Modulation Profiles and Link Rates

On NPR, the modulation profile is not just a speed setting. It picks the channel bandwidth, the TDMA frame length, the raw bitrate and — in practice — the entire character of the link. Changing the profile is not making the network "faster" or "slower" in some abstract way, it is changing how the network behaves.

PacketRF does not invent its own profiles. It selects one of the standard NPR profiles defined by F4HDK and applies the corresponding SI4463 configuration. That is done for the compatibility reasons. If you don't care for compatibility with the original NPR networks, you can invent your own profiles.

The standard profiles

There are two families: the 2GFSK profiles 11–14, and the 4GFSK profiles 20–24. Together they cover symbol rates from 50 kS/s to 500 kS/s and raw rates from 100 kbps to 1 Mbps.

Profile	Modulation	Symbol rate	Raw rate	Usable rate	RF bandwidth
`11`	2GFSK	100 kS/s	100 kbps	71 kbps	100 kHz
`12`	2GFSK	180 kS/s	180 kbps	120 kbps	200 kHz
`13`	2GFSK	300 kS/s	300 kbps	190 kbps	360 kHz
`14`	2GFSK	500 kS/s	500 kbps	300 kbps	1000 kHz
`20`	4GFSK	50 kS/s	100 kbps	68 kbps	100 kHz
`21`	4GFSK	100 kS/s	200 kbps	130 kbps	200 kHz
`22`	4GFSK	180 kS/s	360 kbps	220 kbps	360 kHz
`23`	4GFSK	300 kS/s	600 kbps	330 kbps	600 kHz
`24`	4GFSK	500 kS/s	1000 kbps	470 kbps	1000 kHz

The raw rate is what the radio is keying out; the usable rate is what actually arrives as IPv4 throughput once framing, TDMA overhead and typical scheduling are taken into account. For an operator the "usable rate" column is the more usefull one.

Picking a profile

The right profile is not automatically the highest one.

The higher profiles trade margin for speed. Shorter TDMA frames mean better latency and more raw throughput on a clean link, but they also demand a cleaner path: better signal-to-noise, better filtering at the modem and at the antenna, less RF sloppiness in general. A profile-24 link on a marginal/noisy path will not be a fast or a good link.

The lower profiles trade speed for tolerance. They carry less, but they carry it through worse paths, with more margin against weak signal and multipath, and with more forgiving behavior when something on the path is not quite right.

The practical rule is: pick the highest profile that still behaves well on your real path, not on paper. A few examples:

A long rural backbone link with a weak path: profile 20 or 21 is often the right answer, even though 24 looks better in the table.
A short, strong, well-filtered point-to-point link between two known sites: 23 or 24 will feel noticeably better than the lower profiles.
A network that mostly carries telemetry, control traffic and small IP packets: a moderate profile is almost always the more sensible choice than chasing the largest number.
Mobile use: F4HDK's recommendation is to stay on 11, 20 or 21, because higher symbol rates suffer badly from multipath and the protocol's FEC is not strong enough to make up for it.

Profile and latency

Because the profile sets the TDMA frame length, it also sets the latency floor. A profile-14 link cannot have a fast-mode RTT below 130 ms even on a perfect path, simply because that is the shortest schedule cycle the modulation allows. A profile-20 link floors at around 560 ms. The exact numbers, including slow-mode behavior, are listed in Latency and Ping Expectations.

This is also why a profile change can completely alter how a link feels even when BER does not move.

What to test on a new profile

A profile is not really tested until you have tried more than one packet size. The default ICMP echo of 56 bytes will not tell you much about how the link handles an actual application. At minimum:

56, 252, 253, 300, 800, 1200.

The jump from 252 to 253 is the important one. That is the first NPR segmentation boundary, where one IP packet stops fitting into one radio frame. Crossing it roughly doubles per-packet airtime, and a link that pings beautifully at 252 and badly at 253 is doing exactly what NPR says it should do.

Bottom line

Think of NPR profiles as link operating modes, not as marketing speeds. Wider and faster is great when the path allows it. On a marginal link it is also the fastest way to make the network worse.