The Sounds of Dialup Modems and Related Equipment

The good old dialup modem. The bread and butter of internet access, of transacting data through simple plain old telephone wires. While the majority pay no real attention to how such a device works and just merely use it - I spent some time of my own building up collections of different modems and recording their signals. It turns out - there's some subtle differences between modems, and there's a whole plethora of different connection modulations which sound different. Throughout time - the pursuit for speed has changed the generally accepted modulation in use - where today, we reach our final dial up modulation - V.92. I only just recently changed from dialup to ADSL, and saw it befitting to present this collection of dialup sounds as a tribute to the dialup modem. This is not intended to be a joke as such - I can remember in a comedy TV show, someone boasted that they had downloaded music from the internet which sounded like a dialup connection being made, instead, it is the culmination of many hours of playing around, of line testing modems and reading AT command reference manuals to disable, enable and or force certain modulations to get a recording. All in all, I managed to teach myself many different things about dialup modems, and I could tell just how successful a connection was just by listening to the traning (negotiation) phase. There's some sort of orderly structuredness to most digital modulation techniques - I certainly enjoy the routine-ness of it, and that's also why I do radio and data decoding from the radio. Inspired by Modemsite's own collection of recorded dialup sounds back in High School, here are mine - I hope you enjoy it. Click on the names to hear the recording.

You may also be interested in Modems under Lab Conditions for other High Quality recordings of Modems

You might also be interested in my latest (2016) definitive collection of high-quality V.90 and V.92 modem sounds

V.21 - 300bps

One of the first standards approved data modulations to be used on the telephone network was the V.21 system. (Ignoring Bell 103 that is.) This often involved using acoustic couplers which you placed a telephone handset into - however, this recording was made by forcing a modem to emulate an older modem. Notice the data-tone that is 0.5s on and 2s off - that is only a relatively recent feature - and gives away the fact that this wasn't recorded with a real period-era modem. Back then, people had to manually dial the numbers and then place the phone handset into the coupler. This was essentially a dumb modem, that just literally translated the Frequency Shift Keying signals directly to ones and zeros. This sample file was recorded using a Texas Instruments based Aztech external modem.

V.22(bis) - 1200bps to 2400bps

The successor to V.21 was the V.22 (and then V.22bis) standard. This improved data rates by using Phase Shift Keying instead of FSK. Note that these standards have been outlawed in several countries - I think Australia included - as they use discrete carrier tones which are prone to crosstalk in trunked wires and can cause interference to other communications. Most modems then were still dumb, but this modulation required better coupling, and so directly connected modems began to flourish at about this time. Hayes were beginning development on their smartmodem which gave birth to the commonly accepted AT command set. This sample file was recorded using a Texas Instruments based Aztech external modem.

V.32(bis) - 9600bps to 14400bps

V.32 was widely thought to bring us up to the limit of phone line transmission. I believe QAM was being used, however, less than optimally - which allowed for more bits per symbol to be encoded and hence higher rates. Unfortunately this maxed out at 14400bps and something new had to come to the field to bring us up to a faster speed. This sample file was recorded using a Banksia MyModem 14.4k modem with a Rockwell Chipset. The Rockwell chipsets were the most popular chipset in the modem market, which later had changed names to Conexant - but they all have a distinctive sound in V.90 mode - more on that later.

V.34(bis) - 14400bps to 33600bps

So here comes Trellis Coding, otherwise known as TCM which allows more speed than ever before. Now we have truly (nearly) met the Shannon Limit of a phone line's bandwidth response. As you can tell, this sounds vaguely related to V.90 (which most of us are still using) in that the basic structure of the smart training process is there. V.90 merely builds upon this. This recording was made using a Texas Instruments based Aztech external modem. However, connections made using other modems can sound different - here is one which is made by an Ambient (later known as Intel) Chipset modem. Some modems were less compatible than others and yielded strange slow synchronization to the training signal and or total failure to connect after several retrain attempts. V.34 was also adopted in the Super3G faxing standard which allows fax machines to break from the limitations of the earlier V.17 standard which only allowed them to connect at 14400bps. This is what happens when a V.34 data modem tries to connect to a Super3G Fax machine.

V.90 - 33600 to 56000bps

The most popular V.90 standard is actually based on an interesting hack - that is, the phone service at your ISP is often a digital line, and so that can be exploited to provide better bandwidth on the downstream, while keeping your upstream at V.34 modulation. This tradeoff only works provided your line is very quickly converted to digital from your premesis and sent digitally throughout the telephone switching system without being converted to analog again. This standard worked best for local connections, however, on long distance connections there are times where it just won't work as it's too sensitive to quality of the phone line. In the case of training failure, V.90 modems tend to downshift to V.34 at times (say two consecutive training failures) - and unfortunately, once in V.34 it cannot upshift back to V.90 modulation hence the "it gets slower over time" syndrome. That being said, the V.90 standard was actually born from taking parts of the K56flex standard - which in itself was a combination of Conexant's K56 standard and Lucent's V.Flex standard. This meant that there were some initial compatibility issues which plagued these modems and firmware upgrades were relatively common to make modems standard compliant. The sound linked above is the most common sound which most people will remember - that of a Rockwell or Conexant modem connecting in V.90. However, the Digital Impairment Learning sequence (DIL) which is near the end of the connection, is actually not standards specified - and therefore, different chipsets implement their own DIL signals - for example - this is an Ambient Intel Chipset Modem, this is a Smartlink Modio SoftModem, this is a Texas Instruments Chipset Modem. Between those, we have most of the DIL signals covered as most of them will sound like the ones above - e.g. US Robotics Modems often feature bonging sounds like the TI modem above, and Lucent WinModems often have sounds which are either like the Ambient Tech or the Rockwell Conexant. That being said, there are probably more sounds out there - but these are the most common. Betcha never really noticed that there were differences between modems! Occasionally, some freak occurance can mean that V.90 modems suddenly pause their training and then resume, like in this clip and other times there is just some compatibility issues with descriptor ID's which cause the modems to hang up as soon as the connection is made.

V.92 - 33600 to 56000bps

V.92 is the successor to V.90, but isn't seeing much penetration as dialup is dying out. V.92 offers faster connection training through a feature called Quick Connect, and it offers Modem on Hold (MoH) ability with ISPs that support it, which, in combination with call waiting, allows your modem to suspend your internet session and place or receive calls and then resume the internet connection without any additional phone charges. It also offers the ability to use a more efficient compression algorithm called V.44 which was developed by Hughes Space Systems to improve throughput on compressable content, and also offers PCM Upstream which allows you to have a symmetrical connection of up to 48k upstream and 48k downstream - that is, sacrificing downstream for upstream rate. However, ISPs are reluctant to invest into such infrastructure given the lack of possible return. The audio file was actually recorded from a SmartLink SoftModem in a laptop, calling a Telstra V.92 enabled POP locally. The connection is established, and then I make a request to put the Modem on Hold, which is swiftly denied by the ISP and a connection is re-established. Notice, however, that the ISP has not enabled MoH or Quick Connect at their end. PCM Upstream wasn't observed, however, V.44 compression was. Unfortunately with ISP's total control over what you can and cannot do - most ISPs have chosen not to offer their customers most of the benefits which can be had on a V.92 connection. The difference in a V.90 and a V.92 connection can be heard in the initial negotiation sequence after the data answer tone - a quick warble is exchanged between the two modems which signify they are both V.92 compliant, followed by a decision to go and do a full training sequence (which then sounds just like V.90).

Related Stuff

TTY - around 75bps

Still in use nowadays, sometimes known as a TDD, these are often used by the deaf and hard of hearing to communicate over phone lines by typing. It is essentially a teletype terminal that often uses acoustic coupling and is very similar in theory to RTTY - using Phase Shift Keying. The characterset used is a limited Baudot set which allows for the system to transfer text faster by using less bits per character.

FAX - 2400bps to 14400bps

Also still in use, the humble fax machine also uses a modem, however, its standards are different to those of a data modem. The fax negotiation is broken into phases - Phase A is making a call, Phase B is negotiating the CSIDs and Fax machine capabilties, Phase C is the actual page data, Phase D is the end of page and end of transmission negotiations. You can hear phases B to D in the recording above - which is done in V.17 modulation at 14400bps. Negotiations at lower speeds sound similar, as the Phase B is done using a slow 300bps FSK mode, it's only the Phase C payload data sound that sounds different.

Unrelated and Strange

Just an unusual phone call I had one day to my ISP. Instead of being greeted with a data modem, I get random digital garbage. Something is definitely wrong with their gear.