Below the Noise Floor — Episode 13: "The AetherSDR Voice Chain, Part Two: EQ and Processing"

With a clean baseline signal established in Episode 12, this episode covers the refinements: transmit equalization, speech processing and compression, VOX sensitivity and hang time tuning, and how to save everything as a profile. Part two of two.

Transcript

SPEAKER_00 0:00

Last episode, we built the foundation of the Ether SDR voice chain, audio input confirmed, mic gain set to produce a healthy meter reading without clipping, ALC showing moderate activity rather than pinned at maximum, power output set to a reasonable level. That baseline gets you a clean, functional transmitted signal. This episode covers what comes next: equalization, speech processing, vox refinement, and profiles. These are the tools that take a signal that already works and make it work better. They are adjustments, not corrections. If the foundation from episode 12 is not in place, these tools cannot fix it. But if it is in place, these tools are genuinely worth understanding. I want to set expectations clearly before we start. Everything in this episode is optional in the sense that a signal with none of these things enabled, but with a good baseline from episode 12 is a perfectly acceptable signal. Many operators use modest equalization and some compression and never touch anything else. Some use none of it. All of those are valid approaches. The goal here is to understand what each tool does so you can make an informed decision about whether to use it, not to convince you that you need all of it. Let me start with transmit equalization because it is the most consistently useful of these tools and the one with the clearest practical payoff for HF voice. Equalization. EQ adjusts the relative level of different frequency ranges within the audio. Transmit EQ and Ether SDR does this to your outgoing voice audio before it reaches the modulator. The reason transmit EQ matters for HF SSB specifically comes down to how voice audio is transmitted and received. SSB voice on HF passes audio roughly in the range of 300 Hz to 3000 Hz. Natural speech has a lot of energy in the low frequency range, the fundamental frequencies of the voice, the warmth, and body of the sound. It has less energy at higher frequencies where consonants and intelligibility cues live. On an HF-SSB radio link, with its limited bandwidth and variable noise floor, the low frequency energy often contributes more mud and boom than intelligibility, while the higher frequency consonant energy is exactly what helps a weak signal remain copiable. The conventional transmit EQ approach for HFSSB reflects this. Cut or attenuate the low frequencies, typically below 200 to 300 Hz, to reduce boom and low frequency noise pickup. Boost the mid-presence range, roughly 1000 to 3000 Hz, where consonants and intelligibility cues live. The result is a voice that sounds slightly thinner and more telephonic than natural speech, but cuts through noise and interference better and remains intelligible at lower signal levels. In EtherSDR, the transmit EQ is typically a graphic or parametric equalizer with bands covering the transmit audio range. A reasonable starting point for HFSSB is a gentle cut below 200 Hz, a slight boost around 2000 to 2500 Hz, and the rest left flat. This is a subtle adjustment, 3 to 6 decibels of cut or boost, not dramatic reshaping. The right way to evaluate your transmit EQ is to get a signal report from someone who can hear you. Ask specifically whether your audio sounds natural and clear or whether it sounds muddy, boomy, thin, or harsh. That feedback tells you which direction to adjust. Now speech processing and compression. Speech processing on HF does one thing. It increases the average power of your transmitted voice signal relative to the peak power. An unprocessed voice signal has a high peak to average ratio. When you say a loud vowel sound, the audio level peaks high. When you say a quiet consonant, the level drops low. When you are silent between words, the level drops to near zero. The transmitter's power output follows these variations. The average transmitted power over the course of a sentence is significantly lower than the peak power during the loudest moments. What a distant receiver hears is directly related to your average transmitted power, not your peak power. Speech processing addresses this by compressing the dynamic range, reducing the difference between loud and quiet sounds so that the average level across the whole sentence is higher while the peaks are kept at a safe level. More average power for the same peak power means a stronger, more consistent signal at the receiving end. The keyword is moderate, light compression, a few decibels of dynamic range reduction, improves intelligibility and average power, aggressive compression, 10 or more decibels of reduction, makes the voice sound pumped, distorted, and fatiguing to listen to. The natural rhythm and dynamics of speech are part of what makes voice intelligible, and destroying them in pursuit of maximum average power backfires. In Ether SDR, the speech processor level is typically a slider or numerical control. Start with it set low, 2 to 3 decibels of compression. Make a contact and get a signal report. Increase the compression slightly if you want more punch. Stop increasing before the voice starts to sound unnatural. Most operators find a sweet spot somewhere between 3 and 8 decibels of compression and leave it there. Now Vox. We covered Vox in episode 7 as a PTT option, so this is refinement rather than introduction. The two Vox parameters that need tuning for comfortable operating are sensitivity and hang time. Vox sensitivity determines how much audio level is required to trigger transmit. The correct sensitivity setting is the highest sensitivity at which the radio does not trigger on background room noise. Put the radio in Vox mode, leave the microphone alone, and listen for any accidental triggering. If the radio keys up spontaneously, reduce sensitivity until it stops. Then speak at your normal operating level and verify the radio keys up promptly. Hangtime, sometimes called Vox delay or tail, is how long the radio stays in transmit after your voice falls silent. Too short and the radio drops to receive in the middle of a sentence whenever you pause for breath. Too long and you have to wait several seconds after you finish speaking before the radio returns to receive. A hang time of around 300 to 500 milliseconds is usually comfortable. Profiles. This is one of the most useful features in Ether SDR for anyone who operates on multiple bands or in multiple conditions, and it is also one of the most overlooked. A profile in Ether SDR saves a collection of settings, mode, filter width, transmit EQ, compression level, power output, and other parameters under a name so you can recall all of those settings instantly by selecting the profile rather than manually adjusting everything each session. A practical example. A third profile called FT8 Digital that stores the settings for digital mode operation, no noise reduction, audio input routed from WSJTX, lower power output. When you sit down to operate, instead of working through a checklist of individual adjustments, you select the profile that matches what you plan to do, and all of the settings apply instantly. To create a profile in Ether SDR, configure everything the way you want it. Then find the profile save function in the interface and name it something descriptive. Do this once for each of your common operating configurations. Let me bring everything from episodes 12 and 13 together into a complete picture of the voice chain. The chain flows from microphone through mic gain through the speech processor and transmit EQ into the modulator through the ALC to the power output stage and then the antenna. Setting it correctly means working from left to right, confirm audio input, set mic gain to healthy levels, configure transmit EQ for HFSSB intelligibility, set compression to moderate levels, verify ALC is not pinned, set power output appropriately. Save all of that as a profile. When the chain is set correctly, a contact on 20m SSB sounds like two people having a clear conversation across whatever distance propagation allows. Your voice is recognizable, the signal is strong relative to your power output, and the other operator does not need to ask you to adjust anything. That is what we were working toward across these two episodes, a voice signal you are not embarrassed by. One, you can put on the air with confidence. Next episode is the last one in phase 3. Episode 14 covers multi slice operation and panadapter layout in Ether SDR, running multiple receivers simultaneously, watching two bands at once, and when that capability is actually useful versus when it is overkill. This is below the noise floor.