Kreisky said:thanks
I know SW and I am reading the italian tutorial of CN. ARTA appeared to be more easy in calibration... but
... next days I will measeure impedance with SW and tray to compare results.
Marco
I have used the S.W. before 2 years and i have spent enough time to learn it. Also i have builded a "Jig Box" of mine invention (more flexible from this of Negri) as well a mic preamp for the Behringer ECM-8000 because during this time i had not the EMU0404 (now i have it) instead i used a S.B. X-Fi Platinum. Generally, SW it is some difficult in set up but if this done correctly, the results obtained are very accurate. I remember very well this, by measuring the Z and T&S parameters of a Peerles speaker and comparing these with the original values given in the datasheet of speaker. There was also a nice software (i don't remember it and i must search in my old HDDs to find) for box calculation. With these two softwares we had builded with a friend, a pair of satelites and a band-pass/bass reflex subwoofer with Peerles units. The result was amazing good.
But after these all, i discovered "CIARE" (ah! lucky Italians) which offers ready calculated boxes for its nice speakers. We have made two nice sets of sub+satelittes with CIARE units with very good performance (at least better from our Mission speakers).
SW as well other meas softwares are usefull for estimating T&S parameters of these all cheap but very good units offered in electronic shops without brand and specs (originate from China).
BTW, i looked for ARTA-LIMP but i am not dispossed to install another one software in my PC... there are already enough of them installed in my HDD, and thus NO THANKS.
Regs
Fotios
Kreisky said:
I will try SW and at the same time I hope to find someone wich has the HW as mine.
so lucky that I live in French
So what? You are very lucky because in France you can find nice speakers like Audax, and many others except those made in France. France is the best market for audio devices in Europe. You can also meet from close Sarkozi-Brunni, and many other funny things
Ah! Lucky Marco.
What is that HW again?
Avec salutations amicales
Fotis (as she call me my neighbor Solanz from Sarrebourg)
PS. If anybody interested in high performance software generator - http://www.diyaudio.org.ua/blog.php?page=download&mode=download&id=52
It has some limitations, you can not change resolution, it always 24bits, and it uses windows default sound card. But you can enter any number of frequencies and set amplitude and phase for them. It is helpful, if you would like to measure beyond sound card THD limits.
It could have some bugs, if you've found some, or have some questions, please let me know.
It has some limitations, you can not change resolution, it always 24bits, and it uses windows default sound card. But you can enter any number of frequencies and set amplitude and phase for them. It is helpful, if you would like to measure beyond sound card THD limits.
It could have some bugs, if you've found some, or have some questions, please let me know.
Asus Xonar Essence STX FFT analysis
For the Asus Xonar Essence STX, except that it is a top sound card, we have the opportunity to make an estimation of our FFT analyzers, because Asus it offers a leaflet "Audio Precision Test Report" in which there are all the common measurements taken from an Audio Precision SYS-2722. You can download these measurements from here:
http://dlcdnet.asus.com/pub/ASUS/Audio_Card/Xonar_Essence_STX/e4253_Xonar_Essence_STX_Audio_Test_Report.pdf
I did the test with Virtins M.I. Pro3.1 FFT analyzer (callibrated very carefully according to Essence STX I/O level) with external loopback cables. This is the right way, because in the measurement included as well the ADC converter of sound card. I did the same test with the SpectraPlus, but the results was unbelievable good, thus i rejected those. Moreover, the signal generator of Virtins MI Pro3.1 has an outstanding property: it is independent either from the Windows mixer or the patch mixer of Essence STX. What i mean? In playback window of mixers yet if the Wave In is muted (deactivated) the Virtins sound generator it produces signal in the output of sound card! Instead Spectra Plus generator it is active only if the Wave In it is un-muted.
During my efforts to obtain as much as possible close results with Audio Precision, i found that: 1) The sampling frequency of 48KHz at 24bit (as mentioned by Stefanoo in previous post) it is the most appropriate for good results. 2) In X axis (frequency) i selected linear scale instead log, like in the Audio Precision test. 3) A FFT size of 2048, a test frequency of 1007.8125Hz and a record length of 10000 points can gives almost the same figure like Audio Precision.
In the attached pictures, you must take into account that in the same plot included the input and output performance of sound card, instead in the plots of Audio Precision the plots for input and output are seperate. For this reason the worse results obtained from Virtins FFT analyzer. But the difference it is very negligible: e.g. instead 0,0003% THD+Noise given in Audio Precision plot, in Virtins is 0,0006%! This is a deviation from -110dB to -105dB!
The output of Essence STX it is very quiet, presents only a noise of 3,6mVrms maximum (although the card it is PCI express installed inside the PC). Here is the noise measured with DSO:
Because there are many things which can affect the noise figure of sound card, a signifficant role on this it plays the quality of VGA card which is placed very close to sound card, as well the size of PC case, the quality of Power Supply etc. I attach a photo of inside of PC where the measurements have been done:
Here is the plot of Left channel:
And here is the plot of Right channel:
I leave in you, to make your own estimations.
Regs
Fotios
For the Asus Xonar Essence STX, except that it is a top sound card, we have the opportunity to make an estimation of our FFT analyzers, because Asus it offers a leaflet "Audio Precision Test Report" in which there are all the common measurements taken from an Audio Precision SYS-2722. You can download these measurements from here:
http://dlcdnet.asus.com/pub/ASUS/Audio_Card/Xonar_Essence_STX/e4253_Xonar_Essence_STX_Audio_Test_Report.pdf
I did the test with Virtins M.I. Pro3.1 FFT analyzer (callibrated very carefully according to Essence STX I/O level) with external loopback cables. This is the right way, because in the measurement included as well the ADC converter of sound card. I did the same test with the SpectraPlus, but the results was unbelievable good, thus i rejected those. Moreover, the signal generator of Virtins MI Pro3.1 has an outstanding property: it is independent either from the Windows mixer or the patch mixer of Essence STX. What i mean? In playback window of mixers yet if the Wave In is muted (deactivated) the Virtins sound generator it produces signal in the output of sound card! Instead Spectra Plus generator it is active only if the Wave In it is un-muted.
During my efforts to obtain as much as possible close results with Audio Precision, i found that: 1) The sampling frequency of 48KHz at 24bit (as mentioned by Stefanoo in previous post) it is the most appropriate for good results. 2) In X axis (frequency) i selected linear scale instead log, like in the Audio Precision test. 3) A FFT size of 2048, a test frequency of 1007.8125Hz and a record length of 10000 points can gives almost the same figure like Audio Precision.
In the attached pictures, you must take into account that in the same plot included the input and output performance of sound card, instead in the plots of Audio Precision the plots for input and output are seperate. For this reason the worse results obtained from Virtins FFT analyzer. But the difference it is very negligible: e.g. instead 0,0003% THD+Noise given in Audio Precision plot, in Virtins is 0,0006%! This is a deviation from -110dB to -105dB!
The output of Essence STX it is very quiet, presents only a noise of 3,6mVrms maximum (although the card it is PCI express installed inside the PC). Here is the noise measured with DSO:
An externally hosted image should be here but it was not working when we last tested it.
Because there are many things which can affect the noise figure of sound card, a signifficant role on this it plays the quality of VGA card which is placed very close to sound card, as well the size of PC case, the quality of Power Supply etc. I attach a photo of inside of PC where the measurements have been done:
An externally hosted image should be here but it was not working when we last tested it.
Here is the plot of Left channel:
An externally hosted image should be here but it was not working when we last tested it.
And here is the plot of Right channel:
An externally hosted image should be here but it was not working when we last tested it.
I leave in you, to make your own estimations.
Regs
Fotios
As some others have reported the EMU 0202, other USB audio devices, generate a lot of noise at their output, far above the advertised level. With the high level of noise I was unable to use this for distortion measurements. I decided to investigate on the off chance I could fix it. I found two individual sources of noise, plus some general background noise.
Please note that the numbers in the following are from memory only and are probably inaccurate.
I connected the 0202 up to an HP 332A distortion meter. played a 1KHz tone using the 0202, and got about 3% THD + noise, worse than my Sansa Clip MP3 player. The output of the HP 332A was connected to an oscilloscope. The oscilloscope showed a large repeating “bump” in the noise about every 250ms. RIAA bounces during its measurement and I assume these noise “bumps” are the cause. Searching the internet lead me to check for latency problems, apparently common with both USB and Firewire devices on Windows. This latency is the delay that Windows can take before responding to a request to service the USB, not a frequency latency as we normally have in audio with crossovers. It seems that if the delay gets too long, the 0202 suffers from Windows transfer "glitches" and noise bumps are the result. The latency was measured with the Thesycon’s DPC Latency Checker (DPC Latency Checker). My laptop had about 500us and was supposedly fine for streaming audio and video. As it was snowing outside and these noise bumps could be caused by USB latency problems, I followed their instructions to reduce this latency and was able to get this down to 6us. Rechecking the output on the scope showed the noise bumps were gone and it seems likely they were caused by these USB latency problems. RIAA now measured quite well at a usable low THD+noise of about -110dB and bouncing during the test was fixed. It seems Windows, my network drivers and possibly some of my other software caused this audio problem. I originally could not sample at any setting above 44.1 Khz-16 bit and there was a lot of distortion even then. Now 48 KHz and 24 bit is great with RIAA. Higher speeds and bit levels also worked but I did not write down the details.
On the HP 332A THD + noise at 0.1% (-30dB) was much better, but, still too high for my intended use. The output from the HP 332A, again displayed on the scope, showed a strong approximately 138 KHz signal. As the EMU 0202 measured fine using RIAA, I have to assume that the EMU 0202 filters out high frequency on its input, but it is there on the output. The HP 332A has a very large frequency bandwidth and so its measurement includes this as noise. To be able to use the HP 332A I need to reduce all of the high frequency noise on the output of the EMU 0202. The noise is present at all times, even when no sound is played through the 0202. I used the 332A to null higher frequencies and found that nulling at 138 KHz lowered THD + noise by 3.5 dB. A T-notch filter would look after this, but I need far more. It seems that a high order low pass filter is necessary. One of those high order things I know nothing about.
I'm not very good at this filter thing and would like any comments or suggestions you may have regarding this.
Thanks
Jim
Please note that the numbers in the following are from memory only and are probably inaccurate.
I connected the 0202 up to an HP 332A distortion meter. played a 1KHz tone using the 0202, and got about 3% THD + noise, worse than my Sansa Clip MP3 player. The output of the HP 332A was connected to an oscilloscope. The oscilloscope showed a large repeating “bump” in the noise about every 250ms. RIAA bounces during its measurement and I assume these noise “bumps” are the cause. Searching the internet lead me to check for latency problems, apparently common with both USB and Firewire devices on Windows. This latency is the delay that Windows can take before responding to a request to service the USB, not a frequency latency as we normally have in audio with crossovers. It seems that if the delay gets too long, the 0202 suffers from Windows transfer "glitches" and noise bumps are the result. The latency was measured with the Thesycon’s DPC Latency Checker (DPC Latency Checker). My laptop had about 500us and was supposedly fine for streaming audio and video. As it was snowing outside and these noise bumps could be caused by USB latency problems, I followed their instructions to reduce this latency and was able to get this down to 6us. Rechecking the output on the scope showed the noise bumps were gone and it seems likely they were caused by these USB latency problems. RIAA now measured quite well at a usable low THD+noise of about -110dB and bouncing during the test was fixed. It seems Windows, my network drivers and possibly some of my other software caused this audio problem. I originally could not sample at any setting above 44.1 Khz-16 bit and there was a lot of distortion even then. Now 48 KHz and 24 bit is great with RIAA. Higher speeds and bit levels also worked but I did not write down the details.
On the HP 332A THD + noise at 0.1% (-30dB) was much better, but, still too high for my intended use. The output from the HP 332A, again displayed on the scope, showed a strong approximately 138 KHz signal. As the EMU 0202 measured fine using RIAA, I have to assume that the EMU 0202 filters out high frequency on its input, but it is there on the output. The HP 332A has a very large frequency bandwidth and so its measurement includes this as noise. To be able to use the HP 332A I need to reduce all of the high frequency noise on the output of the EMU 0202. The noise is present at all times, even when no sound is played through the 0202. I used the 332A to null higher frequencies and found that nulling at 138 KHz lowered THD + noise by 3.5 dB. A T-notch filter would look after this, but I need far more. It seems that a high order low pass filter is necessary. One of those high order things I know nothing about.
I'm not very good at this filter thing and would like any comments or suggestions you may have regarding this.
Thanks
Jim
Hi Jim,
Thanks for this information and the heads-up on USB-connected soundcards. When you said RIAA, did you mean RMAA?
Have you tried any of the other audio analysis software packages to see if they exhibit the same behavior? Lots of them are freeware or shareware with a trial time. Visual Analyzer is one you might try. I'm not a computer geek.
Cheers,
Bob
Bob,
Thanks Bob, Yes it was RMAA (RightMark Audio Analyzer ), the latest version 6.2.3. Sorry about that.
I did use other software to play the 1 kHz tone, but only RMAA to check the card. The noise "bump" problem was present with all of them until I "fixed" the USB latency problem. If you don't know how to "disable" your network card and check for the latency you may need the help of a computer geek. Many "kids" will know how to do this. To start them off... you need to right click on "My Computer", select Properties, Device Manager, right click on the network devices, and other things your not using, and disable them. Restart the computer, test using RMAA.
I have not spent much time with analysis software, as I have had issues with sound cards and it seems Windows as well.
Jim
Thanks Bob, Yes it was RMAA (RightMark Audio Analyzer ), the latest version 6.2.3. Sorry about that.
I did use other software to play the 1 kHz tone, but only RMAA to check the card. The noise "bump" problem was present with all of them until I "fixed" the USB latency problem. If you don't know how to "disable" your network card and check for the latency you may need the help of a computer geek. Many "kids" will know how to do this. To start them off... you need to right click on "My Computer", select Properties, Device Manager, right click on the network devices, and other things your not using, and disable them. Restart the computer, test using RMAA.
I have not spent much time with analysis software, as I have had issues with sound cards and it seems Windows as well.
Jim
Bob, I don't think software will make a difference since the input bandwidth, back into the PC, will be band limited by the anti-aliasing filter.
Jim, what exactly are you trying to do. Do you just want to use the 0202 as a signal source? Or do you plan to use it as a source for the external HP analyzer?
What frequency range are you trying to cover using it as a signal source?
If for example you only need to cover 20 kHz with sine waves, then you can place
the low pass filter -3dB cutoff at something like 30 to 50 kHz depending on the
design of the filter. How much attenuation you need above 100 kHz will determine
the order of the filter that you need. I'm not saying that this is the best solution.
What frequency range do you want to cover for analysis of higher order harmonics?
Jim, what exactly are you trying to do. Do you just want to use the 0202 as a signal source? Or do you plan to use it as a source for the external HP analyzer?
What frequency range are you trying to cover using it as a signal source?
If for example you only need to cover 20 kHz with sine waves, then you can place
the low pass filter -3dB cutoff at something like 30 to 50 kHz depending on the
design of the filter. How much attenuation you need above 100 kHz will determine
the order of the filter that you need. I'm not saying that this is the best solution.
What frequency range do you want to cover for analysis of higher order harmonics?
Last edited:
Does this help as a reference for the 0202? I don't see the high level of noise that people are talking about within the audio passband based on these measurements:
A Winner
I find that my laptop when powered from the AC adapter will often produce a whine at the output of the internal and external USB audio devices. It is very easy to hear in normal use and for some reason it goes away if I float the AC plug ground. A ground loop obviously but with my laptop and headphones nothing else is grounded, LOL!
I'm interested in the 0202 or 0404 for distortion measurements but have not purchased one yet and so will consider other units also.
I do currently have an M-Audio Transit which is fine for speaker testing.
A Winner
I find that my laptop when powered from the AC adapter will often produce a whine at the output of the internal and external USB audio devices. It is very easy to hear in normal use and for some reason it goes away if I float the AC plug ground. A ground loop obviously but with my laptop and headphones nothing else is grounded, LOL!
I'm interested in the 0202 or 0404 for distortion measurements but have not purchased one yet and so will consider other units also.
I do currently have an M-Audio Transit which is fine for speaker testing.
Last edited:
I didn't even noticed this thread moved along. My 0404 tests pretty much confirm some of your experiences. Was able to get low thd number only when all the stars were properly aligned. Got the best results with the Virtins MI software, after reading the manual. For any sample rate and fft window they recommend a particular frequency for the generated tone. In any case, I had to to use the ASIO drivers, and the 0404 input and output pots had to be set just right to get low thd. The loop cable makes a difference too, I built my own shielded cable. I think the lowest thd I could measure was something like 0.0004, which is, what, 68dB? I've seen some people showed 0.0001, 80dB? Please correct me if my numbers are wrong here.
But I thought there was room for improvement and cracked the 0404 case open today. I had a quick look at the pcb and wrote an initial impression in a wiki (see my sig). I noticed the following:
* the mic/line input uses a 15532 opamp with a data sheet thd of about 0.001
* there are quite a few dc-dc converters getting the right voltages out of the 5V wall wart. Some people claimed that a linear 5V supply made a difference, but my measured thd was the same with the wall wart or with my 5V shunt regulator. Makes sense, because the 5V in gets chopped up and put together again in all the crazy ways to make 12V, 5V, -5V, 1.8V, 3V, etc.
* checked the positive and negative rails of a few opamps with a 60MHz scope with lowest vertical sensitivity of 2mV/div. The positive rail seemed quiet at about 2mV p-p. The negative rail had nasty 6-8mV p-p spikes with a period of about 6.x uS. Incidentally, the dc-dc inverting converter, a TPS6755, works on a switching frequency of 160kHz. Neither the 15532 opamp, nor the other opamp, the 2068 (which is used in great numbers in the 0404) seems to have such great PSRR, so to me, this would be a starting point of improvement, to fix the negative rail voltage. It would be interesting if other people's 0404 or 0202 exhibit the same symptom. Could be that I just got a bad card.
The other obvious target for improvement could be the input and output opamps. There are definitely better opamps that one could use.
But I thought there was room for improvement and cracked the 0404 case open today. I had a quick look at the pcb and wrote an initial impression in a wiki (see my sig). I noticed the following:
* the mic/line input uses a 15532 opamp with a data sheet thd of about 0.001
* there are quite a few dc-dc converters getting the right voltages out of the 5V wall wart. Some people claimed that a linear 5V supply made a difference, but my measured thd was the same with the wall wart or with my 5V shunt regulator. Makes sense, because the 5V in gets chopped up and put together again in all the crazy ways to make 12V, 5V, -5V, 1.8V, 3V, etc.
* checked the positive and negative rails of a few opamps with a 60MHz scope with lowest vertical sensitivity of 2mV/div. The positive rail seemed quiet at about 2mV p-p. The negative rail had nasty 6-8mV p-p spikes with a period of about 6.x uS. Incidentally, the dc-dc inverting converter, a TPS6755, works on a switching frequency of 160kHz. Neither the 15532 opamp, nor the other opamp, the 2068 (which is used in great numbers in the 0404) seems to have such great PSRR, so to me, this would be a starting point of improvement, to fix the negative rail voltage. It would be interesting if other people's 0404 or 0202 exhibit the same symptom. Could be that I just got a bad card.
The other obvious target for improvement could be the input and output opamps. There are definitely better opamps that one could use.
PB2,
I was using the same settings as the person who wrote "A Winner". You are right that the distortion is quite low in the audio band, if I disable my network card and a few other things. My remaining problem is that my distortion meter measures total noise, and as you see on the "A Winner" site, noise increases quite a bit at very high frequencies (over 20 kHz. My meter measures harmonics and noise up to 500kHz. It seems there is a lot of noise in the very very high frequencies and this adds up. If you measure with a device that has less bandwidth you will not notice. If you are using the 0202 to measure itself things will be fine, if you turn off (disable) a few things. Of course if your computer is newer and faster than mine you may not have to jump through these hoops.
Jim
I was using the same settings as the person who wrote "A Winner". You are right that the distortion is quite low in the audio band, if I disable my network card and a few other things. My remaining problem is that my distortion meter measures total noise, and as you see on the "A Winner" site, noise increases quite a bit at very high frequencies (over 20 kHz. My meter measures harmonics and noise up to 500kHz. It seems there is a lot of noise in the very very high frequencies and this adds up. If you measure with a device that has less bandwidth you will not notice. If you are using the 0202 to measure itself things will be fine, if you turn off (disable) a few things. Of course if your computer is newer and faster than mine you may not have to jump through these hoops.
Jim
Hi Jim,
Just tried something. At the output of the negative rail converter there's a 100uV capacitor from - (output) to +. Yellow little thing SMD, with code 107A on it. Then a tiny resistor with code 0 on it, sort of a CR filter, except that the resistor is just a jumper I think. So I replaced the capacitor with a 470uF OSCON. And replaced the 0 resistor with a small inductor, and then another 470uF OSCON across - and +, basically creating a CLC filter. Indeed this got rid of all the nasty HF spikes on the negative rail.
Running the Virtins MI software and tweaking the input and output volume buttons on the 0404, and the Windows volume button I managed to get 0.0003% (-111.1dB) THD, and 93.3dB SNR. For some reason THD+N is pretty high at 0.0022% (-93.3dB).
In any case, getting rid of those spikes on the negative rail doesn't seem to make so much of a difference, but at least it's less bothersome knowing that they're gone.
By the way I made a mistake in my previous message, I meant to say that before I could measure 0.0004% THD which is not 68dB, but 107.96dB, and 0.0001% THD which is -120dB. Forgot the % :/
Just tried something. At the output of the negative rail converter there's a 100uV capacitor from - (output) to +. Yellow little thing SMD, with code 107A on it. Then a tiny resistor with code 0 on it, sort of a CR filter, except that the resistor is just a jumper I think. So I replaced the capacitor with a 470uF OSCON. And replaced the 0 resistor with a small inductor, and then another 470uF OSCON across - and +, basically creating a CLC filter. Indeed this got rid of all the nasty HF spikes on the negative rail.
Running the Virtins MI software and tweaking the input and output volume buttons on the 0404, and the Windows volume button I managed to get 0.0003% (-111.1dB) THD, and 93.3dB SNR. For some reason THD+N is pretty high at 0.0022% (-93.3dB).
In any case, getting rid of those spikes on the negative rail doesn't seem to make so much of a difference, but at least it's less bothersome knowing that they're gone.
By the way I made a mistake in my previous message, I meant to say that before I could measure 0.0004% THD which is not 68dB, but 107.96dB, and 0.0001% THD which is -120dB. Forgot the % :/
ikoflexer,
Thanks for checking into that. I think my problem may be a bit like the plot in the second post in the following thread:
http://www.diyaudio.com/forums/digi...5-4396-ultrasonic-noise-spectrum-graphic.html
You can see the large broad peak in noise at about 200kHz. If I can get this high frequency stuff down, I may get close to your level.
Jim
Thanks for checking into that. I think my problem may be a bit like the plot in the second post in the following thread:
http://www.diyaudio.com/forums/digi...5-4396-ultrasonic-noise-spectrum-graphic.html
You can see the large broad peak in noise at about 200kHz. If I can get this high frequency stuff down, I may get close to your level.
Jim
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.