It’s a gaffe , there’s no swerving it.... ( unless we have made a gaffe ) .
Now where’s my man servant, Amir the bathroom towels need airing ..
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Review and Measurements of Benchmark DAC3
- Thread starter amirm
- Start date
Now where’s my man servant, Amir the bathroom towels need airing ..
OP
- Thread Starter
- #122
OK, looks like the old instructions were correct too in that if you hold both mute and dim buttons down, it sets both Aux and Mains to calibrated levels (as opposed to one or the other). The trick was holding them down for a lot longer. Either I don't know how to count to three or they don't! 
Regardless, here are the USB linearity tests comparing Aux and Mains, in calibrated mode which outputs nearly 2 volts RMS:
(I stopped the test around -60 dB).
As we see, both Mains and Aux do the same thing. And they even less accurate than uncalibrated mode.
I can also see the differential in level between channels as I watch the output per before.
Regardless, here are the USB linearity tests comparing Aux and Mains, in calibrated mode which outputs nearly 2 volts RMS:
(I stopped the test around -60 dB).
As we see, both Mains and Aux do the same thing. And they even less accurate than uncalibrated mode.
I can also see the differential in level between channels as I watch the output per before.
I asked this earlier, I suppose it got overlooked in all the excitement...
What's the difference between a linearity measurement where the left end of the measurement goes down (-dB) versus one that goes up (+dB)?
What's the difference between a linearity measurement where the left end of the measurement goes down (-dB) versus one that goes up (+dB)?
OP
- Thread Starter
- #124
So, drilling into this, I looked at the analyzer output using high resolution FFT and high averaging, comparing the Topping D50 to DAC3 in calibrated mode:
This is post my custom filtering and attenuates noise and distortion (but doesn't eliminate them).
We see that the D50 in blue is essentially free of harmonic distortion. But suffers from USB 1 kHz packet spikes.
The DAC3 on the other hand, despite the very low amplitude of main signal, has a strong spray of harmonics of 200 Hz. The linearity test shows the sum total of noise+harmonic distortion added to mains signal. So it is affected by the much higher harmonic distortion energy.
Note that the Topping D50 has higher main tone so its distortions in a relative measure (which linearity is), is lower than they seem.
Comments? Thoughts?
This is post my custom filtering and attenuates noise and distortion (but doesn't eliminate them).
We see that the D50 in blue is essentially free of harmonic distortion. But suffers from USB 1 kHz packet spikes.
The DAC3 on the other hand, despite the very low amplitude of main signal, has a strong spray of harmonics of 200 Hz. The linearity test shows the sum total of noise+harmonic distortion added to mains signal. So it is affected by the much higher harmonic distortion energy.
Note that the Topping D50 has higher main tone so its distortions in a relative measure (which linearity is), is lower than they seem.
Comments? Thoughts?
OP
- Thread Starter
- #125
As I explained earlier, a snapshot is made at -20 dB and that is used as a reference evel. Negative values means the sum total of noise+distortion+signal was lower than -20 dB level. Positive values means the sum was higher.
OP
- Thread Starter
- #128
This is using 256 k point FFT so there is 50 dB of process gain/lowering of noise going on there. Actual level then is around -120 dB. I am also using 16 times filtering to smooth the noise floor further.
OP
- Thread Starter
- #129
I will try again
.You have a yardstick. You measure it fully. Then you divide it in half and measure it again and see if it is exactly 1/2 of the original size. You keep doing that over and over again.
The "yardstick" here is programmable and is set by default to whatever level is there at -20 dB. Since linearity errors there are way off in the noise, that is a good reference for what the original signal as lowered by -20 dB should be. Everything else is then relative to that.
The AP will make a single measurement at -20 dB, then will drop it down to -120 dB and see if the measured voltage is higher or lower than than what the math says the level should be 100 dB lower. Then it steps up 2 dB at a time (-118, -116, -114) and plots the results. The outcome of the device is either exactly right or is higher or lower. Higher is positive dB error, negative is lower dB error.
OP
- Thread Starter
- #130
So it is clear, the -20 dB point measurement allows Linearity to work regardless of the actual voltage output of the DACs under test. They are all being tested against their own near full amplitude level and see if they can divide the signal correctly.
OP
- Thread Starter
- #131
I changed it before taking note
. But I think it was in USB 1.1. It is now in 2.0. Quick look at linearity shows nothing changed.
Now I understand.
*shouts a Thank You.
For the approximate 2562 miles we are distant, that's -132dBfs at your end.
Did you hear me?
Oh wait...
2562 / 767.2 = 3.3
It won't get there for another 3.3 hours. Mark your calendar.
OP
- Thread Starter
- #133
Now you see the silliness of others trying to replicate my measurements but go to -140 dB!!!
Ultimately we are trying to see if we can get perfection out of 20 bits of audio reproduction as to have a completely transparent and noise-free "channel."
Does it mean DC offset or very low frequency like < 5Hz can affect the result?I will try again
You have a yardstick. You measure it fully. Then you divide it in half and measure it again and see if it is exactly 1/2 of the original size. You keep doing that over and over again.
The "yardstick" here is programmable and is set by default to whatever level is there at -20 dB. Since linearity errors there are way off in the noise, that is a good reference for what the original signal as lowered by -20 dB should be. Everything else is then relative to that.
The AP will make a single measurement at -20 dB, then will drop it down to -120 dB and see if the measured voltage is higher or lower than than what the math says the level should be 100 dB lower. Then it steps up 2 dB at a time (-118, -116, -114) and plots the results. The outcome of the device is either exactly right or is higher or lower. Higher is positive dB error, negative is lower dB error.
OP
- Thread Starter
- #135
There is a 50 dB filter around the two sides of the 200 Hz so contributions from outside of that are sharply reduced.
The coupling is AC so DC offset should not enter the equation.
I have thoughts. On the topping d50, you have one dac per channel, and on the benchmark dac 3 balanced, you have one dac for hot and cold of each channel, and linearity is great. Do these dacs just not like being combined in other ways?
With the dac 3, they switched to a switching power supply. Is that somehow particularly affecting the unbalanced output?
With the dac 3, they switched to a switching power supply. Is that somehow particularly affecting the unbalanced output?
Last edited:
- Joined
- Feb 23, 2016
- Messages
- 22,735
- Likes
- 41,971
Seems like this should be linearity+noise like thd+noise versus linearity and thd.
OP
- Thread Starter
- #138
Correct. It is the main point I made to AP. That you can't exclude what is actually in the signal the user gets in linearity measurements. Noise and distortion should be included to the extent it doesn't screw up the analyzer too badly.
OP
- Thread Starter
- #140
I don't have a lot of insights as to actual problem here. My guess is that the high-power output stage amplification is at issue here due to seeing a lot of harmonic distortion.
On the power supply bit, I don't see contributions from it. I think they are doing a good job there.
