PureSignal Feedback Calibration

PureSignal Feedback Calibration

The default HW Peak value shipped per board is calibrated for the radio's stock internal feedback path. The moment your feedback chain departs from that — an external directional coupler on an amplifier, a sample tap with non-standard coupling, an inline attenuator pad — those defaults stop being right for your station. This page explains why, and walks through the short bench procedure to find the value that is right.

If you're seeing the Observed bar sit in orange or red even at moderate drive levels, or PureSignal taking a long time to converge, this page is the most likely fix.

What the Observed bar actually measures

Despite the name, Observed is not the raw RF power at the feedback connector. It's the post-IQC TX I/Q envelope — the magnitude of the predistorted TX signal that Zeus is sending to the radio's TX-DAC.

HW Peak is the TX-envelope value that maps to feedback-ADC full scale through calcc's hw_scale = 1 / HW Peak. In other words, HW Peak describes the geometry of your specific feedback chain — radio TX path → PA → coupler → cable → feedback ADC — collapsed into a single number.

So the colored bands on the bar mean:

Band	Class	Meaning
Blue	(none)	Post-IQC envelope < 95 % of HW Peak. Comfortable LUT headroom.
Orange	.warn	Post-IQC envelope within the top 5 % of HW Peak. Brushing the LUT ceiling.
Red	.bad	Post-IQC envelope ≥ HW Peak. calcc is being asked to extrapolate beyond its calibrated range.

Two consequences worth internalising:

1. Amp output power has very little effect on where the Observed bar sits. It tracks predistortion expansion against the LUT scale, not raw RF level. Drivers commonly report "I expected red at 1500 W and blue at 200 W, but they're the same" — that's the expected behaviour.
2. Raising HW Peak to silence the bar without confirming Feedback Level is a foot-gun. It only changes where Zeus draws the warning line. If the feedback ADC was physically clipping, it still is; you've just stopped getting told about it.

The trustworthy ADC metric is Feedback Level — the 0..256 dial in both the popover and the dashboard. That's info[4] from WDSP, the actual feedback envelope as a fraction of ADC full scale.

When the default HW Peak isn't right

The factory defaults assume the radio's stock feedback geometry:

Board	Default HW Peak
Hermes / ANAN-10 / ANAN-100 (P1)	0.4072
ANAN G2 / G2 MkII / Saturn (P2)	0.6121
ANAN-7000 / 8000 / Orion (P2)	0.2899
Hermes-Lite 2 (P1)	0.2330

If any of the following is true, those numbers are unlikely to be right for your station:

• You're feeding the feedback ADC from an external linear amp's sample tap or directional coupler, not the radio's internal coupler path.
• You have inline attenuation between the tap and PS-IN that wasn't there when the default was characterised.
• The amp's coupler has a coupling factor that differs from the typical Apache external-coupler assumption (e.g. a –55 dB tap vs the typical –30 to –40 dB external coupler).
• You've replaced cables or connectors in the feedback line and the cumulative loss has changed.

In any of those cases, the default HW Peak no longer describes the TX-envelope → feedback-ADC mapping that your physical chain implements. The bar will mis-warn, calcc will dwell in COLLECT longer than it should, and convergence may be flaky.

Calibration procedure

You'll need:

• Dummy load on the antenna port (50 Ω, rated for your operating power).
• A way to set drive at a representative normal-operating power. Voice on a real mic in your normal operating mode is fine; a two-tone test or a steady CW dead-key works too.
• A few minutes of bench time.

Steps:

1. Connect to the radio. Tune to a band you actually operate on.
2. Open Settings → PURESIGNAL. Note the current HW Peak value (in HARDWARE section) so you can roll back if needed.
3. Leave PS disarmed. Click the PS button off if it's on.
4. Key MOX briefly at normal operating drive — same drive % you'd use during a real QSO. Watch the Feedback Level number (0..256, the big number in the calibration ring).
5. Read the peak Feedback Level during steady-state TX. This is your ADC headroom indicator:

Feedback Level	Diagnosis	Fix
180–220	ADC is in the sweet spot. ~30 % headroom for predistortion expansion.	None needed — your chain is well-matched.
140–180	Slightly under-driven. ADC has plenty of headroom; LUT is being squeezed into the lower bins.	Either accept it, or raise HW Peak to push the LUT range up.
< 140	Significantly under-driven. calcc will struggle to fill its upper amplitude bins.	Raise HW Peak upward, or reduce inline attenuation if you have any.
220–250	Hot. Working but close to clipping; predistortion expansion may push peaks into hard-clip.	Add 3–6 dB inline attenuation on the PS-IN feedline.
256 (pegged)	ADC is clipping. PS cannot see the true envelope at the peaks. Correction will be unstable.	Add 10 dB or more inline attenuation immediately. Don't try to fix this by raising HW Peak — you'll just hide the clipping in the UI.

6. Adjust HW Peak in small steps (typically 0.01–0.03 at a time) up or down depending on the diagnosis. After each change, arm PS and run a calibration cycle, watching the Observed bar.
7. Goal: Observed bar sits in blue during steady-state correction, touching orange only on extreme peaks. Feedback Level still well clear of 256.
8. Once it sits in blue at your normal operating power, you're done. Save the value as your station-calibrated HW Peak — it's hardware-specific and won't drift between sessions.

If you have to redo this because you swapped the tap, the amp, the feedback cable, or moved to a different radio, the procedure is exactly the same.

When to add a pad vs adjust HW Peak

These are not equivalent. The decision matrix:

• Feedback Level pegging at 256, or > 240 sustained → Add inline attenuation. A 10 dB / 5 W SMA or BNC attenuator (Mini-Circuits BW-S10W5+ or similar) is the standard answer. After the pad lands you in the 180–220 range, you can usually return HW Peak to the per-board default.
• Feedback Level < 180 and Observed bar in orange/red at default HW Peak → Raise HW Peak. Your chain is under-driving the feedback ADC; the default LUT scale is too compressed for what your hardware delivers. Pads would make this worse.
• Feedback Level in 180–220 but Observed bar still in orange → A small HW Peak tweak (5–10 %) is the right adjustment. Predistortion expansion in your chain is pushing the LUT ceiling; giving it slightly more range is correct.

The rule of thumb: let Feedback Level decide. The bar is a derived UI indicator; Feedback Level is the ground truth from the feedback ADC.

Worked example — ANAN-G2 + RF2K-S external –55 dB tap

A typical operator chain that will need recalibration straight out of the box:

• Radio: ANAN-G2 (P2 OrionMkII, default HW Peak = 0.6121).
• Feedback source: RF2K-S internal directional coupler, –55 dB sample tap.
• Inline attenuation between tap and PS-IN: none.

Initial symptom: Observed bar steady red at any output power (193 W tested → 1500 W tested), default HW Peak = 0.6121.

Diagnosis steps:

1. Predicted feedback level from –55 dB tap at 500 W amp output: +57 dBm – 55 dB = +2 dBm at PS-IN. That should be hot.
2. Measured Feedback Level at steady-state TX: ~140 / 256 (~55 % of ADC full scale).
3. ADC is not clipping — the chain has ~6 dB of headroom. So adding a pad would be the wrong move. The default HW Peak is the problem.

Resolution:

• Raised HW Peak from 0.6121 → 0.655 in two steps (0.623 → 0.655).
• Observed bar moved red → orange → steady blue.
• Feedback Level still ~140 / 256 at the same operating power (raising HW Peak doesn't change the ADC reading, only the LUT scale).
• PS converges cleanly, CORRECTING stays on, correction dB settles.

Why the default was off: the RF2K-S internal coupler runs a few dB lower than nameplate, and/or the G2 PS-IN 0-dBFS reference is a couple dB hotter than the typical 0 dBm assumption. Cumulative effect was that the radio's stock LUT scale (0.6121) was a few percent too small for the post-IQC envelope produced by this physical chain. Operator-tuned to 0.655 and it's right for this station.

That value is hardware-specific. If you build the same chain on your bench, expect to land somewhere in the same general neighbourhood, but do the procedure on your gear — don't copy the number blind.

See also

• PureSignal — the main PS panel guide and the STATE / CORRECTING distinction.
• PureSignal on Protocol 2 — P2-specific wiring and 3-state arm sequence.
• RF2K-S Amplifier — RF2K-S panel in Zeus.
• PA Settings — drive / power calibration that feeds the PS chain.