Jump to content

Which speaker controller is the right one for me?


peniku8

Recommended Posts

I'm looking for a system controller for a very specific setup: I need it in a EQ→Limiter→Limiter→EQ topology. Does this exist in some kind of modular dsp or would I have to cascade multiple units for this purpose? On the pc I'd just make a plugin chain as needed. Seems to me like hardware is a decade behind :(

Alternatively: are there speaker controllers which allow me to import an impedance measurement of my subs so they can properly limit according to power instead of voltage, so I don't have to build this myself with the aforementioned chain or EQs and limiters?

Link to comment
Share on other sites

  • peniku8 changed the title to Which speaker controller is the right one for me?

Symetrix or QSC or BIAMP DSPs do arbitrary signal processing with a flow graph approach.  You can download the software and design site files offline to see if it will work for you.  The Symetrix limiters accept side chain input which could simplify your signal flow.   Powersoft amps can limit on real power input to the driver.

I'm not quite sure why no one has developed an amp that senses real time voice coil temp from the copper/alu thermal coefficient yet but that should be possible (and guarantee driver survival from thermal failure).  A low frequency small amplitude excitation could be used to sense DCR.

Link to comment
Share on other sites

1 hour ago, kipman725 said:

Symetrix or QSC or BIAMP DSPs do arbitrary signal processing with a flow graph approach.  You can download the software and design site files offline to see if it will work for you.  The Symetrix limiters accept side chain input which could simplify your signal flow.   Powersoft amps can limit on real power input to the driver.

I'm not quite sure why no one has developed an amp that senses real time voice coil temp from the copper/alu thermal coefficient yet but that should be possible (and guarantee driver survival from thermal failure).  A low frequency small amplitude excitation could be used to sense DCR.

Thanks for the suggestions, but they all seem to target an installation market with the phoenix connectors, I'd prefer XLR since the intended use is live sound.
It looks like every hunt for something like this leads me back to Powersoft. I've looked into lake processors, but they're big buck. So much so, that I'd rather just replace all my amps with Powersoft for ease of use and better reliability (over clone amps).
My clone amp is a dsp amp (10kQ amp section) but nothing but a peak limiter. I'm scared of letting DJs loose on the system.. without absolutely choking it in the first place just for safety concerns.. Maybe I'm just gonna do a few gigs like this until a bit more money came in and then go to Powersoft.. in late 2025 it seems, since Germany just been canceling everything everywhere for the past two years.

Link to comment
Share on other sites

I run peak and 'RMS' (longer time constant) limiters, usually the system is over provisioned and we are controlling output level through the mixing desk though so they are not activating.  I also have a compressor on the input with a long attack and release and medium knee that is adjusted with pink noise to engage when the output limiters are starting to engage such that its impossible to activate the output limiters constantly with music signals (the compressor effectively turns everything down if that happens).  Perhaps its just our priorities with the speakers we bring but I find the most demanding content is live bands on the tweeters where we have observed limiting and once lost a diaphragm.

Its pretty easy to bring the connectors out to patch panels, I use Symetrix 8x8 DSP + DIGIO and wouldn't go back to fixed signal flow.

Link to comment
Share on other sites

I'm not a fan of patch panels and breakout cables. More points of failure and more rack space needed.

Is your input limiter a multiband limiter or does it pull down the top end with the bass?

I usually don't run into tweeter issues, but also broke one once, which was due to old age probably (the surround shattered). My mains are ARCS WIDE, which have some pretty beefy B&C compression drivers in em I think.

Link to comment
Share on other sites

  • 2 weeks later...

My limiter on the input is a simple compressor not multiband its more like a system volume limit, if I want to restrict the maximum volume I just reduce the threshold.

The outline newton might be able to implement your setup? I haven't used one (out of my price range...): https://www.production-partner.de/test/outline-newton-audio-dsp-prozessor/

Link to comment
Share on other sites

I think at this point I'm best off getting some Powersoft amps and the question becomes a theoretical question:

A controller needs to know a speaker's impedance over frequency plot, which the Powersoft amps can easily obtain, since they're directly hooked up to the speaker (it's an amp after all). So what I'd be interested in would be: is there a controller which accepts impedance plots to semi-accurately (temperature related) limit amplifier output power?

If not, you'd have to implement this yourself via EQ→Limiter→reverse EQ, which was the idea behind a free flow dsp.

Link to comment
Share on other sites

I'm not aware of any outboard appliances that offer much in the way of modularity, which is a real bummer.  I've been using my own modular PC-based DSP software for 5+ years now.  I wrote it because I wanted something Linux-based that I could experiment with.  I had been using MiniDSP devices and hated having to use their pointy-clicky Windows-only interface to get my filters installed, especially being that I was already computer-generating the filters.

I'm now working on the successor software, which could become part of standalone appliances some day.  I'm targeting multiple hardware platforms---not just PC but also potentially ARM-based stuff like Raspberry Pi.  I'd like to be able to use it to build fully integrated (DSP+amp) speakers, and the functionality you're looking for is something I desire for that purpose among others.

A key difference with my processor vs. a VST-based processor is that my architecture is tightly integrated.  As latency is pushed lower, extra layers of indirection contribute non-trivial cost if the process graph is complicated.  The downside is that I don't support external modules, which is what VST is all about.  In theory, such support is possible, but it's not my priority now.

I know my work doesn't help you right now, but I thought I'd mention it because I know I'm not the only one who's wanted/needed these kinds of capabilities.  Integrated and outboard devices are even less useful when you need to process many channels and want some kind of matrixing capability.  For example, how do you optimally bass manage 5+ speaker channels (at different room locations) + "LFE" to 4 subs (also different locations) in a home theater?  Answer: with even current state-of-the-art solutions, you don't, really.  Ideally you want separate sub delay (and maybe EQ too) for each signal between each input and output channel pair.  This is especially important in larger rooms or where you want to (properly) do something like a double-bass array.  Ideally, each speaker and the sub(s) its crossed with behave as a coherent source throughout the crossover.  In practice this is very hard to achieve if you can't at least apply separate delay to each combination of input channel and output channel.

Link to comment
Share on other sites

I've seen various PC based crossover/DSP software on DIYAudio and had assumed that the delays would be too great for live audio.  What kind of delay do you get SME?

It should be easy to temperature limit sub woofers as the thermal overload is slow usually over hours and the voice coil is aluminum or copper for which we know the thermal coefficient.  The amplifier should be able to use a low frequency excitation of low amplitude (say 5Hz) to sense the voice coil resistance and by comparison with the cold resistance at a known temperature sense the voice coil temperature (using the thermal coefficient).  This way you wouldn't even need to have an impedance plot of the cab.  For high frequency drivers however this wouldn't be suitable and we would have to do some kind of model fitting to extract the Re term from live impedance plots obtained via dual channel FFT analysis. 

It's not going to happen but the ideal platform for this would be if Powersoft just opened up the FPGA or DSP chip that they presumably have inside their amps which must already have ADCs monitoring output voltage and current....

Link to comment
Share on other sites

You have to be careful with PC based round trip latencies, the audio interface only displays its own latency, which omits USB controller latency for example, which is the main issue when it comes to USB based solutions. Even if the latency shows as 1ms in the interface settings, the USB controller will add another 5-10ms to that afaik.

Use thunderbolt (pcie) for direct communication with the CPU. There are also network based solutions, like DigiGrid, which looks compelling to me (been thinking about getting DigiGrid for the studio for a while now).

Link to comment
Share on other sites

On 2/2/2022 at 6:09 AM, klipsch said:

Nice post SME. Best wishes if you decide to go commercial. The need does exist. I looked at doing some quick and dirty C level functions, but didn't make it a priority with my free time. Instead, I ended up with a great deal on 2 used 10x10 minidsps for my single venue needs. 

Thanks much!  Business-wise, I'm feeling very risk-averse since the pandemic made a mess of everything.  The real killer app that I am developing is for sound quality optimization, and I believe what I can do is a game-changer for audio reproduction, certainly in home listening rooms.  From a business standpoint, that's actually kind of a problem.  I could probably get paid a lot of money to allow the tech to be buried forever.  I don't want that at all, but in the game of markets, I might have that choice made for me if I'm not extremely careful.

The other thing is that I could probably occupy myself for a decade or more with more R&D including more formal scientific work---theory development, listener experiments, etc.  If I could self-fund it and find the right advisor for myself (someone not in audio most likely---I need someone with strong statistics / information theory background), I might do a Ph.D.

I should at least mention here that there is at least one Linux + open source DSP program that is quite capable.  See BruteFIR.  I opted to make my own DSP for various reasons, but with a bit of tinkering, BruteFIR may be usable to some.

22 hours ago, kipman725 said:

I've seen various PC based crossover/DSP software on DIYAudio and had assumed that the delays would be too great for live audio.  What kind of delay do you get SME?

It's complicated.  My personal setup has a pretty high delay because I don't really need low latency.  I'm using 3 x 1024 sample buffers at 48 kHz, which I think works out to 107 ms round-trip plus whatever the interface adds (single digits ms, I think).  I'm using larger buffers because I'm running 12 x 128k FIRs, 12 x 64k FIRs, 8 smaller FIRs, and a few hundred PEQs.  It's also running on a single core mid-level 2009-era CPU.  I have many big opportunities for improvement.  My new code-base supports multi-threading and multi-rate processing, which will improve compute efficiency a lot.

I'll likely have to get away from USB to get the best latencies.  I can't use thunderbolt on my Motu16A in Linux and don't know about other hardware, but the 16A does support AVB.  AVB is arguably the best technical standard for Ethernet audio transport, though it's not the most widely supported because it requires specialized hardware capabilities and quite a bit of supporting software as well.  I've only spent a few hours on it, but I haven't gotten AVB to work on my Motu Ultralite AVB yet.  It passes audio but is very glitchy.  I think I can get it to work with more time, but CPU overhead may end up much higher than I prefer.

OTOH, the AES67 and compatible protocols (which the Motu stuff do not natively support) are simpler and a lot more likely to work with typical Ethernet interfaces.  Looking just now, it appears that Linux support for AES67 has matured a lot!  When I am further along on the software and am ready to spend more time looking at hardware, looking at AES67 capability will be a high priority for me so that I can support a wider variety of hardware.

Just a side note about delay in general is that the filters themselves can introduce non-negotiable delay.  Linear phase filters and crossovers require additional delay depending on how low in frequency, how high Q, and how much accuracy is desired.  Even delaying a speaker in order to better match it to a sub with a long horn expansion or in a different part of the room is contributing very real delay.  Fortunately, I don't think linear phase filtering or crossovers are critical for outstanding sound.  It's something I need to revisit, but as far as I can tell, phase is not terribly important as long as the shift is not too dramatic.

========

If you all don't mind me asking you all, for live sound purposes, how much latency do you all think is tolerable in a DSP appliance?  Lower is better, obviously, but at what point is it really a problem?

Link to comment
Share on other sites

  • 2 weeks later...

Interesting the comments about USB protocol delay been additive to the number you see in the driver, would make using very small buffers quite pointless.  Ethernet should be much better but you might require small frame lengths to achieve the best possible latency.  For live sound I would say around 50mS would be acceptable for FOH but perhaps as low as 10-20mS for any monitoring system.  For FIR filters for speaker correction 512-1024 taps @ 48KHz would seem to be OK for live sound.  I notice the Linea research ASC48 has 768 tap filters @ 96kHz correction filters (~4mS delay).

If you get your DSP software running on ARM you could also potentially use something like Xilinx Zynq to accelerate FIR filters.

 

Link to comment
Share on other sites

On 2/20/2022 at 4:05 PM, kipman725 said:

Interesting the comments about USB protocol delay been additive to the number you see in the driver, would make using very small buffers quite pointless.  Ethernet should be much better but you might require small frame lengths to achieve the best possible latency.  For live sound I would say around 50mS would be acceptable for FOH but perhaps as low as 10-20mS for any monitoring system.  For FIR filters for speaker correction 512-1024 taps @ 48KHz would seem to be OK for live sound.  I notice the Linea research ASC48 has 768 tap filters @ 96kHz correction filters (~4mS delay).

If you get your DSP software running on ARM you could also potentially use something like Xilinx Zynq to accelerate FIR filters.

 

My speaker optimization requires way more than 1024 taps.  More like 128k taps, as I noted above.  ARM support would mean running the DSP on the CPU itself rather than a dedicated DSP chip.  Though I'd like to look into supporting dedicated DSP chips as this could help lower latency further, it's likely to involve some heavy R&D investment and is a low priority.

You make a good point about FOH tolerating more latency than monitors, and I appreciate you sharing these figures, which seem very reasonable to me.  Of course we're talking about how much total latency is acceptable, not how much is acceptable for a single device in the chain.  I'm thinking that for devices intended for live performance monitoring, I want to aim for under 5 ms if possible.  If I can do this, it will probably be for a much smaller number of channels than would be possible with higher latencies.  That's probably an OK compromise. 

I should also be clear that I do not believe in DSP optimizing a speaker and sub(s) separately.  They must be optimized from measurements with the crossover in place or from precise simulations of the crossover.  Otherwise, the result is completely spoiled by the crossover.  If the mid-bass is screwed up, the whole bass response will be subjectively *much worse* for it.

Link to comment
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.

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.

Loading...
×
×
  • Create New...