MiniDSP DDRC-22D DIRAC measurements

[Rja4000]

Hi

After all those talks about AVR and their poor performances, I measured my MiniDSP DDRC-22D Dirac for SINAD, just to check what impact the DSP and Dirac can be in that story.

The MiniDSP DDRC-22D is stereo only.
Note that Dirac Stereo forces the sampling frequency to 96kHz, so that's how I measured it.
Also, this is not Dirac v2 yet.

The DDRC-22D is a purely digital device, so I measured from AES In to AES out, connected to RME ADI-2 Pro fs and with Virtins Multi-Instrument Pro 3.8.
(The RME is just used as an USB/AES converter here.)


SINAD Flat (No correction)

So no impact to speak about. Just that output signal is 10dB less, because it's compensated to allow level correction.

SINAD with correction active

We see an increase in noise floor, up to -125dB, which is around 115dB SINAD at our -10dBFS output level...
Nothing to cry wolf about.

DIRAC Correction FFR

We see some serious correction here.
(Don't worry about FFR artifacts)

Bottom line
In Stereo mode, with proper DSP implementation, Dirac is NOT degrading the SINAD value in any significant way.

 

[RichB]

Many processors are limiting Dirac to 48khz, do you feel that would have an impact on SINAD?

 

[Krunok]

Many processors are limiting Dirac to 48khz, do you feel that would have an impact on SINAD?

This seems like a noise related issue due to input attenuation to make headroom for correction. if thatt is so that would affect any room EQ system depending on how much attenuation is needed due to max filter boost.

 

[Rja4000]

This seems like a noise related issue due to input attenuation to make headroom for correction. if thatt is so that would affect any room EQ system depending on how much attenuation is needed due to max filter boost.

Yes, of course.
It could depend on the DSP type/implementation, I suppose. Not an issue nowadays.
I just wanted to make sure...

 

[Krunok]

Yes, of course.
It could depend on the DSP type/implementation, I suppose. Not an issue nowadays.
I just wanted to make sure...

In any case very nice write-up!

 

[Rja4000]

Many processors are limiting Dirac to 48khz, do you feel that would have an impact on SINAD?

I'm not sure if Dirac 5.1 or more is 96kHz or if it's still limited to 48kHz. Or if that's a limitation due to the DSP power.
The difference, if any, is possibly about some "side effects". You'd better have some margin to manipulate the sound, I suppose.
But I'm just guessing here.

 

[Julf]

The difference, if any, is possibly about some "side effects". You'd better have some margin to manipulate the sound, I suppose.

You need margin in the number of bits (to avoid rounding errors), less so with sample rate.

 

[Rja4000]

You need margin in the number of bits (to avoid rounding errors), less so with sample rate.

And to allow some room above noise for corrections....

 

[audimus]

Do you have a measurement of the RME bypassing the miniDSP altogether for a baseline to compare to the miniDSP with no correction?

At the same input level as what the miniDSP would have provided to the DAC.

 

[Rja4000]

Do you have a measurement of the RME bypassing the miniDSP altogether for a baseline to compare to the miniDSP with no correction?

At the same input level as what the miniDSP would have provided to the DAC.

Well, I can get it, but it will be the same.

The level of noise you see in the miniDSP without correction is basically the dithering noise of 24 bits signal, averaged on 5 successive frames.

 

[Rja4000]

When you read the raw figures, the impact of the DSP correction is actually huge, with a SINAD decrease of 25dB.
But the final SINAD is still good enough not to negatively impact output with any current DAC.

 

[Rja4000]

You need margin in the number of bits (to avoid rounding errors), less so with sample rate.

Sample rate would have an impact on latency, though.

 

[audimus]

Well, I can get it, but it will be the same.

The level of noise you see in the miniDSP without correction is basically the dithering noise of 24 bits signal, averaged on 5 successive frames.

You are probably right but it will confirm that it is indeed the case with the miniDSP or whether it is contributing something else unexpected.

Also a multi-tone test across the spectrum with and without Dirac correction would be useful to confirm what should be expected or if there are unexpected artifacts from correction than what just the 1khz test reveals.

Not saying you should, only if feasible and curious as me. Thanks.

 

[audimus]

When you read the raw figures, the impact of the DSP correction is actually huge, with a SINAD decrease of 25dB.
But the final SINAD is still good enough not to negatively impact output with any current DAC.

Another thing that is a concern is the calibration of the absolute value of SINAD. It is significantly higher than the results from the non-pro version of the RME used here. Why the apparent discrepancy?

The Dirac induced distortion is tolerable only because of this high level to start with. What if a DAC measures in the more typical 110-120db range that is considered quite good and Dirac pulls it down by 25db. Would we still be able to conclude that the correction impact is irrelevant?

 

[Rja4000]

Another thing that is a concern is the calibration of the absolute value of SINAD. It is significantly higher than the results from the non-pro version of the RME used here. Why the apparent discrepancy?

The Dirac induced distortion is tolerable only because of this high level to start with. What if a DAC measures in the more typical 110-120db range that is considered quite good and Dirac pulls it down by 25db. Would we still be able to conclude that the correction impact is irrelevant?

This is purely digital path here.
The same RME SINAD is around 113dB (around 0dBFS) when looped back with analog conversion (DAC->ADC), and this is pretty impressive.
But here, it's just used as an USB/AES converter... which is why the figures are much higher in 'no correction' mode.

So the limit, in real life situation, will be the digital to analog conversion (whatever the DAC, at around -10dBFS 1kHz) , not the DSP.
The 25dB SNR decrease is in digital domain, but digital noise level is still very low after that, so the final analog-level noise will still be mainly set by the conversion, IMO.

Said differently: the correction "consumed" 4 of the 24 bits, but current DACs can only reproduce 20 bits, so that's not an issue.

This is, basically, what I wanted to check.

 

[jmpsmash]

How about the digital volume control? I assume you have it at 0dB? What would the SINAD be if were to be used to control volume?

 

[Rja4000]

How about the digital volume control? I assume you have it at 0dB? What would the SINAD be if were to be used to control volume?

I suppose it is located after the DSP (I can't imagine a reason why they'd want it before).

Yes, the volume is at 0dB.
My DAC, behind it, has a digital volume control.

That's digital volume, so I don't think that makes a difference if it's set in the miniDSP or in the DAC.

If you lower the digital volume, the SINAD of the DAC will be lower, due to the DAC's noise.
So DSP's "noise" will have even lower impact.

 

[Julf]

Sample rate would have an impact on latency, though.

Sure, unless you decrease buffer size - but in most applications latency doesn't matter.

 

[guyinoz]

Hi, maybe an unrelated and novice question but if Dirac reduces input by 10db then is there an impact on the dac and amp going forward. For example, does the dac have to have something specials boost it so that the amp gets say 1.2V ? Or will any dac do this ?

 

[Rja4000]

The reduction is there to make room for some frequency boost.
So you can't expect to "correct" it in the DAC, since you'd probably overload it with corrected signal.