chosen_too
Member
- Joined
- Oct 3, 2018
- Messages
- 94
- Likes
- 48
Thanks! Listening d50 to for two month without any complains, less fatiguing than Allo Boss for me because of more transparent sound.
-
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!
ESS THD ‘Hump’ Investigation
- Thread starter jackenhack
- Start date
Thanks! Listening d50 to for two month without any complains, less fatiguing than Allo Boss for me because of more transparent sound.
Veri
Master Contributor
- Joined
- Feb 6, 2018
- Messages
- 9,599
- Likes
- 12,042
Yep, D50 with THX AAA or Atom is fairly unfatiguing
glad you enjoy yours
graz_lag
Major Contributor
Topping DX7
Veri
Master Contributor
- Joined
- Feb 6, 2018
- Messages
- 9,599
- Likes
- 12,042
Why Mojo? Top performers are: OPPO UDP-205(discontinued/defunct comany), SMSL D1, Chord Qutest. All are in the $1k+ range though lol.
The mojo performed well but in fact still worse than khadas/el dac/d50/enog2/grace sdac... (so many budget contenders really, bless healthy competition).
DX3 is the current best value all-in-one performer dac+amp. The DX7 is great yes, but the original non-s model would be rather hard to find. For DAC-only DX7s performance, the SU-8 is quite a lot cheaper.
Last edited:
miero
Active Member
- Joined
- Aug 1, 2018
- Messages
- 241
- Likes
- 292
Probably the most expensive DAC/CD/SACD player with ES9028Pro chips available (~19000 EUR):
- http://www.accuphase.com/model/dp-750.html
- http://www.accuphase.com/technical_information/dp-750_technical_information.pdf
- http://www.accuphase.com/model/dp-750.html
- http://www.accuphase.com/technical_information/dp-750_technical_information.pdf
maxxevv
Major Contributor
- Joined
- Apr 12, 2018
- Messages
- 1,872
- Likes
- 1,964
The NX4 DSD too being a direct competitor to the Mojo as its also portable.
maxxevv
Major Contributor
- Joined
- Apr 12, 2018
- Messages
- 1,872
- Likes
- 1,964
caguilar91
Member
- Joined
- Nov 6, 2018
- Messages
- 63
- Likes
- 52
Looks like a Bakelite radio from that eraWhy does the aesthetics remind me of a late 70's era product though ?
OP
- Thread Starter
- #149
I liked their earlier design better, with the hand crank and megaphone on top.Looks like a Bakelite radio from that era
caguilar91
Member
- Joined
- Nov 6, 2018
- Messages
- 63
- Likes
- 52
RCA Victor?
Shangri-La
Member
- Joined
- Jan 2, 2019
- Messages
- 90
- Likes
- 32
In lower price range of course... What backs up your statement of khadas/el dac/d50/enog2/grace sdac outperform the Mojo? Not questioning you dont get wrong. Just genuinely curious to know. The Mojo is more 'main stream' and gets a lot more exposure, and that comes at a cost. It was reviewed here but the measurements were different back then so difficult to compare to current products. I totally belive in technology advancement especially DACs and really want to know how new contenders fair against well-known products.
Veri
Master Contributor
- Joined
- Feb 6, 2018
- Messages
- 9,599
- Likes
- 12,042
Just raw distortion characteristics (its jitter is admittedly fairly faultless). Its SINAD is only 102dB which is not quite on the 108-110+ level of competiting products that are also cheaper..
Also read this post if you haven't yet
https://www.audiosciencereview.com/...-thresholds-of-amp-and-dac-measurements.5734/Shangri-La
Member
- Joined
- Jan 2, 2019
- Messages
- 90
- Likes
- 32
Just raw distortion characteristics (its jitter is admittedly fairly faultless). Its SINAD is only 102dB which is not quite on the 108-110+ level of competiting products that are also cheaper..
Also read this post if you haven't yet
I'm a firm believer of measurements do translate to real-world listening experience. And that thread of interpreting measurements is awesome beyond belief. Thank you!
Shangri-La
Member
- Joined
- Jan 2, 2019
- Messages
- 90
- Likes
- 32
Just raw distortion characteristics (its jitter is admittedly fairly faultless). Its SINAD is only 102dB which is not quite on the 108-110+ level of competiting products that are also cheaper..
Also read this post if you haven't yet
Going by that thread, the "ESS hump" is well below lenient limits and really should be a non-issue then.
Veri
Master Contributor
- Joined
- Feb 6, 2018
- Messages
- 9,599
- Likes
- 12,042
Yes, that's pretty much guaranteed except if you listen at ear-piercingly loud levels
as far as I see/interpret it at least.5th element
Member
- Joined
- Feb 2, 2019
- Messages
- 42
- Likes
- 152
I have been working in DIY audio for years and reading this thread prompted me to make some additional distortion vs amplitude measurements of one of my builds.
It is a multichannel DSP unit that uses an ES9018 for 8 of its channels.
In terms of hardware configuration the ES9018 is configured so that its outputs give out a current (it can give a voltage). This differential output current is then fed into a pair of opamps that perform I/V conversion. Naturally you've now got a differential voltage but due to the way the ES9018 (and most current output DACs) works these have a voltage offset about which the bipolar signal swings. To convert this into a single ended output you need to sum the differential output. This cancels out the voltage offset common to the differential outputs. This is how you are supposed to use the DAC to create a single ended output. You are most certainly not supposed to simply connect the hot to a single ended socket.
When measured as configured the ES9018 presented with the hump spoken of. No surprises here really, this is a text book implementation and the ES9018 doesn't really have anything to configure with regards to distortion cancellation, or does it?
One other way in which this DSP unit uses the ES9018 is to drive a pair of headphones. So in one configuration I've got 8 single ended outputs and in the other I want just two.
Now the traditional way to use an ES9018 for just two channels is to parallel up the outputs so that instead of just one DAC per channel, you use 4. This gives you 4x the output current and improves the theoretical SnR of the DAC providing other noise sources don't dominate. Obviously the individual DACs are identical. If you go this route you've hardware locked yourself into a two channel corner, which wouldn't work for me as I needed 8 separate channels. So what did I do?
The ES9018 has the ability to invert the polarity of the outputs of its DACs (register configurations). So when I wanted to use just two channels I told it to invert some of them. Normally I had 8 single ended channels of the same polarity. 4 running the left channel, 4 running the right. So I inverted two of them, one left and one right. Now I had 3 channels running the left and 3 running the right with the same polarity, but also a pair of inverted left + right. Effectively 4 channels were now configured to output stereo balanced outputs. So I added in an extra opamp after each balanced pair to sum these down to single ended again to drive the headphones. 4 channels mixed down to 2.
Now here's the interesting thing, these channels don't really exhibit the ESS THD hump, or at least not to anywhere near the same degree.
First up we have an over-layed plot of all 8 individual SE ended channels. These are post a CS3318 multichannel volume control chip. As these are for a pair of active loudspeakers the volume control chip sets the levels for the different drivers. I didn't reset the individual volumes to equal so there's a spread of about 8dB from the loudest to quietest channel, this is why the traces diverge. One channel, for whatever reason, has a nasty peak present that the others do not, shrug. I don't think this is hump related.
Second we have the same graph but with some drawing in white over the top. One is a freehand sketch showing a rough trend of the bump and then I've placed a straight line there to show what we'd ideally be seeing.
And lastly we have the two channels configured as headphones, with the DAC channels of reversed polarity, then combined with a channel of normal polarity.
With one set of channels there is virtually zero hump and with another pair a much smaller amount of hump.
It would seem that inverting one set of output channels and summing them acts to cancel out whatever mechanism creates the hump. Channel matching will of course be very important in how well the hump cancels out of course. For example on the 8 channel measurement, one stands out with that odd peak up to 0.05% that would not work well for cancellation.
So what is happening here?
Normally speaking you have a DAC output, this is balanced but with a common current offset. You convert the current into a voltage, which then has/needs an opamp following to cancel out the common offset and to sum the opposite signals. Common signals cancel whereas opposite ones add. We end up with a distortion bump. Clearly the bump isn't common to the way the balanced current output is created but in fact the non-linearity is present in both signals and with its phase opposite for the two outputs. So when they are summed you end up with the non-linearity still present.
If you flip the polarity of one of the entire DAC channels then you send your signal through but in antiphase to the non-flipped channel. This would appear to be software related and the DSP part of the DACs front-end just inverts the data entering the DAC. The interesting thing here is that the hump mechanism would appear to be hardware related in that this non-linearity ends up coming out of the inverted DAC channel with the same signal magnitude and sign as the non inverted DAC channel. If you now send these two SE signals through a second opamp, the opposite signals add and the common hump non-linearity gets cancelled out.
So how do you solve this problem for a standard DAC configuration without the additional op-amp added in? This is my best guess.
Take an ES9018, it has 8 DAC channels. For a stereo DAC you have 4 for the left and 4 for the right. Usually all you do is connect the 4hot current outputs together and the 4cold outputs together and end up with a current with 4 times the magnitude, but you still get a hump, all channels are connected with the same phase. My proposal would be to reverse the polarity of 2 of the 4 channels. Then you connect 2hot and 2 inverted cold outputs together, for the 4hot signals and 2cold + 2 inverted hot signals together for the 4cold signals.
The hump non-linearity is essentially equal but opposite on the differential outputs of a single DAC channel, but so is the signal, so everything adds together at the differential to SE opamp. Well then just invert the signal in software to one DAC channel. The hump non-linearity doesn't invert when you tell it to invert the signal because it's a hardware problem. So you then connect the DAC outputs together but hot to cold, cold to hot. The reversal of the paralleled connection means the hump non-linearity now cancels out in the additive paralleling of the current outputs and as you've reversed the signal the actual signals, the things you want, add together constructively. Bye bye hump.
This, in theory, is how I think it could potentially work but some DAC experts would need to chime in with regards to if this makes any kind of actual sense!
It is a multichannel DSP unit that uses an ES9018 for 8 of its channels.
In terms of hardware configuration the ES9018 is configured so that its outputs give out a current (it can give a voltage). This differential output current is then fed into a pair of opamps that perform I/V conversion. Naturally you've now got a differential voltage but due to the way the ES9018 (and most current output DACs) works these have a voltage offset about which the bipolar signal swings. To convert this into a single ended output you need to sum the differential output. This cancels out the voltage offset common to the differential outputs. This is how you are supposed to use the DAC to create a single ended output. You are most certainly not supposed to simply connect the hot to a single ended socket.
When measured as configured the ES9018 presented with the hump spoken of. No surprises here really, this is a text book implementation and the ES9018 doesn't really have anything to configure with regards to distortion cancellation, or does it?
One other way in which this DSP unit uses the ES9018 is to drive a pair of headphones. So in one configuration I've got 8 single ended outputs and in the other I want just two.
Now the traditional way to use an ES9018 for just two channels is to parallel up the outputs so that instead of just one DAC per channel, you use 4. This gives you 4x the output current and improves the theoretical SnR of the DAC providing other noise sources don't dominate. Obviously the individual DACs are identical. If you go this route you've hardware locked yourself into a two channel corner, which wouldn't work for me as I needed 8 separate channels. So what did I do?
The ES9018 has the ability to invert the polarity of the outputs of its DACs (register configurations). So when I wanted to use just two channels I told it to invert some of them. Normally I had 8 single ended channels of the same polarity. 4 running the left channel, 4 running the right. So I inverted two of them, one left and one right. Now I had 3 channels running the left and 3 running the right with the same polarity, but also a pair of inverted left + right. Effectively 4 channels were now configured to output stereo balanced outputs. So I added in an extra opamp after each balanced pair to sum these down to single ended again to drive the headphones. 4 channels mixed down to 2.
Now here's the interesting thing, these channels don't really exhibit the ESS THD hump, or at least not to anywhere near the same degree.
First up we have an over-layed plot of all 8 individual SE ended channels. These are post a CS3318 multichannel volume control chip. As these are for a pair of active loudspeakers the volume control chip sets the levels for the different drivers. I didn't reset the individual volumes to equal so there's a spread of about 8dB from the loudest to quietest channel, this is why the traces diverge. One channel, for whatever reason, has a nasty peak present that the others do not, shrug. I don't think this is hump related.
Second we have the same graph but with some drawing in white over the top. One is a freehand sketch showing a rough trend of the bump and then I've placed a straight line there to show what we'd ideally be seeing.
And lastly we have the two channels configured as headphones, with the DAC channels of reversed polarity, then combined with a channel of normal polarity.
With one set of channels there is virtually zero hump and with another pair a much smaller amount of hump.
It would seem that inverting one set of output channels and summing them acts to cancel out whatever mechanism creates the hump. Channel matching will of course be very important in how well the hump cancels out of course. For example on the 8 channel measurement, one stands out with that odd peak up to 0.05% that would not work well for cancellation.
So what is happening here?
Normally speaking you have a DAC output, this is balanced but with a common current offset. You convert the current into a voltage, which then has/needs an opamp following to cancel out the common offset and to sum the opposite signals. Common signals cancel whereas opposite ones add. We end up with a distortion bump. Clearly the bump isn't common to the way the balanced current output is created but in fact the non-linearity is present in both signals and with its phase opposite for the two outputs. So when they are summed you end up with the non-linearity still present.
If you flip the polarity of one of the entire DAC channels then you send your signal through but in antiphase to the non-flipped channel. This would appear to be software related and the DSP part of the DACs front-end just inverts the data entering the DAC. The interesting thing here is that the hump mechanism would appear to be hardware related in that this non-linearity ends up coming out of the inverted DAC channel with the same signal magnitude and sign as the non inverted DAC channel. If you now send these two SE signals through a second opamp, the opposite signals add and the common hump non-linearity gets cancelled out.
So how do you solve this problem for a standard DAC configuration without the additional op-amp added in? This is my best guess.
Take an ES9018, it has 8 DAC channels. For a stereo DAC you have 4 for the left and 4 for the right. Usually all you do is connect the 4hot current outputs together and the 4cold outputs together and end up with a current with 4 times the magnitude, but you still get a hump, all channels are connected with the same phase. My proposal would be to reverse the polarity of 2 of the 4 channels. Then you connect 2hot and 2 inverted cold outputs together, for the 4hot signals and 2cold + 2 inverted hot signals together for the 4cold signals.
The hump non-linearity is essentially equal but opposite on the differential outputs of a single DAC channel, but so is the signal, so everything adds together at the differential to SE opamp. Well then just invert the signal in software to one DAC channel. The hump non-linearity doesn't invert when you tell it to invert the signal because it's a hardware problem. So you then connect the DAC outputs together but hot to cold, cold to hot. The reversal of the paralleled connection means the hump non-linearity now cancels out in the additive paralleling of the current outputs and as you've reversed the signal the actual signals, the things you want, add together constructively. Bye bye hump.
This, in theory, is how I think it could potentially work but some DAC experts would need to chime in with regards to if this makes any kind of actual sense!
^^^ Very interesting. Perhaps this is what @John_Siau was hinting about in the DAC3 thread without divulging their secret sauce recipe?
- Joined
- Feb 13, 2018
- Messages
- 574
- Likes
- 720
Similar threads
- Poll
