MiniDSP Flex HTx Measurement (FW1.1.17 Update)

[sarieri]

This is a set of measurements of the Minidsp Flex HTx. It was sent to me by a friend. I have no affiliation to Minidsp. This set of measurements should not be directly compared to Amirm's test as I'm using Comos ADCiso here, but it should provide an general idea of performance of this gear.
I'm rather an amateur hobbyist so nothing is guaranteed here. Let me know if you think anything is wrong and I will look into it.

My measurements draw very little causal relation to how good this gear sound but rather how well this is engineered.

DA Performance:

The following measurements are measured from HTx's USB input.

4V SINAD:
Here is the 1kHz FFT at 0dBFS. This is measured with Cosmos APU(Analog Notch filter at 6dB), Scalar, and ADCiso.

Fair enough.

After FW1.1.17 Update, I got this:

Notice the huge drop in H2/H3!!!!

SINAD vs Generated Level:
This is measured with Cosmos APU(Analog Notch filter at 6dB), Scalar(autoranger), and ADCiso.

As expected.

Dynamics Range:
This is measured with Cosmos APU(60dB Preamp), A LPF, Scalar(autoranger), and ADCiso. It has to be measured with an LPF because Cosmos APU's preamp is extremely sensitive to HF noises so they have to be filtered out. A -60dB 1kHz sine wave was measured.

Someone might want to comment the little spikes around 5k. I have no idea what that is. My suspect is that it probabaly has something to do with APU's preamp.
Other than that, 64+60=124dB Dynamic Range. Pretty Good!

SMPTE vs Generated Level:
This is measured with Cosmos Scalar(autoranger mode), and ADCiso.

Our old friend ESS Hump here. An ESS engineer told me 9017s EVB does not have this hump. Should we blame ESS for not teaching all his customers how to get rid of this problem or should we blame Minidsp?

The Hump issue is also fixed in FW1.1.17:

Jitter Test:
This is measured with Cosmos Scalar(0dB as Impedance Bridge), and ADCiso. A 0dBFS 12kHz sine wave was measured.

Ugly, but I guess who cares?

Linearity:
This is measured with Cosmos Scalar(0dB as Impedance Bridge), and ADCiso. A set of 1kHz sine waves were measured.

SINAD vs Frequency:
This is measured with Cosmos Scalar(autoranger), and ADCiso. Without APU notch filter, it is expected that SINAD measured at 1kHz is slightly lower than the aforementioned 1kHz SINAD test.

[Revision Log]:
02/27/2024 First version
02/27/2024 Revision 1: Correct SMPTE Test
03/08/2024 Revision 2: Add SINAD vs Frequency Test. Some speculation on Multitone issue (See post #2).
03/19/2024 Revision 3: Updated measurements of FW 1.1.17. Sadly, the Multitone issue still persist.

 

[sarieri]

Multitone USB:
This gear has 0dBFS Multitone issue. To address the 0dBFS Multitone issue, here is my setting in REW.

Notice the crest factor is 13.3dB so it should not clip at -11.4dBFS. With all output gain set to 0dB and output volume set to 0dB, I get this messy result:

If I toggle channel under test(in this case CH2) to -1dB WITHOUT changing gains of any other channel

Now, if I further tweak all output channel's gain to be -1dB. The issue comes back exactly like before, so I don't need to show you any graph anymore.

Multitone Toslink:

I'm able to get less messy result at -10dB level. With higher gains, I always get messy results.

Update:
Inspired by @mdsimon2 in his thread here: https://www.audiosciencereview.com/...sp-measurements-and-rising-noise-floor.42383/
I figured that Htx is still applying volume attenuation AFTER ASRC. So it is possible that part of the Multitone bug is due to clipping. Sadly, at this moment, there is no way to attenuate the input signal in HTx.

Update 2:
Here is a set of measurements of multitone when EQ/Crossover profiles of LX521 are applied. The purpose of this set of measurements is to check whether the noise floor rise. In this test, four different channels (T, UM, LM, W) are fed into Cosmos Scalar(0dB as impedance bridge) and then into Cosmos ADCiso. Notice the LX521 dsp profiles are attenuated to different levels in four channels(0dB, -4dB, -12dB, -6dB), so different channels are compared to Reference Multitone attenuated to corresponding levels.

This is surprisingly good. The noise floor does not rise except in Woofer Channel. Notice the patterns of the rise in noise floor in Woofer Channel looks similar to the Multitone bug I described above, so there is a chance it can be solved through a firmware update.

 

[sarieri]

ADDA Performance:

Analog input is from D6s at vol96(4V)

4V SINAD:

Dynamic Range:

SMPTE vs Generated Level:
To be updated

Linearity:

Frequency Response:

There seems to be +-0.1dB drift from average. Normal?

 

[sarieri]

DA Performance from SPDIF/Toslink Input:

Toslink input from SMSL PO100pro.

4V SINAD:

Dynamic Range:

SMPTE vs Generated Level:
To be updated

Jitter:

 

[sarieri]

Some random thoughts:

HTx has only around 4.5ms latency and most importantly its latency is fixed. For sure you can achieve better and cheaper results with things like camilladsp but most of those community offerings do not have a fixed latency. With THD+N at this level, it is a perfect substitue for ASP.

Further Investigations on ESS Hump:

After some investigation and measurements on SMSL D6s(which has no Hump measured by Amirm), I’m convinced that ESS Hump is never “solved”. Instead, the drastic rise in IMD is buried under the noise of AP. It still kind of matters since “solving” the problem at least pushes the Hump to lower levels but NO, it’s not solved.

 

[phofman]

HTx has only around 4.5ms latency and most importantly its latency is fixed.

Please is that analog-in -> analog-out latency, i.e. only latency of the actual device, excluding latencies of digital buffers of input/output measurement devices? Thanks.

EDIT: I see, probably from https://www.audiosciencereview.com/forum/index.php?threads/minidsp-flex-htx.49512/page-16 , i.e. analog - analog

 

[sarieri]

Please is that analog-in -> analog-out latency, i.e. only latency of the actual device, excluding latencies of digital buffers of input/output measurement devices? Thanks.

EDIT: I see, probably from https://www.audiosciencereview.com/forum/index.php?threads/minidsp-flex-htx.49512/page-16 , i.e. analog - analog

Yes, it’s Analog in-Analog out latency

 

[MAB]

Jitter Test:
This is measured with Cosmos Scalar(0dB as Impedance Bridge), and ADCiso. A 0dBFS 12kHz sine wave was measured.

Ugly, but I guess who cares?

Thanks and agreed.

 

[howard416]

Thanks for the great work!

In your view, is there a practical fix for the relatively high IMD when utilizing up to all 8 output channels?

 

[sarieri]

Thanks for the great work!

In your view, is there a practical fix for the relatively high IMD when utilizing up to all 8 output channels?

I believe to get rid of the ESS Hump, one has to change few resistor/capacitor value on the I/V circuit and calculate the bandwidth. I have no knowledge on I/V circuit of ESS9017s though. I reported this problem to the Minidsp Engineering team, but honestly, I doubt they will ever solve it. As far as I see it, all their ESS DACs have this problem, so either they do not care about the Hump, or they are not capable of solving the problem. Honestly, it would already surprise me if they ever solve the Multitone issue. I have big doubts on their engineering team.

From a practical point of view, it might not matter and this device is still among the best of its kind. From a technical point of view, however, they took advantage of the ESS technology and that’s pretty much it.

Obviously Minidsp is great at their job.

 

[sarieri]

Thanks for the great work!

In your view, is there a practical fix for the relatively high IMD when utilizing up to all 8 output channels?

It's fixed. Software update, nothing painful.

 

[howard416]

Your note still says otherwise?

"03/19/2024 Revision 3: Updated measurements of FW 1.1.17. Sadly, the Multitone issue still persist."

 

[sarieri]

Your note still says otherwise?

"03/19/2024 Revision 3: Updated measurements of FW 1.1.17. Sadly, the Multitone issue still persist."

IMD Hump is solved. Multitone issue persist. But from my perspective, IMD Hump is much more problematic.

 

[adLuke]

Hello @sarieri
I have tried to follow your discussion here but, alas, a lot of it is over my head. I believe it's a very thorough investigation of the capabilities of the HTx, but my limited knowledge makes my praise of little value
I would like to ask you how the problems you found can affect the quality of sound in a setting where miniDSP HTx is used to receive signals from digital sources (PC, TV, etc) and sends analogue signals to one or more amplifiers.
And how is it possible to alleviate the problems via REW (from PC) or other ways (for TV, fireTV, etc)?
Thanks in advance for your help.

 

[sarieri]

Hello @sarieri
I have tried to follow your discussion here but, alas, a lot of it is over my head. I believe it's a very thorough investigation of the capabilities of the HTx, but my limited knowledge makes my praise of little value
I would like to ask you how the problems you found can affect the quality of sound in a setting where miniDSP HTx is used to receive signals from digital sources (PC, TV, etc) and sends analogue signals to one or more amplifiers.
And how is it possible to alleviate the problems via REW (from PC) or other ways (for TV, fireTV, etc)?
Thanks in advance for your help.

I described two issues in this thread:
1. Multitone Issue;
2. IMD Issue.

Problem 2 is partially solved (I say partial because I found something new that I’d rather not share at the moment). The ESS Hump is gone for sure. Problem 1 is in investigation hopefully by minidsp.

Now, NONE of the problems I found are likely to be distinguishable double blinded from an objective perspective. Also, even though we have the multitone problem, at reasonable low volume level, this issue is pretty much none existent since the issue I addressed is pretty much gone once the volume is below -15dB. You can check Update2 of Post2 to verify this. I believe to fully fix this multitone problem and the other half of the IMD problem, Minidsp has to do something about the SHARC and maybe matrix router. I still believe there is a correlation between IMD performance and subjective sound quality judgment (but at this low level IMD, maybe not).

With that all being said, this device has no opponent in the market. Short FIXED delay(4.5ms). Top notch SINAD (>120dB). Top notch ADDA (114dB). HDMI input. And it’s only $1k USD. Keep in mind that Topping is selling their DM7 at $600 without so many features. Okto Dac8pro is also more expensive without a fixed delay.

Finally, my measurement is from the perspective of a perfectionist not from a hobbyist. I can’t imagine why someone would complain about the sound quality of this.

 

[sarieri]

And I know someone is waiting for PEQs from companies Topping. If you only need two channels then you should probably wait because D90III discrete will have PEQ on all digital input(except i2s) using XMOS chip. But if you need 8 channel PEQ+ADDA+dirac live+a good GUI, forget about it. I don’t foresee Topping releasing it in the next couple of years.

 

[Sokel]

Since you do analog measurements,can you please replicate a couple on the low's channel with couple of filters engaged like @mdsimon2 did in the digital domain?
It would be interesting to see the coloration between them.
(the ones on the end of the thread,with the fixed firmware)

 

[sarieri]

Since you do analog measurements,can you please replicate a couple on the low's channel with couple of filters engaged like @mdsimon2 did in the digital domain?
It would be interesting to see the coloration between them.
(the ones on the end of the thread,with the fixed firmware)

I will look into it tomorrow.

 

[adLuke]

I described two issues in this thread:
1. Multitone Issue;
2. IMD Issue.

Problem 2 is partially solved (I say partial because I found something new that I’d rather not share at the moment). The ESS Hump is gone for sure. Problem 1 is in investigation hopefully by minidsp.

Now, NONE of the problems I found are likely to be distinguishable double blinded from an objective perspective. Also, even though we have the multitone problem, at reasonable low volume level, this issue is pretty much none existent since the issue I addressed is pretty much gone once the volume is below -15dB. You can check Update2 of Post2 to verify this. I believe to fully fix this multitone problem and the other half of the IMD problem, Minidsp has to do something about the SHARC and maybe matrix router. I still believe there is a correlation between IMD performance and subjective sound quality judgment (but at this low level IMD, maybe not).

With that all being said, this device has no opponent in the market. Short FIXED delay(4.5ms). Top notch SINAD (>120dB). Top notch ADDA (114dB). HDMI input. And it’s only $1k USD. Keep in mind that Topping is selling their DM7 at $600 without so many features. Okto Dac8pro is also more expensive without a fixed delay.

Finally, my measurement is from the perspective of a perfectionist not from a hobbyist. I can’t imagine why someone would complain about the sound quality of this.

Thanks a lot for your reply. I had a sense of your assessment but hearing it loud and clear takes all the doubts away.
Thank you very much

 

[danstahl]

I described two issues in this thread:
1. Multitone Issue;
2. IMD Issue.

Problem 2 is partially solved (I say partial because I found something new that I’d rather not share at the moment). The ESS Hump is gone for sure. Problem 1 is in investigation hopefully by minidsp.

Now, NONE of the problems I found are likely to be distinguishable double blinded from an objective perspective. Also, even though we have the multitone problem, at reasonable low volume level, this issue is pretty much none existent since the issue I addressed is pretty much gone once the volume is below -15dB. You can check Update2 of Post2 to verify this. I believe to fully fix this multitone problem and the other half of the IMD problem, Minidsp has to do something about the SHARC and maybe matrix router. I still believe there is a correlation between IMD performance and subjective sound quality judgment (but at this low level IMD, maybe not).

With that all being said, this device has no opponent in the market. Short FIXED delay(4.5ms). Top notch SINAD (>120dB). Top notch ADDA (114dB). HDMI input. And it’s only $1k USD. Keep in mind that Topping is selling their DM7 at $600 without so many features. Okto Dac8pro is also more expensive without a fixed delay.

Finally, my measurement is from the perspective of a perfectionist not from a hobbyist. I can’t imagine why someone would complain about the sound quality of this.

I'm confused by the multi-tone issue...if no filters are applied I would expect no clipping until digital clipping at 0+ volumes. If filters are applied, then the clipping could occur at a negative volume: eg a filter that boosts volume at a frequency by 5db could clip at -5 master volume. Is my understanding correct? In which case I would know at exactly what volume it is "safe" to go to?