MultiNav Pro+ BLE module - does RSSI stay usable for ranging at long distances?

RSSI for that kind of coarse ranging can work but honestly it gets pretty noisy and unpredictable once you're pushing past a few hundred meters - multipath, obstacles, even weather can swing your readings enough to blur the line between "500m" and "2km" pretty badly. If you're okay with occasional false positives you might get away with it using some heavy averaging/filtering on the readings, but I'd sanity check it in a few real outdoor environments before committing to that approach since lab results are almost always way more stable than what you'll see in the field.

@Barbara Moore yeah the previous reply nailed the core problem — RSSI ranging and long distances are kind of fundamentally at odds with each other, and the MultiNav Pro+ makes it worse in an interesting way. Because the module has such aggressive receiver sensitivity and external amplification, you're going to see signal where most BLE modules would just drop out entirely, but that doesn't mean the RSSI values are stable or trustworthy for ranging at those distances. You're essentially reading a very faint signal that's bounced off everything between here and there, and the number you get back is more of a "we received something" than "here's a reliable power measurement."

 

For your specific use case though — coarse zone detection, 500m vs 2km — I'd think about it differently. Instead of trying to use RSSI as a continuous ranging proxy, you could set threshold triggers. Something like "if RSSI drops below X for Y consecutive readings, asset has likely crossed the 500m boundary." Kalman filtering or a simple moving median with a reasonably large window (30-50 samples) helps a lot with the noise, but you still need to calibrate those thresholds in your actual deployment environment because terrain and vegetation will shift them significantly from what you measured anywhere else.

 

Honestly the cleanest solution for your use case would be pairing even a cheap GNSS module on the remote unit and using the data link the BLE modules give you to just transmit the coordinates — that way you get actual zone detection instead of a noisy proxy. The GNSS module in the RFOXiA ecosystem is worth a look if you want to keep it all in one stack. But if GPS on the remote unit is genuinely off the table, the threshold-with-heavy-filtering approach is your best bet, just test it outside in conditions similar to your deployment before you commit.

Sarah's point about multipath is real - I spent way too long chasing RSSI-based ranging on a similar project before I just accepted that past ~300m outdoors you're basically looking at a very fuzzy zone indicator rather than anything like a distance estimate. For your 500m vs 2km threshold specifically I'd prototype it with a sliding window median filter (not mean, outliers will wreck you) and log a bunch of real-world walks at both distances before you commit to any threshold values.

Both Sarah and Karen are spot on - I'll just add that on the MultiNav Pro+ specifically the BLE TX power settings make a big difference to how gracefully the RSSI degrades at range, so if you haven't already played with cranking that up it's worth doing before you decide whether the approach is viable at all. Also +1 to the median filter, I learned that the hard way too lol.

Jumping on what Elizabeth said about TX power - definitely max that out first if you haven't, it genuinely changes how the signal floor behaves at range and gives your filtering algo more to work with before things get totally swamped by noise. One other thing worth trying for your specific 500m/2km use case is logging RSSI variance alongside the median, because in my experience the signal gets *noisier* as you get further out, so high variance itself becomes a useful secondary indicator that you're at the far end of range rather than close in.

Mary's variance tip is underrated - I actually ended up using a combo of median RSSI *and* rolling variance as dual inputs into a simple threshold classifier on a project last year and it made the zone detection way more reliable than either signal alone. One thing I'd flag for your specific 500m vs 2km case though: make sure you're collecting training data across different times of day because foliage moisture and atmospheric conditions can shift your baseline RSSI enough that thresholds that work at noon can fall apart at dawn.

Jessica's point about time-of-day variation is something I'd really stress for a 500m vs 2km split specifically - those thresholds are far enough apart that you might *think* your classifier is solid from a quick afternoon test session and then get burned when morning dew or summer leaf-out shifts everything by 5-8dBm. Honestly if you can leave two nodes deployed overnight and just log continuously for 24h before you settle on any threshold values, do it, it's way less painful than debugging in the field later.

Solid thread - just wanted to +1 the 24h logging suggestion from Mary because that 5-8dBm shift from environmental conditions is not theoretical, I've seen it bite people on exactly this kind of zone-boundary use case where the two thresholds aren't that far apart in RSSI terms even if they feel far apart in meters. One thing I'd add: if you can manage it, try to capture a rain event in your training data too, wet foliage hits different than morning dew and depending on your deployment environment it could be the edge case that tanks your classifier.

Rain event data is such a good call James - I skipped that step on a similar project and spent an embarrassing amount of time wondering why my field results were garbage during a week of wet weather. If you really can't wait for rain, hosing down the vegetation around your test path isn't a perfect substitute but it'll at least stress-test your thresholds a bit before you commit to them.

Charles's hose trick is genuinely underrated as a quick sanity check - I've done the same thing and while it's not a perfect substitute it definitely surfaces threshold fragility before you're standing in a field in the rain wondering what went wrong. One thing I'd add to the rain conversation though: if you're in a wooded or mixed environment, wet bark on tree trunks can be a surprisingly significant absorber at BLE frequencies, not just the foliage itself, so factor that in when you're picking your test path geometry.

Sarah's point about wet bark is one I hadn't really thought about until it burned me - I had a test path that ran parallel to a dense treeline and the trunk absorption was way more of a factor than I expected once things got wet, basically shifted my 500m threshold enough that I was getting false "far" classifications on assets that were clearly still in the near zone. Definitely worth mapping out your path geometry relative to any significant vertical structures before you lock in thresholds, not just the foliage overhead.

David's point about vertical structure orientation is underrated - the difference between running a test path *parallel* to a treeline vs *perpendicular* to it can be surprisingly large in terms of how much trunk absorption you're integrating over the link budget, and most people only discover that after they've already committed to a deployment geometry. If you have any flexibility in where you site your nodes, even a 20-30 degree angle change relative to any dense treelines can meaningfully reduce how much that variable bites you in wet conditions.

Bit late to the thread but worth flagging that if you have any control over node placement, keeping them elevated even a meter or two off the ground makes a noticeable difference at the longer end of the range - ground reflections mess with RSSI consistency more than most people expect and it compounds fast once you're trying to hold a reliable threshold out past a few hundred meters. The MultiNav Pro+ BLE does fine for that "500m vs 2km" kind of coarse classification but only if your variance is tight enough to trust, and height is one of the cheaper ways to buy yourself some margin there.

Richard's point on elevation is real - I spent way too long chasing RSSI variance on a deployment before I realized both nodes were basically sitting at grass level and getting hammered by ground multipath, raising them to roughly chest height on simple pole mounts tightened things up enough that my 500m/2km classification went from "maybe useful" to actually reliable. Even a cheap bit of conduit and a ground stake goes a long way if permanent mounting isn't an option.

Echoing the elevation advice hard - one thing I'd add is that if you're doing your threshold calibration walks, do them with the nodes already mounted at whatever height you're planning to use in deployment, because the numbers you get at ground level won't translate and you'll end up re-doing the whole thing anyway, learned that the annoying way on a similar asset tracking build last year.

Calibration height matching is so easy to skip and so painful to discover later - another thing worth doing while you're at it is running your calibration walks at a few different times of day if you can, because if there's any vegetation in the path even light dew in the morning can shift your RSSI baseline by a few dB compared to a dry afternoon and that'll mess with your thresholds if you only sampled once.