One more note... there was a blue transformer with a metal handle, number 4150 I believe, which was rated for 50 watts. One of the terminals could be pressed down to interrupt the circuit for direction changes. It came in the better sets with die-cast locomotives, so if you're trying to be period-correct that's something to think about. I'm not sure how beefed-up it is "under the hood" to handle twice the wattage. I grew up an MPC kid, but the year I got my first metal steam loco, I moved up to a postwar 1033 and never looked back.
I agree with what everyone has posted. The "round thing" your friend describes is probably Lionel's "Mechanical Sound of Steam" which is a hollow rolling drum filled with pellets. That being said, there were a at least a few different versions of the Electronic Mighty Sound of Steam. Notably, there was an early version with Whistle. Then EMSoS went back to chuff only until 1980, when they introduced a different electronic whistle. During this time there were changes to the circuit itself. The earliest ones used only discrete components: transistors, resistors, etc. At some point toward the late '70s (I'm not sure when) they introduced an integrated circuit. IMO, the early versions without whistle have the "cleanest" sound, which is actually static or "white noise."
Also, there are two methods of modulating the "chuff." One is with a pair of contacts on the smoke generator. That method was used throughout the 1970s. Beginning with the UP 8002 in 1980, the balance of Lionel's Collector steam locos relied on a hall effect wheel on the tender truck to control the chuff rate. I guess this made the wiring easier, but the chuff was no longer guaranteed to sync with the movement of the crosshead.
If you get a loco with EMSoS, please remove the tender shell and re-mount the board with double-sided tape. The foam pad used at the factory generally disintegrates after 50(!) years. If the electronic board shorts to the tender frame (which is grounded), it can ruin the board and kill the chuff. My $.02.