SMSL filters mislabeled on ES9039 DACs

[JIW]

Several SMSL DACs using versions of the ESS ES9039 chip have mislabeled filters. First, I will use the case of the DL200 using one ES9039Q2M chip to show this is the case and correctly identify the filters. Then, I will show that other SMSL DACs using ES9039 chips namely the D-6S, DO100 Pro and SU-X also have mislabeled filters and try to identify the filters.

The case of the DL200
Looking at the below picture from the SMSL DL200 marketing material on the SMSL website, it is clear that at least some filters are mislabeled. For instance, FL7 is labeled "Minimum phase slow roll-off" but clearly has a fast roll-off and FL6 is labeled "Minimum phase fast roll-off" but clearly has a slow roll-off. Since the roll-off for all other filters seems to fit the labels, this may simply be a flaw in the figure.

This is not so, however, as a look in the DL200 manual reveals.

The labels are similar but looking closely at the similar colours of FL1 and FL5, FL1 has a slighty darker blue and is thus a slow filter. Further, @Qu@rk has done some measurements of the DL200 filters in the DL200 thread (Link to post). The 34 kHz spike is a measurement artifact and not part of the DAC output. Clearly, FL1 is slow and FL5 is fast.

Looking at the data sheet for the ESS ES9039Q2M used in the DL200, it is clear that the SMSL has a different numbering of their filters compared to ESS while using the same names. Since SMSL only uses 7 of the 8 available filters, it is unclear which one is omitted.

Looking at the properties of the filters below, it is clear that ESS #4 is a fast filter and ESS #8 is a slow filter since the stop band that is the band with full attenuation begins at 0.55 x fs for #4 and only at 0.8 x fs for #8.

It is also clear, that ESS #2 is SMSL FL3 due to being fast with the shortest pass band to 0.44 x fs and longest stop band from 0.5 x fs. Since ESS #3 has the highest stop band attenuation at 115 dB, it is SMSL FL4. Since ESS #5 is the slow filter with the longest stop band from 0.75 x fs compare to 0.8 x fs of the other slow filters, it is SMSL FL6. Thus, SMSL FL1 is either ESS #7 or #8. This leaves ESS #1, #4 and #6 to correspond to SMSL FL2, FL5 and FL7.

Looking at the frequency responses in the data sheet it is clear that the figures for ESS #1 and ESS #7 are incorrect since they show a slow filter where there should be fast one et vice versa. Note that the figures go to 200 kHz while the other figures go to about 40 kHz.

Note the large stop band ripple for ESS #4. This clearly identifies it as SMSL FL5. This leaves ESS #1 and #6 to correspond to SMSL FL2 and FL7. Since the figure for ESS #1 is incorrect but ESS #1 has less stop band attenuation at 96 dB compared to 99 dB of ESS #6, it is likely that ESS #1 is SMSL FL2 and ESS #6 is SMSL FL7.

Looking at the impulse responses in the data sheet, ESS #1 to #7 behave as expected for linear and minimum phase filters while #8 has an asymmetric response but with some pre-ringing.

Thus, the impulse response can be used to tell if SMSL FL1 is ESS #7 or #8. However, ESS #1 and #6 can hardly be distinguished that way.

@Qu@rk has also been so kind as to measure the impulse responses of the DL200 filters (Link to post). Clearly, SMSL FL1 is ESS #8.

Thus, according to the data sheet, the filters with the lowest latency are SMSL FL2 and FL7.

In summary,
SMSL FL1 - Minimum phase slow roll-off low dispersion (ESS #8)
SMSL FL2 - Minimum phase (ESS #1)
SMSL FL3 - Linear phase apodizing fast roll-off (ESS #2)
SMSL FL4 - Linear phase fast roll-off (ESS #3)
SMSL FL5 - Linear phase fast roll-off low ripple (ESS #4)
SMSL FL6 - Linear phase slow roll-off (ESS #5)
SMSL FL7 - Minimum phase fast roll-off (ESS #6)


Other SMSL ES9039 DACs
SMSL's other DACs that use ES9039 chips where they publish filter frequency response figures are the D-6S using a single Q2M, DO100 Pro using two Q2M and SU-X using two MS Pro.

The filter figure from the D-6S manual seems like a copy of the one for the DL200. Since it has the same list of filters, the mislabeling is probably the same.

Amir has measured the D-6S (Link). This is the filter figure. Apart from only going to 30 kHz, it seems identical to the manual.

@dsnyder0cnn also did some measurements of a D-6S. Filters 2 and 7 are minimum phase as is evident from the impulse responses.

Thus, the D-6S seems to have the same filter mislabeling as the DL200.

The DO100 Pro manual has this figure for the filters. It seems to be a copy of the one for the DL200 but the date and time are different so it is a different measurement.

There are only two slow filters but three are listed. Also, clearly FL3 is ESS #2 which is not even listed. It may well be the case that the filters are the same as for the DL200.

Lastly, from the SU-X manual. Since it uses a different chip to the others, the filters may be different. Apart from no filter, it again seems like a copy of the DL200 figure.

Clearly, filter 5 is mislabeled as well as filter 6 and 7. I have been unable to find the MS Pro data sheet but found the one for the M Pro and Pro. If the filters are the same, filter 1 is correctly labeled as well as filters 3 and 4. Filter 5 is linear phase fast low ripple, filter 6 is linear phase slow while filter 8 is minimum phase slow or minimum phase slow low dispersion. Filters 2 and 7 are the minimum phase and minimum phase fast filters. Since filter 2 has a slightly longer pass band, it is the minimum phase filter and thus filter 7 is minimum phase fast filter.

SMSL DL200 manual: https://www.smsl-audio.com/themes/p...?file=/upload/portal/download/DL200Manual.pdf
SMSL D-6S manual: https://www.smsl-audio.com/themes/p...ml?file=/upload/portal/download/D6SManual.pdf
SMSL DO100 Pro manual: https://www.smsl-audio.com/themes/p...e=/upload/portal/undefined/DO100PROManual.pdf
ESS ES9039Q2M data sheet: https://www.mouser.de/datasheet/2/1082/ES9039Q2M_Datasheet_v0_1_3-3168946.pdf

SMSL SU-X manual: https://www.smsl-audio.com/themes/p...ml?file=/upload/portal/download/SUXManual.pdf
ESS ES9039M Pro and ES9039 Pro data sheet: https://www.mouser.de/datasheet/2/1082/ES9038PRO_Datasheet_v3_7-3074335.pdf

TLDR: For DL200 and probably also D-6S and DO100 Pro:

SMSL FL1 - Minimum phase slow roll-off low dispersion (ESS #8)
SMSL FL2 - Minimum phase (ESS #1)
SMSL FL3 - Linear phase apodizing fast roll-off (ESS #2)
SMSL FL4 - Linear phase fast roll-off (ESS #3)
SMSL FL5 - Linear phase fast roll-off low ripple (ESS #4)
SMSL FL6 - Linear phase slow roll-off (ESS #5)
SMSL FL7 - Minimum phase fast roll-off (ESS #6)

For SU-X assuming data sheet for M Pro and Pro applies to MS Pro:

SMSL FL1 - Off
SMSL FL2 - Minimum phase (ESS #1)
SMSL FL3 - Linear phase apodizing (ESS #2)
SMSL FL4 - Linear phase fast roll-off (ESS #3)
SMSL FL5 - Linear phase fast roll-off low ripple (ESS #4)
SMSL FL6 - Linear phase slow roll-off (ESS #5)
SMSL FL7 - Minimum phase fast roll-off (ESS #6)
SMSL FL8 - Minimum phase slow roll-off or Minimum phase slow roll-off low dispersion (ESS #7 or #8)

 

[nanook]

Thanks a lot for this detailed report. I will check if this in line with my findings on the D-6s (they are heavily wrong, that's for sure) and come back.
The DO400 nicely lines up with the other ES9039xxx based SMSL DACs ;-/

SMSL DO400 Fully Balanced Audio Decoder & Headphone Amplifier

Edit: Corrected comment for #8, Low Dispersion Dirac measurement and spectrum at Fs=44100 Updated slides for PCM filter options. According to my measurements the description of the PCM filters in the menu of the DO400 is wrong for several of the filter settings. Since people experiment with...

audiosciencereview.com

 

[JIW]

Thanks a lot for this detailed report. I will check if this in line with my findings on the D-6s (they are heavily wrong, that's for sure) and come back.
The DO400 nicely lines up with the other ES9039xxx based SMSL DACs ;-/

SMSL DO400 Fully Balanced Audio Decoder & Headphone Amplifier

Edit: Corrected comment for #8, Low Dispersion Dirac measurement and spectrum at Fs=44100 Updated slides for PCM filter options. According to my measurements the description of the PCM filters in the menu of the DO400 is wrong for several of the filter settings. Since people experiment with...

audiosciencereview.com

Interesting. I did not look at it because it has no figure for the filters in the manual only the list. There is a difference, however. Filters 1 to 4 are correct. Filter 5 is indeed the fast linear low ripple as the stop band ripple reveals. Filters 6 and 7 are the fast and slow minimum phase filters. So the correct list is

SMSL FL1 - Off
SMSL FL2 - Minimum phase (ESS #1)
SMSL FL3 - Linear phase apodizing (ESS #2)
SMSL FL4 - Linear phase fast roll-off (ESS #3)
SMSL FL5 - Linear phase fast roll-off low ripple (ESS #4)
SMSL FL6 - Linear phase slow roll-off (ESS #5)
SMSL FL7 - Minimum phase fast roll-off (ESS #6)
SMSL FL8 - Minimum phase slow roll-off (ESS #7)

Also, the data sheet for M Pro and Pro does not have properties, FR and IR for all filters apparently on the MS Pro.

Since the SU-X uses the same chip and uses the same list in the manual, it may have the same order of filters as the DO400.

 

[JIW]

I have found some videos showing the menu of the DO400 as well as the SU-X. They are the same in skipping the fast linear low ripple filter. The SU-X thus likely has the same mislabeling of filters as the DO400. I had my doubts the units with the filter name in writing on the device would have this error but alas they seem to do.

 

[nanook]

Thanks for making this nice overview. This was definitely missing in my pdf.

@SMSL-Mandy: The list JIW provided can be used as input for your engineering when they build a new firmware. They should probably double-check the register values they write to the DAC-chip because there are a few filters that are not easily distinguishable.

It's really a pitty. People do serious A/B listening to different filters and draw conclusions on which filter type they prefer - and in the end just the menu entries (LCD-screen) or manual is wrong because people did not spend 15 minutes to check if the FW-code matches.

As far as I saw in the test of the unofficial review of the Topping D50 III , Topping managed to make the menu entries correct. B.t.w. the Topping D50 III is a very interesting unit. I have not yet seen PCB photos, but from the photos they have on their website, the I/V- stage might be the fully-differential design with good common-mode suppression on the differential output that is shown in the ES9039PRO datasheet.

 

[JIW]

I have looked for more SMSL DACs with the ES9039 on the SMSL website. I found the D400 ES and the VMV D1se using one MS Pro. Amir has reviewed both. Both have mislabeling of the filters resulting from the fast linear low ripple filter being used but referred to as the slow linear.

 

[JIW]

Thanks for making this nice overview. This was definitely missing in my pdf.

@SMSL-Mandy: The list JIW provided can be used as input for your engineering when they build a new firmware. They should probably double-check the register values they write to the DAC-chip because there are a few filters that are not easily distinguishable.

It's really a pitty. People do serious A/B listening to different filters and draw conclusions on which filter type they prefer - and in the end just the menu entries (LCD-screen) or manual is wrong because people did not spend 15 minutes to check if the FW-code matches.

As far as I saw in the test of the unofficial review of the Topping D50 III , Topping managed to make the menu entries correct. B.t.w. the Topping D50 III is a very interesting unit. I have not yet seen PCB photos, but from the photos they have on their website, the I/V- stage might be the fully-differential design with good common-mode suppression on the differential output that is shown in the ES9039PRO datasheet.

Yes, the D50 III filters are labeled correctly.

I have also looked at the manual of the SU-10 with ES9038Pro which Amir reviewed. The list is a bit odd naming the fast linear twice.

However, the figure is all kinds of wrong. The order of the filters is also different from the list.

Strangely, however, in Amir's measurement everything is correct.

 

[nanook]

Either SMSL reuses a faulty code in many of their products, or the documentation provided by ESS is wrong (and Topping did measure and sort this out for the D50 III).

The documentation that ESS gives for the PCM filter (see below).
The "Linear phase fast roll-off low ripple" is missing in this early datasheet of the "PRO".

@IVX: You already made a DAC with the ES9039q2m.
--> Did you come across inconsistencies??

ES9039q2m (version 0.1.3):

ES9039MPRO (version 0.2.1):

 

[JIW]

Not quite. The frequency and impulse responses are in the expected place before the slow linear filter. It being omitted elsewhere and ESS claiming only seven filters are available is strange. But also if SMSL had programmed using the terms in the table, they could not have gotten the fast linear low ripple filter since it is selected by 3'd3 which is missing. It seems more likely SMSL missed the jump from 2 to 4 on the Pro models.

 

[nanook]

The datasheets I have are very (!) preliminary (version 0.x.x , so not even a release). No idea if there are more recent datasheets available.
I cannot really believe they omitted one of the filters in the PRO version. In the end it's just loading a different set of coefficients into the FIR registers.
--> Did anyone come across a more recent datasheet?

Also strange, the "linear phase fast roll-off low-ripple" does not show good attenuation in the datasheet graph. In my measurement it reaches 90dB attenuation as opposed to 60dB in the datasheet.

 

[JIW]

The datasheets I have are very (!) preliminary (version 0.x.x , so not even a release). No idea if there are more recent datasheets available.
I cannot really believe they omitted one of the filters in the PRO version. In the end it's just loading a different set of coefficients into the FIR registers.
--> Did anyone come across a more recent datasheet?

Also strange, the "linear phase fast roll-off low-ripple" does not show good attenuation in the datasheet graph. In my measurement it reaches 90dB attenuation as opposed to 60dB in the datasheet.

I have version 0.2.1 as well as can be seen from the header. All other fast linear phase filters also only reach about 60 dB but are specified to 107 and 118.

 

[JIW]

I found version 0.3.2 on the ESS website (Link). The fast linear phase low ripple has been completely removed but can probably still be accessed as in the Q2M using 3'd3. The fast linear phase filters also show the right amount of attenuation.

 

[IVX]

Either SMSL reuses a faulty code in many of their products, or the documentation provided by ESS is wrong (and Topping did measure and sort this out for the D50 III).

The documentation that ESS gives for the PCM filter (see below).
The "Linear phase fast roll-off low ripple" is missing in this early datasheet of the "PRO".

@IVX: You already made a DAC with the ES9039q2m.
--> Did you come across inconsistencies??

ES9039q2m (version 0.1.3):
View attachment 377698

ES9039MPRO (version 0.2.1):
View attachment 377699

I have abandoned ES9039Pro proto due to too high distortions at 10kHz(H3 -110db), so I use 9039Q2M only(H3 -135db at 10kHz) - better and 10x cheaper.
BTW, the ES9039Q2M filter list is the same as ES9822Pro.

 

[Sir Sanders Zingmore]

That explains the vast number of posts across the internet complaining that SMSL filters don’t sound correct

 

[nanook]

I found version 0.3.2 on the ESS website (Link). The fast linear phase low ripple has been completely removed but can probably still be accessed as in the Q2M using 3'd3. The fast linear phase filters also show the right amount of attenuation.

They had to drop one filter in order to fit the "filter off" into 3 bits.
The filter graphs look a lot different and more reasonable in the version 0.3.2. I will update my comparison.

 

[nanook]

I have abandoned ES9039Pro proto due to too high distortions at 10kHz(H3 -110db), so I use 9039Q2M only(H3 -135db at 10kHz) - better and 10x cheaper.
BTW, the ES9039Q2M filter list is the same as ES9822Pro.

Your measurements on the ES9039PRO are perfectly in line with my measurements on the DO400.
1 KHz: H2, H3 ca. -135dBc
10 kHz: H2, H3 ca. -110dBc

SMSL DO400 Fully Balanced Audio Decoder & Headphone Amplifier

This is useful thanks. I'm also hyper treble sensitive and get ear pain for a couple days after wearing some of my too bright headphones. Would apodizing also be best for this? Thanks. In this case, it makes sense to try "Filter Off".

audiosciencereview.com

PS: Measured with your APU and ADCiso. Thanks a lot - I so happy with these units!

 

[JIW]

They had to drop one filter in order to fit the "filter off" into 3 bits.
The filter graphs look a lot different and more reasonable in the version 0.3.2. I will update my comparison.

No. The bypass is done using other bits in register 90 instead of 88. Also, how could SMSL get the fast linear low ripple filter if it cannot be selected?

The same is the case in the Q2M.

 

[IVX]

Your measurements on the ES9039PRO are perfectly in line with my measurements on the DO400.
1 KHz: H2, H3 ca. -135dBc
10 kHz: H2, H3 ca. -110dBc

SMSL DO400 Fully Balanced Audio Decoder & Headphone Amplifier

This is useful thanks. I'm also hyper treble sensitive and get ear pain for a couple days after wearing some of my too bright headphones. Would apodizing also be best for this? Thanks. In this case, it makes sense to try "Filter Off".

audiosciencereview.com

PS: Measured with your APU and ADCiso. Thanks a lot - I so happy with these units!

you are welcome, however, I said -135db at 10kHz has ES9039Q2M, at 1kHz -150db or about. One of my sample of 9039S shows THD -149db@1kHz 0dbfs! It is Victor's OSC performance level.

 

[JIW]

Since the ES9039 Pro chip may cause some issues, I have looked for implementations from other manufacturers - Topping and Loxjie. Firstly, Topping who use it in both the D70 Pro Sabre and D90 III Sabre. For the former, the manual gives the list of filters as

and the marketing material on the website has this figure

Amir has also measured it and gets

This all seems correct. The D90 III Sabre has the same list of filter but a different default according to the manual

and both the manual and the marketing material on the website have this figure

This is wrong. F-1 is a slow filter but should be minimum phase (ESS #1) which has fast roll-off, F-5 is the linear phase fast filter. Amir has also measured it and gets

This is also incorrect but consistent with Topping's figure. F1 and F2 are the slow minimum phase filter. F3 and F7 are the minimum phase fast and minimum phase filters. Going by the stop band ripple, F3 is probably the minimum phase fast filter and F7 is the minimum phase filter. Assuming F4-F6 are between F3 and F7 in increasing order, i.e. F4 then F5 then F6, F4 is indeed the slow linear filter, F5 is the fast linear filter and F6 is the linear apodizing filter. Clearly, the default filter is not the fast linear phase but instead minimum phase fast or minimum phase.

Curiously, Topping can implement the filters properly on one device but not another. More so, it is the device that was released later that is faulty.

Loxjie use the ES9039 Pro in the D40 Pro. This is the list of filters from the manual

and this is the figure immediately below it

This is correct until FL4 which is the fast linear low ripple and not as stated the slow linear. Instead FL5 is the slow linear, FL6 is the minimum phase fast and FL7 is either minimum phase slow or minimum phase slow low dispersion. This is the same error as on the SMSLs/VMV using the Pro chip. Amir has also measured the D40 Pro and gets

This is identical to Loxjie's figure and thus confirms the error.

Topping D70 Pro Sabre manual: https://dl.topping.audio/usermanual/d70p_s_en.pdf
Topping D90 III Sabre manual: https://dl.topping.audio/um/d90iii_sabre.pdf
Loxjie D40 Pro manual: https://drive.google.com/file/d/1BHviYGLgJvYf_TycT9KZbxpOV1g_uZfH/view?pli=1

 

[JIW]

The filters of the newly released SMSL RAW-MDA1 using two ES9039Q2M are also labeled incorrectly. The figure from the manual is the same as for the DO100 Pro as can be seen from the date and time and the error is the same.

SMSL RAW-MDA1 manual in this post: https://www.audiosciencereview.com/forum/index.php?threads/s-m-s-l-raw-mda1.55037/post-2015381