Field Notes from a Ground Station: Who Feels Satellite Latency the Most

Archi4 min read

satelliteslatencyruralbroadband

Contents
  1. Field Notes from a Ground Station: Who Feels Satellite Latency the Most
  2. The test bench
  3. The clinic that cannot wait
  4. The logistics firm with a 3-minute window
  5. Who is not affected much
  6. What the numbers say
  7. The policy gap
  8. A short list for anyone testing a link

Field Notes from a Ground Station: Who Feels Satellite Latency the Most#

We spent two weeks testing a consumer satellite internet terminal from a rented warehouse on the edge of a rural county. The setup: a dish on the roof, a $600 router, a laptop, and a stopwatch. The point was not to review the hardware. It was to find out who actually pays when the sky link wobbles.

The answer surprised us. Gamers complain loudest. But the people who lose money are the ones running small businesses and clinics where a 700-millisecond delay turns a routine task into a failed one.

The test bench#

We ran three checks every hour: a 100-packet ping to a public server, a 5 MB file upload, and a video call to a fixed endpoint. We logged 312 hours of data. Median round-trip latency was 38 ms. The 95th percentile was 410 ms. During rain, the link dropped for an average of 47 seconds per hour. None of this is unusual for low Earth orbit service.

What matters is not the median. It is the tail. A 410 ms delay once an hour is invisible to a person browsing the web. It is fatal to a point-of-sale terminal waiting on a card authorization.

The clinic that cannot wait#

A nurse practitioner at a rural clinic 40 miles from our test site described the problem in concrete terms. Her clinic uses a satellite link for telehealth visits and for sending lab orders to a hospital system 90 minutes away by road. When latency spikes, the video call freezes. The patient on the other end, often an older farmer with limited transportation, has to reschedule. That is a wasted day for the patient and a lost slot for the clinic.

The clinic does not have a fiber option. The local cable provider stopped building 12 miles short of the town. The satellite link is the only broadband available. The nurse said the clinic budgets for two no-show appointments per week due to connectivity. At $85 per visit, that is $8,840 a year. That is one month of a nurse's salary.

The logistics firm with a 3-minute window#

A small freight broker we spoke to runs dispatch for 14 trucks. Drivers use satellite links when they are outside cellular coverage. The broker's software requires a driver to acknowledge a load assignment within 3 minutes. When a driver's link lags, the assignment expires. The broker loses the load. The driver loses the miles.

The broker estimated that 6 percent of assignments fail this way. On a $2,200 average load, that is $132 per failed assignment. Over a year, with roughly 1,800 assignments, the loss is about $14,000. That is more than the firm spends on satellite hardware and service combined.

Who is not affected much#

A remote software developer we interviewed said the link is fine for his work. He pushes code in the morning, pulls in the afternoon, and does not notice the delays. A streaming household reported no problems. These users have buffers: Git can retry, video can buffer. The people who get hurt are the ones with real-time constraints and no fallback.

What the numbers say#

Our logs show a clear pattern. The latency tail is worst between 7 and 9 p.m., when the satellite constellation is handing off between ground stations and the local network is busy. That is exactly when clinics close and when freight brokers push evening dispatch. The timing is not cruel by design. It is cruel by coincidence.

We also tested a workaround: a local caching server at the clinic. It cut the perceived delay for lab orders by half. The cost was $1,200 for hardware and a day of setup. The clinic has not bought one. The budget cycle is annual, and the request is sitting in a queue.

The policy gap#

Subsidy programs for rural broadband count households, not clinics or small firms. A clinic that serves 900 patients a year counts as one subscriber. A freight broker with 14 trucks counts as one. The latency problem does not show up in the coverage maps that regulators use to allocate money. It shows up in the appointment book and the dispatch log.

, Log the 95th percentile, not the average. The tail is where the damage is., Test during the evening peak, not at 10 a.m., Ask the clinic or the broker what a failed transaction costs. That number is the real benchmark., Check whether a local cache or a retry rule can absorb the delay. Sometimes a $1,200 box beats a $12,000 dish upgrade.

The sky link works. It works well enough for most people. The ones who feel it most are the ones with the least room to wait.

More articles