Accelerating Satellite Communications

At around $0.15/MB transmitted, satellite communication is prohibitive. At that rate, a full-length feature film in 4K resolution would cost a whopping $15K to transmit. No wonder we prefer fiber optics to microwaves.

At Compression.ai, we develop data management solutions that not only compress data streams, but also accelerate applications that use them, both upstream and downstream. We wanted to quantify how much we could speed up satellite communications, so we got our hands on a couple datasets from two operational satellites:
• Landsat 8
• Earth Observing-1 (EO-1)

The Landsat data consists of a single scene of 11 bands, while the EO-1 data has 145 bands.
Both have a spatial resolution of 30 meters.

We ran these datasets through both our lossless and near-lossless compression engines, which are compliant with the latest Consultative Committee for Space Data Systems (CCSDS) standards.

Lossless Compression:
We compared our lossless compression with JPEG2000, the go-to format used today for lossless 16-bit compression.
We found that Compression.ai produced files 15-25% smaller than JPEG2000, with the added advantage of being compliant with CCSDS 123.

The results obtained are shown in the table below:

Near-lossless Compression:
Our near-lossless implementation looks for correlation between the different bands, and predicts a band’s values based on those of the other correlated bands. While fully compliant with CCSDS 123, data compressed with Compression.ai’s near-lossless compression was only ~12% the size of the original (~82% compression).

Quality metrics for the near-lossless compressed data are shown below:

So how do these reductions in file sizes impact the cost of space missions? To answer that question, we decided to look at the operational costs of Landsat 8 specifically relating to data capture: namely storage and transmission costs. Since all Landsat 8 data is freely available on AWS, published S3 costs are an accurate representation of what cloud storage is currently costing the Landsat 8 program. Transmission costs were estimated using Amazon’s Ground Station offering, which allows program managers to save up to 80% of the cost of operating a dedicated ground station.

Extrapolating from a single scene from Landsat 8, we found that:

• Lossless compression using Compression.ai would save the Landsat 8 program assuming a 15-year lifetime:

○ $2.6M in storage costs
○ $2.1M in transmission costs at 260.92 Mbps
○ 2.2 hours daily in transmission time

• Near-Lossless compression using Compression.ai would save the Landsat 8 program assuming a 15-year lifetime:

○ $4.3M in storage costs
○ $3.6M in transmission costs at 260.92 Mbps
○ 3.6 hours daily in transmission time

With FPGA implementations readily available, Compression.ai algorithms are well positioned to save future space observation missions millions of dollars in operational costs. With the expected lifetime of satellites increasing year after year, the importance of reducing operational costs will only increase as newer satellites are put into orbit.

Reach out here for more information on how Compression.ai can help accelerate transmission and downstream applications.