Review and Measurements of Grace Design m900 DAC & Amp

[amirm]

Pretty disappointing the USB low power mode has this problem. Do you think that with low impedance headphone, even at low volumes, fidelty is going to be good enough or just ok ?

No, I think it will be fine.

 

[amirm]

Could you test it as a DAC-only without an external power supply?

I did with the dashboard and it made no difference with or without.

 

[m8o]

P.s. when using it as a dac with outputs from the RCA connections, volume control should be set to 90. 90.5 activates the output opamps.

 

[amirm]

P.s. when using it as a dac with outputs from the RCA connections, volume control should be set to 90. 90.5 activates the output opamps.

I think we had it a lot lower than that to get 2 volts.

 

[m8o]

I think we had it a lot lower than that to get 2 volts.

In that case, from my memory of how Mr. Grace described that it works on Massdrop's forum during their first offering (ie 'drop'), there was a digital volume control that was attenuating the digital signal in play. Maybe that factors into the ultrasonic anomalies you observed? Dunno. Something to experiment with tho I imagine.

This goes back years at this point. But I'll try to find the posts that describe the operation of the mixed-mode digital + analog volume control. One does need to have a Massdrop account to be able to read it. But first, I have to find it.

 

[amirm]

In that case, from my memory of how Mr. Grace described that it works on Massdrop's forum during their first offering (ie 'drop'), there was a digital volume control that was attenuating the digital signal in play. Maybe that factors into the ultrasonic anomalies you observed? Dunno. Something to experiment with tho I imagine.

Performance is excellent in most of the measurements using that same setting. The tests that show poor results are due to ultrasonic tone at 40 kHz.

 

[m8o]

Performance is excellent in most of the measurements using that same setting. The tests that show poor results are due to ultrasonic tone at 40 kHz.

Indeed, they are. I am just spitballing if having the volume set to anything other than 90, i.e. unity gain for the output from the DAC chip, is a factor or source of that...

Massdrop's forum kinda stinks. It's hard to navigate. But I did get lucky with search. Here are some quotes from Michael Grace about the topic...

The RCA line outputs are variable. For reference, a volume setting of "90" is "unity" on the volume attenuator and will result in an output level of 2Vrms for a full scale digital signal (0dBFS). The maximum output level with the volume at "99." is +14dBV (5.1Vrms)
...
(this is a copy of a response to the same question on Head=Fi) The m9XX attenuator is a hybrid design. Most of the volume control duties are handled in the digital domain with 32 bit processing but there are two analog gain ranges. This allows a full 98dB of volume control range, preserves a very low noise floor for IEMs, and allows high peak output voltage for low efficiency planar magnetic phones.

EDIT: What is meant by two analog gain ranges above is digital attenuation with the DAC chip is used for volumes set between 0 - 90. When you turn the volume above 90 to 90.5 you hear a subtle click. The output gain stage is activated and an analog volume control is in play between 90.5 - 99.5.

This is a post from another person ( Caleb Roberts ) on the topic...

I suggest that there may be a minor misconception regarding operating m9xx at an "optimal" volume setting. It is true that setting the volume above 90 raises the noise floor because it raises the gain of the output amplifier. However, the noise floor is constant from 90 all the way down to 0. At the volume setting 90, the digital gain control in the AKM DAC IC is set to 0dB / full scale. I think that it is incorrect to suggest that there is a compromise in signal integrity if the gain is set below full-scale. The digital gain control in the AKM DAC IC is at least as accurate as the volume control in FB2k or other player apps. The AKM DAC IC has a 32-bit datapath and the dynamic range is greater than any analog circuitry and loudspeaker combination you can connect to its outputs. Note too that the channel matching of a digital gain control is far better than a stereo potentiometer. Listeners are often distracted by a bias in perception that louder signals seem to sound better.

All of this is only to say, there is at least some possibility that operating this unit with volume control at a setting other than 90 when using RCA out, 'could' be what is throwing those ultrasonic frequencies.

 

[amirm]

That's very useful info. I can try to play with that and see if that is the problem.

 

[Blumlein 88]

Indeed, they are. I am just spitballing if having the volume set to anything other than 90, i.e. unity gain for the output from the DAC chip, is a factor or source of that...

Massdrop's forum kinda stinks. It's hard to navigate. But I did get lucky with search. Here are some quotes from Michael Grace about the topic...



This is a post from another person ( Caleb Roberts ) on the topic...



All of this is only to say, there is at least some possibility that operating this unit with volume control at a setting other than 90 when using RCA out, 'could' be what is throwing those ultrasonic frequencies.

I like the way they've done the volume here. Pretty much digital to your standard 2v out for 0 db FS, and some useful analog gain which won't compromise SNR for those who could use more. I wish more devices worked like this.

 

[Ron Texas]

As for volume settings on the M900, I thought it worked best below 90, but going over 90 only happened with the line out. With the HD6xx I don't need to go anywhere near 90 on the volume control.

 

[HammerSandwich]

We have that large spike at 40 kHz. The more I increase the frequency, the taller it gets. To wit, the inset is at 16 kHz and by then the 40 kHz tone is at very high levels.

The frequency of the 40 kHz tone is not changing with source frequency. It is its amplitude that increases.

Perhaps I'm having trouble reading the inset, but it looks like enormous H2 at 32kHz with 40kHz gone.

 

[andreasmaaan]

I’m not sure I understand your explanation of the poor multi-tone performance @amirm, as the IM seems to be present right throughout the audio band. To clarify, are you of the view that these IM products within the audio band are actually the IM products of tones outside the audio band?

 

[Tatteredmidnight]

This is a very surprising, and, honestly, disappointing result from a company with a stellar reputation for quality engineering. I would have expected a boring review without the kind of anomalies we are seeing here. At this price point, you shouldn't be giving up performance to budget import units. It would be interesting to see how the SDAC compares at $79.

EDIT: on the new AP scope, I know it was already tested on the old setup.

 

[confucius_zero]

is the m9xx the same?

 

[amirm]

Perhaps I'm having trouble reading the inset, but it looks like enormous H2 at 32kHz with 40kHz gone.

You might be right. Let me fire up the analyzer and check. For some reason when cutting and pasting the graphs in this mode, the AP software deleted the X axis labels after 20 kHz.

 

[Blumlein 88]

Perhaps I'm having trouble reading the inset, but it looks like enormous H2 at 32kHz with 40kHz gone.

Good eye and good catch. There is no high 40 khz tone with the 16 khz input tone. It looks like high second harmonic as you said.
So whatever is happening is a bit more complex.

 

[amirm]

I’m not sure I understand your explanation of the poor multi-tone performance @amirm, as the IM seems to be present right throughout the audio band. To clarify, are you of the view that these IM products within the audio band are actually the IM products of tones outside the audio band?

I don't know for sure. They seem correlated to me seeing how they both showed serious issues with higher bandwidth and higher frequency tones which they both share.

FYI I dropped a note to Grace Design folks. Hopefully they see and respond.

 

[amirm]

OK, confirmed that my testing was at volume = 90 and without the extra gain. The moment I go up one click, SINAD drops 10 dB:

Since all of my testing was at unity gain/2 volt, then this is not an issue.

 

[restorer-john]

I like the way they've done the volume here. Pretty much digital to your standard 2v out for 0 db FS, and some useful analog gain which won't compromise SNR for those who could use more. I wish more devices worked like this.

I don't know why more don't either. The final buffer stage could have some mild variable gain instead of unity. The opamps used these days don't vary their bandwidth product or noise much for mild levels of gain. At least switchable.

 

[amirm]

OK, found the problem. I ran the white noise test and noticed there was no attenuation to speak of in the transition band! Checked the filter setting and it was set to "F4" ("= slow roll off, minimum phase Not linear phase in the pass band. Best for acoustic music without compression and artificially high levels of treble. Will have very low level of ringing caused by transients and ringing will be shifted to after the transient. )

I set the filter to F1 and now we are getting sane outputs:

There is still a rise in low frequencies that is not explained but the high frequency issue is gone.

Multitone issue remains which may be related to low frequency problem above:

Summary: using the F4 filter was causing severe aliasing due to not getting rid of out of band components.