The basic design of a secure GSM phone is to take the voice from the microphone, digitize it and run it through a compression algorithm, before encrypting it and sending it via a GSM data call to the other party. The compression algorithm is also called a codec and does with voice what mp3 does with music – making sure it takes up less data.
CryptoPhones use two different codecs. The original CryptoPhone code is called CELP, running at 8kHz. The output stream of the codec is 4.8kbit/second, enabling it to be transported over a 9,6kbit/s GSM data call. The new CryptoPhone codec, introduced with the CryptoPhone G10i+, is a custom development based on ACELP which provides significantly improved sound quality while reducing the necessary bandwidth usage. The ACELP variant has been specifically optimized for an output bandwidth of only 4 kbit/s, so the complete CryptoPhone stream including all overhead data requires less than 4,8 kbit/s.
The speech and sound quality you can expect is comparable to international phone calls. You should note that the overall speech quality depends on the GSM signal quality, so degradation does happen in low coverage areas. While in unencrypted GSM the sound quality gets bad and you would experience dropouts as the phone moves out of coverage, with the CryptoPhone under the same circumstances the call delay would increase. Simple indicators on the CryptoPhone show GSM signal coverage and call quality / delay.
All calls made with the CryptoPhone are subject to a certain delay in the call, as if your call is routed over a satellite link. Most of the call delay originates from the way GSM networks handle the data calls. The CryptoPhone must use the GSM data call instead of the normal voice call mode to ensure a transparent communications channel between the two CryptoPhones. Because the delay is a side-effect of all GSM data calls there is nothing we can do about it. All available GSM encryption products on the market suffer equally from this delay. The CryptoPhone itself introduces comparatively little delay from the voice encoding and encryption