Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[phusis]

Hi

I'm currently using J River MC15 as a "preamp" in my high quality stereo setup with great results. My Windows 7 Ultimate-based mediaserver(or HTPC) comprises an RME HDSP 9632 soundcard, which is "talking" exclusively to J River MC15 in WASAPI Event Style mode. Music is stored lossless(FLAC) on the harddrive, and the analog outs from the RME soundcard is connected via a dedicated interconnect cabel(hereby bypassing the RME brakeout cabel) directly into the poweramp(that is, with no hardware preamp in between). DSP settings in J River is set to 24 bit, and the RME master gain volume is set to max - whereby attenuation is done solely via the internal digital volume control in JRMC15.

Now my question goes:

How much has been done from the J River team to optimize the internal digital volume in their application to the audiophile segment? Although I find the sound of my setup to be top notch as used per above, could the internal digital volume control in JRMC15 really be looked upon as a "high end" attenuation device as such? - "High end," is in interfering as little as possible.. Furthering more specifically, on a side note: what dictates the so-called wordlength(or bitdepth) of the digital volume control here - the music source itself(normally 16 bit, or upwards), or the max bitdepth value of the soundcard? I'm not technically proficient in this regard, so any elaboration on this matter is highly welcomed!

Maybe above question seems irrelevant when I'm happy with the sound of my setup as is(though it shouldn't or can't be a reference in itself), but I'm just curious - maybe more could be done to optimize this solution, or could be it simply needs a little more exposition so that the high-minded audiophile would see it as a (more than) serious alternative? I believe using the internal digital volume control, in named setup and as described above, is a much better solution than many at first would have it. An analog volume control from a hardware preamp is often regarded as the best way to go, and in some cases it may surely be, but it's a flawed solution still, and I can easily see it losing terrain to a digital alternative - if the latter is implemented properly.

Sorry for the long-ish message. I do hope though my question will be addressed in that I find J River MC15 a wonderful application to anyone interested in high quality music reproduction, and that the possiblities with digital attenuation could well boost its appeal even further.  

Best,
M

[Matt]

Hi

I'm currently using J River MC15 as a "preamp" in my high quality stereo setup with great results. My Windows 7 Ultimate-based mediaserver(or HTPC) comprises an RME HDSP 9632 soundcard, which is "talking" exclusively to J River MC15 in WASAPI Event Style mode. Music is stored lossless(FLAC) on the harddrive, and the analog outs from the RME soundcard is connected via a dedicated interconnect cabel(hereby bypassing the RME brakeout cabel) directly into the poweramp(that is, with no hardware preamp in between). DSP settings in J River is set to 24 bit, and the RME master gain volume is set to max - whereby attenuation is done solely via the internal digital volume control in JRMC15.

Welcome to Interact.  I use a similar power amp setup in my home theater, powered by Media Center.

How much has been done from the J River team to optimize the internal digital volume in their application to the audiophile segment? Although I find the sound of my setup to be top notch as used per above, could the internal digital volume control in JRMC15 really be looked upon as a "high end" attenuation device as such? - "High end," is in interfering as little as possible.. Furthering more specifically, on a side note: what dictates the so-called wordlength(or bitdepth) of the digital volume control here - the music source itself(normally 16 bit, or upwards), or the max bitdepth value of the soundcard? I'm not technically proficient in this regard, so any elaboration on this matter is highly welcomed!

The internal volume, like all other audio processing in Media Center, is done in 64-bit floating point.  So from a practical standpoint, the internal volume control in Media Center is perfect (64-bit offers billions of times more precision than your ear can detect).  An analog circuit could never match this precision.

More on the 64-bit data path here:
http://yabb.jriver.com/interact/index.php?topic=59364.0

[phusis]

Hi Matt

Thanks for your quick response!

The internal volume, like all other audio processing in Media Center, is done in 64-bit floating point.  So from a practical standpoint, the internal volume control in Media Center is perfect (64-bit offers billions of times more precision than your ear can detect).  An analog circuit could never match this precision.

Wow. However, in practicality, isn't the J River internal digital volume in my setup also dependend upon the bitdepth of the soundcard and/or the music source? Sorry if it's a silly question..

My concern is whether there'll be any limitation in resolution at lower volumes.

[Matt]

Wow. However, in practicality, isn't the J River internal digital volume in my setup also dependend upon the bitdepth of the soundcard and/or the music source? Sorry if it's a silly question..

Yes, you're exactly right.

The audio engine does a final down-mix to whatever bitdepth the soundcard requires (24-bit in your case).

However, using the extra bits during processing is still important since no precision (effectively) is lost through any number of processing stages.

[phusis]

Yes, you're exactly right.

The audio engine does a final down-mix to whatever bitdepth the soundcard requires (24-bit in your case).

However, using the extra bits during processing is still important since no precision (effectively) is lost through any number of processing stages.

Ok, I see. So the effective digital volume control wordlength/bitdepth in my setup is 24 bit, regardless of a 16 bit music source? Would you say there's any limitation(less than 16 bits) even when playing at very low volumes in my setup?

[Matt]

Ok, I see. So the effective digital volume control wordlength/bitdepth in my setup is 24 bit, regardless of a 16 bit music source?

Yes.  The input format bitdepth is not relevant as everything gets converted to 64-bit (this still allows a bit-perfect data path if the input and output depths are the same and no processing is enabled).

Would you say there's any limitation(less than 16 bits) even when playing at very low volumes in my setup?

The only trick for me has been to find a setup where, at very low volume, there isn't noise or hiss at the speakers.  With such a powerful amplifier connected, always running at full volume, any noise in the system is also amplified a lot.

[phusis]

The only trick for me has been to find a setup where, at very low volume, there isn't noise or hiss at the speakers.  With such a powerful amplifier connected, always running at full volume, any noise in the system is also amplified a lot.

Hmm, why would you at all times run your amps at full volume, and how does that not translate into hearing at full volume?

[Matt]

Hmm, why would you at all times run your amps at full volume, and how does that not translate into hearing at full volume?

A power amp has no volume control, so it's always at full volume.

[phusis]

A power amp has no volume control, so it's always at full volume.

But that's dictated by the source output, isn't it? I mean, a Class A poweramp always runs at full power, but the actual level of volume and amp "work" is a product in conjunction with source output and speaker load. In my understanding the only thing that can trigger or exposes an amps inherent noise level are very sensitive speakers(requiring little power to generate high level sound) or full output signal. However, at some point a volume control must be inserted, and once that happens the exposition of the noise floor is dependend on how much of the full signal is let through, as well as, again, the speaker's sensitivity. Or maybe I'm misunderstanding something - do correct me if so

[glynor]

Can I just say... This is a fascinating conversation.  Thanks.

[Matt]

The only trick for me has been to find a setup where, at very low volume, there isn't noise or hiss at the speakers.  With such a powerful amplifier connected, always running at full volume, any noise in the system is also amplified a lot.

But that's dictated by the source output, isn't it? I mean, a Class A poweramp always runs at full power, but the actual level of volume and amp "work" is a product in conjunction with source output and speaker load.

Sorry if I wasn't clear.

I'm talking about chassis noise that the source outputs (a soundcard in this case).  

If you hook a cheap soundcard to a power amp and set the soundcard at full volume, you'll hear noise (hiss, static, etc.) in your speakers.

The volume of that noise never changes.  If you play music, the noise is masked by the music.  But the quieter the music (made quiet using Media Center's internal volume), the more noticeable any noise would be.

That's why a good source (soundcard, DAC, etc.) is important.  I've found a few sound cards that are close enough to silent that you only hear a subtle hiss with your ear right next to a giant speaker.

But I'm always on the look-out for hardware that would be even better.  For example, I'd love to find / try an outboard USB DAC that did 7.1 because it seems like it could better eliminate chassis noise.

[phusis]

Oh, now I understand Thanks for clarifying!

That's why a good source (soundcard, DAC, etc.) is important.  I've found a few sound cards that are close enough to silent that you only hear a subtle hiss with your ear right next to a giant speaker.

But I'm always on the look-out for hardware that would be even better.  For example, I'd love to find / try an outboard USB DAC that did 7.1 because it seems like it could better eliminate chassis noise.

I have no problems with "chassis noise" via the RME soundcard, especially after I earthed my poweramp and HTPC. From then on it went practically dead silent, with only the very faintest of noise in the tweeters - the ears almost glued right next them.

I'm actually thinking about replacing the RME soundcard with a Lynx L22 ditto. Something tells me the Lynx could be a worthwhile upgrade, and in its price range I believe it's quite the bargain.

Sadly I'm not well versed in multi channel USB DAC's, and therefore have no recommendations readily at hand. My own setup is 2-channel only.

[mojave]

But I'm always on the look-out for hardware that would be even better.  For example, I'd love to find / try an outboard USB DAC that did 7.1 because it seems like it could better eliminate chassis noise.

I tried the Echo Audio Audiofire 8, but it had no way to control the volume for other programs besides MC (ie games). It didn't sound as good as my ASUS Essence ST with H6 daughtercard, either. The Lynx Aurora 8/FW is another firewire external audio interface that is supposed to sound incredible and has full volume control using hotkeys. However, it is a little out of my price range at over $2000.

[phusis]

So, to sum up on my initial question: can the internal digital volume control of MC15 indeed be regarded a high end attenuation device? Perhaps a more pragmatic approach could be infused into my question, so that to inquire to what extent this solution is, in practicality, viable; had, for example, the gain of my poweramp been significantly higher(say, above 26dB), the internal digital volume control would only have a limited range of attenuation, and thus might hit against a ceiling of limited resolution(that would otherwise be less obtrusive at much lower volumes)? Or is this a mute observation? Meaning, if a 24-bit DAC is used, and the way MC15 is "stitched" together, will there ever be a problem with a lack of resolution at lower volumes(using the internal volume control of MC15), even with a high gain poweramp? Maybe you could provide some options to vary the level of attenuation in the internal slide, so to accommodate different poweramp gains, or would that option compromise sound quality?

And a question to the DSP settings: what would be the most advisable in my situation/setup - setting the bitdepth to '24-bit,' or to 'Source bitdepth'?

[Alex B]

Matt already stated that the internal volume DSP is practically perfect, more precise than any analog volume controller.

The only remaining question is about the dynamic range. 24-bit provides a dynamic range of 144 dB. It is enough.

You can always set the output bit-depth to the maximum that your HW supports.

When the output is set to 24-bit and you play 16-bit files without any DSP, the lowest 8 bits will simply be padded with zeros (= silence) assuming you use an output mode that bypasses the Windows mixer (which would do its own tricks if used).

When the output is set to 24-bit and you use DSP for attenuating the volume any resulting signal that would be quieter than 144 dB below 0 dBFS will be truncated, but that level is several tens of dBs below any practically audible sound level and thus the truncation has no effect to the perceived audio quality.

Even with the 16-bit output mode it is extremely difficult to hear any adverse effect of the truncation or alternatively to hear the faint dithering noise that can be applied for reducing the rounding errors (MC provides both options).

[phusis]

Alex B -

Thanks for your reply.

Matt already stated that the internal volume DSP is practically perfect, more precise than any analog volume controller.

Indeed, but as stated earlier I'm in no way technically proficient on these matters, and therefore wasn't aware whether Matt addressed the issue in very general terms or not. I found my question concerning lack of resolution wasn't addressed specifically, and though it may have been (indirectly) implied in Matt's answer, I didn't see it.

[Chili-Jam]

If you are attenuating the signal in MC what happens is that you feed the internal DAC of your soundcard
with "less bits" than the original source is. So here you go, you are loosing resolution in your DAC. And what is lost is lost and cannot be regained by adding more gain afterwards. In fact you will pick up quantisation errors because your whole signal gets rescaled and has to match integers in the end.

But you shouldn't worry to much if you are happy with your sound. It depends on all your gear if you will notice differences in sound.
But if you really want to go into high-end audio you will need a dedicated external DAC which you feed with an untempered bit-perfect signal. This DAC will deliver unattenuated line signal to a preamp that is
capable of feeding a quality signal to your poweramp. But this way you will spent many times the money of your setup before you will see the improvement.

So in fact you have an efficient setup and you shouldn't worry to much about the way MC is calculating the volume. 

[phusis]

Chili-Jam -

Thanks for chiming in.

If you are attenuating the signal in MC what happens is that you feed the internal DAC of your soundcard
with "less bits" than the original source is. So here you go, you are loosing resolution in your DAC.

I see..

And what is lost is lost and cannot be regained by adding more gain afterwards.

The question of gain, if you're referring to it in the context I'm using it, is merely a matter of the quantity of sound(SPL). It has nothing to do with compensating for lost bits, if you will.

But you shouldn't worry to much if you are happy with your sound. It depends on all your gear if you will notice differences in sound.

I'm happy with the sound quality as is, yes, but my gear(and my ears) is certainly capable of differentiating even very small differences. If upgrades in SQ are to be found at a reasonable cost, I'm most eager to embrace them!

But if you really want to go into high-end audio you will need a dedicated external DAC which you feed with an untempered bit-perfect signal. This DAC will deliver unattenuated line signal to a preamp that is
capable of feeding a quality signal to your poweramp.

The issue here, as I see it, it not so much whether using an internal soundcard or an external DAC, but what happens with the signal on its way to, and at the point of attenuation. That said, what you're saying, is, that when the signal is being attenuated in MC15, it gets "tampered" with in a way that prompts the signal to be modified so to match the so-called "integers in the end"? That admittedly sounds troublesome... I understand this is the cost of (digital) attenuation before the signal reaches the DAC, as opposed to afterwards.

But this way you will spent many times the money of your setup before you will see the improvement.

That's what I fear.

So in fact you have an efficient setup and you shouldn't worry to much about the way MC is calculating the volume.

Of course relative to what degree I'm able to detect the inherent shortcomings of this solution, and MC's calculations.

[...] In fact you will pick up quantisation errors because your whole signal gets rescaled and has to match integers in the end.

So, in my setup, this becomes a question of how well the rescaling is done, and how severe the "quantisation errors" are? I'm guessing this addresses the aspect of "pragmatism" - that is, concerning the practical use, as brought up in my initial inquiries...

[Alex B]

If you are attenuating the signal in MC what happens is that you feed the internal DAC of your soundcard
with "less bits" than the original source is. So here you go, you are loosing resolution in your DAC. And what is lost is lost and cannot be regained by adding more gain afterwards. In fact you will pick up quantisation errors because your whole signal gets rescaled and has to match integers in the end.

No, the quantization errors will not be "picked up" because their volume level is below anything that an analog system can reproduce.

Here is an attempt to visualize how this actually works with the 24-bit output. (In the graph the assumption is that the power amplifier is tweaked to produce a 110 dB SPL in the listening room when a 0 dBFS signal is played.)




Here is a more scientifically correct explanation from an anonymous writer (-found by Google. I think it is well written and factually correct):
http://diwaves.com/tmp/sample_bits.htm

But you shouldn't worry to much if you are happy with your sound. It depends on all your gear if you will notice differences in sound.
But if you really want to go into high-end audio you will need a dedicated external DAC which you feed with an untempered bit-perfect signal. This DAC will deliver unattenuated line signal to a preamp that is
capable of feeding a quality signal to your poweramp. But this way you will spent many times the money of your setup before you will see the improvement.

So in fact you have an efficient setup and you shouldn't worry to much about the way MC is calculating the volume.

Unfortunately the poor audio signal will be distorted in a much greater extent after it has been outputted to the analog realm. The electronic components in the DAC's analog output stage, the amplifier(s) and the speakers are not as good as the digital DSP.

[Chili-Jam]

This is not about the internal 64-bit calculations of MC and remapping the output back to the 16-bit of a CD. The problem arises because to lower your output volume you have to go lower in resolution.

Look at it this way: you are playing a CD that is using it's full resolution (16 bit) without attenuating the volume in MC. Your poweramp will deliver 200 watts rms at 4 ohms. I don't know about the efficiency of your speakers and your listening room but this normally deafening. So i guess you will need to attenuate the total signal lets say 24 dbs which means that you will now send an attenuated signal equal to 12-bit resolution. In other words you started 65000 possible volume levels which you now remap to 4000.
This is where you degrade the signal, the more you attenuate the more you do so. You can have a look at this link to get the idea:

http://www.tonmeister.ca/main/textbook/node593.html

In summary if you attenuate the signal in the digital domain you add distortion, if you do so in the analog domain you add noise.

[Alex B]

So i guess you will need to attenuate the total signal lets say 24 dbs which means that you will now send an attenuated signal equal to 12-bit resolution

This is correct for 16-bit output (unless you use noise shaped dithering which can effectively increase the dynamic range by pushing the dithering noise and quantization errors to the frequencies where human hearing is less sensitive).

However, if the output is 24-bit all sixteen bits will survive from a 24 dB attenuation. As you can see from my graph they almost survive from a 48-bit attenuation too. Only the "quietest bit" in the truncated output contains quantization errors and these errors are about 30 dB below the theoretical human hearing threshold (0 dB SPL) and about 50 dB below the background noise level (20 dB SPL in this example).

Even if the sound device has an unusually high output level, the power amp has an unusually high input sensitivity, and the speakers are very sensitive (and thus you must attenuate by several additional dBs to get the desired result) there is still enough headroom for audibly lossless DSP volume adjustments. (Though, personally I would prefer to use a power amp that has adjustable input sensitivity or just use an integrated amp so that I could calibrate the output level.)

Edit: fixed a mistake in the dB values and a couple of typos...

[phusis]

Alex B -

[...] However, if the output is 24-bit all sixteen bits will survive from a 24 dB attenuation. As you can see from my graph they almost survive from a 48-bit attenuation too. Only the "quietest bit" contains quantization errors and because the signal is attenuated by 48 dB these errors are about 40 dB below the theoretical human hearing threshold (0 dB SPL) and about 60 dB below the background noise level (20 dB SPL in this example).

That, I'm very glad to hear!

Even if the sound device has an unusually high output level, the poweramp has an unusually high input sensitivity, and the speakers are very sensitive (and thus you must attenuate by several additional dBs to get the desired result) there are still enough headroom for audibly lossless DSP volume adjustments. (Though, personally I would prefer to use a power amp that has adjustable input sensitivity or just use an integrated amp so that I could calibrate the output level.)

Fortunately, in my setup, the poweramp only sports a gain of 21dB, and so more or less fits like a glove with regard to the range of attenuation via the internal volume control of MC15 - i.e.: with the RME master-gain in the mixer set to max, and so by-passed. In-use in my setup, with the internal volume control set to max, there's still room for more SPL, but since the level available is more than enough for me I see no reason to adjust the output level in the RME soundcard's DSP, which sits at the lowest level as is.