US-144 Calibration Question...

[DrWho]

Hey etc....you're from the Klipsch forum right?

I think you might be way overthinking the issue....the US-144 has analog volume controls for the outputs and inputs. Just turn down the lineout. The problem with an L-Pad is that it changes your effective output/input impedance in terms of where the LF corner ends up.

Also, when driving a balanced input with an unbalanced output, you will lose 6dB since you're tying the balance leg to ground with the adapater. I should have the opportunity to compare the line input to the XLR input, but I'd wager that the difference is trivial. It should be interesting to compare notes though.

The last set of measurements I did, I just used a loopback from the left output into the left unbalanced input. The ECM8000 was hooked up to the right channel xlr input. My measurements were very comparable to my M-Audio box that I had the opportunity to compare against the Klipsch anechoic chamber setup...that was within half a dB from 100Hz to 10kHz. Within a dB from 20Hz to 100Hz, and within 2dB above 10kHz.

The only source of frequency non-linearities in these devices is the digital anti-aliasing filters in the DACs/ADCs and then the series dc blocking caps. I suppose it's due process to ensure the XLR inputs have the same LF corner frequency as the line inputs, but I'd be surprised if it was very different. It would be pointless for them to not be using the same front-end for both the balanced and unbalanced inputs.

Ok, so I just looked at the spec sheet...unbalanced input impedance is 10k. Balanced input impedance is 2.4k. If the unbalanced input has a corner frequency of 3Hz, then the balanced input will have a corner at 13Hz. I just measured the corner of the line input at 3Hz... I'll bust out an adaptor tomorrow evening and actually measure the LF corner of the balanced input. Anyone wanna take some bets?

[brucek]

Wayne A. Pflughaupt wrote:

In addition to the main output attenuator, you have post-preamp attenuators on the individual channel strips. Both are fully capable of reducing the incoming signal to zero

Look at the circuit again. The line-out of the soundcard will connect directly to the mic-in of the mixer, and the line-out of the mixer connects directly to the line-in of the soundcard. Those mains pots on the mixer are after the mic-in preamp and have no capability to dial down its input. The level into the mic mixer is set by the line-out sweep level adjusted in REW. You want to keep this high, and not dial it down to accommodate the next stage as a mic preamp or the level out of the soundcard would be too low. The S/N would suffer. So, the mic-in pot would stand by itself and would likely have to be dialed to almost zero to work (and would likely be twitchy). You really require a pad to dial that level down.

DrWho wrote:

The problem with an L-Pad is that it changes your effective output/input impedance in terms of where the LF corner ends up.

Well, that's why I chose the values I did, to maintain the bridge impedance while supplying the required drop.

DrWho wrote:

Also, when driving a balanced input with an unbalanced output, you will lose 6dB since you're tying the balance leg to ground with the adapater

But this is a non issue since REW normalizes the soundcard cal file to zero.

DrWho wrote:

I should have the opportunity to compare the line input to the XLR input, but I'd wager that the difference is trivial.

Yeah, I agree, especially with this type of device. I think Waynes concern was that a cheap mic preamp may offer poor low frequency response.

brucek

[brucek]

To prove that I eat my own cooking, I decided to test the pad I recommended on my own XENYX802 mixer.

I used whatever resistors I had in my drawer, and found a 100ohm and a 3.3Kohm - perfect, that calculates to -30dB pad with a decent output impedance (~100ohms) and an input impedance (~3.3K) that comes close to the mixers mic input stage (2.6K).

The XENYX mixer has a dual purpose pot that operates the input level for both the line-in stage and the mic-in stage, so I decided to leave the pot in its central detent position (the same as I always use).

Anyway, I did three measures. First the soundcard cal routine for a straight cable, then second for the mixer included into the line stage, then third, the mixer included for the mic stage. Before I ran the mic in test, I added a divider pad as described above and fed it to an XLR with pin 3 and 1 shorted to remove the negative differential amp and unbalance the input. I have to assume the response of the positive diff amp is identical to the negative amp.

It was interesting when I went to the mic-in test after the line-in test, I didn't touch the mixer controls to see how close the pad came to being right on the money, and it was very close. I only tweaked the mixer gain pot up a bit to bring the input level VU in REW to what it was before.

So, you can see the results below as expected. The mic stage response is down about -0.25dB at 10Hz. You'll note as expected that the mic response line (green) is noisier than the line-in response line (gold) or the soundcard only response line (purple). This could be cleared up if I didn't do such a quick and dirty job. My pad was simply clipped onto two RCA connectors and this is a no-no with a high gain stage such as the mic amp. You should put the pad in a metal box with short leads if you want a real clean line... (I don't really care myself - it's fine the way it is).

I show two scalings of the plot.

The first is with the standard +/-30dB we use for normal plots, so you can see the relative effect of the response curves.

The second is the expanded scale to get a close look..








brucek

[Wayne A. Pflughaupt]

Look at the circuit again. The line-out of the soundcard will connect directly to the mic-in of the mixer, and the line-out of the mixer connects directly to the line-in of the soundcard.

Perhaps I was less than clear, the example I showed before is a USB interface mixer, not like the Xenyx. And unlike the Tascam US-144, it has balanced outputs. So the mic input would be fed from the mixer’s balanced output, not from an outboard sound card.

So, you have a fully balanced signal loop, which should eliminate the mismatched impedance issue John mentioned. And between the variable input pad and the two mains pots, that should address the level issue – right?

Not sure if this would address the noise issue you showed in your graph - you'd be a better judge of that than I. As John noted, at least running a mic pre-amp calibration can show us if its response is reasonably linear, even if we can't get a usable calibration file...

Regards,
Wayne

[brucek]

So the mic input would be fed from the mixer’s balanced output, not from an outboard sound card.

huh? we want to include the mixers response in the soundcard cal file. So instead of feeding the soundcard line-out to soundcard line-in (as per usual), we feed the soundcard line-out directly to the mic-in (not from the mixers line-out)...... the mixers line-out feeds the soundcard line-in.

brucek

[Wayne A. Pflughaupt]

This mixer has a USB output for the computer. It is the soundcard! IOW, it does what your Xenyx/soundcard set-up does, all in one unit instead of two...

Regards,
Wayne

[brucek]

This mixer has a USB output for the computer. It is the soundcard!

Oh OK, I see... huh, it looks just like the XENYX so I assumed it was the same....

[Wayne A. Pflughaupt]

Soo... can an acceptable soundcard calibration be had between the balanced output and the mic input?

Regards,
Wayne

[etc6849]

Yep, that's me. Good to hear from you again. I think you were the one over there that told me about REW and I followed the links in your signature :-)

I am wondering about your statement of shifting the corner frequency. I think you have a good point.

Gain wise, you are just shifting the curve downward right by adding a padding circuit? This should not shift the corner frequency to the left or right since ideally I am adding only resistance.

My understanding is one should look at the gain from the output of the sound card, the gain from the voltage divider (padding circuit plus some cabling) and the gain from the soundcards pre amp. If all are in dB, you can add. If they are in voltage (vout/vin) you would multiply them. Since, with the voltage divider, I'm adding what should ideally be a constant impedance over all frequencies (no reactance, only resistance) I shouldn't see a corner shift left or right; only up or down. Now, if I open up the soundcard and modify the resistors connected to the op amps inside, yes this would shift the corner frequency. I wish I had a an adapter to connect this special padding circuit from the output to the input to verify this though cause you have an interesting point.

Bruce: Thanks for the results. I soldered the padding circuit up last night. I didn't use a metal box to shield the dividing circuit though, but just soldered and used shrink wrap to prevent ground to signal contact. I might use some foil and solder this to the ground later on. I don't know how that would work though as I've never tried soldering foil; I don't think it's worth finding a metal box for yet. My soundcard and mic are arriving today so I will post my results soon.

DrWho wrote:

Hey etc....you're from the Klipsch forum right?

I think you might be way overthinking the issue....the US-144 has analog volume controls for the outputs and inputs. Just turn down the lineout. The problem with an L-Pad is that it changes your effective output/input impedance in terms of where the LF corner ends up.

Also, when driving a balanced input with an unbalanced output, you will lose 6dB since you're tying the balance leg to ground with the adapater. I should have the opportunity to compare the line input to the XLR input, but I'd wager that the difference is trivial. It should be interesting to compare notes though.

The last set of measurements I did, I just used a loopback from the left output into the left unbalanced input. The ECM8000 was hooked up to the right channel xlr input. My measurements were very comparable to my M-Audio box that I had the opportunity to compare against the Klipsch anechoic chamber setup...that was within half a dB from 100Hz to 10kHz. Within a dB from 20Hz to 100Hz, and within 2dB above 10kHz.

The only source of frequency non-linearities in these devices is the digital anti-aliasing filters in the DACs/ADCs and then the series dc blocking caps. I suppose it's due process to ensure the XLR inputs have the same LF corner frequency as the line inputs, but I'd be surprised if it was very different. It would be pointless for them to not be using the same front-end for both the balanced and unbalanced inputs.

Ok, so I just looked at the spec sheet...unbalanced input impedance is 10k. Balanced input impedance is 2.4k. If the unbalanced input has a corner frequency of 3Hz, then the balanced input will have a corner at 13Hz. I just measured the corner of the line input at 3Hz... I'll bust out an adaptor tomorrow evening and actually measure the LF corner of the balanced input. Anyone wanna take some bets?

[brucek]

Soo... can an acceptable soundcard calibration be had between the balanced output and the mic input?

Of course. You would use a partial H-Pad. The partial would be the best in this case. Simply take the values you would use for the simple divider (such as 3.3K and 100ohms for ~ -30dB) and divide the series resistor in half (i.e. 1650 ohms. Standard value of 1800 would be fine or go down to 1600).

The two 1800 ohm resistors would attach from the balanced line-out XLR pin 2 and 3 (or equivalent 1/4" phone), and then the other resistor of 100 ohms is soldered across the 1800 ohm resistor ends. Then feed the new balanced output from the 100 ohm resistor to the pin 2 and 3 of the mic-in XLR.

I could draw a picture, but I'm sure you get it......

This should not shift the corner frequency to the left or right since ideally I am adding only resistance.

It won't, you'll be fine.

I didn't use a metal box to shield the dividing circuit though

Not a problem, just keep the leads short and cover it with some foil or whatever. It'll be fine. I had 3 foot clip leads and it was sitting on my computer and mine worked fine....

brucek

[etc6849]

I've attached the measured file for one of my speakers. I had no idea how bad it would look. Is there something wrong with my setup? I plotted this after calibrating the ECM8000, adding loopback cable on the left output/input. I had the mic input on the card set on max, line out on max. I let REW do 4 iterations with 512k length.

I guess I really didn't realize there were 40 dB swings in my speaker's output. Is this normal or am I missing something? The green is without audyssey multiEQ, the blue is with Audyssey. The speakers are Klipsch RF-83's.

PS: Is there a best way to point the mic? Right now I'm pointing it straight at the speaker near the tweeters height, horizontal with the floor.

[etc6849]

Here's what I get for the soundcard with the padded loopback.

[cinema mad]

the ECM 8000 mic should point up towards the ceiling at ear hight when seated, if you are checking the room response, it is designed to graze like the Audyssey mic....

Cheers..

[JohnM]

The wild swings at higher frequencies are normal and caused by comb filtering, reflections from room surfaces partially cancelling the direct signal from the speaker at frequencies where the difference in arrival times corresponds to 180 degrees of phase shift. To see the underlying response apply some smoothing to the trace.

You will generally get perfectly good results with a single 256k sweep, multiple longer sweeps are useful in environments with higher background noise or if you need a very high signal-to-noise ratio for some post processing of the impulse response.

[brucek]

Is there something wrong with my setup?

No, it's your plot setup that needs to be changed.
For subwoofers, always use the standard Vertical graph axis of (45dB - 105dB) and the Horizontal graph axis of (15Hz - 200Hz) using the Graph Limits button in the top right corner of REW.
For full range, use the standard Vertical graph axis of (45dB - 105dB) and the Horizontal graph axis of (15Hz - upper limits you desire, i.e. 20KHZ - certainly no higher than your soundcard can extend).
For full range, enable smoothing to eliminate the comb filtering. Use a 1/3 octave smoothing.

brucek

[DrWho]

etc6849 wrote:

Is this normal or am I missing something?

This is absolutely normal.

I've gotta run, but you should read up on the Haas effect, Heyser, and Time Domain Spectrometry. Instead of smoothing your response, you should play around with the window on the impulse response.

What you are seeing in the frequency domain are the effects of the comb-filtering from reflections throughout your room. The way human perception works, we are able to filter out reflections and perceive the direct sound separately. There are of course limits to this which Heyser discusses to great extent. The reason you don't want to smooth the response is because the process doesn't differentiate between direct and reflected sounds. Sure, it'll give you a pretty line, but it will provide an inaccurate representation for what is actually being perceived. If you can control the influence of the reflections in the measurement, then you can get a better idea as to what your ears are actually hearing. The trade off, however, is a reduction in frequency resolution, so you can really only filter for the mid- and high-frequencies. This works out OK though because the Haas window is longer at lower frequencies.

I hope that helps at least a little bit - I could go on for hours on the topic, but I think it'd be best to study the original sources (as they are way smarter and usually offer better insight).

Btw, your soundcard calibration looks very similar to mine. I hope you're satisfied with the recommendation.

[facesnorth]

I too recently bought a US-144 (and returned a TB-SRM, xenyx 802 & edirol usb-midi), mainly for these reasons:

1) it works with vista 64-bit (main reason I returned the TB-SRM)
2) I was taken by the all-in-one box which to me seemed more effective (and slightly cheaper) than all the cables, adapters, and seperate units I would have used otherwise.
3) thus it seemed with less equipment I would be getting higher quality, and a more elegant and portable solution.

I haven't had time to set it up or do anything yet. However, after reading this I'm worried it's going to be harder and less effective than I originally thought. I couldn't quite follow some of the posts about what actually was required to calibrate this thing. What needed to be soldered? I wonder if this is something I can do. I don't understand all the talk about padding and what to do with this thing once you make it. It sounds very unelegant which is exactly what I liked about this unit. Plus it sounds imprecise... Is there a way to be confident that I can get this thing properly calibrated and be able to trust my measurement graphs?

Would the EMU 0404 USB 2.0 have been a better choice? I can still return the US-144 if I need to. I'm also thinking of future uses down the road for this thing, and while I don't want to get carried away, I do have some interest in hobbyist sound recording as well.

I also have the ECM8000 mic and a RS analog SPL meter, a BFD, and all the necessary cables. I'm itching to get started already. One last question - I'm also using the Onkyo 905, which does the Audyssey MultiEQ. Should I do all my REW stuff before running MultiEQ and run MultiEQ after? Or should I run MultiEQ and then REW? Or should I not use MultiEQ at all?

[brucek]

I couldn't quite follow some of the posts about what actually was required to calibrate this thing. What needed to be soldered?.......................Is there a way to be confident that I can get this thing properly calibrated and be able to trust my measurement graphs?

The simplified lowdown on this issue is that there was concern that the response of the mic preamp integral to the soundcard, might affect measurements in REW. My feeling is that using the line-out to line-in when performing a soundcard cal test was suffice, and that the mic preamp would have no pronounced effect. I tested my XENXY to get a feeling how much of an effect it would have, and found it to be minuscule, with worst case being -0.25dB at 10Hz. This can be safely ignored.

We didn't get a reading from etc6849 on the differences he found in the US-144 between the line-in vs the mic-in, so I can't advice you on whether you need to include the mic preamp or not in the calibration. I suggest you do not.

The temporary mod requires constructing an attenuator out of wires and resistors to allow the inclusion of the mic amp into the calibrate test. I don't know if etc6849 used balanced or unbalanced. He may be able to tell you about that...

If you ignore the mic preamp in the test, you simply loop line-in to line-out temporarily to run the soundcard cal routine...

brucek

[etc6849]

Thanks everyone for all your help. I've learned a lot from everyone :-) It appears I have a lot of studying ahead to really understand all of the different plots in REW and how to properly use/interpret them. I've been busy lately with work, but I'll post a comparison plot of the US-144's calibration tonight or Saturday.

One thing about the US-144 is I am having problems using it on a dell laptop. Windows XP keeps telling me my usb device is drawing too much current (only when I plug it in), but the device works fine on my vista HTPC. The US-144 will work with my laptop, but sometimes I have to tell XP to reset the current protection on the usb port first.

I'm thinking about returning it if I can't use it reliably with my laptop and getting a xenyx 502 and using it with my HTPC. Also, there is some noise at high volume levels when I use it with my laptop (possibly from the fan inside the laptop?). This all could be just my laptop going bad as it's 4-5 years old, but I don't really know.

[facesnorth]

brucek wrote:

The simplified lowdown on this issue is that there was concern that the response of the mic preamp integral to the soundcard, might affect measurements in REW. My feeling is that using the line-out to line-in when performing a soundcard cal test was suffice, and that the mic preamp would have no pronounced effect. I tested my XENXY to get a feeling how much of an effect it would have, and found it to be minuscule, with worst case being -0.25dB at 10Hz. This can be safely ignored.

Thanks, Bruce.

[brucek]

getting a xenyx 502

The 502 has no phantom power. You need the 802......

[facesnorth]

etc6849 wrote:

I'll post a comparison plot of the US-144's calibration tonight or Saturday.

Great I'd love to see this.

[quote]One thing about the US-144 is I am having problems using it on a dell laptop. Windows XP keeps telling me my usb device is drawing too much current (only when I plug it in), but the device works fine on my vista HTPC. The US-144 will work with my laptop, but sometimes I have to tell XP to reset the current protection on the usb port first. [quote]

I was worried about this unit not have a power brick. I wonder if plugging the laptop in would help at all.

I'm thinking about returning it if I can't use it reliably with my laptop and getting a xenyx 502 and using it with my HTPC. Also, there is some noise at high volume levels when I use it with my laptop (possibly from the fan inside the laptop?). This all could be just my laptop going bad as it's 4-5 years old, but I don't really know.

From the sound of it none of these USB 2.0 devices are without flaws. I'm not sure I made the right decision either, but I'm not sure what is. If you find a device in this price range that works great with vista and is a good all-in-one let me know.

[etc6849]

I definitely do not fully understand the software, but here's what I was able to generate. I used the measure soundcard for calibration option under the settings menu. I had to increase the mic volume slightly for the padding (with resistors) option.

This was generated using a vista HTPC. The green is US-144 with normal loopback mono phono to mono RCA. The light blue is the ASUS HDAV1.3 card with normal feedback config. The dark blue is the US-144 with the mic preamp and padding cable included. I do see some difference, but I don't know how much it matters. The ASUS HDAV has a better response though I think, but it costs twice as much so go figure. I didn't change the default window, but then again I don't know if REW applies a window for soundcard cal measurements.

Hope this helps,

Ellery

[quote=facesnorth;139756]Great I'd love to see this.

[quote]One thing about the US-144 is I am having problems using it on a dell laptop. Windows XP keeps telling me my usb device is drawing too much current (only when I plug it in), but the device works fine on my vista HTPC. The US-144 will work with my laptop, but sometimes I have to tell XP to reset the current protection on the usb port first.

I was worried about this unit not have a power brick. I wonder if plugging the laptop in would help at all.



From the sound of it none of these USB 2.0 devices are without flaws. I'm not sure I made the right decision either, but I'm not sure what is. If you find a device in this price range that works great with vista and is a good all-in-one let me know.

[facesnorth]

etc6849 wrote:

This was generated using a vista HTPC. The green is US-144 with normal loopback mono phono to mono RCA. The light blue is the ASUS HDAV1.3 card with normal feedback config. The dark blue is the US-144 with the mic preamp and padding cable included. I do see some difference, but I don't know how much it matters. The ASUS HDAV has a better response though I think, but it costs twice as much so go figure. I didn't change the default window, but then again I don't know if REW applies a window for soundcard cal measurements.

It's hard to tell what the colors are from the graph. Is the one up top that is closest to flat the ASUS?

[etc6849]

yes. it's kind of a blue green color. i guess i should've used different colors huh?