• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DSP studio monitors

JulianW

Member
Joined
Jan 12, 2025
Messages
14
Likes
2
Hello everyone,


I have a question regarding DSP studio monitors that operate at fixed sample rates like 96 kHz. When listening to music in 44.1 kHz, the signal is typically resampled. During this process, the native reconstruction filter of the original 44.1 kHz signal is replaced by the resampling filter, which introduces its own cutoff and attenuation characteristics for band-limiting.


How do such monitors ensure that 44.1 kHz music can be accurately judged?
 
It’s no different than any another DAC without resampling. Any DAC performs oversampling, and applies a filter, so the asynchronous sample rate converter is no different. There will only be an additional filter applied at the ~48 kHz of the DSP, which doesn’t matter, because there will be no content there with a 44.1 kHz input.
 
It’s no different than any another DAC without resampling. Any DAC performs oversampling, and applies a filter, so the asynchronous sample rate converter is no different. There will only be an additional filter applied at the ~48 kHz of the DSP, which doesn’t matter, because there will be no content there with a 44.1 kHz input.
While your explanation is technically correct in that the resampling filter at ~48 kHz does not affect the frequency content below Nyquist (22.05 kHz for a 44.1 kHz signal), it overlooks a critical aspect: the potential time-domain effects of the filter. Specifically, the pre-ringing and post-ringing introduced by linear-phase filters can subtly impact the perceived sound, even though the frequency content remains mathematically intact.


So the DSP has to settle on a generic impulse response.
 
While your explanation is technically correct in that the resampling filter at ~48 kHz does not affect the frequency content below Nyquist (22.05 kHz for a 44.1 kHz signal), it overlooks a critical aspect: the potential time-domain effects of the filter. Specifically, the pre-ringing and post-ringing introduced by linear-phase filters can subtly impact the perceived sound, even though the frequency content remains mathematically intact.
There is no fundamental difference between this filter and any other DACs filter. They will both introduce similar amounts of ringing, and this is generally inaudible:

 
Last edited:
There is no fundamental difference between this filter and any other DACs filter. They will both introduce similar amounts of ringing, and this is generally inaudible:


There is no fundamental difference between this filter and any other DACs filter. They will both introduce similar amounts of ringing, and this is generally inaudible:

would it possibly make sense to use ascr in an audio player to only feed the monitor with 96 khz? or a halfband filter. I can imagine that the monitor expects a 96 khz signal from external sources and its own internal resampling is only a nice to have
 
I can imagine that the monitor expects a 96 khz signal from external sources and its own internal resampling is only a nice to have
In the vast majority of cases, the DACs of the DSP speaker will have their own clock domain, meaning that the ASRC will always resample to that clock domain, even if the input is already at the same rate. In fact, people claim that it’s not preferred to input the same rate because of this: the performance of the ASRC is claimed to be worse when the input and output sample rate are close. Though I’ve never seen any objective data in this regard.
 
In the vast majority of cases, the DACs of the DSP speaker will have their own clock domain, meaning that the ASRC will always resample to that clock domain, even if the input is already at the same rate. In fact, people claim that it’s not preferred to input the same rate because of this: the performance of the ASRC is claimed to be worse when the input and output sample rate are close. Though I’ve never seen any objective data in this regard.
But if DSP works at 96 kHz, like in the Genelec The Ones, that means the integrated DAC doesn't oversample, right?
 
But if DSP works at 96 kHz, like in the Genelec The Ones, that means the integrated DAC doesn't oversample, right?
No, it does not mean that. A DAC will always oversamples unless it has an external digital filter (which will do the same), which isn't usually done and in most cases not even supported.
 
No, it does not mean that. A DAC will always oversamples unless it has an external digital filter (which will do the same), which isn't usually done and in most cases not even supported.
1736882441600.gif

AK4621EF​

I'm not a professional, but what can you deduce from this diagram?
 
Nothing much, It's a codec consisting of an ADC and DAC, the datasheet doesn't say what the oversampling rate is, but it does have a digital filter, so it must be doing it.
 
Nothing much, It's a codec consisting of an ADC and DAC, the datasheet doesn't say what the oversampling rate is, but it does have a digital filter, so it must be doing it.
Strange. For other AKM codecs or DAC chips, the diagrams always show "X INTERPOLATO" as the processing step. I always assumed that interpolation does not occur in DSP monitors in favor of signal purity. Thank you for the clarification
 
The more highly integrated an IC is, the less detail you tend to get about its inner workings in the datasheet. The intro makes it pretty clear we're dealing with a typical modern-day part though:
The on-board analog-to-digital converter has a high dynamic range due to AKM’s Enhanced Dual-Bit
architecture. The DAC utilizes AKM’s Advanced Multi-Bit architecture that achieves low out-of-band
noise and high jitter tolerance through the use of Switched Capacitor Filter (SCF) technology.
Both (plus the choices of master clock) point towards delta-sigma converter technology.
Enhanced Dual Bit is a 3-level modulator I think, been around since '96ish. It is not uncommon to go with modulators of relatively low complexity in ADCs as they tend to have better inherent linearity (not to mention being cheaper to produce which matters in a midrange part), and relatively high-order noise shaping can be used with few negative repercussions, as ultrasonic noise is being dealt with by lowpass filtering during digital decimation. (If you want an example for why the same approach would make a DAC rather annoying to work with, look no further than the trusty CS4272 codec or its relatives like CS4392.)
I don't think the DAC side has a discrete counterpart but something like the AK4482 should be fairly similar. Just a typical early-2010s-era midrange kind of DAC.
 
The more highly integrated an IC is, the less detail you tend to get about its inner workings in the datasheet. The intro makes it pretty clear we're dealing with a typical modern-day part though:

Both (plus the choices of master clock) point towards delta-sigma converter technology.
Enhanced Dual Bit is a 3-level modulator I think, been around since '96ish. It is not uncommon to go with modulators of relatively low complexity in ADCs as they tend to have better inherent linearity (not to mention being cheaper to produce which matters in a midrange part), and relatively high-order noise shaping can be used with few negative repercussions, as ultrasonic noise is being dealt with by lowpass filtering during digital decimation. (If you want an example for why the same approach would make a DAC rather annoying to work with, look no further than the trusty CS4272 codec or its relatives like CS4392.)
I don't think the DAC side has a discrete counterpart but something like the AK4482 should be fairly similar. Just a typical early-2010s-era midrange kind of DAC.
I had hoped that the DSP would do the basic work and the DAC chip would only be used for pure conversion
 
What is your definition of “pure conversion”? A 100 SINAD seems pretty pure to me…
a dac that is clocked to the dsp rate and goes directly into the zero order hold interpolation and dispenses with a regular previous upsampling step so that I can control how it is rolled off after the nyqvist limit
 
Thank you very much for your time, the matter was unnecessary for me anyway, since a DSP probably performs an ASRC anyway and always resamples everything :facepalm::D
 
Back
Top Bottom