Poll - New firmware: sound upgrade or downgrade?

Mine are tineo

1 Like

We have our Sky Q connected to our NDX2 (optical).

The clarity of sound around TV listening is the best it’s ever been. SWMBO commented that we seem to be listening at lower volume levels since the firmware update.

1 Like

Don’t have the 555 but traded in 333 for Linn Selekt DSM. Not sure how new Linn fits the OC system, but it fits my 300 series like a glove :slight_smile:

1 Like

Of cause it should have been great in the title!

I had a Linn Selekt DSM Edition Hub Organik DAC for just over a year. I bought my Naim three months ago. The sound is much better and I wouldn’t want to go back to Linn.

3 Likes

I’ve also never liked the Linn sound. Lacks energy.

6 Likes

I love Linn (proud owner of an LP12 and DMS Isobarick for 42 years), but never quite got accustomed to their amplification or streamers. Very nice, but something missing that I love in my Naim amplification (for 42 years, totally rebuilt this year by AVOPTIONS, shout out to them, they did a magnificent job!) and now streamer (NDX2). PRAT is a lot of it, but also presence. But to each its own, of course.

6 Likes

Look at a 555 PS for your NDX2, different beast. As a Hi-Cap owner you’ll know the story.

2 Likes

Not in my budget. I got a great deal on a used XPS DR. I have had my entire Naim amplification rebuilt, NDX2, XPS DR, and new speakers this year (I am about to retire my 42 years old Linn DMS Isobarick). That’s enough dough spent on hifi. My car has 225,000 miles (got to make choices in life, and hifi before cars for me), so I can’t reasonably expect not to have to replace it in the next couple of years :).

3 Likes

Fairplay, not in your sig :wink: or I wouldn’t have mentioned it.

Well, thanks for mentioning it in case I win the lottery! It’s always good to know that there is still something much better, in case some time in the future, you want to splurge. So, thanks!

1 Like

If you try to do to much at the same time the processor heats up and that generates more noise and so on which later may translate into harshness etc. In the early days we handled this by writing in assembly language (a text representation of machine code). Today the processors are more advanced and not really suitable for manual coding. They have parallelism, more resources and built-in optimizers so it is better to use a language like C++23 with a code-generator and optimizer that deal with this complexity (like LLVM), including making non-blocking code.

Today we use C/C++ even for DSP-processors. The developer should concentrate on algorithms and architecture.

So it is all about and knowing your processor, getting even timing and make sure you avoid heating them up. And use a language that gives you control.

3 Likes

SHARC DSPs: Programmable Control Over FIR and IIR Filters for Sound Design

SHARC DSPs (Super Harvard Architecture Single-Chip Computers) are highly versatile tools for digital signal processing, particularly in audio applications. These DSPs are programmable to implement both FIR (Finite Impulse Response) and IIR (Infinite Impulse Response) filters, enabling precise control over a system’s frequency response, transient behavior, and phase characteristics. This programmability also allows firmware changes to directly impact how these filters affect audio quality, including transient response and phase distortion, which are critical for shaping the perceived brightness and character of sound.

FIR and IIR Programmability

SHARC DSPs provide the flexibility to design custom FIR and IIR filters. FIR filters are preferred for applications requiring linear phase response, ensuring that transients remain sharp and unaltered. These filters rely on convolution and do not involve feedback, making them ideal for maintaining the integrity of fast, transient-rich signals. In contrast, IIR filters use feedback loops, making them more computationally efficient while achieving a desired frequency response with fewer coefficients. However, this comes at the cost of non-linear phase distortion, which can smear transients and alter their clarity.

SHARC DSPs support these filtering techniques through extensive libraries and development tools, such as Analog Devices’ CrossCore Embedded Studio. Programmers can leverage optimized libraries or create custom filter algorithms in C or assembly. The DSP’s SIMD (Single Instruction, Multiple Data) architecture allows for parallel processing of data, enhancing performance for both FIR and IIR implementations.

Firmware Impact on Transients and Phase

The transient response of a system, which refers to how it reacts to sudden changes in input, can be directly influenced by the filter design programmed into the SHARC DSP. Adjusting the firmware to modify filter parameters or coefficients alters the system’s impulse response—its reaction to a brief, sharp input signal. For instance:

Shorter, sharper impulse responses preserve transients and result in a more natural and detailed sound.

Longer, oscillating impulse responses, often introduced by higher Q factors in filters, can lead to a brighter sound by emphasizing certain frequencies but at the cost of introducing ringing effects.

Phase distortion, a common artifact of IIR filters, can also be manipulated through firmware. By designing the IIR filter to minimize phase shifts in critical frequency ranges or compensating with phase-correcting algorithms, the balance between brightness and naturalness can be fine-tuned.

Practical Implications for Audio Design

In audio systems like high-end DACs and amplifiers, firmware updates to SHARC DSPs can significantly alter the listening experience. For example:

Brighter Sound: By extending the oscillating behavior of the impulse response, the system can accentuate higher frequencies, making the sound appear brighter and more detailed.

Controlled Phase Distortion: Careful manipulation of phase distortion can introduce a subjective warmth or clarity, depending on the desired tonal balance.

These capabilities highlight the importance of SHARC DSPs in sound design. Their programmability allows audio engineers to iterate on filter designs and firmware, optimizing both frequency and temporal domains to meet artistic or technical goals.

2 Likes

Naim use IIR filters, as in their architecture with the SHARC they believe it sounds better than FIR despite some phase compromises that can occur with IIR.
Others may use FIR… when you see the word ‘tap’ that is associated with FIR filters.

The SHARC is used by Naim in terms of filtering for the oversampling filter, and anti aliasing reconstruction filtering… as well a digital audio out down sampler option.

The code for the anti aliasing filter for SHARC as used by Naim is I believe 6 lines of machine code.

Apart from phase nonlinearities, IIR filters also introduce oscillations in the impulse response. The decay of these oscillations has an exponential nature and depends on the filter coefficients. Adjusting these coefficients can make the oscillations shorter or longer, which directly impacts the perception of clarity, brightness, and detail in the sound. However, these oscillations are distortions by nature.

The bottom line is that firmware may play a significant role in sound quality, contrary to the misconception that it primarily deals with security or other non-sound-related issues. Properly designed firmware is vital for optimal filter performance and minimal unwanted distortions, directly affecting the listening experience.

1 Like

Indeed, those 6 machine code operands for the reconstruction filter are key, however the timing of the execution of those operands is important to with respect to the noise profile. This likely extends to software processing beyond within the SHARC.

I had not come across issues with unwanted ripple with IIR filters per se, but the Chebyshev filter will introduce ripple in the pass band in order to allow a steeper cutoff. When the filter ripple is set 0% in a Chebyshev filter for a longer cut off, it’s called a Maximally Flat or Butterworth filter. Naim use a Butterworth reconstruction filter in their NDAC and streamers.

IIR filters are recursively called, so the numeric precision does become important. FIR filters are ultimately governed by the size of the filter response kernel (taps) , bit length and the windowing function used. FIR filters are usually more computationally intensive so are limited by electromagnetic noise generated by its execution.

One can

  1. Extract Firmware Contents:

Use tools like the Firmware Analysis and Comparison Tool (FACT) to unpack the firmware images. FACT automates the unpacking and analysis of firmware files, allowing you to inspect their contents.

  1. Identify DSP-Related Components:

Locate the sections of the firmware that pertain to the SHARC DSP. This may include specific binaries, configuration files, or scripts responsible for DSP operations.

  1. Compare Files:

Utilize file comparison tools such as WinMerge or Meld to perform a side-by-side comparison of the DSP-related files from both firmware versions. These tools highlight differences in code, configuration parameters, and other relevant changes.

  1. Analyze Differences:

Examine the identified differences to understand their impact on the DSP’s functionality. Changes in code logic, filter coefficients, or processing algorithms can indicate modifications in how the SHARC DSP is programmed.

1 Like

Sure, but with respect to Naim they do state in their white papers what they ask the SHARC to do for reconstruction… but they don’t state the formula precisely… but it hasn’t changed since the NDAC.

But… you do raise an interesting point… recording the audio from the DAC config with the unwanted firmware effects, and compare with the same firmware on an non effected device or different firmware and compare the signals.

3 Likes

You two need to get a room :wink:

13 Likes

Right!

Evan if i don’t understand what they say, it’s nice to have them around :smiley:

5 Likes