Recently I tested my Buffalo II DAC from Twistedpear - a 1kHz sine 0dB Test Tone.
There is a "squarelike" but faint distortion in the upper frequencies.
But before running a verdict, I should mention that the testtone is from 1982,
from an YEDS-18 Test disc made by Sony for service personnel.
I analysed the file in Audacity , one sample width shows clipping,
whereas 3 samples show no clipping.
No clue what the sample width practically means.
Assuming the sample is error free and spot on with full scale,
clipping needs at least two samples, because the resulting curve
between the two samples overshoots 0dB.
But the test tone on the YEDS 18 it was made while only 16bit ladder resistor DACs
were common, to my knowledge, clipping with them is more audible.
I analysed the file in Audacity , one sample width shows clipping,
whereas 3 samples show no clipping.
The 60 second sine testtone from the YEDS-18 is too big to be attached.
So here is a downloadlink from Wetransfer, it is valid for 7 days:
WeTransfer
Would be interesting to hear whether the fault is a test-signal from the dawn
of the Digital Age, or DACs that spoil a signal with no headroom while upsampling...
EDIT: Audacity also has a tone generator. The generated sinetone made by Audacity also "clips"
when being analysed with one sample of distance.
So this analyse tool seems to be useless or a 1KHz Tone at full scale clips by default,
as there are overshoots between samples...?
There is a "squarelike" but faint distortion in the upper frequencies.
But before running a verdict, I should mention that the testtone is from 1982,
from an YEDS-18 Test disc made by Sony for service personnel.
I analysed the file in Audacity , one sample width shows clipping,
whereas 3 samples show no clipping.
No clue what the sample width practically means.
Assuming the sample is error free and spot on with full scale,
clipping needs at least two samples, because the resulting curve
between the two samples overshoots 0dB.
But the test tone on the YEDS 18 it was made while only 16bit ladder resistor DACs
were common, to my knowledge, clipping with them is more audible.
I analysed the file in Audacity , one sample width shows clipping,
whereas 3 samples show no clipping.
The 60 second sine testtone from the YEDS-18 is too big to be attached.
So here is a downloadlink from Wetransfer, it is valid for 7 days:
WeTransfer
Would be interesting to hear whether the fault is a test-signal from the dawn
of the Digital Age, or DACs that spoil a signal with no headroom while upsampling...
EDIT: Audacity also has a tone generator. The generated sinetone made by Audacity also "clips"
when being analysed with one sample of distance.
So this analyse tool seems to be useless or a 1KHz Tone at full scale clips by default,
as there are overshoots between samples...?
Last edited:
You didn't say how your Buffalo dac is configured in terms of register settings and or in terms of hardware.
As far as clipping goes, intersample overs are a well known potential problem in modern dacs that was not a problem for all early 16-bit dacs. Those 16-bit dac with all-analog output filters rather than digital filters may have the headroom to handle intersample overs without clipping. In addition, some dac chips sound better if the the digital input to the dac chip is attenuated maybe 3dB - 6dB or so before going into the dac. That should help with intersample overs and possibly help with nonlinearity of some dac chips at high digital input levels. Intersample Overs in CD Recordings - Benchmark Media Systems
As far as clipping goes, intersample overs are a well known potential problem in modern dacs that was not a problem for all early 16-bit dacs. Those 16-bit dac with all-analog output filters rather than digital filters may have the headroom to handle intersample overs without clipping. In addition, some dac chips sound better if the the digital input to the dac chip is attenuated maybe 3dB - 6dB or so before going into the dac. That should help with intersample overs and possibly help with nonlinearity of some dac chips at high digital input levels. Intersample Overs in CD Recordings - Benchmark Media Systems
Frankly your comments are not useful and trivial.
The ES9018 based Buffalo II is more or less a breed from
Brian Donegan and Russ White here at DIYaudio. who funded Twistedpear.
The Buffalo II was developed around 2010.
Almost everybody knows that the Buffalo has I2s and SPDIF input, so 95%
of the sources will be CD, MD, DAT. No USB for what you propose, though I sometimes hook my Buffalo II on my DAW
for mixing purposes where -9dB headroom is a standard anyway.
But this is simlpy becaus my hardware is "dated" (but still perfect for 24bit/96kHz delivery),
and SPDIF was more common a decade ago, now it is USB.
So attenuation gives no benefit here
I personally deliver with -0.5dB for CD, -1dB for cinematic mixes
and even -1.5dB if AAC or MP3 are the delivery formats because overshoots
-in mixing-become very audible with heavy compression.
The ES9018 based Buffalo II is more or less a breed from
Brian Donegan and Russ White here at DIYaudio. who funded Twistedpear.
The Buffalo II was developed around 2010.
Almost everybody knows that the Buffalo has I2s and SPDIF input, so 95%
of the sources will be CD, MD, DAT. No USB for what you propose, though I sometimes hook my Buffalo II on my DAW
for mixing purposes where -9dB headroom is a standard anyway.
But this is simlpy becaus my hardware is "dated" (but still perfect for 24bit/96kHz delivery),
and SPDIF was more common a decade ago, now it is USB.
So attenuation gives no benefit here
I personally deliver with -0.5dB for CD, -1dB for cinematic mixes
and even -1.5dB if AAC or MP3 are the delivery formats because overshoots
-in mixing-become very audible with heavy compression.
Last edited:
I know what Buffalo is. I know ES9018 too. I know how they control the volume, and they way the do it does not prevent intersample overs. Worst case intersample overs occur at fs/4, or about 11kHz if the sample rate is 44.1k. In that case full scale on a typical DAW digital VU meter will be in error by 3.01dB. Thus 3.01dB would be clipped in that case. Some dac designers turn down the digital volume level before the dac chip by 3.5dB to prevent the problem from occuring. Since ESS dacs usually use the digital volume control inside the dac chip, turning down the level there does not prevent intersample overs from occuring clipping. In other words, my comments have been exactly right.
That assumes the worst case phase relationship with the sample clock. Practical signals don't have a phase relationship with the phase clock so a full-scale fs/4 tone should not be a problem, and an over level fs/4 tone is unlikely to happen to have such a phase relationship by chance. An 11kHz tone at full-scale is a tweeter- and eardrum-frying phenomenon I'd suggest!
The issue with intersamples overflowing is usually down to compression causing lots of signal peaks to be scaled to just at full-scale. Inevitably the upsamples of such peaks have to clip.
The solution is to record at correct levels, where peaks don't hit the roof (well only very very occasionally). The loudness wars are to blame methinks.
Today I tested my friends BUFFALO-IIISE PRO (ES9028/ES9038) 2-CHANNEL DAC. He recently installed the upgraded DAC daughter card on the IV converter board. The DAC now clips when the digital signal is above - 6 dB. This isn't subtle. The 1 kHz sinewave has flat top and bottom when the level is above -6 dB. Looking at the waveform at the interface between the DAC board and the IV converter board it is apparent that input to the IV board is clipping. The upgraded DAC chip, 9038, apparently has 50 ohm output impedance rather than the 200 ohm output impedance of the earlier 9028 chip and this supplies more current than the IV converter can handle. I was looking at the 750 ohm resistors in the feedback of the input stage. Those will provide 20 mA if the op amp could reach the 15 Volt rails. Apparently that is not sufficient. I was thinking of changing those 750 ohm resistors to maybe 270 ohm if the op amp can supply that much current. Interested if anyone else has solved this problem. The output voltage from this DAC is very high, 12V p2p at the line out. I didn't disassemble it to verify the IV board has 750 ohm resistors.
Last edited:
After scanning the web page for a while I found the single line of text that solves the mystery. "Gain is customizable with though-hole resistors, which are included for both ES9038 and ES9018/28 DAC chips." So that refers to R1 - R4, which are the 750 ohm resistors. As my friend bought his I to V board before the new DAC was released he didn't get the other value resistor. It looks like 180 - 220 ohms will work nicely. He's going to try the change.
