It's clear in the UI when using the EQ function that it's not possible to use it as a 20 band single channel EQ. This amazes me as it seems like this would be _the_ most useful EQ mode available. Surely people are more interested in overall control rather than non granular EQ for left and right channels! Hmmmm.

Any comments anyone?

M

I would think that to have a 20 band eq, you'd only be able to have a mono output...
I'm guessing this from the current setup :-
4 channels of 5 band eq for front and rear stereo output
2 channels of 10 band eq for stereo output

Jazz
(List 112, Mk2 12 gig #40. Mk1 4 gig #30. Mk3 1.6 16v)

It's not the case that you _can't_ have 20 band stereo EQ, with the EQ applied to both channels equally! If the limitation is there I would guess it's in the Phillips ICE DSP. However, if this is the case I'll be very disappointed. I sincerely hope that if it isn't a h/w limitation that empeg will consider 'fixing' this ;)

Anyone listening?

M

Sorry, perhaps I should have mentioned my assumption in the other post... =)

I am assuming that to get a stereo output that the EQ would have to be split equally. You have 2 completely separate signal paths inside the DSP (or 4 if you set it up that way) and the EQ has to be applied separately to each signal path (else you get channel mixing).

To put it another way, I have a Nord Modular, which contains DSP chips to create sounds. Each DSP has a certain amount of processing power available. The more processing power you use per voice, the less the number of concurrent voices.
So if you have a voice that takes 25% of a single DSP's processing power, you can only play a maximum of 4 of them on a single DSP.

Now if the in car DSP is programmed in a similar way and each EQ band takes (for example) 5% of the total processing power then you can go from 1 channel (or voice from the above example) with 20 eq bands, 2 channels of 10 eq bands or 4 channels of 5 eq bands...

However, I have no evidence that the In car DSP is setup like that... =)

Jazz
(List 112, Mk2 12 gig #40. Mk1 4 gig #30. Mk3 1.6 16v)

If you look at the specs in the manual it says, and I quote, "20 band parametric set as 2 x 10-band or 4 x 5-band".

From this I assume that the DSP has the ability to do 1 x 20-band. It really would be far more useful to me if it did this.

Thanks,

M

----
MKII #080000528 - 40Gb Blue

I don't disagree with the manual, in fact my example supports this (You have 20 bands split equally between a number of channels).
It doesn't say the EQ is 20 band per channel, just that it has 20 bands.

Looking at the DSP data sheet it's actually more restricted than my example...
This equalizer accelerator circuit (EQ) can make a two-channel equalizer of 10 second-order sections per channel or a four-channel equalizer of 5 second-order sections per channel depending on the value of AD register bit TWO-FOUR (see Table 9). It takes an input sample set of 2 (stereo) samples or 4 (stereo front and rear) samples via 4 input registers. It delivers an output sample set of 2 or 4 samples via 4 output registers. All input and output registers are 18 bits wide.

http://www-us.semiconductors.philips.com/acrobat/datasheets/SAA7705H_1.pdf (Url supplied by Tony in a previous thread, I assume it's still the DSP in use)
So it looks like a non starter... =(

Jazz
(List 112, Mk2 12 gig #40. Mk1 4 gig #30. Mk3 1.6 16v)

Right, like Jazzwire just said: Empeg doesn't implement the EQ in software. The DSP hardware does it for them, and the 4x5 or 10x2 limitation is entirely within the DSP and out of Empeg's control. There is no way they could program it to do otherwise.

One of my earliest complaints was: Why offer separate left/right controls at all, why not have 10 front bands and 10 rear bands? The reason is that you can't even do THAT in the DSP, and Empeg has no control over it. All they can do is throw parameters at the DSP.

Now, this brings us to the next point, which is the parametric options in the EQ. For most professional audio work, 10 bands of parametric EQ is overkill. If your EQ is parametric, you don't actually need more than a few bands at most. Think about it. If you set up your ten EQ bands to look like this:

..............|....
..|.........|.|.|..
|.|.|.....|.|.|.|.|
|.|.|.|.|.|.|.|.|.|
|.|.|.|.|.|.|.|.|.|
|.|.|.|.|.|.|.|.|.|

Although that might look like ten bands, it's actually only two bands with low Q values. By overlapping the curves and adjusting the parameters correctly, I could draw you just about any useful real-world EQ curve you needed with only three parametric bands.

Now, before you go whining and saying "but I can't see the curves graphed on the screen", just stop. Because most hardware parametric EQs aren't even graphic at all, they just have three knobs for each band. Audio producers and musicians have been adjusting their parametric EQs by ear since before you were born. Young whippersnapper.

___________
Tony Fabris

As far as I'm aware the 10x2 and 5x4 EQ modes are restrictions of the DSP. If there is a 20x1 mode I've never heard it mentioned. The fact that the EQ is parametric is a substantial compensation for having fewer bands - do you really need to attenuate more than 10 individual band centres in your car?

Look at it this way.. we only have a parametric EQ because we ear bashed Philips for months on end to get the specs (along with some nifty hackery by John). The DSP isn't sold as a parametric EQ, you're supposed to load preset responses generated from Philips developer PC software. I doubt any other manufacturer is using the DSP in this way. The empeg wasn't originally advertised as having parametric EQ - it was a bit of a coup when we got it working.

So, look on the bright side!

Rob

Rob,
First let me say that the eq is one of the strongest features (outside the searching obvs.) but I have to chime in and agree that 10 front & 10 back would be great.

I have two reasons for this:

1. Most people have different speakers F & R, not L&R not to mention 2 different amps (1 front/1 rear).

2. A 10 front/back is more intuitive to the majority of people who don't want/need to understand tweaking the freq & q. If there's 10 instead of 5 bands then they have 2x the chance of getting close on the 1st shot without needing to engage the brain, a task that can be difficult sitting in the back seat of parked car under the hot sun trying to twiddle the EQ!

The manual is skimpy on the EQ description and I booted myself out of the sound menu a few times before I got what I wanted, so a straightforward method is preferable. Yes, the sun _was_ getting to me.

On the importance scale I give this one a 3, far below cleaning/fixing emplode, but hey if 10L/10R is possible, this one shouldn't be that hard ;-).

-Zeke
ps: if i had to explain empeg in a fairy tale it would be "a wonderful little magic box that held all the music you could ever desire..."

yah yah, it's late. I know.
-Z

just say you weren't paying much attention...

Of course, the UI could present a single, 20 fixed band equaliser, and then attempt to find a 10 band parametric setting that closely approximates what the user has set, and program the DSP (left and right) with that. Similarly for front and back. What you program the DSP with doesn't have to be the same as what the user sees!

Richard.

but I have to chime in and agree that 10 front & 10 back would be great.

Sure, a 10 front/10 back would be fantastic. But unfortunately, the Philips DSP simply won't do that for them. I had the same opinion when I first got my Mark1, but then I looked closer at the DSP specifications and realized that they couldn't do it.

In the meantime, see my other post in this thread regarding parametric EQs. When you use the parametric features, even five bands for front and back is overkill.

___________
Tony Fabris

Of course, the UI could present a single, 20 fixed band equaliser, and then attempt to find a 10 band parametric setting that closely approximates what the user has set.

Or, a better option, but one that takes even more programming...

The UI could be re-vamped so that you picked one parametric band at a time and adjusted its settings, and those settings could be graphed. You could do away with the left/right display and (as you said) show only the mono display.

Here is a picture of the best parametric EQ I've ever used. It's part of Logic Audio.



Note that in this example, only the middle 3 bands are activated, bands 1 and 5 are unused.

___________
Tony Fabris

"Here is a picture of the best parametric EQ I've ever used. It's part of Logic Audio."

Wow... you know, it has taken that simple picture to finally illustrate the power of the parametric EQ to yours truly. It never really sunk in what the hell that "Q" stuff really meant until now. Damn it. Now I'm going to have to try and figure out how to use the EQ properly and then I'll probably have to buy different speakers, and a new car with better acoustics...

Also... I understand it will never happen, but the front/rear settings sounds pretty cool. For the last six years I've wanted to give my front speakers just a little nudge in the treble area...

Brian H. Johnson
MK2 36GB Blue
"Born to Lose..."

Wow, OK Tony, I bow to your superior knowledge ;) Looks like I will have more than enough to play with then, considering what you've shown above. Thanks for the enlightenment.

M

----
MKII #080000528 - 40Gb Blue

Empeg haven't programmed the DSP. It is a masked, pre-programmed part, with programmmable features. The number of frequency centres is not programmable; what is being presented by the empeg player is what is available from the DSP itself.

One of the few remaining Mk1 owners... #00015

but hey if 10L/10R is possible, this one shouldn't be that hard

Unforch, it's impossible - because it's hard coded into the DSP's function set.

I got round it by just connecting the Left F/R to my front amp, and Right F/R to my rear amp.

Totally freaks you out when you hear it as it puts left channel instruments at the front of the car and vice-versa.

But you do get 10F/10R

One of the few remaining Mk1 owners... #00015

Tony, look what you did! You just wrecked this man's wallet.

Calvin

Also... I understand it will never happen, but the front/rear settings sounds pretty cool. For the last six years I've wanted to give my front speakers just a little nudge in the treble area...

You can do this right now. The Empeg has the ability to have 5 bands front/rear. Change the "2 Channel" to "4 Channel" in the EQ settings. (Note, don't try it at home, it has to be plugged in the car for the option to be available.)

And as I've pointed out, it's possible to get very fine control with only 5 bands as long as you're willing to spend time tweaking the parametric settings.

___________
Tony Fabris

"(Note, don't try it at home, it has to be plugged in the car for the option to be available.)"

Arg... I keep forgetting that. I usually check stuff in the house before I get into the car. While the Home/Auto behavior is really useful, it is also easy to forget.

I think I have to do more digging on this "Q" stuff. A graphical display, as in Tony's example would be nice... for those of us who think in pictures, not in numbers.


Brian H. Johnson
MK2 36GB Blue
"Born to Lose..."

Empeg haven't programmed the DSP. It is a masked, pre-programmed part, with programmmable features. The number of frequency centres is not programmable; what is being presented by the empeg player is what is available from the DSP itself.

OK, my mistake, I shouldn't have used the word "program", I should have said "set up". But nonetheless, the empeg could still present a single 20 band equaliser, or a 10front/10back equaliser to the user, but use the DSP in 5/5/5/5 mode with frequency centres, attenuations and Q values chosen to produce a close approximation to what the user has selected. Obviously if the UI is presenting a single 20 band display, then the parameters for all four channels would be the same.

Richard.

I really think you would need to look at the PDF of the spec sheet for the device, as what you describe is just not possible. I am not dissing you on this, it just sounds as if you just don't see how the DSP works - there is no "programmability" in the DSP - it is a complete subsytem including DACs and EQ in one, intended to reduce chip count by software processing in one chip.There is no capability for empeg to "re-program" the chip other than to set it up with a set of parameters (the way you would call a subroutine). The ability to switch the band centre selection between L/R and F/B is not part of the API of the chip itself; equally, given the limited processing capability of the chip and the number of output channels, the firmware designers of the chip have chosen specific modes of operation based on available resources, which are selectable, but not programmable - L/R, F/B switching not a pre-programmed API feature.

Do you want me to post the URL?

One of the few remaining Mk1 owners... #00015

Hate to say it ... you already have got a graphical representation of the output of the empeg in the EQ interface. Trouble is, it only says what empeg is outputting, not what your amp/speakers are delivering to you when you listen to it. It also does not give you an indication of the cumulative "envelope" that the output filters are generating for you; you just see vertical bars (no indication of Q/interaction between filter envelopes).

Short of you sampling the output with a calibrated mike and then adjusting your EQ settings to a known envelope (that you can observe from the sampling), the EQ UI tells you very little indeed.

One of the few remaining Mk1 owners... #00015

Totally freaks you out when you hear it as it puts left channel instruments at the front of the car and vice-versa.

Eeeewwwwww!

Here's an easier way to do it. I think that chip is a square shape.... all empeg has to do is rotate the chip 90 degrees when they solder it onto the board.

I don't know why they didn't think of that themselves. I guess it just takes an impartial, outside observer to think of these things.

tanstaafl.

"There Ain't No Such Thing As A Free Lunch"

I think I have to do more digging on this "Q" stuff. A graphical display, as in Tony's example would be nice... for those of us who think in pictures, not in numbers.

You did see my FAQ entry on the subject, right?

___________
Tony Fabris

"you just see vertical bars (no indication of Q/interaction between filter envelopes)."

And that's my problem... For me, some form of visual representation of the "Q" upon the overall curve would be most helpful. Forgive me if I'm wrong, but adjusting "Q" at one point along the graph will affect the curve at the neighboring points?(Dependant on the "Q" strength and distance?) That's a few more mental variables than I'd like to be contemplating while driving.

"You did see my FAQ entry on the subject, right?"

Yes, thank you... and in combination with the display you posted I think I am finally getting it. I just need to read it again and twiddle some knobs...



Brian H. Johnson
MK2 36GB Blue
"Born to Lose..."

It doesn't work that way, because the only input we have to the DSP is left/right audio - there's no way we could "un-rotate" this; you'd always end up with left as front and right as rear (or vice-versa) - so it affects all sound, not just the tuner.

Hugo

Of course I realized that while driving home last night...

Oh well.



-Zandr
Mk.I #150
Mk.II #39

schofiel,

Ok hold on here, let's not turn on the blinders to what rjlov is suggesting. What rjlov is suggesting is indeed possible. Consider the following:

....................
..|.|......|.|.|....
.|||||..|||||||||...
||||||||||||||||||..

This is a twenty band equilizer that the user set up for some reason, the software will take that and discover it can be approximated using 5 Q curves on a parametric. It then computes what those values are and sends the 5 curves to each of the 4 speakers. Thus a 20 band equilizer can be approximated using a 4x5 parametric. When rjlov says what the user sees is not what is going on underneath, he's saying put a standard equilizer interface on top of the parametric dsp and no one needs to know!

So basically before you pull the pdf spec sheet out and bible thump it, consider what is really possible.

Calvin

Unfortunately, if you rotate the chip 90 degrees, according to the different laws of electromagnetics, you'd then have to install the empeg upside down and that would be unacceptable.

Calvin ";)"

The parametric EQ is an audiophile function, and right now it performs as advertised. I would be nervous about advertising a 20 band EQ which in fact is only an approximation of the expected function - audiophile types are very scathing about that sort of thing.

Anyway, it's a lot of mathematics and work for something with rather limited commercial and functional gain. I don't think we'll be pulling John away from RDS for this one.

Rob

you'd then have to install the empeg upside down and that would be unacceptable.

Unacceptable? No, it would be great! We would have the volume control on the left where it should be, and the problem some of us have with not being able to read the first few letters of the info display would be solved as well.

I think you're on to something here, Calvin!

tanstaafl.



"There Ain't No Such Thing As A Free Lunch"

I agree, but I just had to point out that it is possible. The math will kill you though.

Calvin

Don't forget there is a kernel switch to flip the display upside down... =)

Jazz
(List 112, Mk2 12 gig #40. Mk1 4 gig #30. Mk3 1.6 16v)

Don't forget there is a kernel switch to flip the display upside down... =)

Definitely the result of OEM work creeping into the Empeg-Car's source code. Interesting.

___________
Tony Fabris

The real problem with rotating the DSP 90 degrees is that it will start sending imaginary numbers to the DAC.



-Zandr
Mk.I #150
Mk.II #39

I agree, but I just had to point out that it is possible. The math will kill you though.

Well, I survived it (and forgotten - it was 15 years ago)... Let's see, it's 15 degrees of freedom nonlinear fit with just 20 data points. Doable, but slow and requiring good initial guess for parameters.

However, plotting the resulting curve with known frequencies (centroids), Q's (deviations) and gains (amplitudes) should be straightforward, so it should not be too difficult to show approximate shape of the envelope.

All this, of course, assuming we are talking about Gaussians added/substracted to constant response curve (or at least something else easily describable analytically). Needless to say, knowing nothing about DSPs, I don't have a foggiest idea whether they really produce something like this, or something much wildler.

BTW, didn't we already discuss this (graphical feedback was one of my wishes), and somebody said that not enough documentation on DSP algorithms is there to model response function with any accuracy?

Cheers!

Dragi "Bonzi" Raos
Zagreb, Croatia
Q#5196, MkII#80000376, 18GB green

You're right, and with all the imaginary numbers coming out, it would kill the CPU to convert them back to reals since the CPU is only operating in integer mode. ;)

Calvin