|Posted by Amine Slimani on December 6, 2010 at 10:45 AM|
How much can a digital filter affect the sound of the DAC? Up until last year, I was unaware of the importance (and even the role) of the digital filter in the sonic outcome of a DAC. It was until I used the DAC19MK3 and got to compare side by side the HDCD capable PMD100 against the DF1704 that I started to understand the importance of the digital filter. I found that both thePMD100 and DF1704 had strengths and weaknesses which led me to replace the dac19mk3 with the dac19dsp which was supposed to have a superior digital filter (the DSP1) than the off the shelf chips such as the PMD100 and the DF1704. Keep in mind that many experienced listeners consider those digital filters as superior sounding to many built-in digital filters in the more commonly used modern sigma-delta DACs.
My DAC19DSP came with the DSP1 in its V3. In comparison to the older dac19mk3, the DSP1 had even better dynamics and soundstaging than the bright sounding DF1704 and was almost as natural sounding the warmer PMD100.
These past few months, after a few upgrades and tweaks, the resolutionof my system increased but the pleasure I was getting from the system decreased. My system started to sound lightly balanced with a slight hardening of the tone. I was so dissatisfied that I was considering either selling my ALO Beyer T1 and getting a “bassier” headphone such as the Audeze LCD-2 or gettinga tube amp to change the tonal balance of the system.
Then the DSP1 V5 upgrade was released. Many head-fiers reported excellent results with the latest DSP1 V5 which involved a replacement of the DSP1 board with a new one. The reports were slightly confusing over what to expect from the upgrade (sonic wise) but it seemed that the difference was rather big as most people were noticing it. I decided to take the plunge andordered a DSP1 V5 to try for my self.
So what to expect from the DSP1 V5?
Timbre and tonal Balance:
One of the first things you notice with the DSP1 V5 is an increase of extension at both frequency extremes, with no apparent change in the tonal balance. There is more air around the instruments and the increase in deep bass was very noticeable. The increase in deep bass is comparable to what you would get from moving from a low powered amp to a bigger more power amp, or even like adding a well tuned subwoofer to smallish tower speakers.
The change was so big that it changed my mind about the frequency response capabilities of the ALO Beyer T1s. I used to find them limited at both frequency extremes (and especially the deep bass) but it is very different now.
But not only there was an increase in the bandwidth, but also in the richness of tone. Again, with the very revealing ALO Beyer T1s, I used to find the tone a little bit on the dry side, like there was a little bit of simplification of the upper harmonics (the HD650s, being already on the warmside, are not good analyzing tools for such minute differences). With the DSP1V5, it is like the colours blossomed and the tones became richer without turning into an overly warm representation.
In my review of the dac19dsp, I described it as neutral, realistic but perhaps not the most impressive at the frequency extremes. With the DSP1 V5, I would describe the dac19dsp now as being neutral, analytical, impressive at the frequency extremes yet natural and musical. Some of the adjectives might seem contradictory but this what the dsp1 V5 is able to achieve.
Soundstage and imaging:
The second most impressive quality was the increase in soundstaging and imaging capability. While the dsp1v3 was no second class runner in this department, the V5 upgrade improved both in size of the soundstage but also in specificity of imaging.
The size and depth of the sound filed increased but you can also see deeper and more clearly in the farthest recess of the soundstage.
The increase in quality was mostly noticeable on classical recordings.Groups of instruments that were massed and blurred together started being discrete and separted without sounding thin or unnatural.
The soundstaging and imaging capabilities were especially noticeable on the ALO Beyer T1s. In fact, as I am going to explain more in details in theBeyer T1’s review, with the dsp1V5 in place, there was a shift from having a soundstage that is "this or that big" to being transported to the recording venue. You are not only able to guess the size of the soundstage (or room) but you are transported into the recording event and get to feel (on minimally processed recordings) different instruments and singers interact with the recording venue.
I noticed an increase in perceived macro-dynamics that was probably due to a better handling of the deep low bass and better transient response.Overall, the representation of the dsp1V5 felt more impactful than the dsp1v3. There was also an increase in micro-dynamics making a little easier to follow small inflexions on solo instruments and voices.
Though, I have to note that the improvement in dynamics was moderate because the dsp1v3 was already very good in that department (save for the impact).
Transparency and Resolution:
Given the qualities and improvement in the timber, soundstaging and dynamics, one can easily guess that the dsp1v5 has a bumped resolution in comparison to the dsp1v3 I was using previously.
What it means is that you hear less of the digital filters artifacts and your brain can relax more while listening. It is not so much about the amount of details (the dsp1v3 could already dig very deep into the recordings) but it is about how little effort you have to do to listen.
You don’t get more details in the usual way, but you rather get submerged intothe recording itself.
The increase in transparency and resolution benefit both low bitrate MP3s (I don’t have many of those but keep them for reality check) and high resolution files.
PLL: This is a classic PLL feature that allows for the signal to be reclocked. By default, the PLL is engaged, and it does indeed level a little bit the field between the transports but the quality of the transport is still of the outmost importance.
Using the Digital Transport as a source, I found that de-activating the PLL increased a little bit the perceived transparency (maybe fake resolution?) and the soundstage. But there was a loss of coherency. Using Jkeny’s MK1Hiface, the resolution went up a notch, but the soundstage lost some depth and 3D effect.
On long term listening, I always preferred leaving the PLL on. But I can understand that with different transports or different preferences, one can prefer the sound with the PLL turned off. It is definitely something worth trying.
Passband settings: It is set by default at -130db. There are optional settings at -90db and -50db. Without getting into much technical details, the lower passband settings offer a more relaxed and less processed sound with a slight loss of sense of focus.
With my current set-up, and when upsampling is set to 96K, I ended upchoosing the -90db setting which offers an “analog” feel and a greater sense of flow in comparison to the stock -130db.
There is no right or wrong here, and I believe it is a great way to fine tune a system.
Oversampling settings: The PCM1704 DACs were probably not designed to run in NOS mode and it shows through the DSP1 as the higher the Oversampling settings, the higher the sound quality. So far, I have yet to hear a user reporting better results with the lower oversampling results.
Overall, while the DSP1 offers the possibility to try different settings, the stock ones are already very good and most importantly neutral. Most users will probably end up choosing the stock settings but it shouldn’t stop people from experimenting.
Is the DSP1 V5 the ultimate digital filter (for the dac19dsp and otheraudio-gd Ref series)? I personally believe that the latest iteration is aspectacular achievement but it still has a few points it could improve on.
Setting the passband settings at -50db or -90db gets you a more “relaxed” sound than the stock setting but looses on some technicalities. Maybe that with better algorithms (minimum phase or apodizing digital filtersperhaps?), it will be possible to increase both resolution and the “analog” feel.
However, I have to stress that the V5 is a big upgrade over the V3: itis like if the V3 was operating at 16 bits (think jagged lines and big steps)while the V5 was operating at 32 bits (think smooth lines and tiny steps).
So by saying that the filter can still be improved, it is more a personal suspicion based on what I know is possible (newest minimum andapdodizing filters for instance) than the shortcomings I have been to discern so far in my test system. On its own, the DSP1 V5 sounds as close to perfect as any digital source I have listened to (the PMD100 comes in second place).
The second point of improvement is more “real”. Despite the fact, that the DSP1 runs the data in parallel internally and is supposed to be somewhatimmune to jitter, it is still very sensitive to the quality of the transport. So there is some work to be done on the reclocking in order to make the DSP1 less sensitive to different transports.
To sum up, I would say that audio-gd has done some terrific job with their latest DSP1 V5 upgrade. To answer the opening question “How much can a digital filter affect the sound of the DAC?” I would say that it can change it dramatically. It brought my dac19dspto new heights and the change was so considerably big that it changed (positively) my perception of the ALO Beyer T1s. Given that the dac19dsp is the lower end model from Audio-gd that uses a DSP1, I believe that owners of the Reference series (Ref5/7/8/9...) will probably experience a bigger change in scale (ifthe components downstream are transparent enough). It is a highly recommended upgrade for current owners of DSP1 audio-gd based DACs.
Picture of the DSP1 V5: