dcx2496 attenuation

[SleepingRobot]

Hi all,

I've decided to go with the Behringer DCX2496 to do the sub EQ/delay/phase adjustments as I add a subwoofer to the 2-channel system in my home office. It's overkill for just the subwoofer, but it gives me the future expansion capability to bi-amp my main speakers - something I've been wanting to do for a long time.

I'm using 2-channel, non-home-theater gear. I'm planning on running the DCX in the pre-out/power-in loop on my integrated amp. I'll run one output to the sub, and two more to the poweramp inputs to drive the main speakers.

The outputs on the DCX are at pro levels, I've read that I should attenuate those by about 20db to get them back to consumer levels. I have some XLR to phono adaptors with some room in the shell to hold resistors, so my question is:

What value resistors should I use, and what topology? (One resistor in series, 5 in an H-pad, or something else?)


(BTW - I had some fun last weekend playing with REW and getting a baseline
reading on the room before I start adding treatments, EQ, etc. Extremely cool
software! I'll post some graphs when I am done.)

Best regards,
Ron

[tbrooke]

This may be more than you want to know but check out:
http://www.diyaudio.com/forums/showt...threadid=63547

Tom

[brucek]

The outputs on the DCX are at pro levels, I've read that I should attenuate those by about 20db to get them back to consumer levels. I have some XLR to phono adaptors with some room in the shell to hold resistors, so my question is:

What value resistors should I use, and what topology?

Ron, you're a bit confused here. The DCX accepts pro levels and will pass them through as such. This would be a unity gain device. The maximum input level to the DCX is +22dBu. This defines its operating range for both input and output and as a result its dynamic range.

The level 22dBu translates to ~9.75vRMS. This means I can send the DCX any level from 0volts to 9.75vRMS and its output (with no filters etc applied) will swing from 0volts to 9.75vRMS.

You have a consumer device that has an output level that generally would have a maximum output level of ~+2dBV. This translates to about 1.26voltsRMS out. See the problem?

Your maximum level of 1.25volts will pass through the DCX just fine and output at the same level. The rub is that you are not taking advantage of even a quarter of the bits available and you'll have a very poor S/N ratio and dynamic range.

If you are going to mix pro and consumer level devices, you'll need to boost your consumer level output to a pro level before feeding the DCX. Most people use a CleanBox device, although there are others that will do the job.

Remember that all these devices in the mains chain add noise... and its cumulative.....

brucek

[SleepingRobot]

Thanks, I had not seen that post before. Looks like lots of fun stuff to go through.

Best regards,
Ron

[SleepingRobot]

brucek wrote:

Ron, you're a bit confused here.

Not too unusual, I'm afraid.

brucek wrote:

The DCX accepts pro levels and will pass them through as such. This would be a unity gain device. The maximum input level to the DCX is +22dBu. [...] ~9.75vRMS. [...] You have a consumer device that has [...] a maximum output level of ~+2dBV. [...] Your maximum level of 1.25volts will pass through the DCX just fine and output at the same level. The rub is that you are not taking advantage of even a quarter of the bits available and you'll have a very poor S/N ratio and dynamic range.

Ah, thank you for that explanation. That makes the issue very clear.

brucek wrote:

If you are going to mix pro and consumer level devices, you'll need to boost your consumer level output to a pro level before feeding the DCX. Most people use a CleanBox device, although there are others that will do the job.

Remember that all these devices in the mains chain add noise... and its cumulative.....

brucek

OK, thanks for the pointer to that device, as well as the caution on noise. Raising and lowering that level so many times sure runs it through a lot of opamps. I'll have to think about that.

Thanks,
Ron

[Ilkka]

brucek wrote:

Your maximum level of 1.25volts will pass through the DCX just fine and output at the same level. The rub is that you are not taking advantage of even a quarter of the bits available and you'll have a very poor S/N ratio and dynamic range.

I use this device in consumer level system without any S/N ratio or dynamic range problems. I'm sure they would get better with higher input level but IMO you are making it sound like a huge problem. I used to have an analog EQ before this so I know how my system should sound like.

I'm running my whole front end (LCR) and sub through it.

[brucek]

I use this device in consumer level system without any S/N ratio or dynamic range problems

Well, I can only look at specs and make suggestions, so I have to defer to what you say, since I don't own a 2496.

But, as I'm sure you know, the LSB (least significant bit) voltage is directly proportional to the maximum input full scale voltage. In this case it's +/-9.75 volts RMS. So, if the input level isn't high enough, the LSB will not be distinguishable from the noise. The noise floor for the ADC converter is fixed and a function of the bit resolution, which in this case is 24bits. It has a theoretical noise floor of 147dB.

Well, of course the device can't obtain that figure, but it does show an incredible noise floor spec of 112dB in relation to full scale of 22dBu. That's extremely good (almost unbelievable actually ). Well, 112dB translates to about 18.5 bits usable above the noise. Quite good.

This all assumes you have a full scale input signal with a swing of +/- 9.75 volts. As you decrease the input signal, the noise rises exponentially. It's the nature of the math. A full scale of 1.25 volts is only about 13% of the possible full scale allowed. I won't bore anyone with the math, but this translates to about 4 bits being used with a dynamic range of about 24dB. That's horrible.

What can I say.

brucek

[Ilkka]

I know the math but that doesn't really bother me since I know it works just okay in my consumer level system. I don't know how it does it but it doesn't sound any worse than my old analog EQ.

[Wayne A. Pflughaupt]

...but this translates to about 4 bits being used with a dynamic range of about 24dB. That's horrible.

Wow. Before I saw that I was going to say, considering receivers typical 90-95 dBa S/N ratings will be at least 20 dB or more worse than the DCX’s, there’s a lot of S/N to spare. But 24 dB! Obviously, that should be grossly audible, so something doesn’t add up. Maybe the math looks worse on paper than what it translates to in real life.

I don't know how it does it but it doesn't sound any worse than my old analog EQ.

Does that mean your analog EQ sounded bad? What kind was it?

Interesting that the extra A/D-D/A conversion doesn’t appear to be audible. Must be some really good converters. Still, I wish manufacturers would give us receivers with digital pre-amp outputs, and amps that would accept a digital signal. Maybe someday.

Regards,
Wayne

[Wayne A. Pflughaupt]

Probaby opening a can of worms here, but...

As you decrease the input signal, the noise rises exponentially. It's the nature of the math.

Well, what if we follow that to it’s obvious conclusion? With the unit idling – i.e. no input signal - the noise should be so loud it would be audible in the next room. But that isn’t the case, is it?

After a little ruminating and a little research, it seems the discrepancy comes from us commonly (and grossly) confusing S/N ratio with dynamic range. (I’ll include myself here – see my last post!! I’ve seen most of this material before, but somehow today it just clicked.)

The two are related, but they are not the same. S/N is a measurement of residual noise. Dynamic range is the ratio of the loudest undistorted signal compared to the quietest discernable signal – i.e. before it “disappears” into the residual noise.

It’s entirely possible to have a component - an analog component at least - with an inaudible noise floor, but limited dynamic range. In other words, it would clip with a 20-dB signal, but still be dead silent with no signal.

Likewise, it’s possible to have very poor residual noise characteristics with excellent dynamic range. Anyone here have any experience with guitar amps? They’ll play loud enough to make you deaf, but unplug the guitar and you can hear the hiss and noise from across the room.

I caught this tidbit from Rane’s Digital Dharma of Audio A/D Converters white paper:

With 20-bit high-resolution [A/D] conversion, low signal-level detail is preserved. The improvement in fine detail shows up most noticeably by reducing the quantization errors of low-level signals.

Notice, no mention of noise with the presence of a low-level signal.

And this:

Here is what is gained by using 20-bits [over 16]:

• 24 dB more dynamic range
• 24 dB less residual noise [emphasis theirs]
• 16:1 reduction in quantization error
• Improved jitter (timing stability) performance

It should be obvious that when it comes to digital audio electronics, “the math” is referring to dynamic range, not S/N. Indeed, “the math” is dependant on the presence of a signal. The math doesn’t function with no signal. Mathematical formulas don't have much to do when you start with a value of zero.

Getting back to our DCX, obviously it doesn’t make sense that its dynamic rage will be limited to 24 dB if used with a home system. All a lower signal level means is that you’re merely not fully utilizing its available dynamic range. But in no way does that mean you’re increasing its residual noise floor. As we see in the second quote above, that’s a function of the A/D converter: As long as the converter is sufficient, the residual noise specification is there. The dynamic range spec, by its very nature, requires a signal in order to be utilized.

In this regard analog and digital equipment are the same: The residual noise floor is determined by the quality of the component itself, not the level if the incoming signal. You don’t raise the inherent noise floor of a component by feeding it a low-level signal. If that were the case, the noise would come up to increasingly audible levels as we turned down the volume, and there’d be no such thing as “background music.”

This also answers “the great boosting debate” with the BFD: Boosting will decrease available dynamic range, because the incoming signal will have to be reduced. But it doesn’t increase residual noise levels. I imagine this is why no one has ever complained of noise problems appearing with boosted filters.

I suppose the BFD or any other digital EQ could increase noise at the adjusted frequency, in an amount equal to the level of boost, as you have with an analog equalizer, but I'm not altogether sure if even that's possible, given what Rane says, "It is important at the onset of exploring digital audio to understand that once a waveform has been converted into digital format, nothing can inadvertently occur to change its sonic properties." So I suppose any boosted noise would ultimately be residual that's present in the incoming analog signal.

Regards,
Wayne

[Ilkka]

Wayne A. Pflughaupt wrote:

Does that mean your analog EQ sounded bad? What kind was it?

Interesting that the extra A/D-D/A conversion doesn’t appear to be audible. Must be some really good converters. Still, I wish manufacturers would give us receivers with digital pre-amp outputs, and amps that would accept a digital signal. Maybe someday.

Regards,
Wayne

No, it didn't sound bad. I just meant that I didn't notice any difference with the DCX2496.

Well do you notice any difference with normal BFD? It also does the extra conversions.

[brucek]

Well, what if we follow that to it’s obvious conclusion? With the unit idling – i.e. no input signal - the noise should be so loud it would be audible in the next room.

You're confusing noise threshold and signal to noise ratio. The noise doesn't change as you reduce the voltage swing that you feed through a digital system. What changes is the signal to noise ratio. The noise is a constant. If I feed a signal that has a maximum signal that approaches the digital systems maximum, then the softest signal we'll hear will be just above the fixed noise floor of the digital system. If I reduce the input level by half, that softest detail that use to pass through as the LSB of the usable dynamic range will now be lost in the fixed noise floor.

S/N is a measurement of residual noise.

That's completely out of context. Continue on and it says,
stated as the ratio of signal level (or power) to noise level (or power), normally expressed in decibels. The "signal" reference level must be stated. Typically this is either the expected nominal operating level, say, +4 dBu for professional audio, or the maximum output level, usually around +20 dBu. The noise is measured using a true rms type voltmeter over a specified bandwidth, and sometimes using weighting filters. All these thing must be stated for a S/N spec to have meaning. Simply saying a unit has a SNR of 90 dB means nothing, without giving the reference level, measurement bandwidth, and any weighting filers. A system's maximum S/N is called the dynamic range.

Notice, no mention of noise with the presence of a low-level signal

They're discussing the advantage of having more bits in the reduction of quantization error. This creates more voltage steps and as such increases detail. It has little to do with noise floor.

It should be obvious that when it comes to digital audio electronics, “the math” is referring to dynamic range, not S/N. Indeed, “the math” is dependant on the presence of a signal. The math doesn't’t function with no signal.

Sure it does, because we know the maximum signal allowed. Dynamic range is by definition the ratio of the maximum possible signal before distortion compared to the the noise floor.
If we state that the S/N ratio is with respect to the maximum signal (a standard industry measure), then dynamic range and S/N are equal.

I am able to define S/N for any given signal level within the limits of the system that defines the dynamic range. Usually the S/N will be in relation to maximum input (as smart move since it will be the best figure). I can say the theoretical dynamic range of 24 bits is 147dB, but I know there is a noise floor when they define both the S/N ratio and the Dynamic range of the DCX as 112dB at +22dBu.

In the front page of the specs they say the dynamic range is 112dB.
Then they say that the input and output noise is -90dBu (@+22dBu --> 112dB). All the information is there for me to define any signal to noise ratio for the system.
They have told us that the noise floor is at -90dBu. Again I won't show the math (unless you love antilogs), but that translates to a noise floor of 24.5 microvolts. That's real good. Then they say that in relation to the maximum signal allowed of +22dBu (which we know calculates to 9.75 volts), that 24.5 microvolts is a S/N of 112dB. Well, that's true because 20 log of 9.75 divided by 24.5 microvolts is 112dB. That's both the dynamic range and the signal to noise ratio.

OK, now I'm going to take a step back and say that I was too hasty in saying that the S/N of the DCX with a 1.25 volt signal was 24dB. That was dumb. My brain calculator doesn't work very well any more, so I used a real calculator this time....duh. no more napkins. Forgive me....

Since we now know the actual calculated noise floor and the dynamic range and the maximum S/N ratio, then we also have all the information we need to calculate the S/N ratio and the dynamic range of the system when someone decides to feed it a maximum level of only 1.25 volts.... it comes to about 94dB. That's just a bit better than 15 bits. Not quite good enough to completely resolve a 16 bit CD, but fair enough............... Still a bit of a crime for a 24 bit capable machine.

Well do you notice any difference with normal BFD? It also does the extra conversions.

Generally, no one uses a BFD for mains. Just for subwoofer......

brucek

[SleepingRobot]

brucek wrote:

[...] I was too hasty in saying that the S/N of the DCX with a 1.25 volt signal was 24dB [...] it comes to about 94dB. That's just a bit better than 15 bits. Not quite good enough to completely resolve a 16 bit CD, but fair enough............... Still a bit of a crime for a 24 bit capable machine.

Hi Bruce, Thanks for double-checking the math.

FYI I just hooked up the unit into the pre-out/power-in loop without any attenuators, etc. It sounds fine, I find your 15 bit computation very believeable. I am not running this at full level, so the quantization error would be even more. But in the brief exposure so far I have not found anything to be objectionable. (Well, there is a tiny bit of grounding hum when I put my ear next to the speaker, but I haven't tried to get that sorted out yet.) But I'm sure there is room for improvement. Once my sub arrives I'll dig into this again.

brucek wrote:

Generally, no one uses a BFD for mains. Just for subwoofer......

Right, that was a big factor in my decision to go with the DCX2496 instead of the BFD1124P.

Best regards to all,
Ron

[Ilkka]

SleepingRobot wrote:

(Well, there is a tiny bit of grounding hum when I put my ear next to the speaker, but I haven't tried to get that sorted out yet.)

Just use a cheater plug to get rid of it. That's the only "problem" with these Behringer devices...

[SleepingRobot]

Ilkka wrote:

Just use a cheater plug to get rid of it.

Yes, I just didn't have one on-hand last night and I didn't feel like mutilating any extension cords to make one. I'll pick one up on my next trip to the hardware store.

BTW - Are there any turn-on or turn-off transients I should worry about with the DCX? If possible I'd like to run it off of the switched power jacks of my amp, so I can still turn things on and off with the remote.

Thanks,
Ron

[Wayne A. Pflughaupt]

Yup – Can of worms!

brucek wrote:

They're discussing the advantage of having more bits in the reduction of quantization error. This creates more voltage steps and as such increases detail. It has little to do with noise floor.

So they were lying when they said “Here is what is gained by using 20 bits [over 16]: 24 dB less residual noise?"

If we state that the S/N ratio is with respect to the maximum signal (a standard industry measure), then dynamic range and S/N are equal.

If that’s true, then why does Rane indicate that they are separate specifications (at the Dynamic Range link in my previous post)? ” Professional-grade analog signal processing equipment can output maximum levels of +26 dBu, with the best noise floors [established as residual noise or S/N at the “S/N” link] being down around -94 dBu. This gives a maximum dynamic range of 120 dB...

Not to mention, occasionally you’ll find a manufacturer that lists them as separate specs:
ftp://ftp.dbxpro.com/pub/PDFs/Spec_S...CutSheetA2.pdf

You're confusing noise threshold and signal to noise ratio.

Not at all. They are the same thing. In the words of the late, great Ricky Ricardo, “Let me ’splain it to you.”

That's completely out of context.

No, it’s not out of context. You latched on to the last sentence at the expense of the rest of the passage. And please note how they italicize “maximum S/N ratio” at both links, indicating that’s a different animal all together.

Take a closer look at the full text and you can see it’s explaining that the noise floor, threshold, or whatever you want to call it, must be established by a specific means. Pay special attention to the third sentence, “The noise is measured using a true rms type voltmeter over a specified bandwidth, and sometimes using weighting filters.”

Ask yourself, “Why would they employ weighting filters?” Do weighting filters have anything to do with the “maximum signal level before distortion” (see dynamic range definition in my previous post)? No. Weighting filters are for noise floor specifications. Rane notes in their Audio Specifications white paper that an A-weighted filter, for example, can “hide” hum components that make for a bad noise spec: “A-weighting rolls off the low-end, thus reducing the most annoying 2nd and 3rd line harmonics by about 20 dB and 12 dB respectively. Sometimes A-weighting can "improve" a noise spec by 10 dB.”

Thus, anytime you see a S/N ratio figure with a weighting qualifier, that’s a residual noise specification.

On the other hand, you never see weighting attached to a dynamic range specification. In fact, you can’t establish a dynamic range specification without also establishing a S/N (residual noise) spec. This is why you occationally see them together.

So:
S/N = residual noise.
Maximum S/N = dynamic range

If I reduce the input level by half, that softest detail that use to pass through as the LSB of the usable dynamic range will now be lost in the fixed noise floor.

Exactly. Nothing profound there. Reducing the signal by half means much if it is lost in the noise floor (which by the way, considering the capabilities of the equipment we use these days, is inaudible). But all you’ve done there is limit the dynamic range of the signal. That has absolutely no affect on the capabilities (specifications) of the component. You can’t raise a component’s residual noise floor by reducing the level of an incoming signal.

This is true of both analog and digital components. Sure, with digital components, the bit reduction from a low signal means less available dynamic range - which is fine, because the weak signal isn’t calling for more. And it’s really no big deal if the LSB gets lost in an inaudible noise floor. (Bits and dynamic range will increase when the signal does.) But as with analog components, that doesn’t mean you’ve raised the (inaudible) noise floor. (Now, there can be a problem when you pass that now-limited dynamic range signal on to the next component, but that's a discussion for another day.)

Indeed, this is the problem with most of your post, freely intermingling the dynamic range of incoming program signals with that of a component.

I suggest taking a look at the Rane Audio Specifications paper. It goes into greater detail about how S/N ratio and dynamic range specs are established than the links I provided in the last post. You will plainly see that S/N ratio and residual noise are the same thing, but dynamic range is something different.

Regards,
Wayne

[Wayne A. Pflughaupt]

Good read on setting levels on digital audio processors – from people who use them for a living. Several pages, but there isn’t much need to read past the first page.

http://messageboa