Jump to content

Sign in to follow this  
Ittaku

DSP devices with variable internal sampling rate?

Recommended Posts

39 minutes ago, Snoopy8 said:

What DSP software have you tried, Con?

Rew and rephase, roon's DSP/convolution, drc, brutefir, ladspa dsp.  Sweeps and impulse reponses ended up looking good, sounded crap. Perhaps the hardware solution just had the right balance of filters and not too much DSP. They know what they're doing whereas I clearly didn't and was just following guides.

Edited by Ittaku

Share this post


Link to post
Share on other sites
4 minutes ago, Ittaku said:

Yeah thanks, I was just explaining why I abandoned software.

You'd fall in love with Audio Weaver. It's been a while since I've used it but very very powerful. 

 

The dspMusik solution is quite well-featured... you can have this much grunt (https://www.danvillesignal.com/dspblok/dspblok-21469-analog-devices-adsp-21469-sharc-dsp-module) or multiples thereof. Should run to 384kHz. Al's worth a conversation... you'd get on well.

 

When funds allow I'll prob get one.

Share this post


Link to post
Share on other sites
13 minutes ago, Ittaku said:

Rew and rephase, roon's DSP/convolution, drc, brutefir, ladspa dsp.  Sweeps and impulse reponses ended up looking good, sounded crap. Perhaps the hardware solution just had the right balance of filters and not too much DSP. They know what they're doing whereas I clearly didn't and was just following guides.

I am currently using REW and rePhase to generate the impulses for convolution on MinimServer (SnakeOil OS, Intel NUC) using ffmpeg underneath. The results are excellent with a larger and deeper soundstage, better resolution of individual instruments, less smearing. I adopt a light touch approach to DSP and never do any boosting. And as you know, convolution is rate specific and there is an impulse for each of the rates I use. All DSD are preconverted to PCM because DAC chip cannot do DSD.

 

I trialled Roon DSP/ convolution and found Roon at least doubled the CPU load cf MinimServer. That was not the primary reason why I decided not to take up Roon. Currently the Devialet Phantoms are not yet Roon Ready, requiring either Airplay or optical connection. UPnP (MinimServer) provides much better sound quality, especially a  much larger soundstage. 

 

Investigated the X4, very impressed but optical input to Phantoms killed that option.  If I were you, I wouldn't look any further...

Share this post


Link to post
Share on other sites
1 minute ago, Snoopy8 said:

Investigated the X4, very impressed but optical input to Phantoms killed that option.  If I were you, I wouldn't look any further...

What do you mean optical inputs killed that option? The phantoms take optical in? The X4 does optical out, but you won't be using either the crossover component, or the DAC part of the device if you use that (I don't use the DAC myself.)

Edited by Ittaku

Share this post


Link to post
Share on other sites


Just now, Ittaku said:

What do you mean optical inputs killed that option? The phantoms take optical in? The X4 does optical out.

The soundstage shrunk when I compared optical with UPnP.  Have tried several optical streamers, even bought glass optical cable, tried different optical input ports on Phantoms & Dialog (connector). Believe me, I tried very hard to adopt optical and did many rounds of testing and with other people listening.

 

On its own, nothing wrong with optical. But UPnP had this soundstage that I really liked.

Share this post


Link to post
Share on other sites
Just now, Snoopy8 said:

The soundstage shrunk when I compared optical with UPnP.  Have tried several optical streamers, even bought glass optical cable, tried different optical input ports on Phantoms & Dialog (connector). Believe me, I tried very hard to adopt optical and did many rounds of testing and with other people listening.

 

On its own, nothing wrong with optical. But UPnP had this soundstage that I really liked.

Ah I see, I misunderstood. Implementation of the optical interface and signal at all ends is important for sure, whereas there's little to no scope for improvement with different optical cables.

Share this post


Link to post
Share on other sites
Just now, Ittaku said:

Ah I see, I misunderstood. Implementation of the optical interface and signal at all ends is important for sure, whereas there's little to no scope for improvement with different optical cables.

My bad for poor wording in the first place..

Share this post


Link to post
Share on other sites
Just now, Snoopy8 said:

My bad for poor wording in the first place..

Nah, the quest for brevity in text has that effect.

Share this post


Link to post
Share on other sites


On 07/10/2019 at 8:43 PM, Ittaku said:

Resampling to a non-integer multiple sample rate is lossy.

Is it audibly so?  That's an important matter to pin down.

 

We see so much written about 44.1/16 being indistinguishable from 96/24 for human ears.

 

If there is any audible difference it may be at an only "just noticeable" level, and then only with "killer" test samples.   This may partly explain why so many devices do not tie the DSP sample rate to the SR of the incoming stream.

Share this post


Link to post
Share on other sites
6 hours ago, MLXXX said:

Is it audibly so?  That's an important matter to pin down.

 

We see so much written about 44.1/16 being indistinguishable from 96/24 for human ears.

 

If there is any audible difference it may be at an only "just noticeable" level, and then only with "killer" test samples.   This may partly explain why so many devices do not tie the DSP sample rate to the SR of the incoming stream.

Yes audibly. It's not a huge effect, but audible with most samples too, not just killer ones. 44.1/16 is indistinguishable from 96/24, but converting 44.1/16 to 96/24 changes its sound. Changing 88/24 to 96/24 is much harder to distinguish.

Edited by Ittaku

Share this post


Link to post
Share on other sites
9 hours ago, Ittaku said:

Rew and rephase,...

....

They know what they're doing whereas I clearly didn't and was just following guides.

Interesting you should say that!

 

This thread was my initial attempt at following the guide when I knew a lot less. The guide had missing bits which davewantsmoore helped fill in. I now know a little more but still far away from understanding the underlying concepts.

 

On a tangent to this topic, I took your idea (again, can't say I understood the concepts) about upsampling with large number of taps.  In rePhase, I generated the convolution impulse with a FFT length of 2M samples and there was a noticeable improvement (over smaller samples) with more body to the music. Easily runs in MinimServer. In comparison, Roon was heavily loaded on NUC i5 8th gen but still ran ok. 

Share this post


Link to post
Share on other sites
2 minutes ago, Snoopy8 said:

This thread was my initial attempt at following the guide when I knew a lot less. The guide had missing bits which davewantsmoore helped fill in. I now know a little more but still far away from understanding the underlying concepts.

Well as I said, I got what appeared to be "good results" numerically, it just didn't sound good apart from having a relatively nice frequency response. I think there's more of an art form to it than just fixing the numbers. Gee where have I heard that before about measurements?

Share this post


Link to post
Share on other sites


14 hours ago, Ittaku said:

HDP3  which just popped up for sale. I checked the legacy page from deqx and it describes it thus:

image.png.f68be27a6b95f135bb5422e54d4192f0.png

 

The 2.6 is described thus:

image.png.9ff8c2e5c5318d189601d7657443024f.png

 

I think they've always done processing at a fixed sample rate according to the data I can get from the website.

I'm almost 100% sure they used to process at variable rates on my HDP-3 as long as it operated with a digital signal.

I used to have mine switch between 44K, 48K, 88K, 96K.

For a long time I've just upsampled with JRiver to 96K to match the DAC.

I will check later tonight by feeding it 44K.

I am running the latest firmware and software.

 

All analog inputs are sampled at 96K and therefore processed and sent to the DAC at 96K.

That's really what it's specifying above, nothing else the way I see it!

I think the DAC will oversample (note: not upsample) at 96K regardless of the data stream if that's a problem for you.

 

Edited by Satanica

Share this post


Link to post
Share on other sites

Also, you'll need to find one with the digital output(s) if you want to bypass the DAC(s) as it was not a standard option on the HDP-3.

Or perhaps they can still upgrade an existing one.

I think it could be worth getting in contact with them and let them know what you want.

 

Edited by Satanica

Share this post


Link to post
Share on other sites
On 08/10/2019 at 6:53 PM, Ittaku said:

Interesting, thanks. So are those the old models and no longer available? All the products on their website are resamplers.

Correct; old models.

 

The 3 was the DEQX more than 10 years ago but has since been superseded by the 4 and the 5 and the "Pre" variants.  The 3 had a max input of 96k so a sign of its time.

 

There were purchased-separately-add-on digital-out board and 2 versions of a balanced-out analogue board for the HDP3.  Plug-in daughter boards.

 

I had a HDP3 and I'm pretty sure it had 2 families of filters (48 and 96) and (44 and 88). 

All input is resampled to, and processed at, 96 in today's models.

 

The sweep tone, measurement and filter generation were all done in 96k (48k in really early models).  The 44, 48 and 88 filters were then numbercrunched from the 96 version.

 

 

But got me thinking ...

 

 

Get a 96k filter and numbercrunch it to say 44 (as was done in the HDP3)

Have a 44k input and numbercrunch it to 96 (as is done now)

Same conceptual process but in inverse.  Nothing to be gained, is there?  Both a result of non-integer crunching.

 

[Which is why I don't bother getting my MSB transport to integer-resample CDs (I play CDs exclusively) to 88 as the DEQX resamples to 96 anyway so it may as well do it from 44.]

 

 

Edited by aechmea

Share this post


Link to post
Share on other sites
1 hour ago, aechmea said:

I had a HDP3 and I'm pretty sure it had 2 families of filters (48 and 96) and (44 and 88). 

All input is resampled to, and processed at, 96 in today's models.

 

While the current models can accept up to 192K I'm pretty sure they will downsample to 96K anything larger than 96K.

Edited by Satanica

Share this post


Link to post
Share on other sites


6 hours ago, Ittaku said:

Yes audibly. It's not a huge effect, but audible with most samples too, not just killer ones. 44.1/16 is indistinguishable from 96/24, but converting 44.1/16 to 96/24 changes its sound. Changing 88/24 to 96/24 is much harder to distinguish.

Maybe you are worrying too much about the fixed DSP sampling rate, even though you can hear the small difference? I can understand lower end models minimising costs, but DEQX isn't exactly low end. Pity we do not have anyone from DEQX who can explain the choice of using a fixed sampling rate and keeping it for the higher models.

Share this post


Link to post
Share on other sites
1 minute ago, Snoopy8 said:

Maybe you are worrying too much about the fixed DSP sampling rate, even though you can hear the small difference? I can understand lower end models minimising costs, but DEQX isn't exactly low end. Pity we do not have anyone from DEQX who can explain the choice of using a fixed sampling rate and keeping it for the higher models.

Sorry if you'll excuse me for being paranoid when I'm feeding the output into a $70k DAC. The DEQX, minidsp, and many other communities, have no sympathy for users like me.

Edited by Ittaku

Share this post


Link to post
Share on other sites
1 minute ago, Ittaku said:

Sorry if you'll excuse me for being paranoid when I'm feeding the output into a $70k DAC. The DEQX, minidsp, and many other communities, have no sympathy for users like me.

But you already have the X4 and you are happy with its results. Not healthy to be paranoid for the sake of it?

Share this post


Link to post
Share on other sites
6 minutes ago, Snoopy8 said:

Pity we do not have anyone from DEQX who can explain the choice of using a fixed sampling rate and keeping it for the higher models.

If @Ittakugets in direct contact with them I think they will be OK to discuss it.

Edited by Satanica

Share this post


Link to post
Share on other sites
Just now, Snoopy8 said:

But you already have the X4 and you are happy with its results. Not healthy to be paranoid for the sake of it?

This is an academic exercise to see what else is there. The X4 has some troublesome firmware quirks and limitations in its design which I spent a lot of time working around, but now I'm mostly happy. It just saddens me to see that DSP is basically off the radar entirely for true high end usage otherwise.

Share this post


Link to post
Share on other sites
Just now, Satanica said:

If @Ittaku gets in direct contact with them I think they will be OK to discuss it.

I'm assuming I'll get the same sort of response that I got from minidsp - ridicule that it could possibly matter audibly. I'm not keen on going down that path of humiliation again.

Share this post


Link to post
Share on other sites
2 minutes ago, Ittaku said:

It just saddens me to see that DSP is basically off the radar entirely for true high end usage otherwise.

I'm gathering you think that anything that upsamples is NOT truly hi-end performance. I'm shaking my head at this one.

Share this post


Link to post
Share on other sites

Since this is an academic exercise, did you research Trinnov?  Not asking you to contact them.

 

Should they be using fixed sampling rate (& I am guessing they do), it is difficult to argue that they are letting the very high end down because it is their market.

Share this post


Link to post
Share on other sites
4 minutes ago, Ittaku said:

I'm assuming I'll get the same sort of response that I got from minidsp - ridicule that it could possibly matter audibly. I'm not keen on going down that path of humiliation again.

If you really believe what you believe and know what you want then just take any other views on the chin.

Anyway, I think the DEQX guys are a good bunch and would be very surprised of any ridicule and/or humiliation.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

  • Classifieds Statistics


    Currently Active Ads

    Total Sales (Since 2018)

    Total Sales Value (Last 14 Days)

    Total Ads Value (Since March 2020)
  • Recently Browsing   0 members

    No registered users viewing this page.


×
×
  • Create New...