The lab is back together after last years er, humidity problem- several inches of it on the floor, so I took my Emu 0202 down for a quick noise test. I use it for recording vinyl to CD with great results, but never really looked at the analog output. In a word, it's noisy as all get out. I used my old Tek diff plug-in with low pass filters and here's what I saw-
No doubt the noise contaminates the distortion, even with the 30khz LP filter inserted, got about 0.06% at 1khz. 44.1/16 settings only, as I haven't updated the software on the basement Vista machine-
As an ADC I've been pretty impressed, but as a signal source there are too many games one has to play with the settings to extract a good result. It's a question of trust- I want the same answer as my analog equipment. OTOH, I still find the SA valuable, even if the calibration is a bit questionable.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
No doubt the noise contaminates the distortion, even with the 30khz LP filter inserted, got about 0.06% at 1khz. 44.1/16 settings only, as I haven't updated the software on the basement Vista machine-
An externally hosted image should be here but it was not working when we last tested it.
As an ADC I've been pretty impressed, but as a signal source there are too many games one has to play with the settings to extract a good result. It's a question of trust- I want the same answer as my analog equipment. OTOH, I still find the SA valuable, even if the calibration is a bit questionable.
To BFNY
Many thanks BFNY for your encouraging post.
I did an upgrade in M.I. Pro v3.1 recently (free of cost) and i have not re-calibrate completelly the software yet, because the procedure is a bit complex compared with others and because the drivers of EMU ASIO are not working properly in x64. As i said, when M.I. Pro 3.1 launched the first thing it does is the searching for the ASIO driver of EMU. One problem is this, but can be resolved with trick. BTW because you reffer the input impedance, somewhere in the callibration menu of M.I. there is also a utility for its calculation in respect with the original probes that offers Virtins Tec. I have the whole stuff, but no time - to the present - to do a whole and correct setup of MI.
I believe that i will get better results when i will do this.
Moreover, the results presented in the spectrogram are very good.
I have a question: What you mean as "flattop window"?
Thanks
Fotios
Many thanks BFNY for your encouraging post.
I did an upgrade in M.I. Pro v3.1 recently (free of cost) and i have not re-calibrate completelly the software yet, because the procedure is a bit complex compared with others and because the drivers of EMU ASIO are not working properly in x64. As i said, when M.I. Pro 3.1 launched the first thing it does is the searching for the ASIO driver of EMU. One problem is this, but can be resolved with trick. BTW because you reffer the input impedance, somewhere in the callibration menu of M.I. there is also a utility for its calculation in respect with the original probes that offers Virtins Tec. I have the whole stuff, but no time - to the present - to do a whole and correct setup of MI.
I believe that i will get better results when i will do this.
Moreover, the results presented in the spectrogram are very good.
I have a question: What you mean as "flattop window"?
Thanks
Fotios
Re: issues with Emu tracker and THD measurments
I find it hard to understand why you make such a judgement and non buying recommendation when you "don't know if the card is defective or not."
My Tracker Pre is working as it should. It measures really low THD (0.0005) and has good SNR.
Considering its low price,
I highly recommend anyone to try it out.
Sigurd
I find it hard to understand why you make such a judgement and non buying recommendation when you "don't know if the card is defective or not."
My Tracker Pre is working as it should. It measures really low THD (0.0005) and has good SNR.
Considering its low price,
I highly recommend anyone to try it out.
Sigurd
Stefanoo said:
I tried ARTA which uses 64bit processing, and I get the same
THD result iin ARTA as I get in SpectraPLUS 5.0.25.10:
0.00035%.
For this use the "external" signal generator software from Dr Jordan Design. With this it is easy to have the same generator for different softwares (if they allow an external generator to be used).
Sigurd
THD result iin ARTA as I get in SpectraPLUS 5.0.25.10:
0.00035%.
For this use the "external" signal generator software from Dr Jordan Design. With this it is easy to have the same generator for different softwares (if they allow an external generator to be used).
Sigurd
fotios said:
Sigurd Ruschkow said:
Thanks Sigurd for remembering me this nice generator. I have forgot it somewhere in my old PC. As for softwares, MI Pro3.1 allows everything, because its generator/oscilloscope/FFT analyser are seperate modules. I will download the latest version of Dr Jordan gen, now.
Regs
Fotios
Re: Re: issues with Emu tracker and THD measurments
Sigurd,
BTW I was trying to figure out why I had probed such a high output noise but still the SpectrumAnalyzer, yet didn't give me the same low number as you posted somewhere else i used as reference, would give me a -135dB of noise floor.
Then I started tweaking the settings on spectraplus and now Conrad posted, since i have asked him this data, the output noise of his 0202 (thank you Conrad!) and it is like mine: now i can state that the card doesn't have any problem.
Still i was suggesting not to get the card due to the fact that customer service is extremely low (read around on internet and you'll find the same thing).
They deal with problem through the website and i have sent them an email and got an answer after 10days, short, written by a non english speaking person and thus hard to understand and most of all not informative whatsoever in reference to the problem and questions i had.
This doesn't have to do with the fact that the card was broken or not.
Moreover if it was broken i would have sit here with a broken card and not support from the manufacturer at all.
BTW Corad: if you would set 24bit 48KHz you would go down to the numbers you see on the many screenshoots.
By loop analysis with the same parameters you have set, i got pretty mcuh the same THD you got with your nice HP.
If yuo want, you can try downloading the trial version of spectraPlus from their website, it's not that big and should be ok for your line (although it will take awhile since it is 8MB).
Once you have downloaded the program you can sample as much as you want on Vista by following the steps i have highlighted above and then you can compare if the low distortion measured with spectraplus matches your HP.
Just a thought!
Anyways thank you again fo posting the results.
Sigurd Ruschkow said:
Sigurd,
BTW I was trying to figure out why I had probed such a high output noise but still the SpectrumAnalyzer, yet didn't give me the same low number as you posted somewhere else i used as reference, would give me a -135dB of noise floor.
Then I started tweaking the settings on spectraplus and now Conrad posted, since i have asked him this data, the output noise of his 0202 (thank you Conrad!) and it is like mine: now i can state that the card doesn't have any problem.
Still i was suggesting not to get the card due to the fact that customer service is extremely low (read around on internet and you'll find the same thing).
They deal with problem through the website and i have sent them an email and got an answer after 10days, short, written by a non english speaking person and thus hard to understand and most of all not informative whatsoever in reference to the problem and questions i had.
This doesn't have to do with the fact that the card was broken or not.
Moreover if it was broken i would have sit here with a broken card and not support from the manufacturer at all.
BTW Corad: if you would set 24bit 48KHz you would go down to the numbers you see on the many screenshoots.
By loop analysis with the same parameters you have set, i got pretty mcuh the same THD you got with your nice HP.
If yuo want, you can try downloading the trial version of spectraPlus from their website, it's not that big and should be ok for your line (although it will take awhile since it is 8MB).
Once you have downloaded the program you can sample as much as you want on Vista by following the steps i have highlighted above and then you can compare if the low distortion measured with spectraplus matches your HP.
Just a thought!
Anyways thank you again fo posting the results.
As mentioned, I need to update my software and drivers but time is a bit short right now- maybe over the weekend. Regardless, that won't affect the rather startling amount of wide band noise. Fortunately, the way a THD measurement is extracted from the FFT, almost all the noise should be ignored, rather than getting the traditional THD+N. I have to wonder where all this noise is coming from. It's not just clock noise. My guess is since the thing has to work on 5V, they chose some CMOS opamp parts that have good specs, but aren't that quiet. There used to be no shortage of those, though maybe the situation has improved in recent years. I still build with old fashioned dual 15 volt supplies so I don't pay much attention to those sorts of parts. Has anyone managed to find or trace out a schematic for any of the Emu boxes?
Re: To BFNY
The time window explanation is best described in the HP AN-243 here, starting on page 34
http://cp.literature.agilent.com/litweb/pdf/5952-8898E.pdf
You need to use the right window depending on what sort of measurement you are making. If you are concerned with the best amplitude accuracy, you should use the flattop window.
I mentioned input impedance, as I notice some sound cards have fairly low input impedance, i.e. 10kohm single ended.
Some audio circuits will show an increase in distortion when driving say 10k ohm versus 100k ohm. Most audio measurement instruments have 1M ohm input impedance. A really good low noise, low distortion buffer like the Pass B1 circuit, or an AD797 based amp, would likely help most sound cards, and also protect the inputs. I would use at least 100K input impedance.
Bob
fotios said:
The time window explanation is best described in the HP AN-243 here, starting on page 34
http://cp.literature.agilent.com/litweb/pdf/5952-8898E.pdf
You need to use the right window depending on what sort of measurement you are making. If you are concerned with the best amplitude accuracy, you should use the flattop window.
I mentioned input impedance, as I notice some sound cards have fairly low input impedance, i.e. 10kohm single ended.
Some audio circuits will show an increase in distortion when driving say 10k ohm versus 100k ohm. Most audio measurement instruments have 1M ohm input impedance. A really good low noise, low distortion buffer like the Pass B1 circuit, or an AD797 based amp, would likely help most sound cards, and also protect the inputs. I would use at least 100K input impedance.
Bob
Re: Re: To BFNY
Thanks BFNY for the link and suggestions. I have read the concrete article regarding flattop window. As i said i use balanced TRS cables. The inputs have Z = 1Meg according to the tech specs given from EMU. As for the flat top window, indeed there is in the windows menu of M.I. But it gives by far worse results in all parameters. Better results obtained when used a window named Blackman-Harris. As i said, M.I. Pro it is very complex in relation with Spectra Plus. It offers a lot of options for measurements. I must do a carefull study of the new version 3.1 manual. There are a lot of improvements which i don't know yet.
Regs
Fotios
BFNY said:
Thanks BFNY for the link and suggestions. I have read the concrete article regarding flattop window. As i said i use balanced TRS cables. The inputs have Z = 1Meg according to the tech specs given from EMU. As for the flat top window, indeed there is in the windows menu of M.I. But it gives by far worse results in all parameters. Better results obtained when used a window named Blackman-Harris. As i said, M.I. Pro it is very complex in relation with Spectra Plus. It offers a lot of options for measurements. I must do a carefull study of the new version 3.1 manual. There are a lot of improvements which i don't know yet.
Regs
Fotios
PMA said:
Hi PMA
Unbelievable, but M.I. Pro3.1 either in "No Loopback" mode (loop with external cables) or in Loopback which reffered as "Ia<--Oa / Ib<--Ob" gives almost the same results. This has the sense that so EMU0404 has very good ADC/DAC and balanced In/Outs as well M.I. Pro3.1 gives very reliable results if it is calibrated.
I had also an unpleasant surprise. My expensive Hameg HM8130 function generator with which i made the calibrations, presents a high THD (0,165%) in its output! The internal generators of softwares are much better.
Regs
Fotios
Take a look in this
Only to informing any one before he mess-up with sound cards etc.
Please take a look in this nice packet offered from Virtins Technology in an amazing low price (included the shippings) of 250USD
http://www.virtins.com/page2.html#VT%20DSO-2810F
VTDSO-2810F
Fotios
Only to informing any one before he mess-up with sound cards etc.
Please take a look in this nice packet offered from Virtins Technology in an amazing low price (included the shippings) of 250USD
http://www.virtins.com/page2.html#VT%20DSO-2810F
VTDSO-2810F
Fotios
Re: EMU0404usb - 192KHz/24bit
Just now, during reading all posts in this thread again and most carefully, i observed a curious thing regarding the numerical indications of SpectraPlus. Please look at the windows of THD, THD+Noise: For the Left channel the "THD = 0,00047%" while the "THD+Noise = 0,00042%"! What is going here?
Normally, the percentage of "THD+Noise" might be greater from THD percentage. How SpectraPlus obtained a such smallest sum?
The same i observed also in the picture attached from post #18 of Syn08: "THD = 0,00016%" and "THD + Noise = 0,00015%". Take a look please also in this.
Am i lose something here, regarding SpectraPus?
Fotios
fotios said:
Just now, during reading all posts in this thread again and most carefully, i observed a curious thing regarding the numerical indications of SpectraPlus. Please look at the windows of THD, THD+Noise: For the Left channel the "THD = 0,00047%" while the "THD+Noise = 0,00042%"! What is going here?
Normally, the percentage of "THD+Noise" might be greater from THD percentage. How SpectraPlus obtained a such smallest sum?
The same i observed also in the picture attached from post #18 of Syn08: "THD = 0,00016%" and "THD + Noise = 0,00015%". Take a look please also in this.
Am i lose something here, regarding SpectraPus?
Fotios
Attachments
As i can see, no one is interested furthermore for this very intersting issue on this thread. In the past 2 or 3 times, was happened the same. I don'tremember where are these other threads.
You know, the issue - IMHO - regarding sound cards as interfaces it is not their noise floor or their dynamic range or their THD or their bandwidth.
The issue it is THEIR BIG "TIM" DISTORTION regards to sharp signals with fast transients (like square waves) when these injected in their input. That means big overshoot, ringing etc. The question is this: Is this TIM distortion a component of THD? According to mine experiments YES!
You can make an experiment for proof. Put a 1KHz sinus from the internal generator of software and get the THD percentage obtained. Then, simply change the sinus to square wave in generator and you can see a big increase of THD.
You may consider a sampling rate, be it so small, of 48KSa/s by far enough to cover the audio spectrum. But the TIM? You know probably that any DSO with its usual flash front end ADC of 8bit - 100MSa/sec does not affect signals with fast transients. A square wave, is presented exactly as a square wave on the screen. IMHO this is the proof for the bigger importance of a high sampling rate instead of a high bit bepth. For this reason i believe (from those that i know, maybe i am wrong) proffesional spectrum analyzers they offer sampling rates from 100Msa up to 1 or 2Gsa per second.
As for Spectra Plus everyone he knows why it is so much popular.
Regs
Fotios
You know, the issue - IMHO - regarding sound cards as interfaces it is not their noise floor or their dynamic range or their THD or their bandwidth.
The issue it is THEIR BIG "TIM" DISTORTION regards to sharp signals with fast transients (like square waves) when these injected in their input. That means big overshoot, ringing etc. The question is this: Is this TIM distortion a component of THD? According to mine experiments YES!
You can make an experiment for proof. Put a 1KHz sinus from the internal generator of software and get the THD percentage obtained. Then, simply change the sinus to square wave in generator and you can see a big increase of THD.
You may consider a sampling rate, be it so small, of 48KSa/s by far enough to cover the audio spectrum. But the TIM? You know probably that any DSO with its usual flash front end ADC of 8bit - 100MSa/sec does not affect signals with fast transients. A square wave, is presented exactly as a square wave on the screen. IMHO this is the proof for the bigger importance of a high sampling rate instead of a high bit bepth. For this reason i believe (from those that i know, maybe i am wrong) proffesional spectrum analyzers they offer sampling rates from 100Msa up to 1 or 2Gsa per second.
As for Spectra Plus everyone he knows why it is so much popular.
Regs
Fotios
Looking a bit deeper into the performance of the E-MU 0202, here's what I found. After updating my drivers and Visual Analyzer, and using the best signal generator I have (from the Bob Cordell THD analyzer), I can get the ADC side to read about 0.0003% at a kilohertz. The Cordell analyzer gives 0.0002% when looped, so that number is right in the ballpark. The gain controls of the E-MU pass quite a lot of signal at their minimum setting, and that's where best performance is achieved, and obviously you have to use most of the bits. Sample rate was 44.1, FFT was 16384 points. Nothing was gained by increasing the sample rate, though it would certainly be necessary for higher test frequencies. I didn't measure the input s/n ratio, but it's extremely good.
Now we come to the output stage. As before, a full output 1khz signal measures 0.062% on my HP and 0.055% on the Cordell. The difference is mostly the different filter characteristics and how much noise they pass. However, a pure THD measurement from Visual Analyzer gives 0.0005%, so the signal quality is very good, if it weren't for the wretched noise. FWIW, the E-MU shows small harmonics at 3, 5, 7 and 9 khz. Maybe a bit more lost in the fuzz. The Cordell signal generator shows harmonics at 2 and 3 khz, and nothing else.
What I don't understand is how E-MU specifies their s/n ratio for the output stages. I measure unweighted figures of -52dB in a 10khz band, and a terrible wideband figure of -29dB. E-MU specifies in excess of -100dB, though it is A-weighted. I wonder if they do some trick like applying only power through the USB cable, but without a computer connected, and/or maybe disabling the internal clock? The logic might be that they have no control over the associated computer, so it shouldn't be allowed to influence the specs.
Bottom line is the system can certainly make a good THD measurement, if one is careful with setup and signal levels. You need to keep the output near maximum, so an attenuator box would be essential. For anything where noise matters, find another generator!
As for fotios' TIM issue, I think these things are at their best with many averages of sinusoidal signals. It wouldn't occur to me to hit it with square waves. As for music, IMO TIM is a red herring. The decreasing amplitude with frequency and the limited bandwidth of most musical signals prevent it from happening so long as the circuit bandwidth gets out to 20k or so. Where it's a problem is on the test bench because most signal generators don't have adjustable rise and fall time. Anybody who just hooks up a square wave generator to an audio circuit without considering the overload possibilities can expect trouble, but it doesn't mean much musically.
Now we come to the output stage. As before, a full output 1khz signal measures 0.062% on my HP and 0.055% on the Cordell. The difference is mostly the different filter characteristics and how much noise they pass. However, a pure THD measurement from Visual Analyzer gives 0.0005%, so the signal quality is very good, if it weren't for the wretched noise. FWIW, the E-MU shows small harmonics at 3, 5, 7 and 9 khz. Maybe a bit more lost in the fuzz. The Cordell signal generator shows harmonics at 2 and 3 khz, and nothing else.
What I don't understand is how E-MU specifies their s/n ratio for the output stages. I measure unweighted figures of -52dB in a 10khz band, and a terrible wideband figure of -29dB. E-MU specifies in excess of -100dB, though it is A-weighted. I wonder if they do some trick like applying only power through the USB cable, but without a computer connected, and/or maybe disabling the internal clock? The logic might be that they have no control over the associated computer, so it shouldn't be allowed to influence the specs.
Bottom line is the system can certainly make a good THD measurement, if one is careful with setup and signal levels. You need to keep the output near maximum, so an attenuator box would be essential. For anything where noise matters, find another generator!
As for fotios' TIM issue, I think these things are at their best with many averages of sinusoidal signals. It wouldn't occur to me to hit it with square waves. As for music, IMO TIM is a red herring. The decreasing amplitude with frequency and the limited bandwidth of most musical signals prevent it from happening so long as the circuit bandwidth gets out to 20k or so. Where it's a problem is on the test bench because most signal generators don't have adjustable rise and fall time. Anybody who just hooks up a square wave generator to an audio circuit without considering the overload possibilities can expect trouble, but it doesn't mean much musically.
Thanks Conrad for your reply.
It seems like Stefano and Salas have resolved their problem. That is life.
For that you wonder:
"What I don't understand is how E-MU specifies their s/n ratio for the output stages. I measure unweighted figures of -52dB in a 10khz band, and a terrible wideband figure of -29dB. E-MU specifies in excess of -100dB, though it is A-weighted. I wonder if they do some trick like applying only power through the USB cable, but without a computer connected, and/or maybe disabling the internal clock? The logic might be that they have no control over the associated computer, so it shouldn't be allowed to influence the specs."
Maybe one observation that i made during my experiments can give you an answeer: Either of softwares presented, can control by itself the sound card sampling rate and bit depth according to their settings. There is no need for the control panel of sound card for settings. Concretelly with MI Pro, regardless if the EMU control panel it is locked internaly at 192K (what a fake!) when i make the settings in ADC and DAC device to load the ASIO of EMU, the MI Pro unlock the EMU control panel settings!
Also, great care must be done if the USB port used it is 2.0 and not 1.1. In many PCs USB ports are supporting as well the two versions. This must be checked carefully.
As for the TIM issue simply i wonder if is added in THD. For sinusoidals no problem, but in tests of bandwidth with white noise?
A request please: Can you give me your option for the issue that i placed in the above post #57 regarding the THD and THD+Noise numerical indications of SpectraPlus?
Regs
Fotios
It seems like Stefano and Salas have resolved their problem. That is life.
For that you wonder:
"What I don't understand is how E-MU specifies their s/n ratio for the output stages. I measure unweighted figures of -52dB in a 10khz band, and a terrible wideband figure of -29dB. E-MU specifies in excess of -100dB, though it is A-weighted. I wonder if they do some trick like applying only power through the USB cable, but without a computer connected, and/or maybe disabling the internal clock? The logic might be that they have no control over the associated computer, so it shouldn't be allowed to influence the specs."
Maybe one observation that i made during my experiments can give you an answeer: Either of softwares presented, can control by itself the sound card sampling rate and bit depth according to their settings. There is no need for the control panel of sound card for settings. Concretelly with MI Pro, regardless if the EMU control panel it is locked internaly at 192K (what a fake!) when i make the settings in ADC and DAC device to load the ASIO of EMU, the MI Pro unlock the EMU control panel settings!
Also, great care must be done if the USB port used it is 2.0 and not 1.1. In many PCs USB ports are supporting as well the two versions. This must be checked carefully.
As for the TIM issue simply i wonder if is added in THD. For sinusoidals no problem, but in tests of bandwidth with white noise?
A request please: Can you give me your option for the issue that i placed in the above post #57 regarding the THD and THD+Noise numerical indications of SpectraPlus?
Regs
Fotios
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.