TOPIC:

devteam wrote:

Dear All,

Just back from Chinese New Year holidays so we appreciate your patience! Sounds like there are quite few questions, we'd be happy to clarify what we thought we clarified earlier on but here is a good summary. Let's see how we can make it concise as possible so everybody from newbie to advanced can understand.
- The limitation here is "fixed point processing". There isn't any bug but more an inherent signal processing limitations. Some here believe we're on a great scheme to con the whole world and completely ignoring the issue it's not as bad.. We could list list some good AES (Audio Engineering Society) white paper that are a great read for some of you who want to read more but unfortunately, we have limitations with our AES library on papers we can share.. Using the keywords shown below and googling might be a good way to find external information that you'd believe maybe a bit better than us..

- Basically, to make it simple to understand: Low frequency IIR filter + High sample rate + high Q don't work well with each others. It's not a miniDSP thing, it's a math thing. It comes down to the precision of the filters and so called "quantization errors". There is a great read here: www.earlevel.com/main/2003/02/28/biquads/
Hoping it does summarize in simple terms for those of you looking for a technical explanation.
Maybe little you'd know that many products actually will downsample the sub channels so it doesn't run into this issue, run it through PEQ at very low sample rate (e.g. 32k) and upsample again.. On this product we can't unfortunately (one single sample rate), but would have done it to help since nobody would have realized that downsampling + upsampling.. (done everywhere else..
- Unlike a Sharc DSP where we have all the control we need (and have type 1), on a platform like the miniDSP 2x8/8x8, it's running on a fixed point processor with "pre-built filters". They do happen to work with Type2 which unfortunately isn't that great at the low frequency even if we're running them @ 56bit.
- miniDSP did realize the issue back then (pointed out a while ago by NYT) and did issue a work around on the PEQ for the 96k version which "does" help as we're using a separate biquad to help with the precision issue. @ 48k there is indeed a much better performance overall but if you have high Q + High gain, you'd still be running in similar issue.
- While we did mention the issue to Analog devices, one option would be for us to build a custom filter to make this work. We don't have access to the tools, but could make this happen I guess. Something to consider but to be honest, not 100% sure if we'd have access to these tools. Can get back when we get feedback from ADI.

In the mean time, if there is a general consensus that below 20Hz is an issue, we'd be happy to remove it from the GUi in a new rev. That's something we could certainly do. It's good to know that sometimes, there is a very large user community and not everybody uses the product the same way. Many people here been using it fine, others have issues.. It's always hard to please everybody.

Finally, in an effort to keep this community friendly and working toward the same objective, I'm sure that we're all into finding a better experience if we just work together. Thanks for your understanding.

We appreciate the effort, but removing < 20hz would not really be a solution. It's possible to get some filters working, it just takes some time and loopback testing.

That said, how hard would it be to create a 24khz plugin for the x8 platforms?

miniDSP said:
> @ 48k there is indeed a much better performance overall but if you have high Q + High gain, you'd still be running in similar issue.

I assume you mean 'running into a similar issue'.

Removing -20 kHz would be silly because that's the reason for buying a freaking miniDSP in the first place.

My concerns - and you'll admit, I've been the grumpiest of netizens on this topic - would be resolved by you actually saying in your actual selection criteria what your products are actually good for.

Discovering that you think 48kHz is better for my purposes, when you told me to buy 96kHz, is bad advice and in my country I would be entitled to a full refund, it would over-ride your onerous terms conditions and disclaimers.

It's really no secret that the documentation is crap. It's littered with errors and omissions. It's also no secret that you'd better come up with some hard data sometimes more than once to get mini-dsp to do anything but refer you to a spec sheet, which going full circle, is crap.

I will say though, they'll admit when they are wrong and they'll try to fix it. I don't know how much if anything that is worth after one's already spent hundreds of dollars...... Only the customer can decide that.

PdxRealtor wrote:

It's really no secret that the documentation is crap. It's littered with errors and omissions. It's also no secret that you'd better come up with some hard data sometimes more than once to get mini-dsp to do anything but refer you to a spec sheet, which going full circle, is crap.

I will say though, they'll admit when they are wrong and they'll try to fix it. I don't know how much if anything that is worth after one's already spent hundreds of dollars...... Only the customer can decide that.

@ PDXrealtor,

Thanks for your constructive criticism. Is it possible to behave as gentlemen as we try to get everybody to abide as per this forum rule? Thanks for your understanding to keep the positive spirit of this community.

We do work hard to push application notes (more than one a week these days) but we're always happy to improve our documentation. Since you did find many error and omissions, we'd be happy to update the manual with any comments. Please feel free to send us an email to "This email address is being protected from spambots. You need JavaScript enabled to view it."

DevTeam

StewedNKeefed wrote:

miniDSP said:
> @ 48k there is indeed a much better performance overall but if you have high Q + High gain, you'd still be running in similar issue.

I assume you mean 'running into a similar issue'.

Removing -20 kHz would be silly because that's the reason for buying a freaking miniDSP in the first place.

My concerns - and you'll admit, I've been the grumpiest of netizens on this topic - would be resolved by you actually saying in your actual selection criteria what your products are actually good for.

Discovering that you think 48kHz is better for my purposes, when you told me to buy 96kHz, is bad advice and in my country I would be entitled to a full refund, it would over-ride your onerous terms conditions and disclaimers.

@ StewedNKeefed,

You meant removing below 20Hz I guess? No issues below 20kHz.
As we've mentioned before, there are many customers (e..g. even down to commercial implementations such as the LX521 design by Linkwitz lab) that will not run into any issues with the 96kHz. (even with the below 20Hz). It just depends on the gain + Q settings at these very low frequency. But we do agree that we should add that comment for the time being. We've communicated to ADI to see if we can get access to tools. Will revert when we hear back.

If you have any further question, we'd be happy to help. We can see a plug-in purchase of 96kHz based on your ID but not a purchase of a unit. Please start a tech support ticket with your concerns and we can help you to modify the purchase of the plug-in to 48kHz.

Oh yes.... and they have an excellent sense of humor. Which when you think about it is not very gentlemen like to have a sense of humor when people are spending hard earned money to buy their products that aren't sometimes spec'd out correctly.

DT - you know I've had my issues with your product, and you know I went to great lengths to get help so I could figure out some of YOUR product and its specs. Specs that even you did not know, and in fact contradicted many times in our conversations.

So, while I might be blunt in talking with a fellow MDSP owner it is the truth. You can't say I have not contributed a great deal to your knowledge of the hardware and several areas of the documentation.

I got nothing to send to you at this time. I'm officially on MDSP vacation.

> 20Hz

I blush, yes that's what i meant! Thanks for the correction.

Also, changing the advice is a good move. That will help those who come after me.

Third, I'll raise a ticket so you can make good on my situation. Thank you.

nyt wrote:

We appreciate the effort, but removing < 20hz would not really be a solution. It's possible to get some filters working, it just takes some time and loopback testing.
That said, how hard would it be to create a 24khz plugin for the x8 platforms?

@ Nyt, a good suggestion indeed. That's typical that some DSP designers would be using 32kHz. We unfortunately can't set that low of a frequency on the DSP environment. We're in talk with ADI now. let's see if we can push them to implement Type 1 which would solve all problems at 48kHz. Will update of our progress.

DevTeam

@devteam: I am all for gentlemen-like behavior in this conversation, but you should accept that some customers are a little bit dissapointed when a product does not behave as advertised. Don't get me wrong, I really like my 4x10hd, it does the job and the overall sound quality is great. But it also did cause me some trouble. Mainly the power off plop, that was unexpected because nobody has warned me and killed one of my tweeters. And the pop hardware fix, where I almost destroyed my board by accidentally shorting a cap. Yes, this was my mistake, but it only happened, because I had to fix a mistake you have made! Fortunately, my 4x10 is well

I understand the issue is caused by the fixed point arithmetic and you can't change that, but what about this "fix":
- Implement the EXACT SAME math (fixed point with exact same rounding errors) in the preview graph, so the curve in the GUI is reliable and looks like what you acutally get. This would eliminate the need for loopback testing and make the unit very workable again.
- Implement a stability check for biquads. Imo this is VERY important. It's not that hard to implement and could save some speakers. If a biquad is entered, that is instable, the software gives a warning "INSTABLE BIQUAD" and the filter is NOT implemented.

Please do this asap.
Thank You

corri303 wrote:

@devteam: I am all for gentlemen-like behavior in this conversation, but you should accept that some customers are a little bit dissapointed when a product does not behave as advertised. Don't get me wrong, I really like my 4x10hd, it does the job and the overall sound quality is great. But it also did cause me some trouble. Mainly the power off plop, that was unexpected because nobody has warned me and killed one of my tweeters. And the pop hardware fix, where I almost destroyed my board by accidentally shorting a cap. Yes, this was my mistake, but it only happened, because I had to fix a mistake you have made! Fortunately, my 4x10 is well

I understand the issue is caused by the fixed point arithmetic and you can't change that, but what about this "fix":
<strong>- Implement the EXACT SAME math (fixed point with exact same rounding errors) in the preview graph, so the curve in the GUI is reliable and looks like what you acutally get. This would eliminate the need for loopback testing and make the unit very workable again.
- Implement a stability check for biquads. Imo this is VERY important. It's not that hard to implement and could save some speakers. If a biquad is entered, that is instable, the software gives a warning "INSTABLE BIQUAD" and the filter is NOT implemented.</strong>

Please do this asap.
Thank You

Where's the info on the power-off pop? I just leave mine constantly powered to avoid this. Such poor hardware design and QC to let this slip through.

nyt wrote:

Where's the info on the power-off pop?

Send an email @ minidsp asking for pop fix mod instructions.
You have to change/remove some caps and resistors. The pop will still be there, but with much reduced LF energy.
But be aware, those mini smd parts are a b!tch to solder for people with no smd experience (like me). I'm lucky to have access to smd solder equipment at work (dual soldering iron..) and still it was a nightmare.
I can't imagine to be successful with a standard soldering iron.

And, of course, if you ruin your board, that's bad luck.
And you have to buy the parts youself, of couse - or ship it to china and get it fixed, at your own expenses of course

nyt wrote:

I just leave mine constantly powered to avoid this.

I am not a fan of wasting energy, even if it's just a few watts.
And you still don't have any kind of protection in case of PSU failure, accidental unplugging or whatever.

Another forum has a thread on minidsp power switching issues. The thread started 2 years ago. One of the forum members has posted plans for a device that prevents the pop from damaging speakers.

The link is to post #73, page 3.

www.stereo.net.au/forums/index.php?/topi...op-issues/?p=1303953

"Use at your own risk"

If your amp has properly timed output relays and you switch everything via an multiple socket strip, you don't even need that.
The relays are faster than the minidsp's pop.

But again, be aware that this does not provide any kind of protection in case of PSU failure, unplugging etc... so I personally wouldn't count this as fix.

Does the MiniDSP 2x4 HD have the same issues that are mentioned in this thread?