AVR Frequency Response Graphs

**brucek** · **05-14-07, 05:51 PM**

Quote:

Am I missing something?

When you connect a cable or a cable with a device in the loop, it's no different than taking a measurement with a speaker and microphone as far as setting up REW. It doesn't know the difference and expects a proper setup to get the best response and lowest noise floor etc.

You need to do a Check Levels routine, and then a Calibrate SPL routine to 75dB and then Set Target Level routine and then a Measure routine. The only difference is that when you do the Check level routine you may need to adjust the receiver volume along with the REW input level to get the right level.

brucek

**Otto** · **05-14-07, 06:08 PM**

I don't recall specifically doing any set level stuff, perhaps it was close enough. Loopback measured flat, headroom existed. Set preamp to unity gain and let it go...

**fireanimal** · **05-16-07, 10:40 PM**

Here are the results from my new Sherbourne PT-7010A

AT 60HZ CROSSOVER

AT 80HZ CROSSOVER

8-CHANNEL DIRECT MODE

The purple line is the Sub In/Out
The red line is the Front Left In/Out
The green line is the same input as previous graphs for comparison.

I really like the Sub extension in the 8-Channel mode. It seems to be boosting the Sub channel in all modes.

**Otto** · **05-17-07, 07:46 AM**

Hi fireanimal,

Some of those signals look kinda lumpy. Do you get that same phenomenon when using just a loopback cable?

**MACCA350** · **05-17-07, 08:10 AM**

Quote:

Otto wrote:

Hi fireanimal,

Some of those signals look kinda lumpy. Do you get that same phenomenon when using just a loopback cable?

This is a good example of where a loopback trace in the graph is a good idea, because we'd already know the answer.

cheers

**fireanimal** · **05-17-07, 08:15 AM**

I did, sorry I should of included the feedback line on the graph as well. I don't know where that signal is coming from, could be my hundred year old laptop.

**MACCA350** · **05-17-07, 08:52 AM**

Quote:

fireanimal wrote:

I did, sorry I should of included the feedback line on the graph as well. I don't know where that signal is coming from, could be my hundred year old laptop.

did you run create a soundcard calibration file with the feedback loop in place?

cheers

**fireanimal** · **05-17-07, 08:32 PM**

I had some input/output volume problems causing distortion in the signal path. Here are some new graphs. The red line in all the graphs is the reference line from the feedback loop.

Crossed at 60HZ

8-Channel Direct

Here is the graph out to 22,000. Notice how the Channel Input in Small and Large both taper at the end, as well as the direct sub input. Do you think the Channel Input tapering would cause many problems.

I haven't hooked the preamp up to listen to it yet, just picked it up used, for less than the Outlaw 990 I was looking at.
Does it seem strange that the sub output is hotter than the rest. Also notice on the sub the peaks/valleys until it hits 10hz...

Do you guys notice anything else strange, or does everything look pretty good.

Thanks.

**Hakka** · **05-17-07, 09:05 PM**

I am a moron!

I thought the tone controls on my 3803 were global but it seems they can be set independently for each surround mode. Stereo mode is now flat.

Hakka.

**brucek** · **05-17-07, 10:12 PM**

Quote:

Does it seem strange that the sub output is hotter than the rest.

Did you reset the levels with each measure? You should......

**Otto** · **05-17-07, 11:10 PM**

Quote:

brucek wrote:

Did you reset the levels with each measure? You should......

Back to this...

I'll give you this much -- your levels should indeed be reasonable. They shouldn't be so low that your measurements are in the noise, and they shouldn't be so hot as to cause clipping. But I don't really see how resetting levels between measurements does anything but add confusion to the plots by adding another variable between measurements. I think it's interesting to see the actual change to the signal that's imparted by the processor.

If I have a loopback signal that measures flat from 10 Hz to 20 kHz, and I send that through a processor that has a presumed unity gain, and everything is "ideal" and "perfect", I expect to see that same exact signal at the output of the processor. Now, if I change the input used on this "ideal" preamp to a different input (say "aux" instead of "CD"), and I remeasure, I expect to again see the same measurement as 1) the loopback cable and 2) the input I'd used previously.

Of course, no preamp is "ideal" or "perfect", so any difference between the loopback measurement and the measured output of the processor has, by definition, been caused by the processor.

I find it very relevant to compare whatever inputs and outputs of the processor with precisely the same input signal. We can then see differences between processing modes and such. Yes, I realize that some of this isn't important, and we all normalize our signals with respect to each other.

Quote:

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.

I would submit that if you set up your levels with the sub out, with band-limited pink noise, your levels will be set to X, whatever that may be. If you do a loopback measurement from 10 Hz to 20 kHz, you should measure a flat line (assuming your soundcard cal is in place, etc.).

Furthermore, if you set your levels using full-band pink noise from 20 to 20 kHz, your levels will be set to Y, whatever that may be. Again, measure the loopback, and you should get a flat line. You absolute measurement between the two will be different, but that doesn't matter (again, unless your levels are noisy or clipping).

Now, if I send either of those two signals through the preamp (and there's no clipping), the output measured is the response of the preamp. I cannot see why the amplitude of either the bass-managed sub signal will be greater than that of the bass-managed mains signal.

Quote:

brucek wrote:

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.

How will it look if you do a loopback test? Flat?

Now, before you pop off a terse one-liner to all of this, re-read my post, and try to understand what I'm saying. I know you gave up on this discussion with me before, but it's back, and I think it's adding confusion to this thread (for me, if no one else). If I'm really making a mistake in my measurements, I'll be happy to hear about it. But, so far, the thing that makes the most sense to me is to send a constant signal through the system, and measure the output. If I know what I have going in (the loopback test from 10 to 20 kHz), and I can see what I measure coming out the other end, then the difference was created by the processor, not the level set up.

Anyone else have any opinion on this?

Also, I offered this response as a reasonable explanation to some of the difference in the sub measurements (the sub being 6 dB hot). Is this at all plausible in your opinion?

Quote:

Otto wrote:

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.

**MACCA350** · **05-18-07, 03:09 AM**

Quote:

fireanimal wrote:

I had some input/output volume problems causing distortion in the signal path.

What was the problem, too high or too low? Did you set the Sherbourne's main volume(and all trims) to 0db for any of tests?

Quote:

Otto wrote:

How will it look if you do a loopback test? Flat?

Now, before you pop off a terse one-liner to all of this, re-read my post, and try to understand what I'm saying. I know you gave up on this discussion with me before, but it's back, and I think it's adding confusion to this thread (for me, if no one else). If I'm really making a mistake in my measurements, I'll be happy to hear about it. But, so far, the thing that makes the most sense to me is to send a constant signal through the system, and measure the output. If I know what I have going in (the loopback test from 10 to 20 kHz), and I can see what I measure coming out the other end, then the difference was created by the processor, not the level set up.

Anyone else have any opinion on this?

Yes, I agree. checking levels between traces will invalidate any differences in processor output levels, again this is only of importance if we decide not to normalize all the traces

Quote:

Also, I offered this response as a reasonable explanation to some of the difference in the sub measurements (the sub being 6 dB hot). Is this at all plausible in your opinion?

"Otto wrote:
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. "Hakka", I think you have a similar thing going on in this post."

It seems possible, I didn't notice that in the hookup graph, I used just one channel all the way through the chain, I'll change the OP hookup graph to show this. I'll have a play later and see if the different hookup makes a similar difference.

cheers

**MakeFlat** · **05-18-07, 10:18 AM**

Pink noise contains an equal amount of energy in each octave band. Therefore, using pink noise to calibrate level assures that the calibrated level is weighted equally for the entire range of the test. Thus if you use sub pink noise to calibrate, it will mean that the sub's entire frequency range is taken into account. As you can see, it is significant in that the calibrated level is based on peaks and valleys and anything in between. In contrast, if you were to calibrate based on an x hz tone, the resulting level could be either too high or too low overall. We can now see the advantage of using pink noise to test.

Now if we were to use the sub cal pink noise to calibrate the level of full range measurement, it would mean that the mid range and above were not given any weight in the determination of level. If the signals over the entire frequency range stay mostly constant, except at the extreme ends, then whatever level we use to calibrate would be good enough, as long as we use the same calibration level throughout the entire set of tests.

The significance of the test level to which we calibrate REW, is that the transmitting and receiving ends are set to optimum levels for signal detection and measurement. It assures that the test system operates at a level that is within the normal operating bounds of the devices being tested.

**Otto** · **05-18-07, 10:26 AM**

Quote:

Makeflat wrote:

If the signals over the entire frequency range stay mostly constant, except at the extreme ends, then whatever level we use to calibrate would be good enough, as long as we use the same calibration level throughout the entire set of tests.

Exactly. I totally agree that there's value in setting up your levels correctly -- you don't want your measurements in the noise or clipping. But once you've got that across whatever bandwidth of interest, you don't need to recal it between measurements.

**fireanimal** · **05-18-07, 10:33 AM**

Quote:
fireanimal wrote:
I had some input/output volume problems causing distortion in the signal path.

What was the problem, too high or too low? Did you set the Sherbourne's main volume(and all trims) to 0db for any of tests?

The Windows mixer output level was very low, I reset it and used the volume controls on the USB SB Live to set the levels.

All measurments were done with no alteration to the levels. All the trims on the sherbourne were at 0, but I had to set the volume at -15.0, to be at an even level with the -20.0 signal that I was using through REW. If I set the volume at 0, then the input signal would clip. I even did a master reset on the Sherbourne, and everthing remained the same as before.

**brucek** · **05-18-07, 11:56 AM**

Quote:

Now, before you pop off a terse one-liner to all of this, re-read my post, and try to understand what I'm saying.

Yeah, I'm sorry if a lot of my answers are short, it's just that I don't have a lot of time.

It's hard to add to MakeFlat's excellent response. REW has only two levels of level checking pink noise that it uses. They are either a (low and high cut of 30Hz to 80Hz) or (low and high cut of 500Hz and 2000Hz) respectively . The assumption being that you are checking either the mains or the sub channel and would use the appropriate band limited signal that matched.

As a double check that you aren't mis-using these Check Level pink noise limits, the Measurement routines Check Level button introduces pink noise using a fixed low cut of 20Hz with a variable high cut of the End Frequency chosen in the Measurement panel itself.

This final double check can catch a lot of mis-set levels and errors with reference to the actual measured bandwidth and so I will acquiesce my point of setting the levels irregardless of the UUT bandwidth. Since we have agreed to normalize the end result, who am I to argue......

brucek

Advertisement