Articnoise
Headphoneus Supremus
Ok, a few points.
Break-in - about 10 minutes.
Well yes that was to get your attention, and my honest answer is I do not know for sure, except that brain break-in is by far the biggest factor. In October I did an update to the code, getting ready for final production. This update had a SQ change, and it was not small.
I did the change on a unit that had at least 10,000 hours on it, so it was well and truly broken in. But the SQ change from the code took several months for the feeling that SQ was improving to stop, so brain break-in (getting used to the sound and I suspect the brain learning new ways to process the SQ) is very real - at least to me.
I actually have a brand new unit coming next week, so I can test for how much is down to hardware changes.
FPGA - yes its soldered in - and I have to do this. FPGA sockets creates too much inductance, and so decoupling becomes a major issue. At best using sockets it will just sound and measure badly, at worse it will simply not function. But Moore's law is definitely running out of steam. The 7 series FPGA's were "launched" in Sep 2011, and the devices were only really available last year. The ultrascale parts are not suitable for audio, so there are no better FPGA's on the horizon. But lets say Xilinx have a wonder FPGA available tomorrow in quantity, then I can't just drop what I am doing and re-design Dave. It is not a simple process of dropping a new chip in - a replacement to Dave (assuming new display too) would be a one man year of work for me, and with all the other projects happening, this will be many years away.
Code updates - my background in chip design forces you to make hardware that is right first time, which means lots and lots of testing before tape-out. And I treat FPGA design as hardware, not software. But you are limited by two things - knowledge and the FPGA. With the vast resources available in Dave's FPGA, its knowledge that is the limiter. I suspect I will learn a lot from the Davina project, so if this relates to a SQ improvement for a DAC I will put it in Dave. I will only do a code update if we are talking about a significant and worthwhile change in SQ.
Chord's Korean distributor asked for a simple block diagram for Dave so here goes:
And the different code for DSD+ :
Rob
The best way to evaluate the change made by burn in (break-in) is to compare one DAVE that have been used for several hundreds of hours to one that is totally new. By comparing two units side by side we can remove things like brain burn in and audio memory. And now to my point. Who has access to both a well burned in DAVE as a new one (hint hint)?
