The MiSeq's Cameras And The Economics of DNA Sequencers
I’m fascinated by how much scientific instrumentation costs. Why for example does a MiSeq cost so much more than a high end digital camera?
What does it say about our society, that we’ve decided to cost optimize cameras and mobile phones but not DNA sequencers?
Cheap1 DNA sequencers seem like they’d be an obvious benefit to society. But like most scientific equipment… for largely sociological reasons2 challenging.
However sometimes people suggest that it would be technically challenging to make something like a MiSeq for <$10000. I don’t believe this.
In the last post we looked at the excitation LEDs used in the MiSeq and found they cost <$100. Today let’s look at another core part of the MiSeq imaging system, the camera, and figure out how much that costs to make.
The Image Sensor
The MiSeq has two cameras. These appear to be identical and sit in nice milled enclosures. Looking at the sensor under a microscope shows that it’s marked as a IMX038. This is a consumer grade Sony CMOS image sensor that was used in Digital SLR cameras (notably the Nikon D5000). A 4288 x 2848, 13MP sensor.
The IMX038 would normal be sold as a color sensor. A Bayer filter (array of Red, Green and Blue) would be fabricated on top of the image sensor. As far as I can tell this device has no bayer filter. While this would need to be a custom order, all it means for Sony is skipping some fabrication steps, and it’s not uncommon to see monochrome versions of regular sensors.
The PCB and Interface
On the reverse of the PCB we find our old friend the Spartan 6, here it’s the LX45 variant. Xilinx FPGAs appear basically everywhere in scientific equipment. The Spartan 6s are relatively cheap FPGA often used in data acquisition. The FPGA is paired with what looks like 2Gb of DDR2 800MHz RAM.
The camera is interfaced to the compute module via USB. This uses an FX2 USB controller (a common choice). Illumina seem to love USB using it to interface sub-modules wherever they can.
BOM Cost
The Nikon D5000 cost $800 when new and as mentioned above, this used the same sensor we see here. The MiSeq camera is however likely significantly less complicated and capable than the D5000, which is able to capture, process and compress real time video. In contrast, all the MiSeq camera does it grab images and send them on to the compute for further processing.
If instead we look to the BOM cost of the MiSeq camera module we have:
Giving us an estimated total of $335. Let’s round it up to $400. For two cameras and with the LEDs this comes to $900 for the MiSeqs active optical components.
Unfortunately it’s not really the BOM cost that makes the MiSeq expensive and Digital Cameras cheap, but economies of scale. The MiSeq has probably sold ~10000 units total, millions of cameras are shipped every year.
Runs and instruments
Perceived market sizes.
What about compute, fluidics etc? Any thoughts about the cost of these?