Rendering will always be done with the appropriate MQA DAC or renderer setup, even if the MQA master sample rate is 44.1kHz-96kHz.
No, 96k is covered by unfolding. Rendering is for going beyond 96k.
AFAIK.
I think renderer consists of two parts: up-sampling and de-blurring filters, there 16 of them. A decoded stream (MQA core) is Always either 88.2 or 96k output, nothing higher or lower.
Remember the output decoded stream contains a renderer instructions which tell the renderer whether up-sampling is required and also select the required de-blurring filters. three senecio below:
-
If the original recorded sample rate is 44.1k/48k, this sampling rate information is embedded what is called ârenderer instructionsâ during encoding side. In this case thereâs nothing to decode so the decoder will do the âup-samplingâ to 88.2/96k. The output stream is sent to the renderer, it will select NO up-sampling but will select the required de-blurring filter based on renderer instructions.
-
If the original recorded sample rate is 88.2/96k, this sampling rate information is embedded what is called ârenderer instructionsâ during encoding side. In this case decoding takes place and provide an output stream of 88.2/96k. The output stream is sent to the renderer, it will select NO up-sampling but will select the required de-blurring filter based on renderer instructions.
-
If the original recorded sample rate is either 176.4/192k or 352.8/384k, this sampling rate information is embedded what is called ârenderer instructionsâ during encoding side. In this case decoding takes place and provide an output stream of 88.2/96k. The output stream is sent to the renderer, based on render instructions it will select up-sampling either 2x of 88.2/96k (for original recorded sample rate of 176.4/192k) or 4x of 88.2/96k (original recorded sample rate of 352.8/384k) at the same time it select the required de-blurring filter.
For non MQA DAC, it will always display 88.2/96k, if a renderer is added and someone can tap off the final sampling frequency before it goes to DAC chip, then it probably shows higher than 88.2/96k if the original recorded sample rate is higher than this.
Take note MQA DAC always display the original recorded sample rate NOT the final output stream! In some ways Iâll find it âfakeâ because it does not measure the output stream directly.
This is the original sampling rate of the digital master, 192k (uncompressed PCM) not the actual MQA decoded output. MQA is only good up to 88.2/96k, so anything above for this instance is just up-sampling; 2x 96k=192k using the ârendererâ
Ah, my apologies. I know very little about MQA other than occasionally a couple of little blue L.E.D.âs will light up on my recently purchased Explorer². 
Iâll leave it to you professionals and just continue to lurk!
Although I am trying to learn.
1 Like
I recall we had a similar conversation before, on the subject of MQA 44.1kHz that gets decoded to 88.2kHz MQA Core.
1 Like
I think you already have a good grasp on most everything that MQA offers. 
1 Like
It would have been interesting had you included a comparison with the same master, not MQA encoded, and just played through, just to see if MQA offered any difference.
Iâll give you a hint: in the past there have been MQA announcements that a few MQA DAC perform 16x (768kHz) rendering. Think about when 16x render can happen.
Thank you, that makes sense.
So in my test, my statement about the rarity of 176/192k content applies to case 3 in your description, But the filter choice remains meaningful.
I may do some more tests. Or I may not. Not really exciting.
I did that a while ago.
Concluded that many had minimal differences, many others had huge differences that were likely different masters.
In any case, thatâs about the value of MQA. There are many questions:
Is MQA better than Redbook? In that case, the agonizing over bit depth seems irrelevant, taking some bits out of 24 still leaves more than 16, and there is more bandwidth, so it seems all good, unless you dislike the concept.
Is MQA better than or as good as 24/96? Doesnât seem greatly interesting, since there is little or no streaming of 24/96, so if you prefer, just buy 24/96.
My test was simpler, addressed the implementation detail of rendering. If you have a rendering only capable DAC, youâre fine. If you have a full MQA DAC and donât use DSP, youâre fine. My test was about the third case, if you use DSP and your DAC is not rendering capable, is there a reason to change hardware? My conclusion was no. Based on limited material.
Thanks, excellent description.
I find this description about upsampling, which appears to be the consensus, very frustrating. It doesnât resemble in the slightest the original folding diagrams they provided, with regions A, B and C.
I donât think your description of the three categories of render modes is entirely correct. I believe in all three cases the signal is upsampled to the maximum sample rate of the DAC. What changes is the upsampling filter used (although even that I am not 100% sure about). For different master sample rates, a different filter is selected.
That is provided if the DAC chipset can support a maximum input sample of 768kHz. Being more common DAC chipset max out at 384kHz and some older ones can only do up to 192kHz.
A is 16bit coded, B is 8bit coded while C is up-sampled. Itâs simple as ABC!
Yes, but thatâs not what they originally said!
As I recall. I donât have the original diagrams and descriptions in front of me, and I absolutely refuse to dig them out.
But I could swear that C was claimed to involve picking up information and packing it in.
Maybe they meant that âpicking up informationâ was knowledge about the characteristics of the stream, not its actual data. But that is not the impression I got.
Anyway, thatâs just a frustration with the whole rollout and marketing.
The results are ok but not earthshaking. Iâll use it on Tidal, otherwise ignore it.
The only way to verify that is to access the I2C bus to the DAC chipset in the MQA DAC, namely SDATA and SCLOCK. Capture that on a oscilloscope and we can get the results whether it is doing the maximum sample. My speculation on the region C (MQA unfolding graph) that max out at 192k
These are claims, they are very good at it. MQA cut off at 48kHz bandwidth so anything above is NOT captured. A+B happens in the MQA core which always output 88.2/96kHz. C on the hand is done by the renderer which up-sampled 88.2/96k to higher sampling range.