Genelec 8361A Review (Powered Monitor)

[alankila]

For those of us who stream, it is beneficial to ensure that the DSP circuitry of the 8361A and W371A is matched by your streamer (in my case that's both Roon and my Dell XPS-15). This minimizes sample rate conversion. For me, that meant setting both Roon and my laptop to a sampling output rate of 96 kHz (in keeping with the Genelecs).The sonic improvements are obvious and immediately audible.

Gratitude to Genelec UK for the advice!

I personally do that as well, i.e. match the Linux pipewire's sample rate of 96 kHz with the Genelec internal DSP clock. IIRC, The Ones actually has multiple internal sample rates: woofers run off 48 kHz sampled stream, tweeter and midrange off 96 kHz. I don't know the details of how and why, this is just something I read off from somewhere, along with some other details of the process such as that the internal DSP is based on 64-bit floating point samples.

However, there is some doubt in my mind about your claim that it is audible improvement because resampling algorithms are already expected to be audibly transparent, and the additional bandwidth in sound output should be inaudible to a human. I think it is unlikely that there would be problems in resampling from lower sample rate to the internal clock in a professional monitoring product such as 83x1. I agree that it is theoretically obvious that minimizing number of conversions is good, but in practice it still might make no difference. What did Genelec say, exactly?

I personally did not notice any benefit when doing a sample rate switch, I was doing it more because I could and because it is useful to prove that the link between the speakers and the sound card is stable. One issue I noticed with running the higher rates is that S/PDIF signal is less likely to work in lieu of the AES/EBU signal. The signal is apparently much quieter and the higher sample rates makes it more difficult to receive correctly. I was able to get near instantly reliable transmission at 44.1 kHz and 48 kHz, probably thanks to short cable lengths. At 88.2 kHz, sound faded in and out for a few minutes but eventually stabilized. However, at 96 kHz the sound seemed to never stabilize. In addition to this, if the signal ceases, soundcards go to sleep, the digital link shuts off, and that resulted in the synchronization process starting all over again, which made me disable soundcard sleep altogether. Because of the difficulties in reaching the speaker, I decided to buy one of those cheap usb soundcards off Amazon with usb plug on one end and AES/EBU XLR plug at one end, which has worked without issue at every sample rate.

I'll say this much: I did end up improving the default resampling quality on a Linux system. Resamplers use CPU and people are somewhat obsessive and neat freaks on that operating system, and I've read complaints over the decades whenever a sound server uses any noticeable CPU. Of course, it is resampler inside the sound server that is busy converting from source's sample rate to system's fixed sample rate, so the tendency has been to push the default quality down to silence these complaints. For me personally, switching a couple of notch higher quality resampler seemed to make the treble a little less noisy, somehow, but I didn't try to ABX it. IIRC the default had about 60 dB SNR, and while it should have been sufficient, I would still swear that there was a little bit of fizz/noise in the treble and it became completely clear after improving that.

 

[markstein]

For those of us who stream, it is beneficial to ensure that the DSP circuitry of the 8361A and W371A is matched by your streamer (in my case that's both Roon and my Dell XPS-15). This minimizes sample rate conversion. For me, that meant setting both Roon and my laptop to a sampling output rate of 96 kHz (in keeping with the Genelecs).The sonic improvements are obvious and immediately audible.

Gratitude to Genelec UK for the advice!

Are you sure the SRC is bypassed when using 96kHz input sample rate? I would doubt it, without knowing technical details of the Genelec implementation.

 

[Esprit]

For those of us who stream, it is beneficial to ensure that the DSP circuitry of the 8361A and W371A is matched by your streamer (in my case that's both Roon and my Dell XPS-15)

Go with AES/EBU and you don't have to worry about it.

 

[Esprit]

The Ones actually has multiple internal sample rates: woofers run off 48 kHz sampled stream, tweeter and midrange off 96 kHz.

Source?

 

[Dgob]

Go with AES/EBU and you don't have to worry about it.

I'm not sure how I would link my laptop to the AES/EBU on the 9320A. At the moment it's USB.

 

[Behrelec 37]

I have an Apple laptop. Music comes from it via a USB cable to the 9320a. And the 9320a comes out via xlr aes/ebu 110 ohm digitally, the cables are digitally connected to the speakers.
Of course, in the glm software, digital reception for the speakers is already selected on the laptop and the first room correction and about 30 of my own different variations of conversion have also been transferred digitally to the speakers using only one, that is, the best cut diamond.
This is how aes/ebu works. Of course, you can also recycle it through the TV sometimes with fiber optics, but an impedance converter (75/110 0hm) must then be in between.

 

[Esprit]

I'm not sure how I would link my laptop to the AES/EBU on the 9320A. At the moment it's USB.

If you use the 9320A it is the one that comes out in AES/EBU and the whole path is in digital

 

[Esprit]

I have an Apple laptop. Music comes from it via a USB cable to the 9320a. And the 9320a comes out via xlr aes/ebu 110 ohm digitally, the cables are digitally connected to the speakers.

I've asked this before: is the 9320A fully remote controlled via Mac/PC?

 

[Behrelec 37]

You can get it digitally via USB to a 9320 A converter, and then also analogically to the speakers using standard XLR cables, but you are completely dependent on the quality of your laptop's DAC, no matter what manufacturer it is.

The 9320 is a receiving device. It routes the audio signal to multichannel or stereo. But for multichannel, you have to buy separately either a Genelec multichannel converter (no decoder for decoding though!) or a competitor's converter that sends the AES digital signal out to all Genelec speakers with Tascam 25 cables (special cable), depending on how many there are and what kind of configuration it is intended for. But the 9320a device does not have remote control. The remote control can only be connected to the GLM black soap box, i.e. the older GLM box, although that remote control also drains the battery too quickly if you use it every day in everyday home use. Keep this in mind, Sharp. Oh yeah, the 9320a also apparently converts analog signals to digital, if I remember correctly, but it requires getting to know the user manual better.

 

[Esprit]

Who are you replying to?
Thanks

 

[Dgob]

I personally do that as well, i.e. match the Linux pipewire's sample rate of 96 kHz with the Genelec internal DSP clock. IIRC, The Ones actually has multiple internal sample rates: woofers run off 48 kHz sampled stream, tweeter and midrange off 96 kHz. I don't know the details of how and why, this is just something I read off from somewhere, along with some other details of the process such as that the internal DSP is based on 64-bit floating point samples.

However, there is some doubt in my mind about your claim that it is audible improvement because resampling algorithms are already expected to be audibly transparent, and the additional bandwidth in sound output should be inaudible to a human. I think it is unlikely that there would be problems in resampling from lower sample rate to the internal clock in a professional monitoring product such as 83x1. I agree that it is theoretically obvious that minimizing number of conversions is good, but in practice it still might make no difference. What did Genelec say, exactly?

I personally did not notice any benefit when doing a sample rate switch, I was doing it more because I could and because it is useful to prove that the link between the speakers and the sound card is stable. One issue I noticed with running the higher rates is that S/PDIF signal is less likely to work in lieu of the AES/EBU signal. The signal is apparently much quieter and the higher sample rates makes it more difficult to receive correctly. I was able to get near instantly reliable transmission at 44.1 kHz and 48 kHz, probably thanks to short cable lengths. At 88.2 kHz, sound faded in and out for a few minutes but eventually stabilized. However, at 96 kHz the sound seemed to never stabilize. In addition to this, if the signal ceases, soundcards go to sleep, the digital link shuts off, and that resulted in the synchronization process starting all over again, which made me disable soundcard sleep altogether. Because of the difficulties in reaching the speaker, I decided to buy one of those cheap usb soundcards off Amazon with usb plug on one end and AES/EBU XLR plug at one end, which has worked without issue at every sample rate.

I'll say this much: I did end up improving the default resampling quality on a Linux system. Resamplers use CPU and people are somewhat obsessive and neat freaks on that operating system, and I've read complaints over the decades whenever a sound server uses any noticeable CPU. Of course, it is resampler inside the sound server that is busy converting from source's sample rate to system's fixed sample rate, so the tendency has been to push the default quality down to silence these complaints. For me personally, switching a couple of notch higher quality resampler seemed to make the treble a little less noisy, somehow, but I didn't try to ABX it. IIRC the default had about 60 dB SNR, and while it should have been sufficient, I would still swear that there was a little bit of fizz/noise in the treble and it became completely clear after improving that.

I hear and appreciate your scepticism about the audibility of the difference made by matched sample rates here. I can only say that the immediacy of perception for me was on listening (per chance) to Louis Armstrong & His All Stars performing 'St. James' Infirmary' immediately following the change. Not withstanding the obvious potential for psychoacoustic misperception, I heard better spatial accuracy, transient clarity, and overall cohesion with even more stable imaging. I still believe that I am enjoying that improvement after more extended listening sessions.

I realise that this might not ring true for other ears, set-ups or environments.

 

[alankila]

Source?

Sorry. I think it has been in the old Genelec forums that used to have technical answers to various user questions about Genelec products, including details like these that have no real user-visible impact.

The 64-bit DSP is relatively often mentioned by Aki Mäkivirta whenever discussing these speakers and the topic of that comes up, and I found that in multiple youtube videos just by looking for varoius 8351A presentations and similar. I found my old post elsewhere from couple of years ago where I stated that I read from Genelec forums that it is actually both midrange and woofer that run off the 48 kHz clock and only the central tweeter is at 96 kHz. But citing myself as source is not very reliable, and it seems I already had forgot that it's the midrange as well that is running from 48 kHz resampling of the input audio.

 

[Dgob]

Are you sure the SRC is bypassed when using 96kHz input sample rate? I would doubt it, without knowing technical details of the Genelec implementation.

Yes, the dsp processing in the 8361A occurs at 96 kHz.

 

[markstein]

Yes, the dsp processing in the 8361A occurs at 96 kHz.

This is not an answer to my question.

E.g., with miniDSP SHD also running at 96kHz internally, the SRC is always active. Meaning, there should be no audible impact of different input sample rates, other than the risk of intersample clipping which will not occur with your genelec speaker until operating it at 100% volume level.

Edit: I e.g. limit the sample rate in my setup intentionally to 48kHz to
- save DSP computing power (more PEQs available)
- have a brickwall filter at 24kHz to prevent tweeter resonance and intermodulation distortion (though my speaker has such a filter already built-in)
- have a brickwall filter at 24kHz for vinyl playback (lots of high frequency noise in the 30kHz range on some records)

 

[Dgob]

This is not an answer to my question.

E.g., with miniDSP SHD also running at 96kHz internally, the SRC is always active. Meaning, there should be no audible impact of different input sample rates, other than the risk of intersample clipping which will not occur with your genelec speaker until operating it at 100% volume level.

I appreciate the clarification. My understanding from Genelec UK was that matching the input sample rate to the DSP processing rate minimizes unnecessary sample rate conversion (SRC). However, I’ll reach out to Genelec directly to confirm whether SRC is fully bypassed at 96 kHz input and report back when I get a chance.

 

[markstein]

Thank you, that would be valuable information!

 

[Esprit]

But citing myself as source is not very reliable, and it seems I already had forgot that it's the midrange as well that is running from 48 kHz resampling of the input audio.

It all seems rather strange to me...

 

[markstein]

Well, different sample rates for different transducer passbands would make sense to optimize the accuracy of filters with a given number of filter taps. However, synchronization may not be trivial.

 

[Dgob]

Thank you, that would be valuable information!

Genelec’s R&D confirmed that sample rate conversion is always performed, even when the input sample rate matches the DSP processing rate. While this means that SRC wasn’t technically bypassed in my setup, the change I made—aligning my streamer’s output to 96 kHz—still resulted in an audible improvement. The sonic image felt more solid, and the midrange sounded fuller.

I recognize that expectations and psychoacoustics play a role in perception, but these changes were immediate and noticeable. Whether the impact stems from variations in SRC algorithms, digital filtering behaviour, or some other factor, I’ve found this adjustment worthwhile in my system. As Mort demonstrates below, YMMV, as they say.

There are also, of course, other potential benefits from matching, such as: Reducing external SRC complexity; Minimizing processing load in your source devices; Consistency across the chain; Improved stability in digital transmission. I suspect such considerations underpinned Genelec UK's noted suggestion.

 

[Mort]

I did not hear such improvements changing to 96.