Read noise is noise added every time the sensor is read out.
With DSLRs, this happens once per sub-exposure.
It is fixed per frame
It does not average down within a single sub
It only reduces by stacking many frames
This is why how long each sub is matters far more than beginners are told.
Every sub pays a read-noise tax.
If your sub contains very little sky signal, then:
Read noise dominates the frame
Most of your total integration is wasted fighting electronics, not sky
Stacking thousands of short subs is hugely inefficient
This is the read noise penalty.
DSLRs are especially vulnerable because:
Read noise is much higher than modern astro CMOS
Telescopes concentrate light → small field of view per pixel
Sky background rises slowly, especially at at f/5–f/7
Result:
👉 Short subs can stay read-noise dominated for 30-60s or more.
Let’s define:
SN = Sky Noise (from sky background photons)
RN = Read Noise (from electronics)
Read-noise limited
SN (Sky Noise) < RN (Read noise)
Horrible efficiency (100% - 600% total integration penalty).
Transition
SN ≈ RN
Still heavy penalty (~100% integration penalty)
Sky-noise limited
SN ≫ RN
As SN 3-5x RN, Read noise becomes irrelevant
👉 Your goal is always: SN ≫ RN
Because read noise is per sub, not per hour.
Example (simplified):
RN = 5 e⁻
Short sub: SN = 3 e⁻ → RN dominates
Long sub: SN = 20 e⁻ → RN negligible
Once SN > ~3-5× RN, the read-noise penalty collapses.
At SN ≥ 5× RN, read noise is effectively gone.
Read noise stacks as: √N × RN
Sky signal stacks as: N × signal
So if each sub has weak signal:
You stack noise faster than signal
If each sub has strong signal:
Signal overwhelms read noise early
Stacking becomes efficient
This is why 60 × 60s is NOT the same as 6 × 600s on a DSLR.
For DSLRs:
The histogram peak must be clearly off the left wall
Barely detached = still read-noise dominated
Well separated = sky-noise limited
This usually means:
120–300s at >350mm focal length f/5 in UK skies
But the longer the subs, the worse the thermal noise.
Shorter subs are throwing data away. But longer subs suffer thermal noise and banding.
DSLRs have a low QE, high RN and should be used <300mm FL, this increase in sky noise per pixel compensates.
DSLRs can become photon vacuums at shorter focal lengths. 150mm = 4x larger SNR than 300mm. Hence, camera lenses are perfect.
Camera lenses are:
Faster (f/2–f/3.5)
Wide field → high sky signal per pixel = higher signal per pixel per sub
Telescopes are often:
Slower (f/5–f/7)
Smaller pixel footprint on sky
Much slower sky background accumulation
So a DSLR at 200–600mm absolutely requires long subs.
A common myth:
“I’ll just stack more short subs.”
This costs a huge read noise integration tax.
If each sub is read-noise dominated:
You are stacking garbage more efficiently
Integration time explodes
Detail never emerges
Long subs reduce the penalty before stacking even starts.
Target
Minimum sub length
400-600mm FL 180–300s subs. Going deeper risks oversampling.
Guiding is not optional — it is what unlocks efficiency.
Read noise is paid per sub
DSLRs have high read noise
Telescopes produce slow sky signal
Short subs = massive efficiency loss
Long subs push you into SN ≫ RN, where stacking actually works
This is why deep-sky DSLR imaging without long subs is fighting physics.
But remember:
DSLRs fail when treated like cooled cameras.They excel when arcsec/pixel is higher, which means a short focal length.
They succeed when treated like hot sensors with finite thermal budgets:
Use a sub-length where Sky Noise = ~ 2-3x RN, then pay the RN integration tax, rather than letting Dark Noise swamp signal.