Sharing my tune for feedback: GB60, GB25, GB10. DSP408

J

Julian

Noob
Hi all, I am making this thread for two reasons. Firstly, to share my tune in case anyone is interest. Secondly, to get feedback.

For the settings in REW RTA:

Mode: RTA 1/6 Octave

FFT Length: 16384

Averages: Forever

Window: Hann

Max Overlap: 93.75%

Update Interval: 1

Anything wrong there?

Umik-1, Andy's Curves, Jazzi's Spreadsheet, Frogs, DSP408.

My pink noise tracks I think I got from Skizer and burned to a CD. My truck is the last year with a CD player.

My technique is just to do small circles with the mic around my left ear for 10 seconds or so and right ear for about 10 seconds while the RTA runs and then I hit stop, and save the current values.

I did some test calibration with the GB25's first. This is basically just to learn and see how they respond and get a feel for what I was doing.

These mids fire from my A-pillar.



I currently have them set to play from 300hz-3500 hz.

I was getting this massive peak at 700k. It sounded really bad before the tune.




I used a match range wider than the crossover settings:

175 to 4500 hz.

Not sure exactly where this should be. Skizer had said in a video to tune outside the crossover range, so that is what I tried to do.

I set the slopes in the crossover setting in REW to 0 and the speaker type to none to try to eliminate REW from modifying the curves the spreadsheet had generated.

I did a flatness target of 1 dB, I think 2 would probably have been better.

Max boost was set to 12 dB.

This is what I ended up with after 2 iterations:



The first iteration I just tried to remove that 700hz peak with the EQ it recommended in that region.

Then I ran it again once that was fixed.

Much better. It sounded pretty good. I ran out of EQ bands in the DSP, so I just used the ones that would correct the biggest errors.

I realized that I had the crossover set to 4k in the DSP instead of 3500, so I will fix that when I try the tune again.

I've heard some people say there is a lot of skill involved in tuning with a more limited number of DSP bands.

If there is some benefit to upgrading my DSP to something with more bands, like a Helix, please let me know.

I don't know how exact it needs to be for the ears to hear the difference. I have heard 2 dB is audible?

One thing I did notice is that I had pretty dramatic differences between the peaks when the mic was at my left ear vs the right and that they tended to average down a bit as I ran the RTA for 10 seconds per ear.

So this is just preliminary stuff to go over what I did and the result so that I can fix any mistakes in my technique before I proceed with the other channels.

Obviously, the difficulty is going to compound as I try to match left/right and the different component speakers. So I want to make sure the basics are down first.

Appreciate any and all advice!

Thanks for reading.
 
Last edited:
I also want to thank Jazzi for pointing out some mistakes in a YT vid I did, which I will be fixing in a new video. I am hoping to learn some more tips from this thread that I can incorporate into the video. And of course make sure that I am giving accurate info. Not trying to profit from the channel, just to help people learn.
 
I've been at this myself for a few months now. Still learning, I'm not even walking yet. First initial step for me is to take measurements with only saftey crossovers in place with the first mic measurement at 1/48 set in RTA. That way you can see where your speakers drop off naturally. Then use REW to determine where crossovers need to be set. Then I use REW to help me decide what type and slope to use (this is the tedious time consuming part). This is where acoustic vs. electric crossovers come in to play. Once crossovers are determined, then move on to EQ. Looks like your on your way. I'll try not to give bad information. We'll see if we can help each other out
 
Yes, your speakers will have a natural rolloff caused by the environment they are in. Your left tweeter may roll off naturally at LR4 while your right tweeter may roll off BU2. Left side might play down to 2200 hz while right side plays to 2000 hz. Start with left tweeter, take raw measurement with no eq and crossover set to say 1500 hz. Don't play it too loud as to not damage the speaker. Not necessary to play at listening levels for this step. Pull this measurement up in the eq section. Under target settings, target type select "speaker driver" then enter the crossovers you think you want to use. Then click on target level. Make sure your house curve is loaded first. When you choose "speaker drver" then only the overall house curve needs to loaded for all speakers. Now click on "EQ Filtes" down at the bottom you will see "crossover filters" enter what you think the rolloff looks like. Try different types (LR BU BE) , choose HP for tweeter, then choose 3000 if thats where you want your crossover set, then hit enter and close out of "EQ Filters". You will now see a "predicted curve" that may look like your crossover is set to 3500. Keep trying different settings until your "predicted curve" matches your house curve. Once its as close as you can get it then put those crossover settings in the DSP. Now remeasure your left tweeter the enter new measurement back into REW. Now do your EQ process. You will use much less eq filters now. 70% of your eq filters were probably used to get the slope correct. Let the crossover do most of the work. Now you may only need 4-5 filters for eq.
 
Thanks for taking the time to type all that Stycker! I messed around a bit just taking measurements with the crossovers off and guessing the freqs just so I could tune it a bit real quick. Just saw your reply and will try that tomorrow.

This baseline tune is better, but the mids need work. The soundstage shifts for different frequencies and theres a mid band that "smears" and distorts.

This is my mids and midbass playing. My tweeter sails are being made right now. Should be done soon. Getting a head start.
 
Yuck! Lots of room for improvement.

Let me sort the crossovers. Then I will post some graphs after a new tune.
 
I think it would be helpful to take higher-res measurements as well. Try taking your measurements in RTA 1/48 mode. You can always smooth them later while viewing, but if you take measurements in RTA 1/6th mode, you lose a lot of details.
 
1) Take screen shots and not phone pictures.
2) Every window have set to 5db increments on the db scale. Some of yours have 10, some have 5, some have 2.
3) Read this: https://www.caraudiojunkies.com/sho...ustic-Crossovers&p=70354&viewfull=1#post70354
4) Don't use REW's auto EQ. Try doing it yourself. I usually run auto eq, see how many bands it needed, take a screenshot, then try to beat it with a few less bands needed by manually entering the frequency, q, and gain. This is how you can maximize limited bands of EQ.

5) These are my RTA settings. FFT Length is matched to what the pink periodic noise I generate via REW. . If one is changed, the other is changed as well. Rectangular Window is because of pink periodic noise and based on what Justin Zazzi told me to use for that. I don't understand all of that stuff but he does and I trust him. In other words, the window means something and not just randomly changed.

View attachment 15724
 
Last edited:
I created a new pink noise file using REW to use. I changed my RTA settings to match it.

I will post a better photo of acoustic response of the gb25. I had one but I left a filter on in my DSP and then my laptop ran out of juice.
 
Last edited:
GB25 acoustic crossovers:

Passenger:




Driver:


Based on the sharp drop starting at 3k I decided to set my filter to 2500 hz.

I played around with the crossovers in the DSP to try to get a better match to my target and then tuned.
I relied too much on the REW auto filters, but tried to make them a bit more efficient manually since I only have 10 filters in the DSP408.



This is the driver's side after tuning:




And the passenger side:


free image hosting service

And now both together:




I know it could be better, but I am still learning and it was a good learning experience. If there is anything glaring you see, please let me know. I run out of filters because I am not that efficient.

Thanks!
 
Last edited:
I also like the idea of using high resolution in the RTA window like 1/48th, then smoothing later when you get to the main REW window since you can change the smoothing afterwards.

The Max Overlap doesn't need to be that high, could be as low as 50%, but it's a tiny detail that's not too important.

For matching a target, try to match all the way to at least -20dB down below the average spl level, and down to -30dB if you can. This will help you get the phase to be more textbook-like so that your speakers will sum together better.

Using a smaller number of eq filters is both challenging and rewarding. I find the highpass and lowpass filters can be very powerful if you're willing to set them unconventionally and they can reduce the number of eq filters you need later. Similarly, shelf filters are also very powerful coarse adjustments. Learning how to be efficient with the number of filters you use is a great skill, but can be very frustrating initially. Try to look for areas where two (or more) filters overlap eachother and see if you can get a similar effect by removing one.

Jdunk54nl has a lot of good ideas too, please try those.

If you want to make the center image more focused, you would benefit from more filters. You'll want to play narrow-band pink noise tracks (like 1/3rd octave pink noise) and then listen to where the "center" is for that frequency range. Do this for many different frequencies and you'll find trends like 300-600hz pulls left, 600-1300hz is mostly centered, 1300-3000hz is fuzzy, etc. Then you can make small adjustments to the left and right channels to move the noises towards the center that you want.

You can make coarse adjustments for frequencies above ~500hz by changing the output trim of the left and right channels. To move the image to the right, drop all of the left channels by -1dB and increase all of the right channels by +1dB. Lower and increase them by the same amount so the overall loudness stays similar.

You can make adjustments below ~500hz by very subtly changing the time delays. Add or subtract 0.2ms from the left channels and you'll notice the lower frequencies will shift left or right. It doesn't take much. This is easy to notice with a solo deep male vocal like the intro tracks to the Chesky Ultimate Demonstration Disc.

Keep it up! You're doing great.
 
Very interesting points, Justin. I appreciate it.

You can make coarse adjustments for frequencies above ~500hz by changing the output trim of the left and right channels. To move the image to the right, drop all of the left channels by -1dB and increase all of the right channels by +1dB. Lower and increase them by the same amount so the overall loudness stays similar.
Click to expand...
I would always just adjust down the side that was pulling the stage and not also increase the other. I feel over time, my total output keeps going down and I had to bump it overall. This is probably why.

You can make adjustments below ~500hz by very subtly changing the time delays. Add or subtract 0.2ms from the left channels and you'll notice the lower frequencies will shift left or right. It doesn't take much. This is easy to notice with a solo deep male vocal like the intro tracks to the Chesky Ultimate Demonstration Disc.
Click to expand...
I never considered this. Do you feel this is a hard rule, or can you still use EQ to adjust it? Like I notice sometimes adjusting the EQ makes little to no difference until it is a massive change, say 9dB. So in this case, changing time delay may be what helps. But what if 1dB change does shift it? Should you still use time delay instead?

What about when your mids play down to 300? Should you time delay for it's 300-500 response or just EQ?
 
In the EQ window, set the scale to 5db increments, you are still at 10 db increments. You have 6-12db swings in some of your frequency response, you need to try to limit that big of a swing.

I am also going to suggest upping your volume some when tuning. It looks like on the drivers side and when both are playing together, your noise floor is around 30-35db. You can't follow Justin's advice above if your volume is only about 45-50db. Noise floor should stay at the 30-35db, so you need to eq around 70db to be well above that noise floor.
 
Last edited:
jrwalte said:
I never considered this. Do you feel this is a hard rule, or can you still use EQ to adjust it? Like I notice sometimes adjusting the EQ makes little to no difference until it is a massive change, say 9dB. So in this case, changing time delay may be what helps. But what if 1dB change does shift it? Should you still use time delay instead?

What about when your mids play down to 300? Should you time delay for it's 300-500 response or just EQ?
Click to expand...

That depends on what frequencies are pulling your soundstage. Is it below 500hz or above? How good is your EQ matching side to side? Or, is it an all pass filter needed and not eq or time alignment?
 
Thanks guys, I will take all that into memory and improve on round 2! There was a neighbor mowing while I was trying to tune and I had the truck running for part of the time. That must be what is affecting the noise floor. More volume next time then among the other great tips!
 
jrwalte said:
I never considered this. Do you feel this is a hard rule, or can you still use EQ to adjust it? Like I notice sometimes adjusting the EQ makes little to no difference until it is a massive change, say 9dB. So in this case, changing time delay may be what helps. But what if 1dB change does shift it? Should you still use time delay instead?

What about when your mids play down to 300? Should you time delay for it's 300-500 response or just EQ?
Click to expand...

This is not a hard rule since the effects below does not start and stop at an exact frequency, it's a smooth transition somewhere around 300hz to 1khz (ish).

Lower frequency sounds have longer wavelengths and our heads are not a significant obstacle to the sound so the wave wrap around our head and keep going. To locate lower frequency sounds, our ears rely on the relative time of arrival between the left and right ears. The ear nearest the sound will hear the sound slightly before the other ear and so small amounts of time delay work great to shift the apparent source of a sound to the left and right. This is sometimes call interaural time difference (ITD).

Our ears locate the higher frequencies based on the acoustic shadowing of our head because the size of our head becomes significant at the higher frequencies (shorter wavelengths). The ear further from the sound source receives less sound because it's partially blocked by your head so it's quieter, which means you can shift the apparent position of your sound stage by making one side louder than the other. This is often called interaural intensity difference (IID).

So if your midrange driver plays down to 300hz then the left/right location should be affected by both time delay in the lower frequencies and also loudness in the higher frequencies so you'd want to make both EQ and time delay adjustments on those channels. You midbass drivers that play up to 300hz would benefit from slight time delay changes but you'll likely not hear a difference by making one quiet and one louder unless the change in volume is substantial, like you said 9dB or so. Similarly, your tweeters would likely not shift the sound stage left or right from time delay very much.

However, time delay and eq changes should be the same for any particular quadrant of the car otherwise you risk upsetting the balance/summing between the speakers in that quadrant. For example time delay on your front-left tweeter won't help the stage move left or right, but it can help the tweeter blend with the front-left midrange driver in that same corner. So if you change the midrange time delay to shift the stage left/right, you might want to shift the tweeter by the same amount even though it won't help shift the image left/right.

One last trick: sometimes I will adjust time delay by setting a very steep (temporary) lowpass on my front door speakers at like 200-300hz so that I can focus on the lower frequencies only. Then I'll do the opposite and replace that temporary lowpass filter with a temporary highpass filter on the same door speakers so I can focus on the higher frequencies and get the relative left and right loudness/levels correct. Hopefully I'm aligning the "center" for the lower frequencies and the higher frequencies to the same point in the vehicle, then when I take the temporary hp/lp filters off, the center image should be roughly correct although probbably not focused yet.

Then you move to the 1/3rd octave pink noise tracks to focus the image to a pinpoint location at each frequency band and it's like freaking magic. This is the step that when done well, can really separate a good system from an astonishing system.
 
Im thinking of getting a helix ultra. Eventually I want to be able to tune everything flat within 1-2 db. While that may be possible with 10 filters, it will be a lot easier with the helix and I can use my fiio high res player. Plus I might want to add some center speakers someday. Just not sure if its worth the $ for me and not trying to cop out on learning to tune. I will keep going with the dayton as this is a great way to learn. Kind of like learning to code efficiently when you are limited on memory and processor.
 
You have a 3 way right? You shouldn’t need really more than 10 bands plus crossovers for each driver. I think I have 3-5 eq bands for each driver.

I’ve found countless times, over tuning a driver and getting it ruler flat +- 1 dB doesn’t make it sound better. Usually this means you tried to fix things that shouldn’t be fixed.

Your notion is also right. Spending more money isn’t going to fix knowledge. If you can’t maximize the tools you currently have, find THEIR limitations (not your limitations), then you don’t need to upgrade anything.
 
You must log in or register to reply here.
Back
Top