Your modem then only has to recognize which of 256 levels was sent. In theory that gives 56 kbit/s downstream, because some bits are lost to robbed-bit signalling and level spacing.

Why you rarely got that:

  • Only downstream worked that way. Upstream the signal was analog at both ends, so it stayed at about 33.6k (V.34).
  • The levels near zero are close together, so line noise confuses them. Poor lines settled well below the maximum.
  • Regulators limited transmit power, so in practice the cap was nearer 53 kbit/s.
  • Any extra analog-to-digital conversion in the path (some long loops, some rural exchanges) killed the trick completely.

So "56k" was real, but only under conditions most of us did not have.

#1256 · r3l1c_0f_bbs · · Denunciar

I had a 33.6 modem in the house until almost the end. A friend was proud of his 56k and connected at 44 on a good night. We both considered that a victory.

#1257 · packet_loss_poet · · Denunciar

The Shannon limit for an analog phone line is somewhere around 35 kbit/s, so V.34 was close to the wall. V.90 did not break the wall. It walked around it by removing one of the two analog segments. Delightful.

#1258 · carrier_lost_1200 · · Denunciar

Exactly. "Walked around the wall" is the better summary than my entire post.

Inicia sesión o únete para responder.