Topping DX5II Balanced DAC and Headphone Amp Review

[ngs428]

I spoke too soon, my PEQ settings are still flaky

I’ve had my custom EQ profiles saved to the device vanish. Luckily I have them saved as txt files so they are easy to import. I am in the latest TT and 1.67. In conversations with Topping on this now.

 

[ziggurcat]

I’ve had my custom EQ profiles saved to the device vanish. Luckily I have them saved as txt files so they are easy to import. I am in the latest TT and 1.67. In conversations with Topping on this now.

The dummy profile I created weeks ago is still there. Absolutely nothing has happened to it, nor have any of the default profiles returned, and it's the only PEQ profile that is listed. It will not allow me to delete it unless a second profile is created, too. It was created by clicking the + symbol under User Config, and altering the three default filters that are there when the profile is created.

 

[davmol]

Just received mine a few days ago, updated to latest firmware. I'm not using peq functionality because it's connected via usb to my WiiM ultra to provide a headphone option for the main hifi, and this was a good dimensional / aesthetically pleasing option to stack on the ultra.

Well, I'm blown away, picking it up for £220 from topping store (also got through with no vat added - winning!). It really feels like it shouldn't be possible to produce something with this quality and functionality at this price point. I'm running the D50III and A50III stack in my office and that was nearly twice the price in total. Yes the stack has a better finish / quality, but the DX5II wins the functionality and usability race, with the added bonus of less cable clutter to deal with, being a single box. I might be tempted to change the stack at a later date for another DX5II, just for these reasons (obviously they sound no different to each other, both having great specs). Good times!

 

[pieterv1]

Just received mine a few days ago, updated to latest firmware. I'm not using peq functionality because it's connected via usb to my WiiM ultra to provide a headphone option for the main hifi, and this was a good dimensional / aesthetically pleasing option to stack on the ultra.

Well, I'm blown away, picking it up for £220 from topping store (also got through with no vat added - winning!). It really feels like it shouldn't be possible to produce something with this quality and functionality at this price point. I'm running the D50III and A50III stack in my office and that was nearly twice the price in total. Yes the stack has a better finish / quality, but the DX5II wins the functionality and usability race, with the added bonus of less cable clutter to deal with, being a single box. I might be tempted to change the stack at a later date for another DX5II, just for these reasons (obviously they sound no different to each other, both having great specs). Good times!

View attachment 475879

Lovely compact setup! That's pretty much how my configuration would look like. Do you have the DAC connected over USB? How does it react to volume changes from the Ultra? Does the Ultra control the DAC's volume, with visual feedback on the screen?

And what's the name of the cabinet?

 

[davmol]

Lovely compact setup! That's pretty much how my configuration would look like. Do you have the DAC connected over USB? How does it react to volume changes from the Ultra? Does the Ultra control the DAC's volume, with visual feedback on the screen?

And what's the name of the cabinet?

Yes it's connected to the WiiM ultra over usb. Part of the reason I chose the DX5II was also to use the balanced out to the Fosi V3 mono amps. I don't hear any difference vs se/ RCA input (didn't expect to) but I just figured I may as well get maximum value from the purchase, rather than it just sitting there for occasional headphone use.

I have it running in DAC mode (max output), and the volume is controlled by the WiiM, which is where the visual confirmation of changes will be seen. The DX5II has display options for L/R meters at the bottom of the screen, plus a choice of vu (not much good in my opinion) and "FFT" meter (which I quite like), so there is visual feedback of the volume changes but it's in the form of those meters:

Volume control is great, from comfortable at 30 to very loud (but still enjoyable) at 60, towards being uncomfortably loud at 80 (wiim goes to 100). Obviously this will be different depending on amps and speaker sensitivity, but with the fosi V3 mono and my Linton 85th, it's great and I think it sounds incredible (I'm using room correction/ peq on the Wiim, which also made a huge contribution to the sound quality).

The cabinet is the Fitueyes Eiffel RS and I'm very happy with it. The style and build quality is good, and it's hard to ask for more at the price point £160 ish after their discount code is applied. My friend has the larger model and is also very happy with it.

 

[Spiral_Architect]

I got my unit delivered and would like to share some impressions. Just updated to latest firmware.
The screen and the visualizations are incredibly nice. A bit underwhelmed by the fit and finish. The E30/L30 look more solidly build but in the end it doesn't matter.
The Topping Tune software is absolutely abysmal. I have disabled the auto run on my PC hoping that it will not interfere with functionality. I plan to regularly measure the frequency response to check that no strange filters have been added.

I tried to compare the measured frequency response of onboard PEQ vs. Equalizer APO (see below, sorry for the low resolution but it should be readable). The matching is pretty close, considering the slight mismatch in shelving filters definition and rounding to 0.5db of the gains i think it is satisfactory.

Multitone response with this "normal" filter does not seem as bad as shown in the extreme situation that was discussed a few pages ago.
Measurements were made at 48khz sampling rate.

 

[jkim]

Interesting measurements. You mention that the one shown is a worst case scenario. Is it because three filters are cascaded ? Does this perform better in a more typical usage scenario?

Regarding the low shelf, is there a suitable value of the Q factor that matches the equalizer aaPO implementation with Q0.71 ?

Sorry I missed this post.

Cascading multiple peak filters at low frequencies incurs a bit larger effect, but the main factors are the low signal frequency and high sample frequency. As long as a peak filter is at a low frequency and the sample frequency is high, the quantization error increases. However, I don't think the amount of error is not so large as to be an audible concern---it is definitely less than the case of pure 32-bit floating point computation.

As for the shelving filters, it seems that Q = 1.0 is interpreted as Q = 0.71. That is, if you want Q = 0.71, enter Q = 1.0.

 

[Dandelion]

Sorry I missed this post.

Cascading multiple peak filters at low frequencies incurs a bit larger effect, but the main factors are the low signal frequency and high sample frequency. As long as a peak filter is at a low frequency and the sample frequency is high, the quantization error increases. However, I don't think the amount of error is not so large as to be an audible concern---it is definitely less than the case of pure 32-bit floating point computation.

As for the shelving filters, it seems that Q = 1.0 is interpreted as Q = 0.71. That is, if you want Q = 0.71, enter Q = 1.0.

You have to take the butterworth q and divide it by / sqrt(0.5) to get the value to enter in this eq. The quantization issue is fairly unproblematic unless people do crazy eqing, the more relevant problem is the cramping that occurs in these products (same happens in the RME) at 44/48kHz. Honestly I just implore you guys to get a proper software eq, it costs like 20 bucks.

 

[Timbo2]

You have to take the butterworth q and divide it by / sqrt(0.5) to get the value to enter in this eq. The quantization issue is fairly unproblematic unless people do crazy eqing, the more relevant problem is the cramping that occurs in these products (same happens in the RME) at 44/48kHz. Honestly I just implore you guys to get a proper software eq, it costs like 20 bucks.

In my use case it’s not connected to a device that provides EQ. That’s a large reason why the built in PEQ is useful for me even with the many documented issues.

 

[jkim]

In my use case it’s not connected to a device that provides EQ. That’s a large reason why the built in PEQ is useful for me even with the many documented issues.

We should note that a peak filter's cramping can be an issue only in a limited situation that checks all these boxes:

• The sampling frequency (Fs) is 44.1 or 48 kHz;
• AND the filter's center frequency (Fc) is high relative to its Q (or the Q is low relative to its Fc);
• AND most importantly, the user is not aware of the filter's significantly cramped response, hence not trying to compensate for it to obtain the desired response.

These days computing power required by biquad PEQ at a 96k/192kHz Fs is not demanding even for a microprocessor (or dedicated audio DSP core) adopted by an inexpensive DAC. Or for some reason if a user still wants to use a 44.1k/48kHz Fs, there is a workaround like combining a peak filter with a high-shelf filter as shown here:

 

[Timbo2]

We should note that a peak filter's cramping can be an issue only in a limited situation that checks all these boxes:

• The sampling frequency (Fs) is 44.1 or 48 kHz;
• AND the filter's center frequency (Fc) is high relative to its Q (or the Q is low relative to its Fc);
• AND most importantly, the user is not aware of the filter's cramped response.

These days computing power required by biquad PEQ at a 96k/192kHz Fs is not demanding even for a microprocessor (or dedicated audio DSP core) adopted by an inexpensive DAC. Or for some reason if a user still wants to use a 44.1k/48kHz Fs, there is a workaround like combining a peak filter with a high-shelf filter as shown here.

Thanks. That’s a useful as well as appropriately brief video!

 

[Dandelion]

We should note that a peak filter's cramping can be an issue only in a limited situation that checks all these boxes:

• The sampling frequency (Fs) is 44.1 or 48 kHz;
• AND the filter's center frequency (Fc) is high relative to its Q (or the Q is low relative to its Fc);
• AND most importantly, the user is not aware of the filter's cramped response.

These days computing power required by biquad PEQ at a 96k/192kHz Fs is not demanding even for a microprocessor (or dedicated audio DSP core) adopted by an inexpensive DAC. Or for some reason if a user still wants to use a 44.1k/48kHz Fs, there is a workaround like combining a peak filter with a high-shelf filter as shown here:

Ish... If you make a bell or shelf in high frequencies you'll get cramping, even if you're aware of it it's still not the sound that you should get especially if you're using measurements to correct listening devices or rooms. I know Dan Worrall, he talks from a mix engineer perspective, which is different. The point is that the cramping could easily just not be there, they can simply internally oversample operations of the EQ (but that would increase further the quantization noise) or compensate behind the scenes the eq response to neutralize the cramping.

 

[jkim]

Ish... If you make a bell or shelf in high frequencies you'll get cramping, even if you're aware of it it's still not the sound that you should get especially if you're using measurements to correct listening devices or rooms.

Of course, cramping occurs whether the user is aware or not. The point is, if the user is clearly aware of it and wants to correct it, there is a way (actually multiple ways) to do so. The third condition of an affected situation may be clarified like:

• AND most importantly, the user is not aware of the filter's significantly cramped response, hence not trying to compensate for it to obtain the desired response (underlined part added in an edit).

But again, we are still talking about a limited situation. Simple use of a 96k/192kHz Fs is a straightforward solution unless PEQ is implemented in pure float32 or lower-bit precision. The quantization noise level of the Topping DX5 II or the D50 III is practically not an audible concern, although not ideal.

And fortunately I find that onboard PEQ implemented even on cheap dongles these days is free of measurable quantization error. More reviews will come...

 

[Spiral_Architect]

You have to take the butterworth q and divide it by / sqrt(0.5) to get the value to enter in this eq. The quantization issue is fairly unproblematic unless people do crazy eqing, the more relevant problem is the cramping that occurs in these products (same happens in the RME) at 44/48kHz. Honestly I just implore you guys to get a proper software eq, it costs like 20 bucks.

0.7/sqrt(0.5) is essentially equal to one, that is what I tried but still found some differences in measured response with respect to equalizerAPO.
Today I tried to play with the EQ that I need for the HE400se, based on oratory1990 suggested settings. I compared APO with topping trying to match the response as close as possible by hand. There are two low shelf filters that in EQAPO are defined as follows:

APO:
Filter 1: ON LS Fc 30,0 Hz Gain 4.0 dB Q 0,71
Filter 2: ON LS Fc 105,0 Hz Gain 5,50 dB Q 0,71

The result didn't match by just setting Q=1 (as it should be theoretically ?) I even tried with a single filter and it appears that Q=1 Topping is still different than Q=0.71 eqAPO.
I found a very similar response by tuning the gains and frequency a little bit, which is quite strange, probably coincidental ?

"Matching" Topping:
Filter 1: ON LSC Fc 30 Hz Gain 4.5 dB Q 1
Filter 2: ON LSC Fc 130 Hz Gain 5.5 dB Q 1

The rest of the response matches very well and is identical (I have rounded the gains to .5dB also in the APO filter, to remove that confounding factor from the comparison).

I gave a look to the multitone response with this particular filter and it looks fairly good.

For what concerns software stability, so far i did not encounter lost settings or filters that modify themselves...but as a safety measure I have run a cable from the RCA outputs to my PC line in and will periodically check the frequency response. Measurements are noisier than with my old trusted EMU0404 interface but are sufficient for a quick comparison.

 

[Spiral_Architect]

Of course, cramping occurs whether the user is aware or not. The point is, if the user is clearly aware of it and wants to correct it, there is a way (actually multiple ways) to do so. The third condition of an affected situation may be clarified like:

But again, we are still talking about a limited situation. Simple use of a 96k/192kHz Fs is a straightforward solution, unless PEQ is implemented in pure float32 or lower-bit precision. The quantization noise level of the Topping DX5 II or the D50 III is practically not an audible concern, although not ideal.

And fortunately I find that onboard PEQ implemented even on cheap dongles these days is free of measurable quantization error. More reviews will come...

Is a simple free software EQ like EqualizerAPO free from this cramping effect ? If so, I think that if measured frequency response matches that reference it should be safe to assume that everything is working as intended ?
Also upsampling everything to 96khz is not particularly problematic with most music players. For stuff like games and general system/media sounds I just don't care.

Software EQ like equalzier APO was nice. However it has its shortcomings like when I switch from phones to speakers on the dac/amp...windows has no notion of this, so I needed to remember to turn on and off the EQ. Painful and error prone. This is why I prefer this hardware solution where PEQ is associated to each output and is standalone. Even if I play some streaming music via bluetooth from my phone I have the proper EQ. It's a significant advantage for me.

 

[jkim]

0.7/sqrt(0.5) is essentially equal to one, that is what I tried but still found some differences in measured response with respect to equalizerAPO.

Make sure other settings are identical to each other. In my measurements, the response of D50 III's low-shelf filter with Q = 1.0 is identical to that of EQ APO's with Q = 0.71.

EDIT. I found that you used the "LS" rather than "LSC" command in EQ APO. The "LS" command has no Q factor, and the frequency is not interpreted as the "center frequency" but as the "corner frequency." The two frequency definitions are different.

Is a simple free software EQ like EqualizerAPO free from this cramping effect ? If so, I think that if measured frequency response matches that reference it should be safe to assume that everything is working as intended ?

Most software and hardware PEQ solutions in consumer audio, including EQ APO, are based on the old RBJ biquad formulation; so, they all yield EQ cramping when the sampling rate is low.

 

[Spiral_Architect]

Make sure other settings are identical to each other. In my measurements, the response of D50 III's its low-shelf filter with Q = 1.0 is identical to that of EQ APO's low-shelf filter with Q = 0.71.

Ok i'll try again with a single low shelf filter in the chain. I tested that with an older topping firmware i don't know if it plays a role. With the newest firmware I only tested the combination of filters not the shelving alone. What other settings are there ? Topping tune has none besides Q and gain, are there different options in EQ apo that might not match ?

EDIT: thanks for the info, that explains why I needed to apply a frequency shift...

Now they match perfectly. Used LSC in both.

Most software and hardware PEQ solutions in consumer audio, including EQ APO, are based on the old RBJ biquad formulation; so, they all yield EQ cramping when the sampling rate is low.

Ok. I'll look into some of the material for the sake of learning but for this specific usage in the end there is no "ground truth" unless you are in full control also of the headphone measuring pipeline and you know precisely what exact frequency response you want, with or without countermeasures to compensate for this effect. Here I am just copying presets that come from measurements and were calculated and tested with "some" EQ implementation that might or might not had cramping, so I will never know. I have to trust the suggestion based on measurements not performed by myself and if it sounds good enough..."probably" I am not deviating too much from the intended frequency response.

 

[Minato]

Some time ago a user posted photos of the label on their DX5 II with a grammatical error and a power consumption rating of 15W.
On my DX5 II the label has been updated and now reads "Power Consumption: 30W".
In reality the power consumption is much lower. The 30W refers to the maximum power of the internal AC/DC, 15V - 2000mA, from the teardown photo.

 

[jmdesignz2]

Some time ago a user posted photos of the label on their DX5 II with a grammatical error and a power consumption rating of 15W.
On my DX5 II the label has been updated and now reads "Power Consumption: 30W".
In reality the power consumption is much lower. The 30W refers to the maximum power of the internal AC/DC, 15V - 2000mA, from the teardown photo.

View attachment 476196
View attachment 476197
View attachment 476198

Huh. I just got mine and it has that old label. It’s silver

 

[Viper780]

My white one also has the old lable - which hardware revision do you have?