Elecraft K4 Compression with CESSB

Rick Miller, N1RM


[This is a brief version of a longer article that appeared in the June issue of the PVRC Newsletter. The complete article can be found at https://pvrc.org/Newsletters/2022_06.pdf ]

Single sideband is a very efficient way of modulating RF with audio.  All the RF power coming out of the transmitter directly contributes to the reproduction of the audio signal in the receiver with no watts wasted in a carrier or redundant sideband.

The average power of the human voice, compared to its peak, is low – often several dB below peak power.  Your 100-watt peak transmitter may only put out a few watts of average power when operating SSB voice.

Speech compression increases the level of quieter parts of speech so that the average power is greater and more easily heard. It is a simple concept, but it can be difficult to implement without the output signal becoming distorted or growing in bandwidth (“splattering”).

Starting in beta release 29/30 of the K4 firmware, Elecraft included Controlled Envelope Single Sideband (CESSB) processing in their speech compression. CESSB, introduced by Dave Hershberger, W9GR, in 2014[1], incorporates some clever signal processing to permit significant compression without excessive distortion. In his article, Dave stated that an increase in average power of up to 4 dB is possible with this approach.



Test Setup

The test signal was a 4 second audio clip calling CQ, recorded on a computer using Audacity software. The audio output from the computer was fed to the analog line input on the K4D. The line input level was set for full output power with minimal ALC action to isolate the CESSB processing from ALC.  Noise gate and ESSB were both off, the Tx Equalizer was flat, and the Tx Bandwidth was set to 2.8 kHz. The 100 watt peak RF signal was taken directly from the K4D antenna output, attenuate and fed to a digital oscilloscope and a spectrum analyzer.

RF Waveform Comparison

Figure 1 shows the effects of CESSB on the output RF envelope in the time domain for compression settings from zero (off) to 30 (maximum).  These waveforms show the dramatic effects of K4 compression.

Figure 1 - CESSB effects on FR Envelop
Figure 1 - CESSB effects on RF Envelope


Peak and Average Power Measurements

With the peak output of the K4D set to 100 watts (50 dBm), 800 instantaneous power measurements were taken during the test transmission of the audio clip and those samples were averaged. The results in Table 1 show that at maximum compression, the K4D CESSB implementation achieves the 4 dB average power improvement predicted in the original QEX article.

Table 1 - Average power measurements

CESSB Compression Performance

Spectral Behavior

Very high compression levels aren’t much good if they cause the signal to distort and “splatter” outside its normal bandwidth. While it’s clear that the K4 CESSB compression is effective at increasing average power, it was also important to see how well it constrains the signal’s bandwidth. Figure 2 shows spectrum measurements taken using the voice test clip. There was no measurable increase in occupied bandwidth [2] when the compression level was changed from 0 to 30.

K4D CESSB Compression - Voice Signal

So, how does it sound?

The time domain waveforms of the compressed signal make it look like the RF is clipped, which would create lots of splatter.  The spectrum plot of the signals, though, shows no widening of the signal bandwidth even at the highest compression levels. It looks like Elecraft’s implementation of CESSB is very good.

I used R29 in the WPX SSB contest with the compression level set at 20. I received more than the usual number of unsolicited “loud” comments, and also more than the usual number “great audio” comments.

I listened to the signal on the radio’s monitor, which includes the compression processing.  I found that up to a setting of around 20, the signal got much more punch with no significant effects on audio quality (to my ear).  Full compression at a setting of 30 seems to be right at the onset of perceptible distortion. Of course, all these observations are purely subjective.

Bottom Line

In my opinion, the speech compression processing introduced in Beta release R29 is extremely effective. It made my station louder without degrading audio quality and without increasing my transmitted bandwidth.


[1] Controlled Envelope Single Sideband, QEX – November/December 2014, http://www.arrl.org/files/file/QEX_Next_Issue/2014/Nov-Dec_2014/Hershberger_QEX_11_14.pdf

[2] Occupied Bandwidth (OBW) is defined as the bandwidth containing 99% of the total integrated power of the transmitted spectrum.




Back to blog

Leave a comment

Please note, comments need to be approved before they are published.