(10-06-2017 00:59)Black_Jack писал(а):  Да код для FPGA писать можно и студента посадить. Там в том ЦАПе самое ценное это математика, а в ней самое главное - как сделаны цифровые фильтра. Это самое охраняемый секрет. (Я как-то спросил у Тексасов в саппорте по поводу типа интерполяции в их ASRC. Так вот - ни "гугу", хотя на менее каверзные вопросы - ноу проблем)

Так, PS не дуже ділиться подробицями по програмуванню. Ось найшов що коротко пише Richard, розробник проги BitPerfect:

"The DirecstStream’s design eschews the conventional wisdom regarding the design of a high-end DAC, which says you go out and buy a chip-set from one of the established vendors such as ESS, TI, and Wolfson, and build yourself a DAC around it. Rather like you would build a computer, having chosen an appropriate CPU from Intel’s catalog. PS Audio has taken a different tack. They have avoided using a DAC chip entirely, and have built their entire core converter functionality around a FPGA which switches the output voltage between two seriously stable voltage levels, representing the ‘1’ and the ‘0’ of a DSD bitstream. This is quite a smart approach, but it requires every single input data stream to be converted to a 1-bit format, and the one they have chosen is 1-bit / 5.6448MHz, otherwise known as DSD128.

As it happens, the chip-based solutions don’t quite do it this way. Instead of a 1-bit format, they use a 3-bit, 4-bit, or even 5-bit format, and there is a good reason for doing that. The devilishly complex Sigma-Delta Modulators (SDMs) used to perform these conversions are fundamentally unstable when configured with a 1-bit output. This instability can be avoided entirely by using a multi-bit output format. But the actual D-to-A conversion using those multi-bit bitstreams is more complex. In effect, PS Audio has traded the electronic complexity of a multi-bit high-speed D-to-A converter for a design approach they can get their considerable electronic design chops around. All that needs to be done is to navigate a way around the 1-bit SDM instability problem. As it happens, there are various techniques to mitigate this problem, although it cannot be conclusively eliminated. But by taking the right measures it can be beaten into submission, otherwise SACD/DSD couldn’t be made to work.

Inside the DirectStream, all input formats - DSD128 included - are converted to a single 30-bit 10MHz format, and from there converted to DSD128 which is fed to the core converter. There are a couple reasons for this. The first is that digital volume control can be easily implemented in this intermediate stage. The second - behind which hide an advanced degree’s worth of technical matters - is that the analog output filters in the output stage assume a certain implementation of the DSD128 data stream to which an arbitrary incoming data stream might not adhere. [My thanks to Ted Smith of PS Audio for clearing that up for me.]"
Посилання тут