Help with ARTA Software

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I was having a strange problem using the XMOS USB2.0 Eval board with ARTA. In Spectrum Analyzer mode, it measures superbly, less than 0.0045% THD. But in impulse mode, I can't get it to do a full 20 Hz - 20 kHz sine sweep. The sweep doesn't produce any sound until after 1 kHz. Now, if I set the averaging to 2 or more, then the second sweep begins at 20 Hz, but the first one still comes out as no sound until above 1 kHz. And funnily enough, the pink noise seems to work fine.
Sounds like there is some delay between the sound device driver being opened, and the sound card starting to work ?

Are you using WDM or ASIO ?
Also, ARTA starts doing all sorts of funky things, like showing two impulses, one after the other, when really there should be one averaged impulse from 2 sweeps.
Are you sure that's not analogue passthrough enabled from line in to output causing an echo ? Harder to spot in a loopback test but with a speaker/mic there will be obvious discrete echos.
I've put it down to the async clock on the XMOS that ARTA cannot get a good handle on. I read in the documentation that is designed to work with Synchronous cards only.

Any thoughts?

Oh btw, same problem with Holm. So, it's definitely the soundcard.
Doing single channel measurements on an asynchronous card is going to be inherently problematic, is there some reason why its not synchronous ? A hardware limitation, or just a driver limitation ?

You might have better luck with the dual channel mode in ARTA, although that has limitations of its own.
 
At some point the input will exceed the acceptable input for all of the audio interfaces, and they will begin clipping. When they do your distortion will go up, and no longer represent the input signal.
That's true, but many sound cards can't reach 0dbFS without a small amount of clipping - for example the harmonic distortion whilst still low will jump up by maybe 10-15dB at FS compared to what it is at -2dBFS.

Also I've sometimes found that the full scale output and input don't match - for example on the Audigy 2 ZS the line output with output volumes set to maximum actually exceeds the clipping point for the line input also to 0dB. So a simple analogue loopback test will clip on default mixer settings.

In the case of this sound card the line input level slider controls an analogue attenuator before the ADC so by turning that down a couple of dB digital clipping is avoided in this situation. (It can also be set for gain, increasing the sensitivity beyond the normal 2.8v)

The analogue inputs have a clipping threshold that is (from memory) about 6dB higher than the ADC, so you can apply up to 6dB of attenuation in the input slider before the analogue stages start clipping before the ADC...

On some sound cards the input level slider is simply a digital attenuator, so in these cases it should be left on maximum.

Theres a lot of gotchas and tricks with different sound cards, it takes a little while to learn the strengths and weaknesses of a given card, so its very important to do plenty of loopback testing at different sample rates, with the mixer app set to different settings and so on...
 
If reduced the Master output and record levels I can reduce the harmonics. But I thought the goal was to increase the levels to peak the 1khz tone somewhere close to -3 dBFS. I thought I read that somewhere in the ARTA manual.
You want to maximise SNR by having the input signal be near maximum when possible, but not at the expense of introducing distortion. Only a few types of acoustic measurement require large dynamic range mind you.
It would seem to me that one would have to establish a consistent reference level from one measurement to the next.
Yes. Part of the calibration process is to find what mixer settings work best with your sound card, then taking note of the exact mixer settings, and either leaving them alone, or verifying that they're still correct before a measurement session.

If you share the PC with other uses (playing music, gaming etc) then you may need to do an analogue loopback test before starting a measurement session to check your calibration is consistent, as some other app may have messed with your mixer settings. (Skype, games, etc)
 
Thanks everyone for your comments and help.

On many analog spectrum analyzers you can place markers at key spots on the trace. The markers display precise measurements that can compare with other markers.

My next question is, does ARTA (or any of the other popular software programs) provide this capability?

I know you can move the cursor to a point of interest, annotate the reading and then move to another point of interest.


Anyone?

I am just exploring what ARTA can do (or can not do).
 
I'm not sure Dennis, it doesn't seem immediately obvious and I don't think I've encountered it before, but then I've not been looking either. I know of what you speak though because my electronics spice simulator will allow you to set markers and then it will tell you the difference between them etc. Maybe someone else can chime in if they know of a similar feature in ARTA. I mean you can set overlays but the cursor wont compare one from the other.
 
@ODougbo

This is the external sound card I bought: M-Audio Profire 610

I received it a few days ago and have been testing it (playing with it is a better term).

I have had less than adequate results so far. I can not get the input and output levels correct without resulting in clipping or high distortion. I am beginning to think it is my Windows 7 Pro 64Bit drivers from M-Audio that is causing my problems. I am currently looking for a laptop with XP Pro to borrow from one of my friends.

One of my other concerns using this card on my laptop is that the laptop uses a miniature 4 pin firewire port. This requires that I use the included power adapter.

So I cannot recommend this card yet.
 
Rats! OK I'll buy a cpu in a box -- now what would do the job?

ASUS make a broad range of no nonsense, well designed, audio cards. The top of the range essence ST/STX have fantastic performance as can be seen from their reviews over at stereophile. I have the ST and can only confirm what stereophile concluded.

Now these are reasonably expensive and the other ASUS cards basically use the same interface hardware but with cheaper conveters/analogue stages. There's no shortage of reviews when it comes to finding out how their cards perform in a loop back test so you could go hunting for which of their cards will give you the performance you are after. Note that most of them perform extremely well, some though aren't so impressive, but then you do get what you pay for.

Besides ASUS you've got RME, some of their products are supposed to be excellent. SoNic_real_one has had good results with an E-MU 1820 but those seem a bit harder to find. The ESI Juli@ card is the way to go if you want to DIY in you own ADC/DAC units as they are modular and come with expansion connectors that will give you easy access to power and clock lines.
 
E-MU 204

I was very dissatisfied with the Profire 610 External Card I purchased so I sold it.

I just received a new E-MU 204 External USB Card. I am running it on my Dell Laptop with Win 7 64Bit OS.

I like what I see so far.

Please take a look at my screen grabs. The first is with ARTA Generator on. The second with the same levels and the generator turned off. I adjusted the levels for the lowest THD + Noise.

How do they look? How do I determine the noise floor of this card? I am just not sure what I am looking at.

.
 

Attachments

  • E-MU 204 Initial Loop Back Test @ 1Khz 3.jpg
    E-MU 204 Initial Loop Back Test @ 1Khz 3.jpg
    221.2 KB · Views: 335
  • E-MU 204 Initial Loop Back Test Gerator Off.jpg
    E-MU 204 Initial Loop Back Test Gerator Off.jpg
    219.1 KB · Views: 311
That looks like very decent performance. The thd at 0.0004% makes the card suitable for a lot of things. Try firing STEPS and doing a 20-20k measurement, you can also do a measurement of amplitude vs distortion too.

The noise isn't amazingly low but it's low enough. The fact the adc/dac only seems to show second and third harmonics is good also, but perhaps if the noise floor was lower they might be apparent. Of course the Noise + thd value is degraded simply because the signal level has been reduced for the best thd value.
 
On many of these sound cards, the generator (DAC) is not as good as the ADC. Usually performance can be improved with a good external low distortion sine generator. HP339A, Tek SG505, etc.
Also note there will be high frequency, out of audio band noise (~MHz) from the sound card source as well. You'll easily see it on a 100MHz or greater scope

Also the software used to generate sine wave can have an influence. I have not played with this for a while, so can't say how ARTA does in this area, but it looks pretty good. There was an older thread here with measurements back around 2009 see my comments and measurements starting here

http://www.diyaudio.com/forums/soli...u-tracker-thd-measurements-9.html#post1746046

Note well the improvements by using averaging, large block sizes (131,072) frequencies like 777.1 Hz or multiples when sampling at 96kHz
Bob
 
Thanks 5th & BFNY. All great information.

My concern is setting this up consistently each time I may need to take amplifier measurements. I cannot achieve the same results each time I start the program. It is finicky.

I am still not sure how to determine the noise floor of the E-MU with the sample screen shots I provided above.

What is the noise floor measure in? dBFS, dBV, dBu.



@ BFNY - I have this forum post in my favorites. I keep picking away at all the information there. Thanks. I tried your suggestions. I did not see any difference in THD or THD + N by changing the sample frequency. It makes since what you have said, and I will keep playing with these setup suggestions.

I have an HP 239 oscillator (same built in generator as the HP339). This is a great piece of equipment. I will try this tomorrow for an input.

I sure do like analog test equipment. I like to connect a few cables, push a few buttons, turn a few knobs and the read the results. I just can not afford a good working HP Spectrum Analyzer at the moment.
 
From those two graphs I would say that the noise would be at -103.8dB.

Apart from consistency one thing that is important in high resolution measurements is making sure that everything is connected in a way that will keep ground loops at bay. This shouldn't be too much of an issue with a laptop and external USB interface though as the whole thing can be battery operated and thus floating and create no loops.
 
@5th - Thanks. That is exactly what I thought the noise floor was. So I was correct in my set up. First establish a decent spectrum scan and then turn off the generator. What ever the RMS value is in the lower left is the noise floor. Easy.

What units of measurement is noise normally referenced?

Here is my STEPS scan. I am still working on the "amplitude vs distortion" scan. But I am close.

.
 

Attachments

  • E-MU 204 STEPS - Initial Loop Back Test.jpg
    E-MU 204 STEPS - Initial Loop Back Test.jpg
    139.1 KB · Views: 280
That's just the frequency response (which looks good btw). Run the sweep again but then increase the vertical scale so that 0dB is near the top and then say something like -130dB is at the bottom, then you'll get the harmonics plotted and we can get a good measurement of distortion vs frequency. You can alternatively click on the % button and it'll open a new window where distortion is shown as a %. I prefer the % window generally but the decibel scale csn also be useful.

Amplitude vs distortion is under the record tab, which I'm sure you've found. You basically set what frequency to run the test at (if thd is flat vs frequency from the previous measurement then the frequency you use here isn't important, 1k is nice to use though) then off you go. You can pick how many STEPS :)D) it will take when doing the measurement and the amplitude it will start and end at. Now depending on how you've got the input sensitivity configured in the setup, this could be any value/voltage. Strictly speaking the value itself isn't important all you want it to do is stop at the maximum level that the card can output and start a reasonable way down from this. What this does, is show you any trends in the system that you should be aware of, like sudden rises in distortion etc that you'd want to avoid when making measurements.
 
What units of measurement is noise normally referenced?

.

Noise is normally expressed as normalized per unit of frequency. If you look at op-amp and other discrete transistors, you'll see specs like nV/sqrt (Hz).

In simple terms, noise is a function of bandwidth. The more bandwidth, the more noise.
When you use a measurement tool like a frequency limited analog voltmeter after a notch filter (the typical analog THD+N analyzer i.e. HP339A) the bandwidth determines the total noise. If you examine closely audio specs on noise, this sort of thing is well documented. There are specs that define filters to band limit frequencies for many noise measurements.

When you use a measurement tool like a parallel filter frequency analyzer, things are different. You really need to be involved in the industry or study it closely to understand this. With a parallel filter frequency analyzer, the noise you measure is related to the bandwidth of the many parallel filters. The more narrow the filters, the smaller the noise floor you'll see on the frequency display. The filters can be constant bandwidth (FFT) or constant percentage bandwidth (octave, 1/3, 1/12 octave, etc)
In general, that's one of the reasons why I recommend using large sample block sizes, as above. It makes the FFT "filters" extremely narrow, and allows you to concentrate better on the distortion numbers.

Started doing this stuff with FFT's in '81 using a Nicolet Mini Ubiquitous 444 FFT box. Still doing it at GHz frequencies.
Bob
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.