MQA is Vaporware

[GoldenOne]

On 3/24/2019 at 7:37 PM, PeterSt said:

 

Hey Dirk - I think John announced not to know that.

 

From a Command Prompt:

 

mqascan inputfile

(inputfile can be a flac file)

 

Use your imagination at looking at the output. It's not difficult.

@John Dyson @Miska

So i've been trying to get this to work.

I have a video and article upcoming on MQA, and have actually had a few tracks of mine published in MQA, including with some test files including an impulse response, RMAA test sequence, white noise and a bunch of other stuff hidden inside that have allowed me to get a closer look at what's going on early in the process and with the true 44.1khz and hires masters available for direct comparison.

I'm currently trying to get this script to work to make a pseudo-MQA file and trick the DAC into thinking it's MQA, but have had no luck. I was able to get the MQAscan file to run using the file John Dyson posted, but the mqbgen I get nothing regardless of input. No error or anything, just nothing happens.

I'm not able to get it working by making on a linux system either. Same thing, command executes with no error and nothing happens.
Has anyone been able to get this working to create a file that the DAC recognises as MQA?

 

[GoldenOne]

I am currently working on a video, and I got some tracks published on tidal, in MQA. These tracks contained various test signals such as impulse response, square wave, white noise, multitone, and even the entire RMAA test sequence.

After collecting all evidence and documenting all the discovered issues (some of which have not yet been discussed elsewhere), I contacted MQA to give them a chance to comment prior to this video (and post here) going live.

They have not responded to me, BUT, they have had my tracks pulled from Tidal, and have spoken to the publisher I used to have them block me.

The tracks are still up on some sites such as Deezer, and I have the MQA encoded files saved as well as analog recordings of a full-decode. And I will share these in the final post and video.

But for now, I'm just posting this quickly in order to inform people that MQA's response to me reaching out was to attempt to censor me, and I want to ask that a few people confirm the track's presence before it is fully gone.
If you go in roon and search for "The Callout" by GoldenSound, you will still see that it shows as being on Tidal, in MQA in their DB. Though I'm sure this will not be for long.

The video and post, as well as the tracks themselves, should hopefully prove quite useful for anyone wanting to look further into what MQA is doing.

[GoldenOne]

8 minutes ago, Racerxnet said:

And you did record/save the tracks for further investigation?

I did yes. 

I have the original masters that were submitted, one of which was 44.1khz, one of which was 88.2khz.

 

And I have the MQA encoded versions of both. As well as the versions streamed from tidal on the "hifi tier" (which as it turns out are the same file but without mqa flagging) 

 

These will all be shared in the post. The censorship won't change anything :) 

[GoldenOne]

Just now, The Computer Audiophile said:

I can’t believe nobody did this previously. Brilliant. 
 

This is classic MQA Ltd. Try to censor or in my case talk more often and louder. 
 

I’ll ask the rhetorical question I always ask, does a company with nothing to hide act this way?

 

It’s crazy (at least to me) what Bob Stuart, Mike Jbara, Ken Forsythe, and others will do for a dollar. The MQA stink ain’t coming off those guys. 

Given how they dodged questions and attempted to desperately divert the conversation in your 2018 rmaf talk (which I discuss a little in the video, hope you don't mind), I didn't really expect anything different. Hence why I made backups :) 

[GoldenOne]

1 hour ago, Cebolla said:

 

MQA-CD file tracks should be provided unscathed on the TIDAL HiFi quality (16bit/44.1kHz only) connection as they are already distributed at 16bit/44.1kHz,, but the hi-res MQA ones (which in your 88.2kHz case are distributed at 24bit/44.1kHz) should be corrupted as they're provided mangled to 16bit/44.1kHz.

That's what I thought too. And it certainly used to be the case. 

 

Unfortunately now though it seems that the 16/44 file is just the MQA file but without the client unfolding it. 

 

Its bitperfect to the 'masters' version and this was the case for any other track I've tried. 

Ive actually stopped my subscription because of this 

 

I pay tidal for lossless. If they no longer offer that for any track marked master, I don't want tidal. 

 

 

Ill double check with some other files that I know were definitely from 48/96khz masters as I can't imagine those would be the same

[GoldenOne]

Just as an update, MQA did respond to my email (AFTER the posts about censorship were made here and elsewhere).
Most of the content in their response was either marketing talk, or easily counter-able (or outright easily disproven) claims. 

My video is done, including my response to their response. Its up on my patreon atm as I put things on there early but it'll be up on here as well as a written version very soon.

[GoldenOne]

I've made a long video exploring MQA using files I had published, in MQA, on Tidal.
I'll be writing up a written article this evening to post here and will include the original masters, MQA encoded versions, and unfolded MQA versions so that anyone else can look at them and test further.

MQA did respond to me and their response is discussed at the end of the video (timestamps in playbar/description)
 

 

[GoldenOne]

1 hour ago, JoshM said:

Not sure whether this is better placed in this thread, the thread on @GoldenOne’s video, or the ASR thread, but:

 

1) Over at ASR, John Atkinson is claiming that Golden’s test is invalid since everyone knows a lossy encoder (we’re now apparently conceding that MQA is lossy) can’t handle the signals Golden encoded. This is the same claim Amir made, and if you look at John’s “likes” at ASR, he’s “liking” every post criticizing Golden’s test and defending MQA. 
 

2) Amir has just given a response that completely negates the point of every measurement he’s ever done:

 

 

 

Well that is.....unexpected.

People really are treating this far too much like a team sport and trying to "win". 
I didn't make this video to 'win', or for any kind of recognition. I made it to demonstrate that MQA's claim of being lossless and an equal/better alternative to native FLAC is false.

If people are conceding that MQA is lossy then that's my main aim achieved (with the second being that this video gives manufacturers enough community backing to be able to say no to MQA and not be effectively forced to implement it lest they lose sales). No, these tests wouldn't work for a lossy encoder, which is exactly the point!
How "good" of a lossy encoder MQA is we probably will never know unless they open things up more. But honestly I don't really want to know, the point is that it IS lossy despite their marketing implying it is not, and I'd rather stick to lossless given as it's from all evidence provided, better, and doesn't have a license fee attached.

If MQA were as good as FLAC, and truly lossless, it would not have "broken" with the tests I threw at it.
It did, and therefore until MQA provides concrete evidence as to why these issues are a worthwhile tradeoff for whatever supposed benefit it provides, I don't want it and people shouldn't be comfortable with how much it's spreading.


The fact that amir has suddenly 180'd from his usual stance that gear can be evaluated solely using steady state signals and ideal test conditions, to now saying that this testing is invalid because it doesn't represent music, is rather odd.....

[GoldenOne]

It may be of interest, but I had a conversation with someone in the industry recently who shared a little insight.

 

Bob Stuart apparently created a new format originally for what was the Ponoplayer.

Bob then later began marketing it standalone as MQA after Neil refused to incorporate it into the PonoPlayer.

 

Additionally, and I'm intentionally leaving details out for now until a few things are confirmed but, it might actually be the case that MQA doesn't have the full intellectual property rights to their product and the way they are currently operating may not be legal.
GPL is an interesting area....

This is from Neil Young's book "To Feel the Music" (Thank you Rusty Shackleford for posting):

“Stuart had been attending our monthly development meetings at Neil’s ranch as the team’s audio expert. But as we progressed with the development of the player, it was Gallatin who designed the player’s electronics, while Stuart provided suggestions and comments along the way. Stuart’s main contribution was to be his software.

In these meetings Stuart would describe his software in general terms with little specificity. He explained it as something very complex that involved encoding and decoding using both software and hardware. But he was always reluctant to provide specific details about when we’d see it.

I could sense that Neil was becoming frustrated and impatient, because that software was critical to completing and testing the player. It was not only Neil; the entire development team had begun to wonder whether we’d ever see anything. It made us all very nervous.”

“To signify that a file was authentic, Bob Stuart came up with the idea to put a blue light on the Pono player that would go on when a Pono recording was being played, letting the user know that the recording was Pono: the highest resolution, a pure, unaltered file.”

“Finally, in our November monthly meeting, Stuart said he was ready to discuss the terms for Pono using the software. Hamm flew to the UK to meet with him and the investors in Meridian: the Richemont Group, a Switzerland-based company that owns a number of European luxury fashion brands.

The terms they proposed to Hamm included monthly payments, royalties for each player sold, more stock, and no exclusivity. The terms were much more onerous than Pono could afford and made no business sense based on normal industry standards. Not only would the software not be exclusive to Pono, but it also restricted what Pono could do with it. For example, if Pono was sold or licensed its player to be built or sold by another company, then his technology could not be included.

During these negotiations, Hamm explained our economics and tried to negotiate a more favorable arrangement. Discussions and negotiations continued for several months and included Hamm, Elliot, Neil, Cohen, and Stuart and his investors, but they never were able to come to an agreement.

When I interviewed Stuart for this book, he thought that Pono management had been unreasonable by not accepting his terms, because of the value his software would provide. Stuart felt its value was much greater than what Pono believed it to be. Stuart’s software eventually became the basis for a proprietary compression technology called MQA.”

“Charley dismissed Stuart’s technology as solving a problem that didn’t exist and therefore no longer needed solving. We had no reason to shrink the files at all, since memory and file size were not the issues they had been years earlier. Like Neil, he was opposed to a new proprietary music format that added new restrictions to the music files and was controlled by for-profit companies.”

“Bob Stuart contacted our lawyer, Rick Cohen, and said we were disclosing confidential information on our Kickstarter website. He was referring to an image of the inside of the Pono player that showed the prototype circuit board, including the programmable memory chip that was to store his software—the software that hadn’t arrived.

I thought his complaint was unfounded because there was nothing that was proprietary in the image that would indicate anything related to his technology, and nothing that he designed. In fact, his backing out of our arrangement required us to design around the chip in order to get our early prototypes to work. The chips simply just sat on the boards unused.”

[GoldenOne]

2 hours ago, Luistergek said:

@GoldenOne at first many thangs for your article “I published music on Tidal to test MQA”

I heard from a HiFi dealer that the MQA story is even worse. According to him one of his “Manufacturers” claimed that they are forced to have the MQA decoder chip in the digital path when playing non MQA encoded music. 
Of course, this doesn’t have to be a problem, but he also told me the MQA decoder chip isn’t lossless when “processing” non MQA encoded music. And this would make an honest comparison with the same dac impossible. 
This delayed there version of a DAC supporting MQA regarding high demands from customers in Asia.
I don’t know if he was referring to the filtering like you mention talking about the ifi GTO filter or something else.

In My opinion we must be able to check this as a community by capturing the digital I2S stream from the MQA decoder chip to the actual DAC chip. This way we can also recover the decoded MQA stream in the true digital format. 
Please share us your thoughts on this.
 

So I've actually had conversations with not one but TWO manufacturers since the video release about exactly this.

Apparently, MQA is becoming very pushy about manufacturers keeping external processing/upsampling in the signal path at all times. 
This is a direct quote from a manufacturer to me recently:

"Now, the MQA file, un-unfolded, is usually a 44.1kHz or 48kHz file, while unfolded it is played at the highest sample rate supported by the player that is a multiple of these rates (so e.g. 352.8 / 384kHz / 705.6 / 768kHz). There are few to no DACs out on the market that can handle a switch from e.g. 44.1kHz to 705.6kHz without an interruption to the sound, well, I should say I am not aware of any such DAC and certainly none of our company's devices have one, not even sure if it's theoretically possible. This is the reason for the interruptions to the sound many of our users are reporting when beginning to play an MQA track, seeking through it, skipping to another track, basically anything other than playing a whole MQA album in sequence.


Now, MQA understandably do not want to want this to happen to their audio, so the recommendation to all of us, is to resample all incoming audio to their highest 44.1 / 48 multiple, regardless of whether it's MQA, in a manner consistent with MQA unfolding so that no break in the sound would occur.


If we followed this recommendation, we would have to abandon the [name redacted for company anonymity] for bit perfect rendering of PCM audio at different sample rates."

So not only does MQA want manufacturers to add upsampling before the DAC chip itself, they want manufacturers to use THEIR filter....

This was also mentioned here by @Doug Schneider: 

I am guessing this is why ifi in newer products like the iDSD Neo is ONLY using the GTO filter.  (which as shown is basically identical to the MQA renderer filter, and in fact probably IS identical. White noise test shows a slight difference but that might just be from added noise in the track).
This means that with a product like that. You are effectively doing MQA rendering no matter what in order to prevent any interruption. Degrading performance of all standard PCM content just to keep MQA happy.....

[GoldenOne]

I can't imagine that this price model adjustment will change anything regarding the actual files that are being streamed. But once it's out here in the UK i'll test it

[GoldenOne]

ASR has now locked the MQA discussion thread.
 

Prior to this, one of the moderators requested a phone call with me. They said (and no I am not joking) that there were too many people criticising Amir for the MQA related comments he'd made and that they couldn't keep up with the moderation, so they wanted to lock the thread and wanted me to start a new one, in which they would not allow anyone to bring up the statements Amir had made regarding MQA. Because ASR/Amir needed to "remain neutral".
 

I told them that I was unwilling to do this and that censorship is not a good plan. Firstly it will just raise more questions and make ASR/Amir look suspicious, and secondly it's just morally wrong.

If ASR/Amir wants to remain neutral, then Amir shouldn't be making statements in any direction regarding MQA.
 

If they were going to lock the thread I was not going to help them make it more PR-Friendly by posting a new one and saying that it was all ok. Because it isn't.
No one should be immune to criticism, doesn't matter if it's their forum or not. And if you don't want to be criticised, you should say nothing. (And certainly shouldn't run a forum)
 

Even if there are people being un-necessarily hostile, which to be clear, some were, and that should absolutely be handled, if more moderation is needed, appoint more moderators.

Censoring discussion and preventing people from criticizing the forum owner is just wrong. And doing it now, when the owner of the forum says something in defence of a company that is known to have attempted to censor other forums, makes things look damn fishy.

I made quite clear that in my opinion, the best way for ASR to remain neutral would be to leave the thread up, as-is. And simply have Amir not say anything further on the topic.
Let people continue the discussion and have the forum owners/staff remain neutral from here on.

[GoldenOne]

I will be posting a response to the MQA blog post in the next couple days.
Normally I wouldn't as I really have no interest whatsoever in stirring the pot. But given the size and nature of the company in product in question, as well as the response itself, I feel it's important to do so.

I find it odd that they mention my description of their response as 'marketing', and yet they have careful marketing-speak responses in this blog post such as:

"MQA has never made false claims about ‘losslessness’."
This is a very careful choice of words. MQA are saying they did not make 'false' claims of losslessness. Not that they never claimed MQA was lossless.

"3: Provenance: MQA files are delivered losslessly"
Again, a VERY careful choice of words. The MQA files are delivered losslessly. The information within them however is not lossless.

They are appearing to answer some of the points whilst seemingly doing so in a manner that a lawyer could later argue to have meant something else.

Additionally....I'm fairly certain that the screenshots in Appendix 4 are fake. They are not of 88.2khz core-decoded output. They are screenshots of 44.1khz test files. I'll discuss this in the response video, but they've left a few things in that are somewhat easy to spot for anyone familiar with looking at these sorts of files/test signals in adobe audition.

[GoldenOne]

36 minutes ago, Archimago said:

@GoldenOne creates other videos

I am making a short video to respond to their recent blog post. But only because I feel it's important to address and deconstruct some of the clearly VERY carefully worded statements in that post. I'd much rather be making other stuff which is less drama-involved and much more fun :)

Additionally....the screenshots in appendix 4 are either fake or intentionally mislabelled. They are not 88.2khz core decoded output.

Looking at the impulse response one for example.



1) The spectrogram view is showing only up to 22khz, not 44khz, indicating it is showing a 44.1khz file, not an 88.2khz one as claimed.

2) Even if the 'unfolded' impulse did not show the wonky ringing that it did in my files, it would still have some ringing of some sort. Their own paper and other info shows this. And this means that the adobe audition waveform view would have some information below the centre line like this regardless of how zoomed-out it was:

Instead, theirs has only a straight upward line. (And we can't really tell much more cause they haven't zoomed in close enough to let us see the individual samples)
This implies they have literally just taken a screenshot of a 44.1khz digital 1 sample impulse file. Not the result of core decoder output or reconstruction/deblurring/unfolding.

3) Building upon the previous point, if we are to assume that the MQA encoder has indeed reconstructed a true, single sample, digital impulse, which a true digital single sample impulse represents all frequencies simultaneously, they are reconstructing an illegal signal, and doing the very thing that they claim caused my tests to "Fail" and that they themselves say MQA cannot do.
This is not possible and if it were, it'd be being used to make billions in all other areas of signal processing, data transfer, econometrics etc. You've invented a 'perfect' reconstruction system for which audio would be VERY far down the list on potential use cases for.

[GoldenOne]

13 minutes ago, danadam said:

Er... that looks like a logarithmic axis, I'd say it goes to 44k.

I think you may be right actually. Just playing about with the scale there now and whilst I cannot get mine to show as theirs does it does look like it could be going up to 44k.

I retract that suggestion.

Though the lack of any impulse ringing is still the biggest issue. The only three situations it would show as is in their screenshot:

1) It's just an 88.2khz impulse file. Not unfolded/reconstructed from anything

2) It was reconstructed using cubic/polynomial interpolation (which MQA does not use)

3) They have reconstructed a true digital 1-sample impulse which represents all frequencies upto 44.1khz at full scale. Which they themselves say the encoder cannot do. (And as previously mentioned would imply they've created a 100% 'perfect' reconstruction system for which audio would not be even a blip on the radar in terms of use cases for)


To avoid any doubt MQA could literally put the entire debate to rest by just releasing that one file.....

[GoldenOne]

3 hours ago, FredericV said:

88.2 MQA decoding includes an upsampling step of the baseband 1X signal from a 24/44.1 distribution file to 24/88.2 and then add the missing ultrasonics to the upsampled 24/88.2 version, recovered using lossy origami.

which means that the impulse response should look like test05b in this post, which looks exactly like the plot from Craven's filter
 

In our case we did x2 from 96K to 192K, but doing 44.1 -> 88.2 or 96 -> 192 is basically doubling the samplerate, and the impulse is no longer one sample, but spans several samples.

If we just see one pole at 88.2, that can't be correct, unless he is condensing the time on the X axis in such way it appears as a pole, but in reality it's not.

In audition regardless of how much the time axis is condensed it still displays negative values if they are present.

This is a digital 1-sample impulse:

This is an impulse upsampled with XXHE ArcPredict You can see there is some negative value content even when zoomed out:

 

 

Using a minphase filter which produces the ringing pattern shown in MQA's own AES paper and elsewhere, we see this:

The ONLY time that the impulse would appear as it is in the MQA provided screenshot is if it were a cubic interpolation like this:

 

Given as MQA is not using cubic interpolation, the only option left is that the screenshot is likely fake. It's just a screenshot of a digital impulse response file, not an unfolded result

[GoldenOne]

26 minutes ago, idiot_savant said:

@GoldenOne - From a technical point of view, just consider the MQA encoder is adaptive - MQA say this themselves, so it’s entirely possible it automatically adapts itself to test tones without any “white glove treatment”, so as @The Computer Audiophilesays, be careful!

To play devils advocate, I do think your test tones broke the stated MQA acoustic models, which seem to be done on a track by track basis. 
 

transparency would be good here - so if MQA could provide you with some real data, that would be helpful for one and all?

 

your friendly neighbourhood idiot 

Yep. But then this is the problem, there is zero transparency so we simply don't know.
It relies on faith.

If they wanted to put the entire debate to rest they could literally just release the impulse file they claim to be screenshotting.

[GoldenOne]

 

[GoldenOne]

6 minutes ago, botrytis said:

 

Thank you - you have answered them w/o snark. This is what we need from MQA which I doubt they will do...

 

 

 

That was the aim :)
I (and others) just want honesty and transparency. 

I won't be the one to resort to pettiness and bickering. I think their responses somewhat speak for themselves in terms of how lacking they are in actual truth and how they have resorted to ad hominem.

[GoldenOne]

Well that didn't take long......

 

 

[GoldenOne]

5 minutes ago, lamode said:

I see that BS labelled your video as "a libellous manifesto". What a joke. I wouldn't be surprised if he threatens legal action to shut you up but rest assured that this is would be an act of desperation and a bullying tactic. If anyone needs a good lawyer, it is BS himself, after committing acts of fraud on any customers, partners and investors from which MQA may have received payment based on his false claims.

 

 

I would sincerely hope that wouldn't happen. I don't think they would have any sort of case given as I've been very clear as to what was my opinion vs what was demonstrable fact. 
And absolutely nothing I've put out is a lie or knowingly false.

Not to mention, I think suing me would from a PR standpoint be the worst possible thing they could do.
Their bob talks post was not particularly well thought out given how demonstrably false or misleading some of the statements such as their claims of losslessness are. But a lawsuit to try to bury this would be shooting themselves in the foot to put it lightly.

I've given them several easy-outs if their claims are true. A lawsuit would only prove to people that they are not telling the truth and are having to seek alternative ways of shutting critics up.

Plus....what I've put out in videos so far is FAR from all the information I have available.
Manufacturers have provided me with some VERY interesting information.....

[GoldenOne]

7 hours ago, gfkeenan4 said:

My biggest complaint on GoldenSounds video was its lack of repeatability. For example, when he recorded the output there is no mention on how he recorded unfolded audio.

If I was recording the 88.2khz information, this was recorded from the core-decoded output of roon or tidal as described in the process here: 

It does not matter which is used as the decoder is the same and both are bitperfect output. It provides a 100% bit for bit identical output from each player each run.

If Recording the "full decoded" output, balanced line-out from the idsd diablo was fed into the RME ADI-2 Pro FS R ADC running at 768khz.

 

 

7 hours ago, gfkeenan4 said:

I did actually post my comments/findings on GoldenSounds youtube and it was removed. Was going to email him, but after he removed my comments

I don't believe I've removed your comments. I've removed two comments so far, one of which was accusing the whole discussion of being race-related on account of Jay-Z being black, and the other was just hurling abuse.
Other than that I've not removed any comments and if there's something I disagree with if I see it I respond to it, not remove it. The whole issue I'm protesting against is lack of transparency. Deleting anything I disagree with would completely go against that.

Your comment may have been picked up by youtube and held for review though, if you send me your youtube username I'll check in the held comments and approve the comment. (I don't have comments set to manual review but sometimes youtube holds stuff just because it contains a link or something.
Eg: 

 

 

7 hours ago, gfkeenan4 said:

How did he download the MQA file? I did it via this: https://github.com/yaronzz/Tidal-Media-Downloader

but its not clear how he did it. Was he just recording the output from the Tidal player?

I used that tool yes. I also checked by recording the output of both roon and tidal to compare and both players as well as the downloaded file were bit for bit identical (aside from the silence either side in the manual recordings of player output) so it shouldn't matter which method is used to obtain the file as long as it is bit perfect.

 

 

7 hours ago, gfkeenan4 said:

He also switches from comparing graphs that are BIN dBFS to dB SPL in a few places which really throws of repeatability. I also found a few points he switches from sampling rate to frequency domainsinterchangeably. For example, in his 44.1kHz file, he tries to find his 40kHz test tone. 

If the scale was changed it was because of a change in software, for example spek vs deltawave. But I think the info should be the same given as the levels weren't altered so SPL should be equal to dBFS in all comparisons (other than if comparing a digital to analog recording which will be a couple dB out but I didn't do that in the video.
The original files including the dithered and undithered files and MQA results of both (the dithering did not alter the result, they both show the same issues) are available in the OP if you wish to take a look at them yourself using any software/analysis of your choosing, and have been since the video was released.

I'm not sure what you mean about finding the 40khz test tone in the 44.1khz file, unless you're meaning the aliased tone, in which case yes, the fact that the tone IS present in the 44.1khz file is the problem in itself.


I would add though that to be honest, the analog recordings of "full decoded" output were simply to show that the full decode didn't 'fix' any of the issues shown in the core-decoded audio. Recording of analog output is insufficient to properly analyze if MQA is lossless because it's inherently going to be affected by the device itself, clock drift and run to run variation.
Hence why I used 88.2khz information so that I could check it against a digital core-decoded output with no potential of alteration by a DAC.

It's also worth mentioning that as shown by @mansrthe 2nd/3rd unfold don't really exist. The core-decoder is doing some adaptive stuff but the 'renderer'/2nd and 3rd unfold are just fixed upsampling filters.
 

[GoldenOne]

I put a bull in a China shop that advertised "shatterproof China" 

 

Maybe if you don't want claims of losslessness to be disproven dont make the claim. 

 

I and others aren't doing the same thing for stuff like mp3 cause they don't make demonstrably false claims..... 

 

 

[GoldenOne]

2 minutes ago, opus101 said:

 

But its been amended now to 'better than shatterproof'. So when bull contacts the china, its the bull that gets shattered?

Well it is shatterproof. Better than shatterproof in fact. 

 

But you can't test that cause bulls are outside 'natural' tea drinking conditions. 

And they aren't going to allow you to actually test the shatter resistance so you just need to drink tea from it and make a guess yourself. 

[GoldenOne]

If your argument that your product is faulty but only because it was tested and therefore its the testers fault, you need to rethink. 

 

In other news, I don't need fillings cause I've not seen the dentist.