AVR Frequency Response Graphs

[MACCA350]

Hakka, did you set your global channel trim levels, or just the mode you were in at the time? maybe the particular mode you were in for testing had different trim levels for the sub, because that is a pretty big boost on the sub channel. It seems that you may have a tone control active in the top end for stereo mode, since the direct mode trace was flat.

cheers

[MACCA350]

Here's a couple of things I've been thinking about, I know, I think too much

I'm thinking that we can set the 0db reference level when we loop the soundcard, this way we can see any attenuation from input to output(with the receivers at 0.0db) and have a reference point between receivers. Or maybe set the main fullrange signal at 0db(at the 1kHz point) but also keep a soundcard trace for reference(I like this idea better).

I'm also thinking of changing the graph limits to
Top = 6
bottom =-18
left = 2
right = 20,000(fullrange graph)
right = 200(sub graph)
This will give us a vertical scale with multiples of 3db

Maybe we can set out some colors for particular traces. The first four that REW use are Red, Green, Blue, Purple. We could set them to something like this.

Red = Soundcard reference trace
green = Mains fullrange trace
Blue = Mains with crossover engaged trace
Purple = Sub with crossover engaged trace
And keep all colors and traces for the sub graph so it will basically be a zoom-in of the fullrange graph. Also make sure we use no filtering in the traces.

If we can pick a common mode for the tests to be performed in, or just say whichever one performs the correct bass management for the tests.

What I'm hoping for is identical looking graphs and identical testing methods(atleast for the first two graphs) so we can accurately and easily compare those receivers and pre/amps tested.

Once we've worked out the best and easiest method we can write up some instructions so that anyone can perform the test and post accurate and comparable results. We may need to start a new clean thread for just the results(kind of like a reference thread to post just the comparable graphs and keep this thread for all the discussion and other graphs), once we've nutted all these things out.

Anyway its some food for thought, I'm really interested to see what kinds of variations we get, It just makes it hard to compare the different graphs.

cheers

[Otto]

MACCA350 wrote:

Here's a couple of things I've been thinking about, I know, I think too much

I'm thinking that we can set the 0db reference level when we loop the soundcard, this way we can see any attenuation from input to output(with the receivers at 0.0db) and have a reference point between receivers. Or maybe set the main fullrange signal at 0db(at the 1kHz point) but also keep a soundcard trace for reference(I like this idea better).

Yeah, I was thinking the same thing, namely becuase I screwed up my mesaurements slightly. I did not normalize all my plots from any initial measurement, I normalized each one to a relative point within each measurement. It's not much of a big deal that it's wrong, as they are still relevant because each is relative to itself, but you're right: all these measurements should be shown with respect to the soundcard loopback measurement after the soundcard cal is done (which should be a flat line). We can leave the cable measurement in each plot for reference, or just "know" that the flat line would be a 0 dB if it were in the picture. Perhaps leaving it in is better, then we somewhat know for sure. Everyone just has to be careful to always enter the same "offset" to all measured data.

I'm also thinking of changing the graph limits to
Top = 6
bottom =-18
left = 2
right = 20,000(fullrange graph)
right = 200(sub graph)
This will give us a vertical scale with multiples of 3db

I like the idea of 3 dB increments on the X-axis, but I think that -18 is too shallow to show us all the effects of the slopes. I'm not trying to push what I did just because I did it, but I like the way that my 12 to -30 dB plots came out. They are in 6 dB increments, which, to me, are mostly as usable as 3 dB increments. As long as it's not 4 dB or 8 dB increments, since we are so interested in 3 dB and 6 dB points (as you know).

I think I'm going to move this thread from the BFD forum to the Receiver|Processor forum.

[MACCA350]

Otto wrote:

Yeah, I was thinking the same thing, namely becuase I screwed up my mesaurements slightly. I did not normalize all my plots from any initial measurement, I normalized each one to a relative point within each measurement. It's not much of a big deal that it's wrong, as they are still relevant because each is relative to itself, but you're right: all these measurements should be shown with respect to the soundcard loopback measurement after the soundcard cal is done (which should be a flat line). We can leave the cable measurement in each plot for reference, or just "know" that the flat line would be a 0 dB if it were in the picture. Perhaps leaving it in is better, then we somewhat know for sure. Everyone just has to be careful to always enter the same "offset" to all measured data.

Ok, so we agree to keep a 'soundcard loop' for reference. With this I see two options:

1) Set the 0db line to the 'soundcard loop trace' (in this case we shouldn't need to have a trace for it since the 0db line is the trace, but then again )

2) Set the 0db line to the 'main fullrange trace' and keep a 'soundcard loop trace' for model to model comparison purposes (this will make finding -3/-6/etc points on these traces easier to see(atleast the 'main fullrange trace'). Since we all calibrate our systems, the difference between the 'soundcard loop trace' and 'main fullrange trace' is irrelevant but it will still be interesting to see variations.

What would be handy is if REW was able to set an offset automatically based on a 'zero set' trace, like running a trace and then tell REW to 'set this one to zero' and all the rest will automatically be set with the same offset

I like the idea of 3 dB increments on the X-axis, but I think that -18 is too shallow to show us all the effects of the slopes. I'm not trying to push what I did just because I did it, but I like the way that my 12 to -30 dB plots came out. They are in 6 dB increments, which, to me, are mostly as usable as 3 dB increments. As long as it's not 4 dB or 8 dB increments, since we are so interested in 3 dB and 6 dB points (as you know).

Yeah, I'll agree to that I'd love to keep the -3db line, but I can see that we may need to keep more range in the graph.

Top = 12
bottom = -30
left = 2
right = 20,000(fullrange graph)
right = 200(sub graph)

Any other suggestions?

cheers

[Otto]

MACCA350 wrote:

Ok, so we agree to keep a 'soundcard loop' for reference. With this I see two options:

1) Set the 0db line to the 'soundcard loop trace' (in this case we shouldn't need to have a trace for it since the 0db line is the trace, but then again )

2) Set the 0db line to the 'main fullrange trace' and keep a 'soundcard loop trace' for model to model comparison purposes (this will make finding -3/-6/etc points on these traces easier to see(atleast the 'main fullrange trace'). Since we all calibrate our systems, the difference between the 'soundcard loop trace' and 'main fullrange trace' is irrelevant but it will still be interesting to see variations.

I think option 2 is better for the reasons you cite (being able to easily see the cutoff frequencies).

What would be handy is if REW was able to set an offset automatically based on a 'zero set' trace, like running a trace and then tell REW to 'set this one to zero' and all the rest will automatically be set with the same offset

Good idea.

Any other ideas?

Not at the moment. I should be measuring the AudioControl Maestro M2 sometime within the next week or so. I'll see if anything comes to me as I go through the process again, and I'll stick to the stuff we've agreed to so far.

Thanks.

[brucek]

What would be handy is if REW was able to set an offset automatically based on a 'zero set' trace, like running a trace and then tell REW to 'set this one to zero' and all the rest will automatically be set with the same offset

When I do this type of measurement I first set up the Check Levels routine and then execute the Calibrate SPL meter routine to 75.0 dB and then execute the Set Target Level routine on the looped cable itself.

If it returns (for example) +75.6dB as the target, then I do a measurement with the cable to ensure the flat line at +75.6dB. Then I simply enter -75.6 into the Trace Offset feature and click the Add Offset to Data feature and then Save the measurement. This creates the cable reference and the measurement at 0dB for future reference and comparison.

Then I insert the UUT (unit under test) and do a measurement (with multiple sweeps x8 to obtain the best S/N ratio) and go through the calibrate routine as described above to get the zero reference and then save.

Every time I change an end frequency, cutoff or pink noise reference, etc, I redo the Check Levels and all the other items mentioned above.

brucek

[Hakka]

MACCA350 wrote:

Hakka, did you set your global channel trim levels, or just the mode you were in at the time? maybe the particular mode you were in for testing had different trim levels for the sub, because that is a pretty big boost on the sub channel. It seems that you may have a tone control active in the top end for stereo mode, since the direct mode trace was flat.

cheers

Stereo mode had the sub channel at 0db, pure direct was set at +0.5db, all other trims were the same for all tests. I will check the tone controls but I'm pretty sure they are set flat and tone bypass is engaged.

I have a Y cable coming out of my sub preout into the soundcard but REW should only use one channel so I doubt that has anything to do with the boost.

Hakka.

[Otto]

I got up very early this morning and measured the AudioControl Maestro M2 pre/pro. I borrowed this pre/pro from another Shack member to see if it could work in my system. The Maestro M2 has a good sound, but I found that its feature set is lacking, especially on Room 2 controls, so I won't be able seriously consider it for permanent use in my system.

Anyway, here are some quick measurements. It seems that they are using a 4th order crossover for both mains and sub.

[MACCA350]

Otto, did you work out why there is a 6db boost on the sub channel when compared to the mains?

cheers

[Hakka]

If anyone has a 380x sereis denon I would be interested to see the results of a stereo mode full range measurement.

I double checked all tone controls and they are not engaged.

Hakka.

[brucek]

Otto, did you work out why there is a 6db boost on the sub channel when compared to the mains?

I would think that the check-levels etc has to be reset every time a new measurement is taken to account for the different response to the band limited pink noise used.

Barring that, I see little value in being concerned about the differing levels. Is it not the response that we're really concerned about here? Why not normalize everything to 0dB.

brucek

[MACCA350]

Hakka wrote:

If anyone has a 380x sereis denon I would be interested to see the results of a stereo mode full range measurement.

I double checked all tone controls and they are not engaged.

Hakka.

For the Denon 3805, See post no.2
The red line is the 'stereo mode full range measurement'

cheers

[Otto]

brucek wrote:

I would think that the check-levels etc has to be reset every time a new measurement is taken to account for the different response to the band limited pink noise used.

Really? I'd think that it wouldn't matter much what the target was, or the calibration with respect to an SPL meter. All REW is doing at this point is sending a frquency sweep to a device and reading the results. There's no clipping or other malfunction occurring that I know of.

Barring that, I see little value in being concerned about the differing levels. Is it not the response that we're really concerned about here? Why not normalize everything to 0dB.

Agree.

I think that the 6 dB increase in the sub may be caused by the fact that I'm using both left and right inputs to the preamp. They are summed and filtered to create the sub signal. Since I measured only one of the L/R outputs, we're really missing half of the response of the mains. If both L and R were present, they would combine to be in line with the sub. The sub's output should be considered with respect to both left and right mains. MACCA, I think you have a similar thing going on in this post.

brucek is right in that it doesn't matter much, as we all normalize our signals by using an SPL meter anyway. One thing we are learning here -- there's definitely some liberty taken on the implementation of bass managment from manufacturer to manufacturer. I'm sure some of these differences would be audible. Also, it's obvious that implementations are sometimes just plain wrong. Interesting stuff...

[MACCA350]

brucek wrote:

I would think that the check-levels etc has to be reset every time a new measurement is taken to account for the different response to the band limited pink noise used.

Barring that, I see little value in being concerned about the differing levels. Is it not the response that we're really concerned about here? Why not normalize everything to 0dB.

brucek

My original intention was to see how the receiver modifies the input signal.

It was decided to use the main fullrange signal as the zero point(whereas Otto used the soundcard loop as zero piont BTW did you have a 20kHz graph for the Maestro ).

If you look at post no.2 the only difference between the main fullrange(Red) and main with crossover(green) was setting the mains to small to engage the crossover, so the difference in level is a byproduct of the 3805's processing.

I understand that this is not critical information because in the end we calibrate the system within a room, but it was discussed earlier and we thought it was interesting enough to keep in in the graph, but all this is still up for debate.

It has brought up a question of why is there a difference between receivers level, compare the AudioControl Maestro M2 pre/pro(#33) to the Denon 3805(#2) and the Denon 3803(#25)

But if everyone wants all traces normalized then the only difference we'll see is bass roll-off and crossover operation.

But maybe we can set the 'main's crossover' trace at 0db, to make the identification of crossover order bass roll-off easier(since this is the most common setup mode)

cheers

[brucek]

I'd think that it wouldn't matter much what the target was, or the calibration with respect to an SPL meter. All REW is doing at this point is sending a frquency sweep to a device and reading the results. There's no clipping or other malfunction occurring that I know of.

But if I send a band limited pink noise signal for a subwoofer (with low cut of 30Hz and high cut of 80Hz) and then set up the levels at the bass managed sub out, the level at the sub out and mains out will not be the same. The mains pink noise has a different set of cuts and would require a new level check.

Either way, I would just normalize everything to 0dB so the response is revealed...

brucek

[MACCA350]

Otto wrote:

Really? I'd think that it wouldn't matter much what the target was, or the calibration with respect to an SPL meter. All REW is doing at this point is sending a frquency sweep to a device and reading the results. There's no clipping or other malfunction occurring that I know of.



Agree.

I think that the 6 dB increase in the sub may be caused by the fact that I'm using both left and right inputs to the preamp. They are summed and filtered to create the sub signal. Since I measured only one of the L/R outputs, we're really missing half of the response of the mains. If both L and R were present, they would combine to be in line with the sub. The sub's output should be considered with respect to both left and right mains. MACCA, I think you have a similar thing going on in this post.

That was Hakka's graph, and I think you are right.

Once a well defined method is set out these setup differences will not be there.

brucek is right in that it doesn't matter much, as we all normalize our signals by using an SPL meter anyway. One thing we are learning here -- there's definitely some liberty taken on the implementation of bass managment from manufacturer to manufacturer. I'm sure some of these differences would be audible. Also, it's obvious that implementations are sometimes just plain wrong. Interesting stuff...

It is something to consider, I took to long to post my previous post, and didn't see this one, where I mention this: 'But maybe we can set the 'main's crossover' trace at 0db, to make the identification of crossover order bass roll-off easier(since this is the most common setup mode)'.

I'm not sure which way to go on normalizing, but we'll see what everyone thinks.

cheers

[MACCA350]

brucek wrote:

But if I send a band limited pink noise signal for a subwoofer (with low cut of 30Hz and high cut of 80Hz) and then set up the levels at the bass managed sub out, the level at the sub out and mains out will not be the same. The mains pink noise has a different set of cuts and would require a new level check.

Either way, I would just normalize everything to 0dB so the response is revealed...

brucek

But thats why all levels and trims etc, in the AVR are set to 0db, so we have a level field.

cheers

[Otto]

MACCA350 wrote:

But thats why all levels and trims etc, in the AVR are set to 0db, so we have a level field.

cheers

Correct. I didn't try to set up any levels in the "channel trims" in the pre/pro. All levels were at 0, and I didn't really do anything with the band limited pink noise. Just in and out, through the pre/pro.

I don't think I swept up to 20 KHz. I was having some trouble with my soundcard cal at high frequencies. I'm not sure if there's actually a problem, or if it's just some noise that I've been hearing in the analog outs of my cheap soundcard. I'm considering going to a better sound card.

Sorry if I didn't set up correctly, I didn't go back and read the setup stuff. I like MACCA's idea of seeing what the actual contribution of the system is, though, so I think it's a good idea to set one reference and then adjust all traces by that offset only. Obviously, I think there's also benefit in normalizing to 0 db, as I did just that for my second plot yesterday. We appear to still be going through some set up trials, and that's OK. I think we'll continue through this for a while, and then perhaps MACCA can edit his initial post to summarize all the set up stuff (there's some already there, but I think it doesn't include some of the stuff we talked about after that first post).

That's it for now. Have a good day.

[MakeFlat]

Retracted by me.

[Otto]

I think if I just leave all channel trims at +/- 0 dB and don't adjust anything else, it should make no difference. All I'm doing is just sending a sweep signal in and measuring what comes out the other end of the receiver/pre/pro. There's no BFD in the circuit, and there's no house curve applied.

[MakeFlat]

Otto wrote:

I think if I just leave all channel trims at +/- 0 dB and don't adjust anything else, it should make no difference. All I'm doing is just sending a sweep signal in and measuring what comes out the other end of the receiver/pre/pro. There's no BFD in the circuit, and there's no house curve applied.

Okay - cancel my previous post. Now I see that FR=0dB, FL=0dB and SW=0dB on the trims. It is now looking like the 6dB boost on the sub channel is there by design. So now I can see why normalizing the sub level is the right thing to do.

[brucek]

But thats why all levels and trims etc, in the AVR are set to 0db, so we have a level field.

I don't think I made my point very well.

The sub cal pink noise used in the Check Levels routine is band limited from 30Hz to 80Hz. If I set up my levels out of the sub port with that as the source to 0dB, then I measure from 2Hz to 300Hz for example, the level will drop. I suppose if you used Full Range pink noise to set the levels it would work, but that isn't the case. My feeling is to reset all the measurements to a 0db level with the offset feature and then compare responses.

brucek

[Otto]

I think I understand what you're saying...

I've never used pink noise for any set up when measuring the pre/pros directly. I just set everything to 0 dB and let the signal go straight through the system.

[brucek]

But how do you set the levels. Don't you use the Check Levels routine?

[Otto]

I didn't set the levels per se. I left REW as it was set up previously, ran the loopback, saw reasonable results and there was no clipping. From there, I just sent the signal through the pre/pro. The response of the pre/pro shouldn't change based on the levels I set in REW. Am I missing something?