Rob Watts
Member of the Trade: Chord Electronics
- Joined
- Apr 1, 2014
- Posts
- 3,144
- Likes
- 12,403
Hugo is one year old today.
Sunday 20th Oct I did the first listening test, (with the fully tested and de-bugged code) on Hugo, and its fair to say I had my socks blown off. After listening for 30 minutes, I rushed a quick email at 7.59 am to Matt at Chord, saying "it sounds fantastic". That email didn't really express my excitement as to how much better it was, as I was in a hurry, and wanted to listen some more.
It was actually a profound shock; I knew it was going to sound better, but not the change in musicality - I was hearing things I had never heard from audio before, in particular the timbre qualities of instruments, and the way it resolved the starting and stopping of notes, and I was not sure where this performance was coming from.
One year on, and I now know the improved musicality was principally due to the re-writing of the code from the bottom up, plus the benefit in have much more space from the Xilinx Spartan-6 FPGA, allowing me to do things I always wanted to do. The FPGA design for earlier DAC's was a hotch potch of Verilog (this is code that is used to create the digital design) and schematics with logic gates and macros. Some of the schematics were sourced from 20 years ago. So all this was re-written and tested into Verilog, over a 6 year period, and Hugo just happened to be the first product that had this major upgrade completed.
And what have I learned? That the brain/ear is capable of amazing things, and extremely small errors (errors way below the threshold of audibility) are capable of upsetting the brain's processing of sound and thus become very significant. That you can't make assumptions as to what is an audible error or not, so I find that I now do much more listening tests in a new design. But the most profound lesson has been about musicality itself. Now my design process has been about making things sound more transparent, as I felt this was the only way of making true progress. I don't try to create a particular sound; I want the sound of nothing at all, and I do this by discovering errors that have a subjective consequence, and then find a way of reducing these errors. Now my goal is to make audio sound more musical - in the same way that live un-amplified acoustic instruments sound musical. But Hugo is the first product I have designed in my 30 odd year career where the strategy of making it as transparent as possible has paid off enormously with musicality. Hugo is extremely transparent; but it has this bizarre ability to communicate about what is musically good about a recording. I find I am far less fussy about the provenance of a recording - 1930's mono recordings can sound very musical for example, whereas before I would hear what was wrong, and that would make it not listenable. With Hugo, you hear exactly what is wrong, but you also hear exactly what is right too.
So happy birthday Hugo!
Sunday 20th Oct I did the first listening test, (with the fully tested and de-bugged code) on Hugo, and its fair to say I had my socks blown off. After listening for 30 minutes, I rushed a quick email at 7.59 am to Matt at Chord, saying "it sounds fantastic". That email didn't really express my excitement as to how much better it was, as I was in a hurry, and wanted to listen some more.
It was actually a profound shock; I knew it was going to sound better, but not the change in musicality - I was hearing things I had never heard from audio before, in particular the timbre qualities of instruments, and the way it resolved the starting and stopping of notes, and I was not sure where this performance was coming from.
One year on, and I now know the improved musicality was principally due to the re-writing of the code from the bottom up, plus the benefit in have much more space from the Xilinx Spartan-6 FPGA, allowing me to do things I always wanted to do. The FPGA design for earlier DAC's was a hotch potch of Verilog (this is code that is used to create the digital design) and schematics with logic gates and macros. Some of the schematics were sourced from 20 years ago. So all this was re-written and tested into Verilog, over a 6 year period, and Hugo just happened to be the first product that had this major upgrade completed.
And what have I learned? That the brain/ear is capable of amazing things, and extremely small errors (errors way below the threshold of audibility) are capable of upsetting the brain's processing of sound and thus become very significant. That you can't make assumptions as to what is an audible error or not, so I find that I now do much more listening tests in a new design. But the most profound lesson has been about musicality itself. Now my design process has been about making things sound more transparent, as I felt this was the only way of making true progress. I don't try to create a particular sound; I want the sound of nothing at all, and I do this by discovering errors that have a subjective consequence, and then find a way of reducing these errors. Now my goal is to make audio sound more musical - in the same way that live un-amplified acoustic instruments sound musical. But Hugo is the first product I have designed in my 30 odd year career where the strategy of making it as transparent as possible has paid off enormously with musicality. Hugo is extremely transparent; but it has this bizarre ability to communicate about what is musically good about a recording. I find I am far less fussy about the provenance of a recording - 1930's mono recordings can sound very musical for example, whereas before I would hear what was wrong, and that would make it not listenable. With Hugo, you hear exactly what is wrong, but you also hear exactly what is right too.
So happy birthday Hugo!
