It’s not bad, but ASRCs are not good either. But their idea is that you can adjust the conversion ratio on the fly even by tiny fraction of Hz. However, this is not needed for the kind of upsampling done by HQPlayer. (asynchronous meaning there that input and output rates are not fixed nor related by any static ratio)
They are many times used for reclocking, but the down side is that jitter gets “baked in” the data and then you cannot remove it afterwards.
You can sometimes recognize such from DAC J-test measurements where the noise floor is a bit elevated and not flat around the fundamental.
Perhaps because I dont really understand it, Id never really messed about having different filters for 1x & Nx. I’ve always had the same.
Anyway, its early days yet and still need to judge what ive done by running through different genres but im liking the result of having different filters.
I am seeing that the CPU/Cores are working harder but nothing thats concerning. Can anyone explain beyond the HELP narrative what the 1x & Nx actually do, and what are the benefits/disadvantages of splitting filters? I guess having the option means it can be advantageous?
Since for example 96k content has upper frequency limit of 48 kHz, you can afford spending more of that bandwidth for the filter roll-off. And thus using more relaxed, not as steep filter for those higher rates than you need for 44.1k content.
Since at 96k rate, the same filter would be already less than half long in time, but using more relaxed one means that it can get even shorter in time.
So let’s say you want to use poly-sinc-gauss-xla for 1x content (44.1k and 48k), but similarly poly-sinc-gauss-long should be perfectly fine already - or even better - for Nx content (>= 88.2k). And you save some CPU cycles as by-product.
The ext filters are linear phase long filters designed to have a steep roll off. Ext 3 is longer and steeper than ext 2. Steepness at 1x is more significant than at higher sampling rates.
You might still prefer ext 2 at 1x or be interested in shorter filters with less ringing. They are available but that is not what the ext family are designed for.
Is this information published as graphs somewhere? Impulse and frequency attenuation. For 1x and 2x. I’ve done some measurements myself, but it would be nice with a full set of data for all filters. The descriptions in the manual are very vague.
On a seperate note, for Nx music, it’s important to remember there’s very little music content above 30kHz, even if you can hear up to that and your amp and tweeters can play it.
Most stuff above 30kHz is just noise/hash/HF garbage. There’s good reason to use earlier roll-off filters to filter that crap.
Sure, in relation to ext3, I guess anything is short. But in the post, Jussi on one hand recommends shorter and more relaxed filters for >96kHz. Then he suggests a long filter (where a short version is available). I guess Jussi needs to come in and clarify. Again measured graphs would help compare.
I use the fanless chassis of i9-9900KS+16Gb ROM+ASUS ROG M11H, and I don’t have a GPU installed. I am running hqplayer embedded OS on the USB Key. Should I install 4.28.0-x64amd or 4.28.0-x64gen? If I oversampling to DSD256 using ASDM7ECv2, what are the hardware limitations? If I increase the GPU (what model do you recommend?), what aspects of performance will be improved?
Hi, you are right at the same spot as I am. I cannot even reach your filter, it stops at DSD7 256 + fs. I have put together a quite serious PC some ½ years ago. But it seems if it is CPU cores that does the trick, perhaps it is time to leave Intel and go for AMD?
I have this same question. Same hardware and wondering if I need to keep my AMD 5700 XT in my system or just rely on the CPU (i9-9900K). AMD GPUs seem flakey from various posts here and not likely to help with HQP performance. But could be wrong. Waiting for an answer on this as well.
My cpu is not near yours, I do have a fan on mine though. Intel i5 10th Gen 6 cores, no GPU. I can run all day using poly-sinc-ext2, ASDM7ECv2, DSD 256 for 1x and Nx.
I’m using HQPlayer Embedded 4.28.0.2 x64 gen. The trick for me to use ASDM7ECv2 sucessfully at 256 was to set multicore=1 in /etc/hqplayer/hqplayerd.xml
New HQP user here! Tried for few weeks and decided to purchase license last weekend .
Very happy with poly-sinc-gauss-short / NS9 / 8x pcm filter to my D90SE. Before settling down with this combo, I tried a couple of filters and found that the filters I like more are all without apodizing. (The apod counter also rarely goes above 5 on the albums I listen to, fortunately.)
So here comes my question, maybe for @jussi_laako . It’s just out of curiosity. I understand that there is no harm using apod filter on music that doesn’t need apod correction after reading the entire thread. But I imagine adding apod correction must trade something else off, otherwise there will not be the non-apod version filter in the first place. Given two similar filters like gauss-xl vs gauss-xla, what would we gain if using the non-apod version?
.