Comparing (i) masthead LNA with coaxial switch for reference samples, (ii) Nooelec SAWBird H1 LNA with inbuilt 50 ohm switched reference resistor, (iii) 8M bandwidth signal using part away from hydrogen line as reference.
Comparing LNA at masthead with coaxial switch in front of it, with Nooelec SAWBird H1 LNA with inbuilt 50 ohm resistor which can be switched in and out for generation of reference samples, and with 8M wide bandwidth signal covering 1420MHz hydrogen line and where a part of the signal away from the hydrogen line can be used as reference.
Nice — I ran the numbers and put them side-by-side so you can see the exact tradeoffs. I used your SAWbird H1 numbers (Nooelec SAWbird H1 typical NF ≈ 0.8 dB) and the CX relay loss you quoted earlier (~0.20 dB at L-band). Where the “built-in switch” case is uncertain I show a handful of realistic loss assumptions (best → worst) so you can see sensitivity to that parameter. Sources used for reference values: Nooelec SAWbird datasheet, Tohtsu/CX relay specs and cable attenuation curves.
Assumptions (explicit)
- SAWbird H1 LNA typical NF = 0.8 dB (F = 10^(0.8/10)).
- External mechanical relay insertion loss (CX-520/Tohtsu family) ≈ 0.20 dB at 1.42 GHz.
- “Built-in switch” loss is not specified (depends on design). I evaluate several plausible cases: 0.05 dB (very low-loss internal transfer), 0.10 dB (good), 0.20 dB (typical), 0.50 dB (poorer or extra losses).
- Calculations use Friis/passive-loss rules: passive loss before the first amplifier multiplies the LNA noise factor. Conversions done exactly (F = 10^(NF/10); NF_total_dB = 10·log10( factor·F_LNA ); Trec = (F_total − 1)·290 K).
Compact results (NF and equivalent receiver noise temperature Trec)
| Scenario | Loss placed before LNA (dB) | Resulting system NF (dB) | Equivalent Trec (K) |
| Baseline — SAWbird alone (no added loss) | 0.00 | 0.80 dB | 58.7 K |
| External mechanical relay (CX family) immediately in front of mast LNA | 0.20 | 1.00 dB | 75.1 K |
| Built-in switch — very low loss (best case) | 0.05 | 0.85 dB | 62.7 K |
| Built-in switch — good (likely) | 0.10 | 0.90 dB | 66.8 K |
| Built-in switch — typical (same as external relay) | 0.20 | 1.00 dB | 75.1 K |
| Built-in switch — poorer (0.5 dB) | 0.50 | 1.30 dB | 101.2 K |
(Values shown to 2–3 significant figures; arithmetic computed exactly from the Friis/passive loss equations.)
Short interpretation of the table
- Putting a 0.20 dB passive loss before the SAWbird raises system NF from 0.80 → 1.00 dB, which corresponds to an increase of ~16–17 K in equivalent noise temperature (roughly 59 K → 75 K). That matches the order-of-magnitude you were told by SARA.
- If an internal switch is engineered to be very low loss (≈0.05–0.10 dB) it will be noticeably better than a 0.2 dB external relay — but the benefit is modest (a few to ~10 K).
- If the inbuilt switching introduces losses comparable to typical coax/relay losses (0.2 dB or more) then you get no advantage over the external relay case. In the worst case (0.5 dB) the penalty becomes material (Trec ≈100 K).
The “8 MHz wideband receiver — off-frequency reference” option
What you described (collect an 8 MHz-wide chunk of spectrum and use a region away from the hydrogen line as a reference) is not a pre-LNA calibration method — it’s a software/spectral reference technique. Important points:
- Noise figure / immediate sensitivity: if you do not insert extra passive components before the LNA, the system NF stays at the LNA’s value (0.8 dB in your case → ~59 K). So from a pure sensitivity standpoint this is the best (no NF penalty).
- Calibration quality: using an off-line frequency as your “reference” assumes that (a) the receiver gain and bandpass are extremely stable across the band and over time, and (b) there is no RFI or spectral features in the reference region. If those conditions hold you can track gain drifts without inserting a physical dummy load. In practice environmental/antenna impedance changes and standing waves affect the antenna response with frequency, so off-frequency referencing often leaves residual systematics that true antenna-vs-load switching would have measured directly.
- Practical trade: wideband reference can avoid the NF hit entirely and simplify hardware, but it is less direct (it does not measure the antenna match/thermal noise) and typically yields poorer absolute calibration accuracy unless the instrument and environment are very stable and free from RFI. It’s a great complement (and low-loss) — many groups use wideband methods together with occasional hardware reference measurements (noise diode or load switching) to get the best of both worlds.
Recommendation (practical, for hydrogen-line work)
- If you need the best absolute calibration (measure actual antenna temperature): put the LNA at the mast and use a low-loss mechanical switch or a very low-loss built-in switch (target ≤0.1 dB). If the LNA vendor can guarantee an internal switch insertion loss ≤0.1 dB, that’s slightly better than an external 0.2 dB relay. If the vendor cannot guarantee that, treat it like the external relay case.
- If you want to avoid any pre-LNA loss and can tolerate slightly worse absolute calibration accuracy: use the wideband (8 MHz) receiver approach without inserting the relay before the LNA. Combine this with periodic calibration sequences (e.g., disconnect the LNA briefly or inject a calibrated noise diode after the LNA via a coupler) to tie your spectral reference to absolute scale.
- Hybrid (often best in practice): mount a masthead LNA (SAWbird), avoid permanent pre-LNA passive loss where possible, use the wideband off-line reference for continuous monitoring, and perform occasional physical reference measurements (very infrequent mechanical switching to 50 Ω or noise-diode injection) to anchor the calibration. That keeps NF low while giving you absolute checks.