Your illustration of extra robustness is the presence of two transformers and two division walls. Can you show us the measurements for this device? We should be able to see the negative influence of the transformers on the noise measurements. I would argue that our approach to keep the power supply outside and away from the DAC delivers much superior signal to noise ratio. Here are the e20 measurements: http://www.exasound.com/e20/e20Measurements.aspx
You are comparing an old exaSound DAC with a current model. Your photo brings back the sweet memories for the e20, It was our second DAC, released on September 17, 2012. At that time it had support for DSD128. On February 13, 2013 with the MK II release it became the first DAC available for purchase on the audiophile market to support quad DSD at 11.2896MHz. On July 12, 2013, with the MK III release the e20 became the first DAC to support DSD at 12.288 MHz. We've managed twice to release ground-breaking technology before Merging and Mytek. There was a third world's first - on March 22, 2014 with driver update we were the first to release DSD256 on OS X with custom Core Audio and ASIO drivers.
Looking at the e20 and the heavy-metal device, you can see some other differences. In the small form factor we've managed to pack a top-class headphone amplifier. See the measurements.
Look in front of the big FPGA chip, the two small chips are the Galvanic isolators. In addition the air gap on the PCB prevents dust particles collected on the PCB surface to compromise the isolation. Recently It was mentioned on this forum that Galvanic isolation is the Holy Grail of Computer Audio. The e20 was worthy of this definition back in 2012. I would be curious to find how the newly introduced Galvanic Isolation technologies are better than the one used on the e20. The e20 and all its successors always had galvanic protection without the use of external USB gadgets.
You can see the three low-jitter clocks (two of them next to the FPGA are made to order for us). They support the asynchronous operation. The clocks, together with the FIFO buffer (the chip next to the FPGA) are tightly integrated with the drivers to implement the true asynchronous operation that allows us to do error correction over the USB connection. With all this in place, our devices are quite insensitive to the "audio" qualities of your computer. It doesn't make much difference if you use a spinning hard drive or SSD, all-from-memory playback or barebones Foobar, with audio-optimizer or not. If your computer can provide a steady USB stream, the sonic fidelity will be almost indistinguishable between a 10 years old HP desktop and custom-built computer hardware.
In our opinion the challenges that computers bring to audio reproduction are best solved by specialized hardware outside the computer. Trying to fix computer timing accuracy and to reduce computer noise by tweaking the operating system is like taking painkillers. The root cause is not addressed. Having custom hardware architecture built around the DAC chip to bring the best of it and drivers that are tightly integrated with the hardware and the firmware makes our devices unique.
The 6 black boxes (there are two more behind the front PCB next to the headphone connector) are the relays that cut off the line outputs and the headphone outputs in case of emergency. For best bass extension down to 0 Hz all our DACs are DC-coupled. This can be dangerous if the DAC is directly connected to a DC-coupled power amp. For example if you accidentally disconnect the DAC power cable the transients can destroy your speakers. Our firmware will detect the disaster and will have enough operation time before critical power loss to cut off all outputs for graceful, pop-free, bang-free shutdown.
To give an example of the impact of the quality of the USB connection, our old devices used to be able to play without interruption when the USB cable is quickly disconnected and reconnected. This is possible because of the FIFO buffer. Newer exaSound DACs and the new generation drivers don't allow this anymore. We deploy the same shutdown measures when the USB cable is accidentally disconnected. (Don't try this at home.)
Regarding your assessment of the output stage, the proof is in the sound and in the measurements. Ours is true-balanced all the way from the DAC output to the XLR connector. The unbalanced output comes last with an extra buffer.
I can agree with your assertion that our devices look less heavy compared to other DACs. This is because most of the development budget goes into new development. The e20 PCB that you bring as an example is made with manual placement here in Canada. We can't compete on cost with the Chinese manufacturers, we can't use transformers with silver wire. Our approach is to invest in developing what we believe is the best technology to deliver natural, detailed and dynamic sound.