Dilution of Precision

Four friends, one terrible spot to stand

Suppose you are blindfolded in a field and four friends shout your distance to them so you can figure out where you are. Great plan — unless all four friends are standing shoulder to shoulder in the same corner. Now every one of them is telling you almost the same thing ("you're about forty meters that-a-way"), and the tiniest slip in any shout sends your guess sliding wildly along that one direction. You are precise about how far, and clueless about which way.

Spread those same four friends around the whole field, though, and each shout pins a genuinely different direction. The same small slips now pull against each other and cancel out. Identical measurement quality, wildly different confidence — and the only thing that changed was where everyone was standing.

That, in one field, is dilution of precision. The satellites are your friends, and their arrangement in the sky either sharpens your fix or smears it.

The idea, before any mathematics

In pseudorange you saw that a fix comes from combining measurements along the directions to the satellites. DOP asks a follow-up question that has nothing to do with how good the measurements are: given where the satellites sit, how much does a bit of measurement error get amplified into position error?

The rule of thumb is delightfully visual:

• Satellites spread across the sky → their directions disagree strongly → errors fight each other → small position uncertainty → low DOP (good).
• Satellites bunched together → their directions nearly agree → errors reinforce along one axis → large, lopsided uncertainty → high DOP (bad).

And here is the part that surprises people: only the direction to each satellite matters, not the distance. A satellite twice as far away in the same direction gives you the same geometric information. DOP is a story about angles, not ranges.

Seeing it

Picture the uncertainty as an ellipse drawn around your true position. Good geometry squeezes that ellipse into a small, round dot — you are equally sure in every direction. Bad geometry stretches it into a long, nervous cigar pointing along the direction the satellites failed to cover. The shape of your doubt is literally the shadow of the satellite geometry.

Try it yourself

The green receiver sits at the center. Five blue satellites surround it, each casting a dashed line of sight. The orange ellipse is the shape of your position uncertainty implied by the current geometry, and the panel reports HDOP, TDOP, and GDOP live.

Start from the tidy default spread, then drag the satellites into one corner and watch the ellipse bloat and lean. Try the opposite too: spread them as evenly as you can and chase the DOP as low as it will go. And test the claim that distance is irrelevant — slide one satellite straight toward or away from the receiver and notice the numbers barely flinch.

Geometry qualityExcellent

HDOP (horizontal)

0.89

TDOP (clock)

0.45

GDOP (geometric)

1.00

Satellites

5

Status

ok

Reset to good geometry

Drag the blue satellites. Spread them evenly around the receiver and the orange uncertainty ellipse tightens into a small circle (low DOP). Herd them into one corner and it stretches into a long, nervous cigar (high DOP). Only the direction to each satellite matters — sliding one straight in or out barely moves the numbers.

Play until the punchline lands: you never touched the measurement quality. Every change in that ellipse came from geometry alone.

Now the mathematics

Recall the linearized pseudorange model from the previous concept. Stacking one row per satellite gives the geometry matrix $H$, whose row for satellite $i$ is its unit line of sight plus the clock term:

$$H_i = \left[ \; e_{ix} \quad e_{iy} \quad 1 \; \right]$$

If each measurement has error variance $\sigma^2$, the covariance of the least-squares solution is:

$$\mathrm{Cov} = \sigma^2 \, (H^\mathsf{T} H)^{-1}$$

Every DOP value is a square root of a piece of the dimensionless matrix $Q = (H^\mathsf{T} H)^{-1}$:

• $\mathrm{HDOP} = \sqrt{Q_{11} + Q_{22}}$ — horizontal position
• $\mathrm{TDOP} = \sqrt{Q_{33}}$ — the receiver clock
• $\mathrm{GDOP} = \sqrt{Q_{11} + Q_{22} + Q_{33}}$ — position and clock together

So the whole accuracy story factorizes cleanly:

$$\sigma_{\text{position}} = \mathrm{DOP} \cdot \sigma_{\text{measurement}}$$

Measurement quality ($\sigma_{\text{measurement}}$) is the receiver's and the signal's business. Geometry ($\mathrm{DOP}$) is the sky's business. When satellite directions bunch up, $H^\mathsf{T} H$ becomes nearly singular, its inverse blows up, and DOP with it — the mathematical echo of "every friend is shouting the same thing."

How this appears in engineering

Receivers report DOP constantly, and it drives real decisions. A navigation system may refuse to publish a fix when HDOP exceeds a threshold; a survey crew will schedule observations for windows when the satellite geometry is favorable. This is exactly why your phone struggles in an urban canyon: tall buildings mask whole swaths of sky, leaving only satellites peeking through a narrow strip overhead — a clustered geometry with ugly DOP. The receiver hasn't gotten worse; its view of the sky has. Adding satellites from other constellations (GPS and Galileo and BeiDou) helps mainly by refilling the sky with well-separated directions, which is a DOP argument dressed up as a marketing one.

Check your understanding

• Two receivers take equally noisy measurements. One reports HDOP 1.2, the other HDOP 9. Which fix do you trust, and by roughly what factor?
• In the interactive, why does sliding a satellite radially (toward or away from the receiver) barely change the DOP, while sliding it sideways changes it a lot?
• Why can adding a fifth satellite right next to an existing one fail to improve accuracy, while a fifth satellite on the opposite horizon helps enormously?
• If your measurement error were truly zero, what would the position error be under a dreadful DOP of 50? Why?

Where to go next

You now hold the three pillars of a satellite fix: trilateration turns distances into a position, pseudorange explains what the receiver truly measures and why the clock joins the unknowns, and dilution of precision explains why the arrangement of the satellites decides how much you can trust the answer. Together they are the geometric heart of how anything, anywhere, knows where it is.