Impact of confusion noise at 144 MHz and 6″ resolution. The blue crosses and circles show the noise measured from our stacked Stokes I residual images with and without image-plane deconvolution of sources that would be detected and deconvolved if the uv-data were imaged together. The orange circles show the measurements from our simulations that are fit with σ_IRMS = 0.9σ_VRMS + 5.7 × 10⁻⁴σ_VRMS² + 3.96. Here σ_IRMS is the measured Stokes I sensitivity (root mean square noise) and σ_VRMS is the sensitivity without confusion, both are in micro Janskys per beam. In red are the measurements from our stacked Stokes V maps. The y-axis shows the total noise measured from stacked DDFacet residual images whereas the x-axis shows the thermal noise estimated from the weights of the Stokes V images for datasets included in that stacked image. The red line shows where these are equal and the large black cross is the sensitivity of our final ELAIS-N1 6″ resolution image.