CH Precision DIG IN HD Board Upgrade – part 2
In-depth technical description
As I was curious as to why the decision was made to move from three filter settings to one, as well as what precisely sets the new DIG IN HD board apart from the non-HD one, I asked mechanical and software design engineer Loris Stehle for more info. He was more than happy to explain and his entire response can be read below.
DIG IN HD Board differences
There are 3 main differences introduced by the digital in HD board, together with the D1’s digital out HD board
The physical link (connectors and cables) between a D1 and a C1 (or an I1) has been improved
The new CH Link HD connectors and cables are better shielded than the older ones. The characteristic impedance of this new link is also better controlled and more stable across frequencies. So this link is extremely well suited for high bandwidth signals. Bear in mind that we are not only talking about “slow” audio signals but audio master clocks that are 24MHz square waves, whose harmonics extend close to the GHz range (several hundreds of MHz)!
The electrical conditioning of these D1 to C1 (or I1) signals has been improved
To reduce HF noise due to high frequency digital signal pulsing, without impairing jitter performance, not only does our CH Link HD carries all these signals in low-voltage differential signaling (LVDS) form (as the older CH Link already did), but the electrical specifications of the driving stage (rise time, current limiting, output impedance) have also been fine-tuned to obtain better results.
The synchronous upsampling technique to bring any incoming sampling rate audio data to 705.6kHz or 768kHz has evolved
A C1 equipped with the older Digital input board provided the choice of three different upsampling settings: minimum phase, linear phase apodizing and linear phase sharp. But in reality, all 3 settings worked on the same principle to synchronously upsample the incoming signal (by a 2, 4, 8 or 16 ratio): zero pad the original signal (insert 1, 3, 7 or 15 zero-valued samples between each input samples), and low-pass filter this signal. The name of these upsampling settings refers to the type of low-pass filter used for this second operation. We noticed that despite having a slightly non-constant group delay, the minimum phase upsampling filter provided the most natural sound of all 3 filters. But the other 2 may sound better to some users, depending on the type of music. That’s why we originally kept all 3 filters available, with the minimum phase being the default setting. But since the C1 was first introduced, we did some research on completely different synchronous upsampling techniques. A very interesting one is the spline interpolation: instead of zero-padding the original signal and low-pass filter it (which not only creates new sample values but also modifies the original ones), we compute polynomial curves that pass through the original points. All these polynomial curves temporally connect to each other at their boundaries, not only point-to-point, but their slopes are also identical when they connect (we make sure the curve and their derivative up to a certain order match at their boundaries). We tried many polynomial orders, and length of overlap, and compared them by ear before selecting the one that we implemented in the Digital In HD board of the C1. The whole team agreed that this upsampling technique, with the carefully selected parameters, gathered the best of all previous upsampling filters: the flow and naturalness of the minimum phase, with the speed and resolution of the linear phase. That’s why we decided to keep the spline as our champion upsampling algorithm alone. As this upsampling technique is the evolution of our former HiD upsampling, we called our spline interpolation “Polynomial Equation to Enhance Resolution” (PEtER).
All info by Loris Stehle, mechanical and software design engineer at CH Precision
Part 1 – Part 2