Metering the IMD of EMU0404 with Multi-Instrument Pro-3

I try for 4 days to examine if a PC sound card based software it is worthy and how much for measurements. I declare before continue that according to my oppinion the better way for estimating the performance of an audio device it is the use of a good generator of 10MHz (rise time < 10nsec), a good DSO of 50MHz, a dummy load and 2 or 3 caps. We can buy all of these with 1500 to 2000 Euros. Instead, a proffesional RTA it costs from 3000 Euro and above. A different approach it is the NI LabView software of which the cost varies between 700 to 1400 USD. Also an appropriate data aquitition device for a so much expensive software varies between 700 to 1500 USD. So, it is the same case with the RTA.
The only solution that remains it is the use of the sound card of our PC and an appropriate software. After enough research, i find that the Multi Instrument Pro-3 of Virtins Technology from Singapore was the most complette software with an excellent price (the basic module which contains the scope, the analyzer and the generator it costs 200 USD). I downloaded the trial version (full operated) before 10 days and i try to examine it thoroughly. It contains everything you need for ready measurements of commercial grade such as ready signals for measurements according to DIN, SMPTE, CCIF and other international standards. The only problem it is the sound card with its limitted posibilities in bandwidth (it does not matters if the 192KHz is inaudible according to the erroneous oppinion of some people) and for this reason exactly a cheap Pico Scope of only 12bit (but with 1Ms/s) it costs much more from the better sound card in the market.
The last 4 days i have an EMU0404 in my hands (i don't think that there is a better sound card on the market). It is external USB2.0 and can locked internally on 192Ks/s (at least the 1/5 of the sampling rate of the cheapeest PC scope) and its depth it is 24bit. To obtain its maximum performance must be connected in a clear 64bit system. In 32bit systems it is limitted at 96Ks/s at 16bit. For this reason, its drivers for Win x64 are very stable.
Well, i checked this conjuction's performance with only one measurement. The IMD (intermodulation distortion) because it is most valuable from THD for one reason. There is ready in the M.I. Pro-3 the appropriate signal for measuring the IMD according to SMPTE. For informing, it is a mix of two sine waves the one of 60Hz and the other of 7KHz in an amplitude ratio 4:1 (12db) i.e. if the 60Hz it is 1Vpp the 7KHz it is 0,25Vpp. The estimation of its value it is of big importance - from audible view - because the measured difference between these two frequencies, which represents this distortion, they are into the audible spectrum instead the harmonics produced from the two frequencies are unimportant because they are outside from the audible spectrum. I hope to i am comprehensible.
I made the following preparations with big carreful to avoid any false in the measurement.
First i disabled completelly any audio device and any relative driver of windows from my PC. Thus the windows mixer disapeared completelly from the task bar. The only device and driver remained was of the EMU0404.
The measurement executed with a clear externally loop back from the outputs of EMU to its inputs with balanced cables.
First, with a signal of 7KHz the input level calibrated at -14dB. After this the outputs returned to the inputs, the sampling rate and the bit depth in M.I.Pro-3 placed at maximum in the generator and in the FFT analyzer. To get reliable results, the samples taken from FFT must be placed from 65536 and above. To not exceed the sampling rate of EMU i selected finally 131072 samples (although it works without problem and with 262144 and above but the results may be fake). This setup needs very patience, because although the use of a powerfull PC (for domestic use at least) it takes 2 to 3 sec for each averaging cycle.
The resulted curve presented bellow. Because the numbers are a little indistrinct in the plot i also write them:
A: Peak frequency=60Hz, IMD=0,0045%(-86,9dB)
B: Peak frequency=60Hz, IMD=0,0043%(-87,3dB)
Vertical scale: -150 to 0 dBV
Horizontal scale: 0 to 100 KHz
FFT segments: 1
Resolution: 0,732422 Hz
Averaged Frames: 4
Well, i think that the results are satisfying. For the preparation of this simultaneous test so for the EMU0404 as for the Multi Instrument Pro-3 i have spend 4 days. After this, maybe i will revise my thought about the measurements with such type means. In the negatives of the M.I. Pro-3 may i add the absence of a TIM (transient intermodulation distortion) measurement routine.
I quote bellow the plot

[IMGDEAD]http://www.eal.gr/ADC2_1.jpg[/IMGDEAD]
 

Pan

Member
2002-09-19 3:52 pm
Sure the 0404-USB seems to be the best bang for the buck on the market, but there are audiointerfaces/converters with even better performance.. for a premium price as well one should add.

However.. looking at your graph it seems like the resolution could be better. Is it a limitation in the Virtins instrument or the settings/windowing?

You could also check out ARTA and RMAA as alternative to Virtins (depending on the specific needs).

Indeed a good audiointerrface with good software is today a very powerfull analysis combination to a fraction of the cost that this grade of performance use to cost ten years ago or so.


/Peter
 
Pan said:
Sure the 0404-USB seems to be the best bang for the buck on the market, but there are audiointerfaces/converters with even better performance.. for a premium price as well one should add.

However.. looking at your graph it seems like the resolution could be better. Is it a limitation in the Virtins instrument or the settings/windowing?

You could also check out ARTA and RMAA as alternative to Virtins (depending on the specific needs).

Indeed a good audiointerrface with good software is today a very powerfull analysis combination to a fraction of the cost that this grade of performance use to cost ten years ago or so.


/Peter

Hi Peter

At least a discusion about this subject of electroacoustic measurements started. Thank you verry much for the interest. I try for 2 weeks to open a discusion about. I had opened before 12 days a thread about, but a moderator he throwed my post in electronics and parts section! As you can see the subject SPICE simulation it is more valuable and it is permanent on the top of solid state page because it has the name of J. Curl which started this thread. The subject of Electroacoustic Measurements (methods, standards, gear etc) it is not so important as the SPICE!
Anyway, i try disinterestedly by spending money - time - work to inform the members about my experience so as to help them to not make unnecessary expence for buying useless things.

For the EMU0404, if i had informed for it before one year, then i woold not have spend pointlessly 170Euros to by a S.B. X-Fi Platinum. At least me from as long i searched i haven't find a better from EMU. Of course they are some high price interfaces such as LYNX but they don't offer a better sampling rate or bit depth, but only better S/N. The real thing it is the incorporated ADC-DAC in the interface. For example all the good DSO incorporate double 8bit ADC. Its type is the secret. They are flash type which comprised from a series of real comparators and a register clocked at some hundrends or thousands Ms/sec. For this reason yet a pocket PICO scope it costs from 230Euro and above. Instead the ADC used in all the consumer electronics it is of type Succesive Approximation and it is by far most cheaper because the expensive series of the real comparators replaced from a DAC driven by a microcontroller which clocks the register. Take a look in wikipedia and you understand better the time lag of SAR in comparison with the flash type.
As for the Virtins software, it operates without problem in my x64 system with the C2D of 3GHz and the 8Gb DRAM in 192Ks/s at 24bit. Instead in my other P4 WinXP system it is limitted at 96Ks/s at 16bit (a further increasing cause hung-ups). The plot presented it is the better taken but according to SMPTE standard (60Hz + 7KHz mix sinus). Because as i remembered you prefer the mix of 19KHz + 20KHz which is the standard of CCIF2 and which gives better results from SMPTE i taked three new plots including also and the DIN standard (250Hz + 8KHz sinus). I repeat that the 3 measurements quoted bellow related only with the performance of EMU0404 from input to output and only. Thus the generator has placed in No Loopback mode, which means that EMU it accepted signals only in its external inputs and no one measurement has been taken from an internal loop in EMU which can give ammazing results! (0,0001% IMD, -140dB!!!). So with two balanced cables (because EMU has balanced in/outs with +6dB increased dynamic range) the outputs connected to inputs. From a first look it is obvious that the CCIF2 method gives the better results. The DIN and the SMPTE gives worse and almost same results. Something expected because the IMD it is the difference between the two fundamental frequencies plus the harmonics included between them. All the CCIF measurements have the two fundamental frequencies in a distance of 1KHz between them in all instances, instead the DIN and SMPTE fundamental frequencies have distances of the order of some thousands Hz. The choice is yours. These all for the resolution querie of you.
For the two other softwares, i can't found the ARTA. As for the RMAA it hang-ups in 192Ks/s at 24bit on the x64 system. BTW why this guy of RMAA look for money now? Before few months he offered his software as free of cost.
Now the plots.

1) CCIF2 standard IMD meas.

[IMGDEAD]http://www.eal.gr/CCIF_IMD1.gif[/IMGDEAD]

2) DIN standard IMD meas.

[IMGDEAD]http://www.eal.gr/DIN_IMD1.gif[/IMGDEAD]

3) SMPTE standard IMD meas.

Thanks for the interest and the advices again

Fotios

[IMGDEAD]http://www.eal.gr/SMPTE_IMD1.gif[/IMGDEAD]
 
lumanauw said:
YMEC program also nice

Hi lumanauw

You mean the DSSF3 analyser? I tried also this before few days but although it runs at 192Ks/s i can't seen its actual bit depth. Maybe it is locked according to sampling rate? Anyway it is not hung-up at x64 by no way and sure this is a positive point. Nevertheless its presentation of plots and its flexibility and the menus are some old in comparison with Virtins Instrument Pro-3. Also its cost is bigger.
BTW i have to make a remark about these virtual instruments. The main module that make the meas. it is their oscilloscope (yet if it is hidden or disabled it operates behind the scene always) and from its proccesing the FFT take the data for calculations. I observed that, as much we increase the time base of scope (i.e. if we put a sinus of 1KHz the time base moved automatically at 10msec so we can see a non aliasing and clear signal on the screen. But the same time, the FFT gives bad results. So, if we increase the time base to 100msec then the results of FFT are much better. And the FFT gives its best if the time base increased at 1sec. Of course the appearance of the 1KHz wave on the screen of scope at such time base it is really disgusting, but i think that the FFT then it takes the appropriate number of data to give its true results. I am not sure if my thought it is correct by explaining this phenomenon, but it is existing in reality. Any thought about this will be appreciated.

Thanks for your interest and for the informations.

Fotios
 
NUMBER OF SAMPLES

I ask your forgiveness but i forgot to refer that the better results achieved (as in my plots above) with 65000 to 131000 number of samples. There is a little delay of course for averaging (about 3 sec per cycle) but it is suggested. Also, some times we can use the single triggering mode (one shot) to take one sample per time, and we can select finally that with the better results. Generally up to 4 samples are enough to anticipate possibly spurious signals or any incident (due to extremelly high sampling rate of 192Ks/s - 24bit) crackling noise.
The input level for good results must be placed at 80% on the M.I. Pro-3 level bars or at -6dB in the level bars of EMU.

Fotios
 

Pan

Member
2002-09-19 3:52 pm
Hi Fotios!

Ah, that resolution looks better. I assumed that it was about FFT settings.

The EMU card seems very nice.


Here is the adress to ARTA;

http://www.fesb.hr/~mateljan/arta/


I use ARTA and RMAA (+ a couple of other softwares) with a Echo Audiofire4. For faster signals I use a Picoscope 3224.

Have some problems with clicking sounds (short interupts) with the Echo though and I'm looking at the 0404-usb as an alternative.


/Peter
 
Pan said:
Hi Fotios!

Ah, that resolution looks better. I assumed that it was about FFT settings.

The EMU card seems very nice.


Here is the adress to ARTA;

http://www.fesb.hr/~mateljan/arta/


I use ARTA and RMAA (+ a couple of other softwares) with a Echo Audiofire4. For faster signals I use a Picoscope 3224.

Have some problems with clicking sounds (short interupts) with the Echo though and I'm looking at the 0404-usb as an alternative.


/Peter

Hi Peter

Thanks for the link. I checked the ARTA, and realy seems to be good and probably very easy in handling. Also its author may upgrade it because very soon al PC systems will upgraded in 64bit. In comparison with RMAA it is by far most better.
All these days i checked almost any sound card based virtual instrument with the aid of EMU0404. Generally i am not impressed so much from the results. From the other side, i supposed that maybe a cheap USB oscilloscope make the job better because its sampling rate of Ms/s and the software offered free. Moreover i have my DSO a Hameg HM507 but to unlock the FFT from its software may i pay 80Euros in the factory; and who cares about its FFT! It is elementary. To this i tried the demo of Pico Scope but also this is almost the same except the direct THD, IMD and S/N measuremets offers. I checked also more expensive (such Tektronix, Agilent, HP etc) scopes which of software also supports the Lab View. My conclusion it is that these all are not specialized instruments for audio measurements. The audio it is only a little part of its software and they offer only ellementary measurements such as the Pico Scope. As for LabView, i checked the version 6.0 of one friend because my Hameg DSO provides the driver for communicating up to LabView 7.0. This software it needs enough time to setup, and don't offer any ready routine to make a special audio measurement. It is very tiring. Moreover its cost it is prohibitive!
Lets from the begining. Our aim it is to get a system for complette (included also sound measurements) audio measurements with as much 400 to 500Euros. The only way it is the sound card. The second it is a worthy software. According to my oppinion the Virtins M.I. Pro-3 (from Singapore!) seems to be like the most complette at this time. For example, except the others it offers ready the Crosstalk measurement. Also it comprises all the necessary measurements for speakers and sound. It offers also all the necessary to measure yet the electroacoustic parameters of an unknown loudspeaker. Also an impedance meter, an LCR meter and so on. It seems like that nothing has omitted from its menu. And probably it runs without problem in 64bit systems with 192Ks/s at 24bit. I believe that this software it is new on the market and its owners they offer it in a very cheap price according to its possibilities. I made a contact with the technical advisor of Virtins, and he told me that they are in the finall attempts to join it except sound cards (because their limitted bandwidth as said to me the advisor very sincerelly!) with real data aquisition devices such as the N.I. I hope to this.
One significant remark that i made, it is that these all softwares are by far most flexible from any DSO offered. From this point and after you may jump in systems with prohibitive cost for the budget of our small labs (such as Audio Precision, B&K, Sound Technology etc).
Also we must never forget that, for making a concrette measurement, first of all we need the know-how unless yet the expensive analysers they are useless. Can you tell me how much people engaged with audio (included many members in this forum) it is in position to make the right setup for the simple crosstalk measurement yet with inexpensive means such as the sound cards and the cheap softwares about which we discuss all these days? How many from all these experts in SPICE? I bet anything that the most they don't know these basic procedures of audio measurements OF THE PRACTICE!!!

Fotios
 

Pan

Member
2002-09-19 3:52 pm
I didn' kow Virtins could do acoustic stuff. Did they add that recently? I talked with them a year ago or so and if I remember correct they didn't have it back then but I suggested they should add that.

But is there really anything Virtins can do that you can't do with ARTA + RMAA?

I don't know if crosstalk can be measured with ARTA (but I think that should be possible) but RMAA is quick and easy for that.

EMU 0404-USB + ARTA and RMAA = No brainer for me. about 200Euro for top notch performance. The day you feel this set up is a limitation.. add more powerful software if needed.

Oh one thing. There is no Scope in ARTA BUT in the spectral analysys window there is a button you can click that open up a scope wiev of the data captured. Zoom functions and stuff in this window.

Also there are fre download scope software available on the net.


/Peter
 
Pan said:
Ohps.. seems like the 0404-USB can NOT record at 192kS/s, allthough it can play back.

http://www.digit-life.com/articles2/proaudio/emu-0404-usb.html

OTOH if the lab bench has a DSO that can do fast stuff maybe that doesn't matter much.


/Peter

I read the article, but i can to confirm you that the above recorded plots taken with 192Ks/s at 24bit. Also as i said before, in such high speed of sampling and resolution we must don't exceed the averaging of 4 frames which are moreover enough. And the better way it is the record by the single triggering mode and afterwards we can take the most reliable plot. The author of this article, does not refers if he use a 64bit system such mine builded before 2 months. The C2D E6850 of 3GHz that i use in my system and the Intel M.B. DQ3JO with its 1333 FSB that i use also, they was not existent before 1 year when it writed this article. The EMU0404 has big dependence from the power of CPU and the M.B. bus speed.

Fotios
 

Pan

Member
2002-09-19 3:52 pm
Hehe, Fotios I forgot that one of your graphs showed 192kS/s.

But.. how come you can't use more than 4 averages?

Is that a limitation of Virtins (should be) and not the 0404USB?

Also one channel out, one channel in should be no problem for USB2.0 which on paper are as fast as firewire400. Firewire can handle eight channel in, eight channel out in 192kS/s I believe.


/Peter
 
Pan said:
Hi Fotios!

Ah, that resolution looks better. I assumed that it was about FFT settings.

The EMU card seems very nice.


Here is the adress to ARTA;

http://www.fesb.hr/~mateljan/arta/


I use a Picoscope 3224.




/Peter


Quick question. I bought the Picoscope 4224 and I wish I hadn't. . It's going back.
Duff software, crashing and all the bugs and worst the autoranging Y axis scale make 0dB ref a plain headache. How's your 3224 behaving ?

Richy
 
Sound Cards VS Oscilloscopes

Hi

I don't know if your issue it is the FFT analysis... if yes, there is another one thread in which you can find many interesting informations and discussions. I must do a search because it is 4-5 month old. To the present, the only that i remember well, it is that the bit depth of an ADC it is of bigger importance than the sampling rate in FFT analysis of audio devices, because the Dynamic Range.
Dynamic Range is related with bit depth by the following formula:
D.R. = 20 X log ( bit depth in Decimal Format / 1 ).
Accordingly, an 8 bit ADC has D.R. = 20 X log ( 256 / 1 ) = 48dB
a 16 bit has D.R. = 20 X log ( 65536 / 1 ) = 96dB and a 24 bit has D.R. = 20 X log ( 16777216 / 1 ) = 144dB.
If we have an audio device with THD<-96dB then only an ADC with 24bit depth can gives trustable results in FFT analysis. The same if the audio device has a S/N ratio > 96dB.
Because there is not any DSO that includes an ADC with bit depth bigger than 12bit (or 16bit modified like in Picoscope 4224 or 3224 in the cost of the reduced vertical resolution) instead there are many sound cards of lower cost (from 100 to 200 Euro as much) with ADCs of 24bit, for FFT analysis Sound Cards (we talk of course for economic solutions, not for expensive instruments like Audio Precision) are better from any DSO, either stand alone or PC based.

Regs
Fotios
 
I have an 0202 and am very happy with the measurement performance, though the output is quite noisy. This can disappear in a system measurement, but it bothers me.

My real problem with these sound card/software combinations is how to verify the results. IMHO you need access to conventional high quality ($$) instrumentation to compare readings, and then must always use the exact software setup.

I've seen too many cases where changes in buffers, sample rates, hidden scope settings and PC settings affect the results. It doesn't inspire confidence that any given measurement is truth.

A thought- we need a reproducible calibration circuit that gives a known spectral output. Something anybody could build and confirm that their soundcard measurement system agreed with everybody else's. Ideas?