LDAC measurements

[mike7877]

I've seen LDAC measurements done on some mobile DACs, but far from all of them.

I'm interested to know if the reason that it's not often tested is because results are always THD+n being -93dB +-3dB.
To my memory, that's about what I've seen, when I've seen it.

Redundant? Overlooked? Overlooked on purpose because nobody cares about Bluetooth because USB's right there?

When out and about, there's definitely no way anyone's going to discern -93 from anything better. Maybe once they've reached their destination with a completely silent, reflection treated room and high fidelity amplification and speakers, but not on the street or in a vehicle of any kind... That's where LDAC would be used most IMO. It's where I'd use it.

Asking because I recently got a Topping G5 so I'm wondering, as long as it's been properly implemented, its LDAC performance will be in the -90s?
No future DAC will improve upon LDAC THD+n being in the lower -90s?

 

[ocinn]

Amir measured 112-113 on LDAC. See: https://www.audiosciencereview.com/...c3-review-bluetooth-receiver-bt-codecs.23740/

Multitone performance is very poor however: https://www.audiosciencereview.com/...receiver-bt-codecs.23740/page-12#post-1039526

 

[Trell]

I've seen LDAC measurements done on some mobile DACs, but far from all of them.

I'm interested to know if the reason that it's not often tested is because results are always THD+n being -93dB +-3dB.
To my memory, that's about what I've seen, when I've seen it.

Redundant? Overlooked? Overlooked on purpose because nobody cares about Bluetooth because USB's right there?

When out and about, there's definitely no way anyone's going to discern -93 from anything better. Maybe once they've reached their destination with a completely silent, reflection treated room and high fidelity amplification and speakers, but not on the street or in a vehicle of any kind... That's where LDAC would be used most IMO. It's where I'd use it.

Asking because I recently got a Topping G5 so I'm wondering, as long as it's been properly implemented, its LDAC performance will be in the -90s?
No future DAC will improve upon LDAC THD+n being in the lower -90s?

@VintageFlanker have made several measurements of LDAC and published that on ASR, as I recall. Edit: I recalled wrongly that he published, as he wrote in post #8 below.

As for LDAC performance he says the following in a post:

>>>AFAIC, I measure every BT codecs available when I can. I made quite a few THD+N measurements with LDAC (both Analog and Digital domain) and it's often on par with what you get from regular digital inputs, for sub-110dB SINAD DACs. And about 112dB+ SINAD in digital measurements (still 1kHz sine, WAV file, 24b/44.1kHz).

I happen to have a CD Player/DAP, planned for review someday, which gives me about the same results with USB-in and LDAC (≈104dB in that case). Haven't seen a "broken" implementation so far. Still have to measure aptX HD, tho.<<<

 

[Rja4000]

It is now possible to stream LDAC from Windows without VM, dongle, or other shenanigans.

I appologize if this is old news, but I just stumbled across it. There is a bit of software called Alternative A2DP Driver that lets Windwoes transmit LDAC (or other codecs) directly. This assumes, of course, that the computer already has bluetooth capability. The software lives here...

www.audiosciencereview.com

 

[voodooless]

I’d love to see a diff with actual audio. Getting a sine wave with low THD isn’t really hard for most lossy audio codecs. Audible artifacts start arising with much more complex musical arrangements.

 

[Trell]

I’d love to see a diff with actual audio. Getting a sine wave with low THD isn’t really hard for most lossy audio codecs. Audible artifacts start arising with much more complex musical arrangements.

Shouldn’t a multi tone measurement give useful information here?

 

[voodooless]

Shouldn’t a multi tone measurement give useful information here?

Not quite. Those are just static tones. It’s easy to compress: just make list of the tones frequency, volume and phase, and you can recreate them on the flip side easily. So you’ll need actual dynamic content.

This doesn’t means these tones are useless. They can tell something about the noise floor, low level artifacts and a few other things. But they certainly don’t tell the whole story.

 

[VintageFlanker]

@VintageFlanker have made several measurements of LDAC and published that on ASR, as I recall.

I did quite a few experiments indeed, but didn't published any on the matter. @Rja4000 published an extensive review about BT codecs.

My second-next product to review (NODE X) happens to support aptX HD. It will sure add another interesting BT codec to my list.

Getting a sine wave with low THD isn’t really hard for most lossy audio codecs.

Haem. Neither SBC or AAC (nor regular aptX) do a good job at it.

 

[Elitzur–Vaidman]

Not quite. Those are just static tones. It’s easy to compress: just make list of the tones frequency, volume and phase, and you can recreate them on the flip side easily. So you’ll need actual dynamic content.

This doesn’t means these tones are useless. They can tell something about the noise floor, low level artifacts and a few other things. But they certainly don’t tell the whole story.

I can clearly discern differences in compression/codec qualities when listening to the group clipping. I assume it's because many of their songs contain a lot of pseudo-random (or random) noise. I don't have the best terminology to describe their sound, but the name gives some of it away.

I'd love to do a delta comparison at some point, but I don't have a quality ADC yet.

 

[Trell]

I did quite a few experiments indeed, but didn't published any on the matter. @Rja4000 published an extensive review about BT codecs.

Yeah, I recalled wrongly.

 

[ocinn]

Not quite. Those are just static tones. It’s easy to compress: just make list of the tones frequency, volume and phase, and you can recreate them on the flip side easily. So you’ll need actual dynamic content.

This doesn’t means these tones are useless. They can tell something about the noise floor, low level artifacts and a few other things. But they certainly don’t tell the whole story.

And yet LDAC still butchers static multitone. A test emulating real music won't be any better.

On a side note: I counter the point BT is even very good for convenience in any circumstance you would care about audio quality. In the car? You'd probably want to be charging your phone anyways, so a USB connection to the head unit is better, or in analog head unit cars, you can use a splitter and an Apple dongle or other DAC to send it aux.

On the go? Unless you are married to TWS headphones, what convenience benefit is there to having a Bluetooth DAC/amp in one pocket with your phone in the other? Plugging a dongle into your phone (plenty of high-performance ones on the market, and plenty of small form factor DAC/amps with built-in battery) is far better and barely more inconvenient.

At home? WiiM Mini - $80.

The only form of convenience I can imagine is if I was on an ultra-budget non-streaming chip amp, couldn't afford a wifi streamer like above, and I just wanted background music in the house without tethering my phone. But in that case, who cares about performance metrics, outside of dropout-distance? I don't think current BT technology has any place in high-fidelity audio where objective performance is a part of the greater conversation/intent.

 

[voodooless]

And yet LDAC still butchers static multitone. A test emulating real music won't be any better.

I’d say it does pretty well, actually. ABX it, I don’t think it will be easy.

 

[Trell]

I don't think current BT technology has any place in high-fidelity audio where objective performance is a part of the greater conversation/intent.

On the other hand very many people uses BT for listening and likes it. I could try persuade my teenage child to switch from AirPods Pro to something like TRUTHEAR x Crinacle Zero IEM but I don't think I'll succeed.

 

[VintageFlanker]

And yet LDAC still butchers static multitone. A test emulating real music won't be any better.

I will try to replicate that someday.

 

[ocinn]

On the other hand very many people uses BT for listening and likes it. I could try persuade my teenage child to switch from AirPods Pro to something like TRUTHEAR x Crinacle Zero IEM but I don't think I'll succeed.

is your child's listening "high-fidelity audio where objective performance is a part of the greater conversation/intent."

I highly doubt it. So yes, in your case Bluetooth is great.

I’d say it does pretty well, actually. ABX it, I don’t think it will be easy.

I agree with you, I meant "butchered" as a technical comment of the medium, not if it's audible.

 

[antcollinet]

Isn't the problem with measuring lossy codecs is that they are designed to change the signal. But they are designed to do it in a way that is not audible. (In the ideal situation) or where the aduiblity is very much less than the actual change.

So you'll measure signal distortion - but won't be able to measure the audibility of this without building into the measurement the characteristics of the human auditory system that the codec uses to compress the signal.

 

[voodooless]

I will try to replicate that someday.

In theory you don’t even need any hardware for this. The codec is opensource. You can find command line tools for encoding on GitHub. For decoding there is a library as well. A few lines of code should make it happen

 

[kemmler3D]

Audible artifacts start arising with much more complex musical arrangements.

True story. I find that SBC and basic APT are good in theory but are absolute murder on transients, especially unipolar.

 

[amirm]

Isn't the problem with measuring lossy codecs is that they are designed to change the signal. But they are designed to do it in a way that is not audible. (In the ideal situation) or where the aduiblity is very much less than the actual change.

So you'll measure signal distortion - but won't be able to measure the audibility of this without building into the measurement the characteristics of the human auditory system that the codec uses to compress the signal.

There are two classes of codecs here: perceptual and mathematical. AAC is the former. Normally it is an excellent codec so it is a puzzle for me as to why it performs so poorly. It may be forced into running in low latency mode which is the enemy of perceptual codecs (it can't smooth the peaks and valleys in encoding with buffering). The others attempt to extract redundancy and with it, reduce bitrate. Original requirement from Bluetooth SIG was that the codec had to be very lightweight and free to license. This led to only one viable proposal from philips which was SBC. While at Microsoft, we had a much better codec but could not meet the requirements for such low computational complexity. LDAC works in similar manner to SBC in that it is not a codec based on psychoacoustics.

In general though, all of these codecs should have nailed the simple measurements we have thrown at them, and produce transparency. Whether it is implementation or design, something seriously is wrong with them. I remember when we were working on WMA codecs, I threw simple sine waves and sweeps at the codec and was puzzled by response errors. Reported them to my DSP/Codec team and they fixed them all. Their focus had been purely on music encoding and simple checklists were not done to find these problems. I think the same issue exists here.

 

[Jeromeof]

I often wonder with BT codecs if they are being forced down to lower bitrate when measured. I used to monitor / measure BT performance with older intel Mac ( as a developer their was an excellent apple Bluetooth tool available which would show in real-time the change in bitrate and the real codec being used ) - just trying it in different environments was revealing. It would jump significantly.

I often think people think they are using a high bitrate but actually it has switched to a lower bitrate and even possible still using SBC when they think they are using a better codec.