TOPIC:

I have started to play around with my 4x10 Hd setup and have some questions regarding the DSP:

When setting the gain structure; will the output clip if the input reads close to 0 dBFS and I add gain (boosts) with the EQs? If so, what´s the best procedure for setting up the system? Should I first set my filters as I want them and then feed the DSP maximum gain from my source and lower the level of the input fades of the DSP and check the output meters so they don´t clip (a peak hold function would be welcome)?

When signal aligning drivers; is the delay of an output (from the miniDSP) constant regardless of filter settings or does it vary depending on numbers of filters used and how (frequency, gain and Q)? If it varies with filter usage, it might be tricky to align drivers since one thing affects the other.

Would it be theoretically possible to make a 48 kHz version of the 4x10 plugin in order to get more filters/compressors and better resolution in the low frequency range? Would it be possible to redistribute the possessing power so that instead of having 5 filters for each output, one could have 10 filters for the first output and none for the fourth output (since I don´t currently use it)?

Would it be possible to get better accuracy for the frequency selection in the filter section? 1 Hz steps is sometimes not enough in the lowest octaves. It would be better if the frequency resolution was logarithmic since you seldom need 1 Hz step resolution in the highs.


Some requests for upcoming updates of the software:

* Phase flip on inputs (that way you can make a preset for “BBC style” MS listening: one side phase reversed and then summed to mono; very useful for studio application).

* Peak hold function for metering. Would speed up setting gain structure. And why not a global clip indicator (also with hold function) for the entire chain if clipping can occur within the DSP due to aggressive boosts).

* Copying EQ settings to other channels (I´d like to have the same filter settings for the digital and the analog inputs).

* Bypass all filters in a section (instead of just one).

* Clear all filters in a section.


Sorry if these questions have been answered previously (please direct me to the relevant threads if so), but I didn´t find anything when doing a quick search.

Hi, I'll answer some of these since I went thru the same process

Noob wrote:

When setting the gain structure; will the output clip if the input reads close to 0 dBFS and I add gain (boosts) with the EQs?

The answer is that it might. It depends on the amount of boost, where you put it, and the nature of your input signal. If it's a "Peak" type filter, it's unlikely to cause clipping unless the Q is low and the gain is high. If it's a shelving filter it's easier to do, unless the frequency of the shelf is quite low or high. With shelving filters it is "safer" to use a cut instead. Most systems have "excess gain" anyway.

When signal aligning drivers; is the delay of an output (from the miniDSP) constant regardless of filter settings or does it vary depending on numbers of filters used and how (frequency, gain and Q)? If it varies with filter usage, it might be tricky to align drivers since one thing affects the other.

I have always assume that the delay is constant for all outputs within a miniDSP board. I would be surprised if that is not the case. The delay would only be an issue if using other hardware in "parallel" - in that case you would have to measure it.

Would it be theoretically possible to make a 48 kHz version of the 4x10 plugin in order to get more filters/compressors and better resolution in the low frequency range?

This already exists, in the form of the 10x10 plugin.

Hope this helps.

Hi john, thanks for taking the time and answering.

john.reekie wrote:

The answer is that it might. It depends on the amount of boost, where you put it, and the nature of your input signal. If it's a "Peak" type filter, it's unlikely to cause clipping unless the Q is low and the gain is high. If it's a shelving filter it's easier to do, unless the frequency of the shelf is quite low or high. With shelving filters it is "safer" to use a cut instead. Most systems have "excess gain" anyway.

Hm... So you´re saying that the DSP has built in headroom (how much)? That would suggest that if only using cuts, the S/N ratio is higher than it could be unless you compensate by boosting the output somehow in the DSP?

If that is correct, then the only way of knowing is to feed the system a full range sine sweep (after setting up your filters) close to 0 dBFS (like 0,1 dBFS) and check the input and output meters. The problem is that there’s no peak hold so the sweep needs to be quite slow or you might miss a clip indication.

john.reekie wrote:

I have always assume that the delay is constant for all outputs within a miniDSP board. I would be surprised if that is not the case. The delay would only be an issue if using other hardware in "parallel" - in that case you would have to measure it.

I would hope so but at the same time it doesn’t make sense if the output was not affected (delayed) if adding filters (unless the clever folks at mninDSP has thought of this and made the delay fixed regardless, I hope this is how it works). Yes, I will measure it (the alignment of the drivers); naturally.

john.reekie wrote:

This already exists, in the form of the 10x10 plugin.

Oh, so I can use the 10x10 Hd plugin for the 4x10 Hd and get at least on more filter and better low frequency accuracy then. Good info, thank you!

Hi, "it depends" means "it depends on the input signal and the boost that you apply." So no, I didn't sy what you said I said. You obviously understand enough to figure out the rest for yourself. Cya...

john.reekie wrote:

Hi, "it depends" means "it depends on the input signal and the boost that you apply." So no, I didn't sy what you said I said. You obviously understand enough to figure out the rest for yourself. Cya...

Sorry, perhaps I didn´t understand whay you said correctly.

john.reekie wrote:

The answer is that it might. It depends on the amount of boost, where you put it, and the nature of your input signal. If it's a "Peak" type filter, it's unlikely to cause clipping unless the Q is low and the gain is high. If it's a shelving filter it's easier to do, unless the frequency of the shelf is quite low or high. With shelving filters it is "safer" to use a cut instead. Most systems have "excess gain" anyway.

I understood it like you said that it´s unlikely to cause clipping unless low Q and high gain. If you feed a 0,1 dBFS signal and then add even a small amount of gain; it will clip unless there is some headroom in the DSP or the DSP uses float point math (and then I don´t know what logic to apply when trying the set a good gain structure).

Now if there is some hidden extra headroom, the S/N ratio will undoubtedly be higher than optimal unless you compensate for the low output to the DAC (assuming you haven’t used the extra headroom already).

OK look, why don't you just try it?

john.reekie wrote:

OK look, why don't you just try it?

I will naturally, but it wouldn’t hurt to know in advance how the system works now would it?

No of course not. Perhaps it is me that did not understand your question. If it is related to the specifics of the algorithm used by miniDSP then you would have to ask them directly.

@ Noob,

Quite a lot of questions here and John already answered most of them. In general, when setting up a system, you need to worry about the so called Signal to Noise Ratio (SNR) and Headroom.

Setting up a system too close to 0dBFS (i.e. max the ADC/DAC can handle) doesn't make sense since you'd have no headroom left. The internal structure of the DSP is 28bit but 0dBFS is 24bit. So you do have some internal headroom but ultimately, 24bit (0dBFS) is your limit.
@ the meters, the 0dBFS we show is the one of the ADC/DAC.

As for the little "wishlist" you have, it's always good to hear some feedback but know that some of them are just not technically possible. Others can be added to the list for a future upgrade and considered how much work they require to be implemented. We always do our best to see how we can help. Sometimes, some work arounds (e.g. 0.5Hz resolution) can be done with other ways (e.g. biquad). Not sure I'd be able to hear/measure a 0.5Hz though..

Hoping this info helps.

devteam wrote:

@ Noob,

Quite a lot of questions here and John already answered most of them. In general, when setting up a system, you need to worry about the so called Signal to Noise Ratio (SNR) and Headroom.

Setting up a system too close to 0dBFS (i.e. max the ADC/DAC can handle) doesn't make sense since you'd have no headroom left. The internal structure of the DSP is 28bit but 0dBFS is 24bit. So you do have some internal headroom but ultimately, 24bit (0dBFS) is your limit.
@ the meters, the 0dBFS we show is the one of the ADC/DAC.

As for the little "wishlist" you have, it's always good to hear some feedback but know that some of them are just not technically possible. Others can be added to the list for a future upgrade and considered how much work they require to be implemented. We always do our best to see how we can help. Sometimes, some work arounds (e.g. 0.5Hz resolution) can be done with other ways (e.g. biquad). Not sure I'd be able to hear/measure a 0.5Hz though..

Hoping this info helps.

Hi and thanks for taking the time.


So the delay of each output is fixed regardless of the filter settings (EQ / X-over even if bypassed)?

Yes, I´ve tested some and have noticed that the output indeed clips if boosts are used (no or limited built in headroom I O W, make perfect sense to me) so I used the slow sine sweep at 0,1 dBFS to set my gain structure ones the filters are set up (for all 4 presets).

One feature I would really find useful (if technically possible) is if you could use the leds on the front panel to indicate clipping (at any stage, perhaps input clip on the leds to the left and output clip on the right) and a user definable headroom value (warn about clipping at X dB before actual clipping). This would make the 4/10-10 Hd even more attractive for custom installs for clients.

Perhaps your right about higher resolution might not make a big difference but if you´re trying to battle a high-Q dip at 18,5 Hz, it kind of feels odd to having to resort to 18 or 19 Hz. After all, is it not a bit odd to have only 10 steps of resolution in the first octave (10-20 Hz) and 10000 steps in the last (10-20 kHz)? A logarithmic resolution distribution would have been more natural I think. Even if biquad can resolve this eventual problem (resolution), I find it hard to use the biquads since you always need to adjust the filters and confirm with more measurements.

Regarding using the 8x8 plugin instead of the 2x8 plugin for the 4-10 Hd; are there any benefits of using 96 kHz sample rate at all (assuming the SRC is not terrible at down sampling from higher rates)?

Would it be technically possible to make a 2x8 plugin with for example only 3 bands of EQ on two of the channels and 9 bands on the first two channels? Or even better; to be able to freely assign on what channels you want how many bands (and naturally understanding that there are limited number of total bands available).

- Not sure I understand your point when you say you used a 0.1dBFS signal to setup your system, but in general, make sure to keep "headroom". 5 to 6dB is the least we'd recommend. Else without knowing it you'll overdrive the DSP (increased THD).

- LED on Front panel: Yes a good feature but unfortunately not possible on the hardware (i.e. lack of front panel). Maybe on a different product in the future.

- Resolution: The step is fixed across the range. That's the GUI works across all our platform I'm afraid.

- 48 vs 96kHz. Well, that's more that everybody asks for higher sample rate.. If you can hear the difference, then that's valuable. A lot of elements comes into play and not just sample rate when it comes to system design. We don't want to say 48 sounds the same as 96kHz, but in general, it's not rare that people would hear a lot more the difference between a properly tuned system rather than a 48 vs 96k A/B test. Our 2cts.

- As for your last question of a custom plug-in, building a plug-in is a very long process 1~2month due to the complexity of these systems. There are already a lot of requests for some features so to be honest, not sure if that custom configuration you're asking for (i..e never been asked before) would go ahead. But we can always add it and see in the future if few people ask for it (process for releasing a feature).

DevTeam

devteam wrote:

- Not sure I understand your point when you say you used a 0.1dBFS signal to setup your system, but in general, make sure to keep "headroom". 5 to 6dB is the least we'd recommend. Else without knowing it you'll overdrive the DSP (increased THD).

- LED on Front panel: Yes a good feature but unfortunately not possible on the hardware (i.e. lack of front panel). Maybe on a different product in the future.

- Resolution: The step is fixed across the range. That's the GUI works across all our platform I'm afraid.

- 48 vs 96kHz. Well, that's more that everybody asks for higher sample rate.. If you can hear the difference, then that's valuable. A lot of elements comes into play and not just sample rate when it comes to system design. We don't want to say 48 sounds the same as 96kHz, but in general, it's not rare that people would hear a lot more the difference between a properly tuned system rather than a 48 vs 96k A/B test. Our 2cts.

- As for your last question of a custom plug-in, building a plug-in is a very long process 1~2month due to the complexity of these systems. There are already a lot of requests for some features so to be honest, not sure if that custom configuration you're asking for (i..e never been asked before) would go ahead. But we can always add it and see in the future if few people ask for it (process for releasing a feature).

DevTeam

I meant -0,1 dBFS

By sending a slow sine sweep at -0,1 dBFS through the DSP, one can set the internal levels (after the filters are set up) so that the output never clips.

So how about the individual channel delay; is it constant regardless off filter settings or does it depend on filter usage?

Regarding bands per channel: wouldn’t it be relatively easy to modify the software so that one could route the output of channel 3 to the input of channel 4 (internally within the DSP) in order to get 5-6 more bands of EQ (assuming you only need 3 pairs of outputs) for a channel?

I´m going to try the 8x8 plugin for my 4x10 Hd to get 6 bands per channel and increased resolution, but I really think that many users would enjoy a software that lets the user decide how to distribute the available processing power in order to maximize the setup for the individual need (and perhaps also select the sample rate).

Don´t get me wrong; you guys are making super products at an attractive price point but there´s always room for improvement … right?

One thing I would really like to see added to your product range is a modular power amp solution. Say 3-4 different amp modules (like 180, 400, 700 and 2k watt) that one can combine freely and att to a readymade rack unit. That would complete the product range in a nice way I think.

devteam wrote:

@ Noob,

Quite a lot of questions here and John already answered most of them. In general, when setting up a system, you need to worry about the so called Signal to Noise Ratio (SNR) and Headroom.

Setting up a system too close to 0dBFS (i.e. max the ADC/DAC can handle) doesn't make sense since you'd have no headroom left. The internal structure of the DSP is 28bit but 0dBFS is 24bit. So you do have some internal headroom but ultimately, 24bit (0dBFS) is your limit.
@ the meters, the 0dBFS we show is the one of the ADC/DAC.

I´ve been playing around a bit with the 2x4 balanced board (not connected to any speakers but only to my measurement gear) and have made the following observations:


There is no "internal headroom", since if you feed the board a sine signal just below clipping (input) which reads as -6 dB (and not -3 dB as it should if RMS for a sine wave) on the meters in the software since these for some strange reason are RMS (not very useful for indicating a digital clip. A peak meter with a peak hold function would certainly be a welcome addition): the output will clip even if you only add 1 dB of boost with the EQ.

So, when setting the gain structure, I guess the safest bet is to feed the miniDSP a steady sine and adjust the gain of your sending gear (or change the input sensitivity of the board) so that the input meter (RMS …) never goes above -6 dB (again, it should indicate -3 dB for a sine wave. The level meters seem to be incorrectly calibrated using a 20 log instead of 10 log. If I drop the input level 10 dB from my measuring gear, the input meter in the miniDSP drops by 20 dB so there is obviously something not right here) and if you add any filters; adjust the output (or "input") gain in the DSP to compensate for the gains made by added filters (even if you only use cuts you still want to reduce the gain a couple of dB since there might be some slight boost around the corner frequencies) in order to prevent the output to clip.

Finally, it seems as if the delay of the individual outputs indeed are fixed and does not change when adding EQ or X-overs (as expected; all good). Total delay of the 2x4 balanced board (assuming no additional delay is added in the DSP) is 1,088 ms, not bad at all.


Please update the software so that the input meters show the correct values, and peak instead of RMS (and a peak hold function) so that setting up can be performed in a reliable manner.

Devteam:

Is it true that you´ve been aware of this problem with the level indicators for more than a year and still no fix? If so, it worries me a bit. Even if you know about it and are working on a solution; why not at least inform your current customers about the issue and how to use the level meters until the issue is addressed? I wonder how many users are running their boards into clipping without even knowing since the indicators does not indicate clipping at the correct level.

Don´t get me wrong here; I really like your stuff, but I think the communication could be somewhat better if the bug fix response time is slow.

Noob,

There are a number of threads related to proper gain structure, meter issues, IIR filter delays, etc, that answer all your questions.
Why not do a little searching on the forum, vice giving devteam a hard time?

Dave.