Assisted GPS vs Cold Start on Trackers

Illustration

A cold start has no usable satellite orbit data, so the receiver must collect it from the sky. Assistance supplies orbit data from a download or from the receiver's own prediction. One u-blox datasheet lists a 28 s cold start against 1 s for a hot start or AssistNow Online, in lab conditions. Verify on your board.

Every fix attempt costs battery. How long the module takes to reach its first fix depends on how much valid data it holds when it starts. This guide uses one public datasheet as an example to show how cold and hot starts differ, what assistance changes, and which numbers you must measure on your own hardware. We use this module only as an example.This article contains no Duobay price and no measured result from us.

How do cold, warm and hot starts differ?

A cold start begins with little or no valid orbit data, so the receiver has to collect it from the satellites. A hot start begins with valid data and can fix quickly. A warm start sits between the two. The datasheet gives numbers for cold and hot, and for three assistance methods. It gives no warm start figure, so a warm figure is something to measure on your board.

The datasheet has other constellation columns with different values. We show one column as an example. Warm start: vendor did not publish in this table, so measure it on your board.

What do the assistance methods change?

Assistance gives the receiver orbit data it would otherwise have to collect from the sky. It can come from a download or from the receiver's own prediction. The older u-blox user guide compares three types by how long the data stays valid, how often you load it and how large it is. The datasheet says the legacy Online and Offline services are replaced by Live Orbits and Predictive Orbits, so read this table as the older generation (datasheet revision history, as of October 31, 2026).

The guide describes Autonomous mode as the receiver generating, without any host interaction or online connection, an orbit representation that stays usable much longer than the broadcast data was intended for. That is why a tracker can skip downloads for later starts. The guide is an M8-era advance-information document, so confirm the current service names and limits for your module in the current vendor documentation.

How do you compute the saving per day?

Per attempt, the datasheet lab values give 28 s minus 1 s, which is a 27 s saving between a cold start and a hot start or AssistNow Online. That is a lab operating point. For your product, build the budget from your own variables.

time_saved_per_attempt   = t_cold − t_assist                  (seconds)

energy_saved_per_day     = F × (t_cold × P_acq − t_assist × P_acq − E_assist)
                         = F × (P_acq × (t_cold − t_assist) − E_assist)

units: F in attempts per day, P_acq in watts, t in seconds,
       E_assist in joules, energy in joules per day

Check the algebra on the lab times: with t_cold = 28 s and t_assist = 1 s, the time term is P_acq × 27 s per attempt. If E_assist is larger than that, assistance costs more energy than it saves on that attempt. The datasheet does not publish E_assist, and it depends on the connection you use, so it is yours to measure. Run the formula once per start type your tracker really begins from, using the data-age rules in the table above to decide which one applies.

Inputs, and which are yours to replace

What are the common traps?

These follow from the datasheet footnotes and the guide. They are engineering reasoning, and none describes an incident at any company.

• Treating the lab table as your field result. The datasheet figures are commanded starts with all satellites at -130 dBm at room temperature. A tracker in a coat pocket or under a car seat sees different signals, and the vendor did not publish a field figure. This follows from the stated test conditions.
• Counting only the downloaded data age. The Online-style data is valid for 2 to 4 hours in the older guide. A tracker that sleeps longer than that between fixes must download again, so the saving depends on the report interval you choose. This is arithmetic on the validity window.
• Forgetting the download cost. The datasheet says the Online figure depends on the speed and latency of the aiding link. If the radio must wake and connect to fetch the data, the time and energy of that connection belong in the same budget as the fix. Whether the net result is a gain is not published and needs a measurement on your board.
• Reading the Offline row without its storage note. The datasheet notes that external memory may be required for Offline data. A design with no spare flash may not be able to use it. Check this against your bill of materials early.
• Using an old guide for a new module. The user guide cited here is an M8-era advance-information document. The current datasheet says the legacy Online and Offline services are replaced by Live Orbits and Predictive Orbits. Use the guide to understand the idea, and use the current documents for your module before you write firmware.
• Assuming a warm figure. The datasheet table has no warm start row. Do not fill the gap with a number from another source. Measure it.

What should you measure on your own board?

No datasheet gives start-time distributions or energy per fix for your antenna and your enclosure. We have not run these measurements for this article, and no result appears here. Use the table as a test plan and a log template.

Set each pass line according to your product requirements, and record the basis for it. Write it down before you run the test.

What should you ask before you choose?

• Which start type does my tracker actually begin each fix from, given how long it sleeps between reports?
• Does my report interval fit inside the validity window of the assistance method I plan to use?
• Does my design have a data connection available at the moment I need assistance, and what does waking it cost?
• Does the method I want need external memory, and is that memory in the bill of materials?
• Which service names and documents apply to my module today, and on what date did I read them?
• Where do I store the last known position and time across power loss, and does the module keep its own data in backup power?
• What is the longest fix attempt I will allow, and what does the product do when it is reached?
• Which numbers in my power budget are measured on my board, and which are still datasheet values?

This article leaves out the 3GPP positioning protocols, because we have not checked their original text for this guide. Read the 3GPP specifications directly if your design uses them.

If you want a plan for positioning and battery on your tracker, tell us about your tracker. You can also read about our LTE and GNSS tracker development.

Sources

All pages accessed October 31, 2026. Vendor documents are revised, so re-read them for your module before you decide.

• u-blox, MAX-M10S data sheet (UBX-20035208, R08): https://content.u-blox.com/sites/default/files/MAX-M10S_DataSheet_UBX-20035208.pdf (accessed 2026-10-31)
• u-blox, Multi-GNSS assistance user guide (UBX-13004360, R05): https://content.u-blox.com/sites/default/files/products/documents/MultiGNSS-Assistance_UserGuide_%28UBX-13004360%29.pdf (accessed 2026-10-31)