• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Topping D90SE Review (Balanced DAC)

dsnyder0cnn

Active Member
Joined
Dec 14, 2019
Messages
216
Likes
215
Nice @dsnyder0cnn. Suggest changing the Topping D90SE to Filter 5 "Fast-Roll Linear" (best IMO) to remove all that rippling in the frequency response from the apodizing filter (Filter 1).
Thanks for the tip. I plan to measure the amplitude and time-domain behavior of all seven filters…perhaps this evening. I'll definitely share plots here once I have done so.
 

dsnyder0cnn

Active Member
Joined
Dec 14, 2019
Messages
216
Likes
215
Finally got around to measuring the Topping D90SE and Gustard X16 filters in the stack setup. Here's the amplitude response for the seven different Topping filters:

Topping D90SE - 44.1 Filters.png


The colors that REW ended up picking are not great, but these are in order, from top to bottom:
  • Mode 1 - Fast roll off apodizing
  • Mode 2 - Slow roll off minimum
  • Mode 3 - Fast roll off minimum (default)
  • Mode 4 - Slow roll off linear
  • Mode 5 - Fast roll off linear
  • Mode 6 - Brick-wall
  • Mode 7 - Fast roll off corrected minimum
Here's the same plot for the Gustard's three filter choices:

Gustard X16 - 44.1 Filters.png


From the plots, it's clear that this is a proper subset of the Topping filter options:
  • L-FAST (same as Mode 5 - Fast roll off linear)
  • M-SLOW (same as Mode 2 - Slow roll off minimum)
  • H-FAST (same as Mode 7 - Fast roll off corrected minimum)
I'll just compare the time-domain response of these three, but if you're interested in looking at the others, feel free to download my REW *.mdat file here.

Here's Topping vs. Gustard (in that order) for Fast roll off linear / L-FAST. I'm showing the step and impulse response. For these plots, I'm looking at the length and degree of pre/post-ringing:

Topping D90SE - Mode 5 44.1.png

Gustard X16 - L-Fast 44.1.png


Not identical, but pretty darn close. Here's the same for Slow roll off minimum / M-SLOW:

Topping D90SE - Mode 2 44.1.png

Gustard X16 - M-Slow 44.1.png


Finally, here's Fast roll off corrected minimum / H-FAST:

Topping D90SE - Mode 7 44.1.png

Gustard X16 - H-Fast 44.1.png


Not sure how audible any of this stuff is, but of the three, H-FAST (Mode 7 on the Topping) appears to have no pre-ringing, while post-ringing is not terrible. From what I understand humans tend to be more sensitive to pre-ringing (which does not occur in nature except on Star Trek) than post-ringing, so this seems like a reasonable tradeoff.

At the end of the day, you get to choose which you prefer if you have one of these DACs. Again, feel free to download my REW data file to see the rest of the Topping filters. If you want to use my limits settings to match the captures here, select "%" instead of "dBFS" on the "Impulse" tab and use these limit settings:
Screen Shot 2021-09-24 at 1.04.22 AM.png

For the rest of my listening, I'll leave the Topping on Mode 7 and the Gustard on H-FAST. The amplitude response looks funky (when zoomed in), but I can't hear to 20 kHz anyway, so a half a dB of roll-off at 16 kHz doe not bother me in the least. Enjoy.

Edit: I've fixed the zip file containing my REW *.mdat file. Did this late last night and mistakenly used `gzip` instead of `zip`. Sorry. Also, I created these plots using the latest beta version of REW. I've not tested, but I would not be surprised if the production release of REW fails to open this file, which kind-of sucks. It only takes a few minutes to create an account on AV NIRVANA so that you can download the beta, but sorry for the inconvenience.
 
Last edited:

rlwings

Member
Joined
Nov 22, 2020
Messages
76
Likes
45
Finally got around to measuring the Topping D90SE and Gustard X16 filters in the stack setup. Here's the amplitude response for the seven different Topping filters:

View attachment 155337

The colors that REW ended up picking are not great, but these are in order, from top to bottom:
  • Mode 1 - Fast roll off apodizing
  • Mode 2 - Slow roll off minimum
  • Mode 3 - Fast roll off minimum (default)
  • Mode 4 - Slow roll off linear
  • Mode 5 - Fast roll off linear
  • Mode 6 - Brick-wall
  • Mode 7 - Fast roll off corrected minimum
Here's the same plot for the Gustard's three filter choices:

View attachment 155338

From the plots, it's clear that this is a proper subset of the Topping filter options:
  • L-FAST (same as Mode 5 - Fast roll off linear)
  • M-SLOW (same as Mode 2 - Slow roll off minimum)
  • H-FAST (same as Mode 7 - Fast roll off corrected minimum)
I'll just compare the time-domain response of these three, but if you're interested in looking at the others, feel free to download my REW *.mdat file here.

Here's Topping vs. Gustard (in that order) for Fast roll off linear / L-FAST. I'm showing the step and impulse response. For these plots, I'm looking at the length and degree of pre/post-ringing:

View attachment 155339
View attachment 155340

Not identical, but pretty darn close. Here's the same for Slow roll off minimum / M-SLOW:

View attachment 155341
View attachment 155342

Finally, here's Fast roll off corrected minimum / H-FAST:

View attachment 155343
View attachment 155344

Not sure how audible any of this stuff is, but of the three, H-FAST (Mode 7 on the Topping) appears to have no pre-ringing, while post-ringing is not terrible. From what I understand humans tend to be more sensitive to pre-ringing (which does not occur in nature except on Star Trek) than post-ringing, so this seems like a reasonable tradeoff.

At the end of the day, you get to choose which you prefer if you have one of these DACs. Again, feel free to download my REW data file to see the rest of the Topping filters. If you want to use my limits settings to match the captures here, select "%" instead of "dBFS" on the "Impulse" tab and use these limit settings:
View attachment 155348
For the rest of my listening, I'll leave the Topping on Mode 7 and the Gustard on H-FAST. The amplitude response looks funky (when zoomed in), but I can't hear to 20 kHz anyway, so a half a dB of roll-off at 16 kHz doe not bother me in the least. Enjoy.
This is awesome! Thank you very much for doing it! - So for us less technical folk would you say that the best filter to use is #7 or am I missing something here? (John seems to say #3 is best).
 

dsnyder0cnn

Active Member
Joined
Dec 14, 2019
Messages
216
Likes
215
This is awesome! Thank you very much for doing it! - So for us less technical folk would you say that the best filter to use is #7 or am I missing something here? (John seems to say #3 is best).
I'd be interested in hearing John's reasoning behind recommending #3. I did not show the plots for #3, and, it looks like my ZIP file is corrupted (sorry, will fix soon).

IIRC, there was quite a bit more pre-ringing in #3 than #7. I would say that the frequency response and extension is the best for #3, but my hypothesis is that time-domain differences have a greater impact on perceived sound quality than small differences in amplitude response, especially above 12 kHz. I could be totally wrong, but this seems to correlate well with personal subjective preferences.

Edit: With my zip file now working, here's what I see for Mode 3 on the D90SE:


D90SE - Mode 3.png


So, definitely more pre and post-ringing than Mode 7. Again, I confess that I have only hypothesised that there's a correlation between pre/post-ringing and personal preference. I'd love to study this further, but on paper, Mode 7 / H-FAST "looks" better. :)
 
Last edited:

Matias

Major Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
2,640
Likes
4,336
Location
São Paulo, Brazil

Evgeniy

Member
Joined
Aug 10, 2018
Messages
94
Likes
40
Location
Ukraine, Odessa .
In review https://translate.google.com/transl....ptt.cc/bbs/Headphone/M.1624709833.A.2B5.html
author wrote,
"On the analog side, the DAC is replaced by an ESS9038PRO. All 11 OPA1612s are used. The 1612 of the upper and lower clamp 9038 is used for linear voltage stabilization, instead of LDO ES9311 provides AVCC_R/AVCC_L 3.3V. The middle eight 1612s use 8CH respectively as I/V and then connect in parallel. Increase the output current and win loudly (I/V is not the same as the male board) The leftmost 1612 is used as LPF, which is responsible for converting to single-ended RCA, which means that the balanced terminal is I/V and straight out of XLR"

Hmm, XLR OUT don't have buffers and LPF ?
why?
 

Eddy H.

Member
Joined
Jul 29, 2021
Messages
49
Likes
29
In review https://translate.google.com/transl....ptt.cc/bbs/Headphone/M.1624709833.A.2B5.html
author wrote,
"On the analog side, the DAC is replaced by an ESS9038PRO. All 11 OPA1612s are used. The 1612 of the upper and lower clamp 9038 is used for linear voltage stabilization, instead of LDO ES9311 provides AVCC_R/AVCC_L 3.3V. The middle eight 1612s use 8CH respectively as I/V and then connect in parallel. Increase the output current and win loudly (I/V is not the same as the male board) The leftmost 1612 is used as LPF, which is responsible for converting to single-ended RCA, which means that the balanced terminal is I/V and straight out of XLR"

Hmm, XLR OUT don't have buffers and LPF ?
why?
What does this mean exactly?
 

MC_RME

Senior Member
Technical Expert
Manufacturer
Joined
May 15, 2019
Messages
485
Likes
2,074
It means broken translation. Apart from that John answered that earlier: the I/V stage is the 'buffer', which isn't needed additionally in this concept of the D90SE.
 

rlwings

Member
Joined
Nov 22, 2020
Messages
76
Likes
45
I'd be interested in hearing John's reasoning behind recommending #3. I did not show the plots for #3, and, it looks like my ZIP file is corrupted (sorry, will fix soon).

IIRC, there was quite a bit more pre-ringing in #3 than #7. I would say that the frequency response and extension is the best for #3, but my hypothesis is that time-domain differences have a greater impact on perceived sound quality than small differences in amplitude response, especially above 12 kHz. I could be totally wrong, but this seems to correlate well with personal subjective preferences.

Edit: With my zip file now working, here's what I see for Mode 3 on the D90SE:


View attachment 155428

So, definitely more pre and post-ringing than Mode 7. Again, I confess that I have only hypothesised that there's a correlation between pre/post-ringing and personal preference. I'd love to study this further, but on paper, Mode 7 / H-FAST "looks" better. :)
Thanks! - Hey did you did you see this yet? > https://archimago.blogspot.com/2021/09/measurements-review-topping-d90se-dac.html?m=1

According to this guy filter #5 would be best, lol... So confusing for us guys that just want to be told which filter to use. :)
 

dsnyder0cnn

Active Member
Joined
Dec 14, 2019
Messages
216
Likes
215

rlwings

Member
Joined
Nov 22, 2020
Messages
76
Likes
45
I did. He favors frequency extension, but I maintain that time-domain response contributes more to realism...especially for old guys like me who can't hear past 14 kHz anyway. :)
Makes sense, thanks. I wonder if choosing a certain filter would make a difference to guys like us that use Dirac room correction which claims to correct the time domain\impulse response. - Would choosing one filter over another be better for calibrating with Dirac algorithms? - Hmm...
 
Last edited:

rlwings

Member
Joined
Nov 22, 2020
Messages
76
Likes
45
I'd be interested in hearing John's reasoning behind recommending #3. I did not show the plots for #3, and, it looks like my ZIP file is corrupted (sorry, will fix soon).

IIRC, there was quite a bit more pre-ringing in #3 than #7. I would say that the frequency response and extension is the best for #3, but my hypothesis is that time-domain differences have a greater impact on perceived sound quality than small differences in amplitude response, especially above 12 kHz. I could be totally wrong, but this seems to correlate well with personal subjective preferences.

Edit: With my zip file now working, here's what I see for Mode 3 on the D90SE:


View attachment 155428

So, definitely more pre and post-ringing than Mode 7. Again, I confess that I have only hypothesised that there's a correlation between pre/post-ringing and personal preference. I'd love to study this further, but on paper, Mode 7 / H-FAST "looks" bette
I did. He favors frequency extension, but I maintain that time-domain response contributes more to realism...especially for old guys like me who can't hear past 14 kHz anyway. :)

Sorry for being so annoying but I really want to choose the best filter for the time-domain. I am not well versed in electronic testing and reading charts so I am a bit unclear as to which filter you are suggesting as the best time-domain filter... Would you be kind enough and tell me which number it is... Thank you very much! :)
 

Eddy H.

Member
Joined
Jul 29, 2021
Messages
49
Likes
29
I would like to see the multi tone test with the USB, optical, and coax individually to see the difference.
 
Last edited:

dsnyder0cnn

Active Member
Joined
Dec 14, 2019
Messages
216
Likes
215
Makes sense, thanks. I wonder if choosing a certain filter would make a difference to guys like us that use Dirac room correction which claims to correct the time domain\impulse response. - Would choosing one filter over another be better for calibrating with Dirac algorithms? - Hmm...
Good question. Fortunately (or not), time-domain errors with loudspeakers in a room tend to be an order of magnitude or so greater than what we're seeing with DAC filters, so filter choice when calibrating with Dirac is unlikely to matter. If it were me, I'd still choose the DAC filter with the least pre-ringing. :)
 

dsnyder0cnn

Active Member
Joined
Dec 14, 2019
Messages
216
Likes
215
I would like to see the multi tone text with the USB, optical, and coax individually to see the difference.
Yeah. @amirm didn't say which input he used for the multi-tone test, but I assume it was USB. I would not expect to see a significant difference from the other inputs.
 

rlwings

Member
Joined
Nov 22, 2020
Messages
76
Likes
45
Good question. Fortunately (or not), time-domain errors with loudspeakers in a room tend to be an order of magnitude or so greater than what we're seeing with DAC filters, so filter choice when calibrating with Dirac is unlikely to matter. If it were me, I'd still choose the DAC filter with the least per-ringing. :)
So which filter did you find had the least per-ringing? - (Sorry, I am not good at understanding graphs and tech stuff, lol)
 
Last edited:
Top Bottom