Why are consumer EQ devices rare?

[Tks]

In most cases, you wouldn't want to use EQ both before and after DA conversion. I was simply comparing the two ways of doing it and saying the former was more optimal.



I suspect this is simply because few (if any?) DAC chips have onboard EQ. For that, you need a separate DSP chip and supporting circuitry. RME includes such a chip in their Adi-2 DAC etc., but this involves extra cost and extra work incorporating the DSP chip into the device. Most manufacturers don't bother with this, presumably since the two markets don't generally overlap greatly.

Ahh thats what I also thought (in terms of doing it only once, I thought for a moment you said to do it twice would be optimal, pardon my illiteracy).

See that whole "DSP chip" thing is what I don't understand. Why not just grab some low-powered CPU or FPGA and have it done all essentially with software so-to-speak. Like forget about dedicated hardware acceleration. Decades ago I could imagine this being a problem. I don't understand what the cost would be associated with simply throwing a general purpose sort of chip from ARM, and having a software engineer write-up a GUI and simply have the EQ done the same way it would be done on a computer like Windows (purely software)?

As for markets not overlapping. That is I think at the core of what perplexes me. I don't see how people wouldn't want EQ on nearly any audio processing device (where possible). To cut down on devices that need need to be daisy chained and whatnot? Like I understand I can have the software EQ if my source is Windows (use Equalizer APO for example) but with that I need to have my computer networked to the audio devices, and I would have to get on the computer and not have easy (relatively) access to EQ from the comfort of a couch for example.

Also, when have high cost audio devices ever cared about what "crosses over"? EQ in my mind could be another thing they add to their marketing checklist of specs to stand out among the crowd. I'm here wanting to get the RME over waiting for things like the DX7Pro simply because of creature comforts it provides. Basically what I am asking, are some of these companies asleep at the wheel in terms of progress with respect to product offerings?

In my use cases, the only thing an RME ADI 2 DAC is missing (but the DX7Pro does have) is wireless, and Balanced-Out for headphones. Slap that on, and expand the EQ points from 5, to 10 for instance, and that's the end of the paradigm shift DAC's can undergo for the forseeable future in my book. Luckily with Balanced, I have the 789 coming in August, but still need to figure out what I'm going to do with respect to wireless.

Also these "DSP chips" with respect to EQ... is EQ THAT HEAVY of a load on processors and such where implementations of EQ's is avoided because it's such a massively difficult thing to provide with a DAC for example?

 

[kevinh]

I think while there is Eq available, what is missing is a good solution for program material eq on the fly.
Something like the old Cello Palette. Low Q Tone Control:

 

[RayDunzl]

So you can for instance use the MiniDSP in the following scenario?:

(for example) Send out signal from the digital source (like a computer, or whatever), have it first reach the MiniDSP(and have it do it's EQ), then the DAC, and then the AMP/end device like speakers or headphones.

Exactly.

Source digital ->optical or coax -> miniDSP -> optical/coax -> DAC -> amp/speaker

PC is used to set up the miniDSP, not required after unless you want to change something.

What is FIR and 6144 taps? Totally clueless on that as well.

FIR see answer attempt above...

Taps are memory locations in the DSP that hold a segment (first in, first out) of the stream of samples coming in. 6144 taps means this miniDSP can use up to 6144 samples to which the filter algorithm is applied when calculating what the next output sample will be.

 

[Xulonn]

I use a miniDSP OpenDRC-DI - digital in/out - in front of the DAC

I have known for many decades that speakers and their interactions with rooms are by far the biggest variables in home audio systems, and I've toyed before with the possibility of getting a MiniDSP OpenDRC-DI, which at $325, is a bargain. Am I correct in assuming that this unit, unlike the model that Amir tested, works strictly in the digital domain and might not have the same faults that Amir measured in the miniDSP 2x4HD?

From your charts, it looks like the biggest improvement is in the bass to mid-bass region. Unlike the "unmeasureable" differences in sound quality that audiophools claim to hear, the corrected differences in your chart are huge - up to 10-11dB! (Although I do see a bass dip in the corrected plot of about 6-7dB at 45Hz - was the unit not able to correct that artifact?)

I assume that you have switched the correction on and off with different kinds of music, and I'm curious as to what you heard. Your system info shows that you do indeed have full-range speakers that go down to 20-30Hz, unlike my bass-limited small monitors. However, even with limited-bass speakers like mine, mid-bass bloat can muddy the sound.

Others here at ASR have said that they did not like what DRC did to the mid and high frequency range. Is there an option to limit the DRC filter to the bass region?

 

[mitchco]

As one of the posters already mentioned, software digital signal processing eq has replaced consumer hardware eq period. Often digital eq are part of software music players like JRiver Media Center for eample: https://wiki.jriver.com/index.php/Parametric_Equalizer

Or you can get software "plug-ins" called VST's: https://en.wikipedia.org/wiki/Virtual_Studio_Technology JRiver Media Center can accept these plugins, as does almost any Digital Audio Workstation's (DAW). https://www.kvraudio.com/ is the largest marketplace for plugin's.

DSP is so advanced in the pro audio industry that entire mega dollar hardware machines like 24 track tape recorders have been completely replaced by software: https://www.uaudio.com/uad-plugins/special-processing/studer-a800-tape-recorder.html including their characteristic sonic signature. And goes for virtually any hardware device used in the studio, all replaced by software DSP.

Further to what @RayDunzl is saying, consumers have access to powerful and sophisticated software DSP that can control the frequency and timing response of loudspeakers in rooms, with an incredible level of precision and accuracy. Software like Dirac, Audiolense, Acourate are examples of this. Some of these products can limit the their correction to bass frequency only and cancel low frequency room reflections. Some are so sophisticated and powerful it can take a small book like the one in my sig to explain and walk through the majority of it's capabilities as they are full on custom DSP Designers.

 

[Xulonn]

Some of these products can limit the their correction to bass frequency only and cancel low frequency room reflections.

Do you know of any bass-only units, that like the miniDSP OpenDRC-DI, work strictly in the digital domain?

 

[mitchco]

Sorry, no I don't. My product knowledge is with the PC versions of Audiolense and Acourate DSP software. You could contact the authors of the software, as they would know what other platforms and devices are supported.

 

[RayDunzl]

Am I correct in assuming that this unit, unlike the model that Amir tested, works strictly in the digital domain and might not have the same faults that Amir measured in the miniDSP 2x4HD?

The OpenDRC-DI has no analog jacks. In the past, there were three models, others had analog, looks like they were discontinured, though there is now an 8 analog out device. https://www.minidsp.com/products/opendrc-series

There was the Balanced Out OpenDRC-DA and the balanced in and out OpenDRC-AN - https://www.minidsp.com/support/legacy-products.

The 2x4HD has different internals, more IIR/PEQ, and even less FIR capability, but has 4 outputs, so can be a cheap crossover solution.

From your charts, it looks like the biggest improvement is in the bass to mid-bass region. Unlike the "unmeasureable" differences in sound quality that audiophools claim to hear, the corrected differences in your chart are huge - up to 10-11dB! (Although I do see a bass dip in the corrected plot of about 6-7dB at 45Hz - was the unit not able to correct that artifact?)

The room is open on the left rear, it tends to make a null at the listening position due to phase cancellation around 48Hz. The correction software listens to left, and right speakers, but not both

I assume that you have switched the correction on and off with different kinds of music, and I'm curious as to what you heard. Your system info shows that you do indeed have full-range speakers that go down to 20-30Hz, unlike my bass-limited small monitors. However, even with limited-bass speakers like mine, mid-bass bloat can muddy the sound.

The difference is more pronounced at higher volume levels. Upper bass hides the lower bass if it is unbalanced, as the second harmonic (maybe more) of the low instruments becomes too pronounced.

Others here at ASR have said that they did not like what DRC did to the mid and high frequency range. Is there an option to limit the DRC filter to the bass region?

My ear holes don't (as far back as when I was 8 years old) register high frequencies, so it is difficult for me to comment on that. Others have not complained. Sounds "right" to me.

In the case of AcourateDRC and the OpenDRC-DI, the IIR filters handle bass and makes broad strokes above that, the FIR handles the little wiggly stuff above a few hundred Hz and limits itself to +/-2dB or so.

 

[Kal Rubinson]

Do you know of any bass-only units, that like the miniDSP OpenDRC-DI, work strictly in the digital domain?

Not quite the same but you might look into the DSPeaker devices.

 

[LTig]

Something like the DCX and more properly the DEQ (with graphic equalizer as well as 10-point parametric EQ) is what I was wondering as to why they're so rare, and why not many devices like them exist (the RME and miniDSP are the closest, and only ones I know of).

It's difficult to adjust them properly even with measuring equipment. That's why typical audio consumer EQs were of the graphic type. People just used them as finer adjustable tone control because you cannot use them to EQ the room anyway.

Since most consumer EQs at that time did not improve the sound to say the least they got a very bad reputation among audiophiles (the Cello Palette was a notable exception as well as the professional stuff from Klark Teknik, but they were horribly expensive). Unfortunately the progress in the last 10 years has not had a significant effect regarding hardcore audiophiles though, especially the analog only fraction. They prefer to change tubes or the pickup to modify the sound were a little EQ would do the same ...

 

[LTig]

FIR = Finite Impulse Response. In DSP, it's the most common way to process the amplitude and phase of the filter function convolution with the signal. The more taps, the more accurate (i.e., approaching the infinite impulse response), but you hit diminishing returns relatively quickly.

And FIR filters are almost unusable for AV systems because the delay of the filter usually is too long. That's the reason why some active speakers with digital crossover offer different types of filters for audio only and audio with video.

 

[mitchco]

@LTig true enough. However, if you are into FIR filters, it is likely you are into computer audio. And if you are into computer audio, then software media players like JRiver Media Center compensates for the FIR filter delay while watching movies and lip sync is perfect. Agreed about the downside of active speakers in that scenario... In addition, most active speakers don't have dynamic loudness control, so when listening below reference level, it does not compensate for our lowering ear sensitivity to bass as the volume is turned down. JRiver Media Center has a loudness control to compensate...

"Stereo" systems were once relatively simple, but with the advent of digital audio and DSP, one needs to think about the overall system design with the choices available today to meet one's requirements.

PS. I had the Klark Teknik 31 band eq along with their RTA way back in the early 80's is how long I have been playing with eq in audio. Lots of fun!

 

[edechamps]

I don't understand what the cost would be associated with simply throwing a general purpose sort of chip from ARM, and having a software engineer write-up a GUI and simply have the EQ done the same way it would be done on a computer like Windows (purely software)?

It's definitely possible to build something like that yourself, and I believe some people have done it in the past. I suspect a Raspberry Pi based system could do something like this for dirt cheap (especially if you only need digital I/O). The main problem, though, is latency. General-purpose OSes like Windows or Linux are fundamentally not designed for low latency (real time) operation. There are large (10-20 ms at best) delays involved in each step of getting data from the input of the device, processing it, and then getting the data out. There are use cases where such delays are unimportant or can be compensated for, but there are also many use cases where they are unacceptable.

 

[mansr]

It's definitely possible to build something like that yourself, and I believe some people have done it in the past. I suspect a Raspberry Pi based system could do something like this for dirt cheap (especially if you only need digital I/O). The main problem, though, is latency. General-purpose OSes like Windows or Linux are fundamentally not designed for low latency (real time) operation. There are large (10-20 ms at best) delays involved in each step of getting data from the input of the device, processing it, and then getting the data out. There are use cases where such delays are unimportant or can be compensated for, but there are also many use cases where they are unacceptable.

Latency isn't an issue for music playback. Most of what you listen to was recorded years or decades ago. A few more milliseconds isn't going to make a difference.

 

[LTig]

[..] The main problem, though, is latency. General-purpose OSes like Windows or Linux are fundamentally not designed for low latency (real time) operation.

It's not that bad with Linux. The OSADL maintains a real time linux kernel, and they regularly push improvements back into the mainstream kernel.

There are large (10-20 ms at best) delays involved in each step of getting data from the input of the device, processing it, and then getting the data out. There are use cases where such delays are unimportant or can be compensated for, but there are also many use cases where they are unacceptable.

It's a little bit different. Moving data between user space (where the program runs) and kernel space (where the I/O is done) is not the cause of the long delays which limit the performance. The real bottleneck is that sometimes the kernel decides that there is a very important task to execute, and then it keeps the I/O of all other processes on hold, and this causes a long delay. On a non-realtime OS this delay is undefined. You can test your code for a long time and still not be sure that your buffers are big enough to cover these long delays. On a realtime kernal you still have longer delays but their maximum length is defined.

The OSADL runs several CPUs in a rack and measures the maximum delay. The last time I looked (2 years ago) on an ARM CPU this can be around 50 µs. A standard linux on the same ARM CPU may suffer delays of 2 ms (my experience; for a living I write control code for an embedded ARM CPU in a high end spectrometer where one CPU must be able to feed the hardware with a short term maximum data rate of more than 1 GB/s).

 

[edechamps]

Latency isn't an issue for music playback. Most of what you listen to was recorded years or decades ago. A few more milliseconds isn't going to make a difference.

Which is exactly why I said "There are use cases where such delays are unimportant". Music playback is one of these use cases. But if you want a versatile system on which you want to, say, play video games for example, end-to-end audio delays of 40+ ms can be noticeable.

The OSADL runs several CPUs in a rack and measures the maximum delay. The last time I looked (2 years ago) on an ARM CPU this can be around 50 µs. A standard linux on the same ARM CPU may suffer delays of 2 ms (my experience; for a living I write control code for an embedded ARM CPU in a high end spectrometer where one CPU must be able to feed the hardware with a short term maximum data rate of more than 1 GB/s).

Sure, I have no problem believing that, with dedicated and tightly controlled hardware and for specialized applications, you might be able to fine-tune Linux down to 2 ms. I'm sceptical you'd be able to go that low when you need to go through the Linux audio stack twice and exchange very small buffers with something like an off-the-shelf USB audio interface - the drivers might not even let you use a buffer size that low. I'd love to be proven wrong, though.

 

[Tks]

It's definitely possible to build something like that yourself, and I believe some people have done it in the past. I suspect a Raspberry Pi based system could do something like this for dirt cheap (especially if you only need digital I/O). The main problem, though, is latency. General-purpose OSes like Windows or Linux are fundamentally not designed for low latency (real time) operation. There are large (10-20 ms at best) delays involved in each step of getting data from the input of the device, processing it, and then getting the data out. There are use cases where such delays are unimportant or can be compensated for, but there are also many use cases where they are unacceptable.

Yeah I was thinking of going with a Pi + some addon board from Hi-Fiberry or something of the sort. Just learning too get a custom solution with properly programming on a small screen that could be within and enclosure is going to take some research.

 

[andreasmaaan]

See that whole "DSP chip" thing is what I don't understand. Why not just grab some low-powered CPU or FPGA and have it done all essentially with software so-to-speak. Like forget about dedicated hardware acceleration. Decades ago I could imagine this being a problem. I don't understand what the cost would be associated with simply throwing a general purpose sort of chip from ARM, and having a software engineer write-up a GUI and simply have the EQ done the same way it would be done on a computer like Windows (purely software)?

Just speculating here, but I wouldn't assume that high-end DAC manufacturers by and large are interested in writing software to control a DSP or implementing a UI for it. Most of them take a ready-made DAC chip and design the electronics around it, put it in a case, and sell it. Software is a whole separate area of expertise. And although the chip used for DSP would not be expensive, the electronics for the UI would be (LED screen, buttons to control it, etc.).

I think also that if you’re in the market for a DAC that costs more than a couple of hundred dollars, you’re not so likely to be in the market for EQ.

 

[RayDunzl]

I think also that if you’re in the market for a DAC that costs more than a couple of hundred dollars, you’re not so likely to be in the market for EQ.

That makes me an outlier once again...

 

[Willem]

I agree with Ray. Room interaction is the biggest problem in any reproduction chain, and the more so, the more low frequency output you have. So precisely more ambitious systems need sophisticated equalization.