To speculate, there's 24/192 PCM and DSD64 DoP support for all non-USB inputs, Toslink included; driverless 24/96 for USB 1.0; and 24/384 PCM DSD128 DoP for USB 2.0. The volume control spans a 96dB range in 0.5dB steps. That's 192 discrete values with claimed ±0.1dB accuracy. This is from a resistor ladder and low-resistance analog switches to avoid mechanical parts. Full-scale output is 6Vrms unbalanced, 12Vrms balanced. The white Oled display packs 128x64 pixels. Weight is 3.5kg, dimensions are 26x25x6cm WxDxH (feet excluded). The universal IEC inlet accepts 100-240VAC because it fronts a low-noise SMPS. Normal operation consumes less than 20 watts, standby is a Brussels-happy 0.5 watts. As to galvanically isolated number crunching, the proprietary COS algorithm runs a 4'096-tap 32-bit filter array which upsamples all data to either 176.4kHz or 192kHz with a linear-phase FIR delay embedded on 3'648 MMAC/sec capable DSP. Incoming data enters a 1sec-buffer (unless defeated by rear toggle) before being processed by the algorithm. Clocked by a <1ps crystal oscillator with dedicated reclocker circuit, it is then spit out analog at the line and headfi outs. Phew. That was one data-packed paragraph. As usual, Maggie had just the right emoji for me.


Of course with audiophiles being audiophiles—nosy critters all—the previous paragraph still left holes relative to converter chip/s used; to the output stage and its devices (discrete, op-amps); to how many gain stages there were; and to the operational wattage ratings into our most common load impedances. The balanced headfi output suggested a possible front-to-back balanced circuit. The exclusively single-ended line outputs did not. Likely the volume control was a stereo not quad affair too. Off went another email to Maggie. I was curious what kind of emoji would accompany her response. Was I pushing my luck?

No emoji this time, just "some questions were too technical for me so I copied the tech's words to avoid errors. Feel free to ask if anything isn't clear. We went with the same PCM1792 converter chip as in the D1 but now only use one, not two. The headphone amp is an LME49600 from TI/NS which we picked for its super-low THD+N and capability to supply up to 250mA. There are two LME49600 per channel, one non-inverted, one inverted."
Load (Ω)
output voltage (Vrms)
THD+N (dB@1KHz)
SN/R
16
2.5
-104
-117
32
4
-104
-117
50
5
-104
-117
120
8
-105
-119
300
12
-104
-120


COS Engineering's location in Taipei; D1 without cover super-imposed.

"Gain is set at 3. Volume is dual mono. Into purely resistive loads, THD+N@1kHz driven balanced were shown in the table. The buffer option switch has the same purpose as it had in the D1." About that, I was hazy. What specific scenarios might benefit from a buffer defeat? Now I got this 'no problem' emoji. "It should be switched off for video playback to ensure synchronization." As for the rest, the H1 appeared to be a single-ended circuit from input to volume control after which op-amp outputs went balanced. That and the outsourced switch-mode power supply explained how COS could afford such nice case work and display yet come in at close to one forth the D1's sticker whose circuit is fully balanced, discrete and uses a linear power supply. As to their opamp, its notes refer to pin-selectable bandwidth modes, one a low-current 110MHz mode at 7.3mA, the other 180MHz at 13.2mA. Either has a steep nominal 2000V/μs slew rate. When I asked Maggie whether the new chassis would birth a stereo or mono class D amplifier, she responded with an enigmatic "top secret". Expecting a ninja icon if I pushed more, I stopped. This chassis simply seems tailormade for it.