While both were data transfer protocols serving streams at 300/1200 baud, why did they choose different frequencies, marks, modulation, and anything else?

I understand that they serve different purposes and cannot be identical, but failed to figure out underlying benefits to make choices diverting from existing standard and hardwares, especially when they tried to draft a unified standard for computer hobbyists.

Ref: Computer Users Tape System (CUTS) - Sol-20


PS: Similar examples also find differences in frequencies, like Bell 103/V.21, Bell 212/V.22.

See: Historical Modem Protocols, Modem standards


4 Answers 4


They are different because they are meant for different use cases so they both work well for what they are meant for. One is suitable for real time data transmission between two distant equipment over phone lines or radio or similar, and the other is suitable for storing data on cheap cassette tapes.

It would be useful to use the tape interface as the modem, but how to send data between two devices in real time is very different from storing the tones on wobbly tape and playing them back.

Bell 103 uses FSK modulation, which is both expensive to implement and is fairly sensitive to frequency changes which makes it not reliable enough for data storage due to cassette recorder speed tolerances. To be compatible with the system, you would need to implement detection of four frequencies as you would need to be compatible with the two originating frequencies and the two answering frequencies. Besides the Bell 103 is only a modulation scheme, it defines how to convert logic levels to tones and back at up to 300 baud (300 bits per second), so it does not take part at all what do those logic levels and transitions between them mean. Typically that data stream would UART frames, maybe 8N1 or 8N2 setup, at 300 bits per second, but it could be any other protocol. As there is no relation between the tone phase and bit edges it is completely asynchronous system. Bell 202 just expanded the data rate to 1200 baud to allow 1200 bps in one direction, but Bell 212 uses DPSK to send 1200 bps in both directions at 600 baud.

The KCS simply uses sine tones or square wave pulses, with each bit lasting a constat amount of pulses for constant duration of tone, while the tone frequency determines if the bit is logic 0 or 1. The encoded bits thus contain an embedded clock signal so the modulation is synchronous with the bits which allows huge tolerances in the tape speed. The detection in hardware is very simple, the playback signal just needs to check if voltage has gone up above some threshold of or gone down below some threshold and detect the pulse lengths and counts in software to decode bits. And basically the recording signal generation is even simpler, the software just has to toggle an output pin as per the transmitted bits in a timed loop. The KCS also defined that each byte must be stored with framing of start bit, 8 data bits, and two stop bits.

The KCS protocol allows easily improving the data rate as the cassette tape has a bandwidth of maybe 10-15 kHz, while phone lines have maybe 3 kHz usable bandwidth and Bell 103 FSK modulation is a standard that can't be broken to allow faster transmission or detection of bit changes, so Bell 202 FSK could be used, but it again uses different frequencies than Bell 103 and it would not be useful to be only Bell 202 compatible so implementing both Bell 103 and 202 would be even more expensive.

  • "Bell 103 uses FSK modulation" - so does KCS/CUTS. "expensive to implement and is fairly sensitive to frequency changes" - the Pennywhistle modem cost less than a KSC adapter and was able to deal with large variations in speed (as low as 50 bps or so and as high as 450) and large amounts of line noise. "but how to send data between two devices in real time is very different from storing the tones on wobbly tape and playing them back." - playing 300 bps modem tones into a tape works perfectly fine as a transfer system, I used to do that in the 80s with a CAT and my Akai deck. Commented Nov 6, 2021 at 14:35
  • Would the Pennywhistle modem be able to decode a signal that instead of using 1,070 and 1,270 used 970 and 1,170? That's a tape speed change of less than 10%, which would admittedly be far worse than what a good tape mechanism should do, but well within the tolerance level required for reliable operation when using more widely separated frequencies for ones and zeroes.
    – supercat
    Commented Oct 4, 2023 at 15:41


Three reasons:

  • A cassette interface only needs to fit a single transmission channel at a time, while a modem needs to hold two.

  • Cassette tape is made for a different (broader) frequency range than the phone network.

  • It's all about easy separation of whatever signals are present and suppressing everything else.

Number of Channels

A cassette interface is by definition unidirectional. There is only one transmission at a time from a single sender (*1). Thus only a single pair of signals (frequencies) is needed, which can be spaced equally across the available bandwidth.

A modem needs to hold two channels, each a pair of tones. As a result the phone bandwidth needs to be divided into two equally wide channels, each holding one transmission.

In all practical implementation this will be done by a filter only letting thru whatever half of the bandwidth should be received (*2).

Bandwidth (Frequency Range)

While it is true that a telephone line is defined for a frequency range of roughly 300 to 3400Hz, while a cassette tape is good for 40..12,000 Hz, this is not really an issue, as frequencies for (most) cassette interfaces are kept well below 3000 Hz.

A real world channel isn't a nice box of equal size, but is shaped by filter functions dampening a signal different over the frequency range. As closer it gets to the 'edges', as more a signal is lowered. In reverse this means the best transmission (recording) is right at the middle (*3). So the least dampened (aka best) signal frequency for a phone line is about 1000 Hz to 2700 Hz. Let, for simplicity, assume the same for a cassette recording (*4).

Third: Signal Separation

Next, not all frequencies are present at the same time.

A modem holds two channels, Originate and Answer, each can transmit one of two signals, Mark and Space, at a given time. Thus at maximum two frequencies are present at a given time. This table shows the resulting nine combinations:

Case  Originate   Answer
 1      None       None
 2      None       Mark
 3      None       Space

 4      Mark       None
 5      Mark       Mark
 6      Mark       Space

 7      Space      None
 8      Space      Mark
 9      Space      Space

So the goal is to make each of these nine states recognizeable with the least doubt possible. It's easy if no signal or only one signal is present. for all other cases, we want to have the two signals as far apart as possible while itself being apart from cut off. So let's enter the Bell 103 frequencies to see how they did it:

(Assuming the 300..3400 Range)

Case  Originate   Answer  Separation
 1      -          -      -
 2      -          2025   1325/1375
 3      -          2225   1525/1175

 4      1070       -      770/2330
 5      1070       2025   770/ 955/1375
 6      1070       2225   770/1155/1175

 7      1270       -      970/2130
 8      1270       2025   970/ 755/1375
 9      1270       2225   970/ 955/1175

We see that at any given time any signal has is at least 700 Hz from any channel cut off, as well at least 700 Hz from each other. This will quite good satisfy the need of least dampening while at the same time maximum separation. Great for good detection, isn't it? ​

For a cassette recording things are way less complicated. There is only one sender active at any given time and it can use the whole spectrum, assuming again 300..3400 Hz, 1330 Hz and 2370 Hz would make a great pair.

Making it 1200 and 2400, as chosen by CUTS, is not only close, but as well easy to remember and implement.

Long story short: Same transmission channel but different use case.

Bonus Answer:

Similar examples also find differences in frequencies, like Bell 103/V.21, Bell 212/V.22.

Nop, they are not due the same fact, as between modem and cassette, but regarding different bandwidth available. ITU standards are meant to operate across different networks. Some countries use(d) smaller phone channels with an upper end below 3000 Hz, to keep equal distance to cut of filters, frequencies had to be closer. That's why all ITU frequencies are lower than BELL frequencies for the same purpose. On the other hand, signal distance was kept the same to enable multi standard modems

*1 - Handover between sending (writing) and receiving (reading) is done by rewinding and listening to a different pair of lines :))

*2 - Originate frequencies by the station called, Answer frequencies by the station calling.

*3 - Well, not really true, especially not for recording devices as their filter function is based around an off middle peak, but for this issue we simply go by middle point.

*4 - As the makers of CUTS did the same :)

  • 1
    Greate answer! It seems choosing 1200 and 2400Hz was a questionable decision considering of second-harmonic effect.
    – Schezuk
    Commented Nov 4, 2021 at 11:55
  • 1
    @Schezuk Not really an issue. this is short distance and energy in harmonics falls fast, so loudness adjustment is your friend :))
    – Raffzahn
    Commented Nov 4, 2021 at 12:09

Modems were made to transfer data over a distance measured in kilometers.

Cassette interfaces were made to transfer the data a couple of meters at most.

Modems were made for industrial use. Robustness (of data and equipment) was a key factor. Price wasn't so much.

Cassette interfaces were made for home use. Price was the main factor. Robustness was made as good as possible, but you were expected to verify that what you saved was saved.

Modems had to follow the rules of telecom traffic (not interfere with other things on the line etc)

Cassette interfaces could do what they liked as long as it didn't interfere with other people in the household or neighbors listening to the radio (or later watching the TV)

  • While V.21 requires that modems must tolerate ±12Hz drifts, CUTS paper requires drift tolerances much larger than 5%, which are far more robust than a Bell 103 modem can be and quite confusing.
    – Schezuk
    Commented Nov 4, 2021 at 10:57

In contrast to the answers posted earlier, I will disagree:

There is no technical reason they did not use Bell 103, it was purely an issue of timing and history.

To start with, there is much verbiage in previous posts about having two channels and this means... something. However, both standards are FSK, so technically a tape is nothing more than one half of a modem, either the answer or originate side of it. That automatically makes it cost less, and in contrast to the claims of modems being expensive, the Pennywhistle of 1973 cost less than early KSC adaptors.

There are also comments about different purposes and thus different encodings. These do not hold up to even basic scrutiny when one considers that both are FSK so that's obviously no different. But it holds even less water when one considers all the other completely different standards like HITS, which was PCM at 2000 Hz, or other systems that used FSK at completely different frequencies, like Atari.

If you can use completely different different systems and/or frequencies, then the medium is clearly not driving the decision. But don't take my word for it, here's the statement right from the HITS article:

"The frequency of the tone burst is not critical."

So, then, what really happened?

Everyone got together for the KC meeting in late 1975 and picked some nice sounding numbers.

That is all.

Among the group was Lee Felsenstein, who had invented the Pennywhistle modem. That does raise the question why he didn't suggest this, considering it would mean the existing hardware could be immediately adapted. Given this, I've emailed him that question.

  • True, but: The quality of filters - which is what's needed to distinguish both frequencies, is dependent on parts used. Putting the frequencies as far apart as possible (and likewise away from the devices cut off) means lower quality (higher tollerance) parts can be used. Having them 1200 Hz apart instead of 200 Hz is a big deal here. That's why, BELL 103 asks for +/- 5% tolerance, which needs a more sophisticated design than hat KCS required. So yes, there are very basic reasons. Lee Felsenstein suggesting KCS shows that he was well aware of the gained freedom by more separation.
    – Raffzahn
    Commented Nov 6, 2021 at 15:35
  • The Pennywhistle was hardly sophisticated, and was cheaper than a KCS card in spite of having both channels. This is not the reason. Commented Nov 6, 2021 at 19:56
  • You're aware that your argument uses unrelated items? The price something is sold later on is not necessary related to the considerations during design. Setting the used frequencies six times further apart do improve the signal/interference ratio, or do you deny this?
    – Raffzahn
    Commented Nov 6, 2021 at 20:07
  • It is true that both systems use FSK, but they differ how they use it. A Bell 103 system uses very narrow band for data transmission, the tones for "0" and "1" only differ by 200 Hz and thus decoding data has to be with two band pass filters to determine if tone for logic 1 or logic 0 has more energy. Tape interfaces could just use a single bit under MCU control to basically output a small positive or negative voltage level for recording to tape, and the the input circuitry to take in tape audio simply determines if the voltage level is positive or negative and MCU can receive that bit.
    – Justme
    Commented Nov 6, 2021 at 20:39

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .