Issues with Emu tracker and THD measurements

[Salas]

Will do that too, but your Emu PC card has to have an edge over the USB box with its various bells and whistles even if they are family.

 

[Salas]

You have most likely a problem with the PatchMix. Use the default profile and mute down anything not absolutely required.

Did the muting, gained 1.5dB better noise floor, THD the same. This is fantastic performance for my standards. I am chuffed. Maybe it can do better with a faster 64bit PC with lots of fast mem. It also does well for noise floor up high to 96kHz when sampling at 192k.

Thanks all you guys for your expertise!

 

[fotios]

EMU0404usb - 192KHz/24bit

The PC used is: Intel C2D-3GHz (E6850), 8GB DDR2 sync, FSB 1332MHz. The OS is WinXP x64.
EMU is connected in loopback mode via two balanced cables from out to in. The results presented are not the better. I have obtained better results when the EMU is a little warmed. After half an hour, those are worse. This caused from the analog ICs used in inputs and outputs, as well the volume pots... by no way i can obtain the results of the PCI EMU of Syn88. This because i have installed only the drivers of EMU and its software mixer sliders (only 3) are in max place, so i can adjust volume only from the physical pots of EMU which is a bit difficult. Also, if you are patient and have a good PC, by selecting 1,048 millions FFT samples, then there is an improvement in results.
Anyway, see the results in the picture

An externally hosted image should be here but it was not working when we last tested it.

 

[Salas]

Still my loop is single ended so I may expect a bit more with balanced?
Its a 15cm loop non the less.

 

[Stefanoo]

You have most likely a problem with the PatchMix. Use the default profile and mute down anything not absolutely required.

mmm.....what do i need to mute?
Looking at spectraplus' s website they ran tests and put them on the web on EMU tracker pre.
They get as somebody else THD 0.00045% and SNR 107dB.

mmmm.....might there be some problems with my soundcard?
But then i see salas that got the very same result...so.....

might it be something on the set up?
I ran it Single end.
Maybe they used balanced mode (it's not specified)

link

www.spectraplus.com/Downloads/EMU Tracker Pre test report.pdf

 

[Stefanoo]

0.00041% and -150dB noise floor from 300Hz to 24kHz with the USB one at -15dB like your level. That's the best I will ever get with tracker pre, and its super for my needs. Buttons and USB cable give the edge to your PC card. Thanks Syn08.

what settings did you use to obtain 0.00041%?
what did you mute?

 

[syn08]

Still my loop is single ended so I may expect a bit more with balanced?
Its a 15cm loop non the less.

Welcome to the real world!

You just got another good reason why I will never use a sound card for sensitive measurements. True, my place is known as horrible when it comes to noise, in particular mains frequency, so you may do a little better.

Here's the wonderful EMU you just praised, in single ended mode, loop is 6ft. (certainly required to do measurements on the bench). Note that I can do, in the same place, with my pro analyzers, and the same loop length, only -140dB noise floor single ended (1601 points, so much more RBW though) but absolutely no residuals.

The last reason for not using sound cards for measurements is the AC coupling only. I care about PSD and noise well under 20Hz. I also care about low frequency distortions, not possible to reliable and precise measure. These cards are full of low end/cheap electrolytics.

Edit: Wrong picture, level was not compensated for the single ended loss. It's ok now.

 

[fotios]

Still my loop is single ended so I may expect a bit more with balanced?
Its a 15cm loop non the less.

I think this is not the issue... i mean the length of cables. As you probably know, balanced outputs formed from an inverting IC which follows a non-inverting IC. From these two ICs we get the Balnced signal (hot - cold). Balanced inputs formed from a subtracter IC which in the best case it is buffered from inputs via two voltage followers IC. Subtracter as you know, converts the balanced signal to single ended. If we use mono Jack plugs, then in output the inverter IC output is grounded. In input, the - input is also grounded. Although we are used to do this in our work (between efx or eq units) because is there not any audible effectwe ignore it for small length cables. But... maybe in this measurement system this plays some role. I used from the begining balanced cables, and i can't inform you for the effect (if exists) of using unbalanced cables. Do an experiment, it is simple.
I will also mute the unused inputs in windows mixer. The meas shown in my picture was taken with all inputs unmuted. I will inform you if there are better results.

Fotios

 

[fotios]

Yes, the mutting of unused inputs was some effective. In the right channel i got:
THD arround 0,00020%
THD+Noise arround 0,00021%
SNR arround 114 dB.
These values obtained, are very close to PCI EMU of Syn88 presented in his first post.
I think can be better because the little instabillity presented by the EMU usb. Looking the spectrogram running, i observe allways the three parameters to get gradually better values untill some moment, and after it again gradually to get worse values. But, for example, THD never exceeds 0,00050%.

Fotios

 

[Stefanoo]

i still don't understand: what did you exactly mute? and where did you mute those things? through the mixer?

mmmm....i can't reach that low THD, what setting did you use on spetra for the specific case of THD 0.00020?


I did notice that numbers go down till a certian point where all of a suddent they go up....maybe it is due to the fact that the PC is elaborating some other datas for Win...i don't know!

 

[Stefanoo]

Welcome to the real world!

......

The last reason for not using sound cards for measurements is the AC coupling only. I care about PSD and noise well under 20Hz. I also care about low frequency distortions, not possible to reliable and precise measure. These cards are full of low end/cheap electrolytics.
.....

...and i do also care about the price......welcome to the real world

i mean the systems you are talking about are running over ...way over $1K......

i do do prefer real instruments for the fact that they are more versatle and they are protected against over load...wider bandwidth and in general lower noise.


But how much? i mean....with your $350 you can read 0.00015% of distortion...and i don't think...at least in audio....you wanna add one zero more....as it really doesn't matter.....in case it is just cool to have such a powerful intrument...that's all.

Nevertheless bandwidth is needed...anyways....you can sample at 192KHz and EMU has nearly flat passband within 10Hz-80Kz...which is a lot.

don't get me wrong if i would have the possibility at the moment...i would buy a nice HP spectrum analyzer or why not...a nice compact Audio Precision system....but i can't...and the soundcard is what i can efford right now in instruments.....but....good for you that you have such a high class piece of equipment...i wish i would have access to....

 

[Salas]

Me I muted those on the pic but I only got 1.5dB better noise floor. I don't know if I understood well. The 0.0004% vicinity I reached was with -15dB level on the generator, and for non delayed channel. The displayed S/N ratio worsens with less db than your -9dB setting as expected.

 

[Salas]

.....but....good for you that you have such a high class piece of equipment...i wish i would have access to....

If someone really works with those for a living, would be burning a sound card each week. The ergonomics and protections and features make them dedicated machines precise workhorses. Its just fortunate that the PC boom gives us DIYers the performance part at least for the odd project. I have burned my laptop's out just bcs a guy slipped a crocodile clip on a tube amp I was fixing, the amp was off power, and from all places it chose to land on B+ charged PSU cap! External cards for me now on. They are replaceable too.

 

[Salas]

might it be something on the set up?
I ran it Single end.
Maybe they used balanced mode (it's not specified)

The adapters?

 

[syn08]

...and i do also care about the price......welcome to the real world

i mean the systems you are talking about are running over ...way over $1K......

I paid (then on EBay) for my HP3562A DSA about the same as for a Lynx L22 sound card. Agreed, you need guts, some luck and to be prepared to take some ****. My 3562A died after about 6 months (dry switching PS caps) but repairing was $10 and a visit to a local store to get new caps. Works like a charm since.

i do do prefer real instruments for the fact that they are more versatle and they are protected against over load...wider bandwidth and in general lower noise.

And have calibrated attenuators, BTW.

But how much? i mean....with your $350 you can read 0.00015% of distortion...and i don't think...at least in audio....you wanna add one zero more....as it really doesn't matter.....in case it is just cool to have such a powerful intrument...that's all.

The EMU1212M was $150 Canadian, or about 120 US I agree, you get what you pay for BTW, the numbers I got in the graphs are most likely wrong. I measured a ultra low distortion oscillator and compared with two calibrated instruments. The sound card says 3ppm, my instruments say 6ppm. Not that it really matters, but just to figure out how much you may trust you those PC screen numbers. To add insult to injury, I guess you noticed the results dependence on the input levels

Nevertheless bandwidth is needed...anyways....you can sample at 192KHz and EMU has nearly flat passband within 10Hz-80Kz...which is a lot.

The EMU won't work at anything over 48KHz sampling. Everybody is waiting for new Windows drivers for quite some time.

Any sound card I was able to test won't do anything decent over 10KHz. So they are essentially very good for measuring THD and spectra for a 1KHz input. Which won't tell you much about the amp quality. It won't also measure LF noise. And relying on 80KHz BW (-1dB? -3dB? Who knows?) is not really a good idea.

But I agree a sound card can be a good low cost solution. It's only about managing expectations; I've seen here and in other places audio equipment measured via sound card and presented as supa-dupa performant. They may be so, but it's not a sound card and a PC that will ultimately prove that.

 

[syn08]

Me I muted those on the pic but I only got 1.5dB better noise floor. I don't know if I understood well. The 0.0004% vicinity I reached was with -15dB level on the generator, and for non delayed channel. The displayed S/N ratio worsens with less db than your -9dB setting as expected.

Don't you use the EMU PatchMix? I don't know how good is the integration with the poor Windows control panel. I doubt you can mute the digital channels (if you have any).

 

[Salas]

I will on the laptop which is only for measurements. Now I just check it on the PC (got Emu yesterday). So I don't want to fully install stuff in the general purpose machine that runs router, firewall, all my applications, etc etc. Just familiarizing a bit. Thanks guys, I got to know almost all there is to know about Emu Tracker and more (VS dedicated, Spectraplus manipulations) in just one day! It was a nice lucky incident that Stefano had to open that thread just 2 days ahead of me.

 

[Stefanoo]

I paid (then on EBay) for my HP3562A DSA about the same as for a Lynx L22 sound card. Agreed, you need guts, some luck and to be prepared to take some ****. My 3562A died after about 6 months (dry switching PS caps) but repairing was $10 and a visit to a local store to get new caps. Works like a charm since.



And have calibrated attenuators, BTW.



The EMU1212M was $150 Canadian, or about 120 US I agree, you get what you pay for BTW, the numbers I got in the graphs are most likely wrong. I measured a ultra low distortion oscillator and compared with two calibrated instruments. The sound card says 3ppm, my instruments say 6ppm. Not that it really matters, but just to figure out how much you may trust you those PC screen numbers. To add insult to injury, I guess you noticed the results dependence on the input levels



The EMU won't work at anything over 48KHz sampling. Everybody is waiting for new Windows drivers for quite some time.

Any sound card I was able to test won't do anything decent over 10KHz. So they are essentially very good for measuring THD and spectra for a 1KHz input. Which won't tell you much about the amp quality. It won't also measure LF noise. And relying on 80KHz BW (-1dB? -3dB? Who knows?) is not really a good idea.

But I agree a sound card can be a good low cost solution. It's only about managing expectations; I've seen here and in other places audio equipment measured via sound card and presented as supa-dupa performant. They may be so, but it's not a sound card and a PC that will ultimately prove that.

Hey....you are a lucky man.....

you paid 120USD for the 1212M which is the best soundcard around...then you paid few bucks for the HP3562 (which BTW it's a great machine!)....well...what can I say? good for you!

with regard to the fact of being lucky or not.....honestly I got tired to deal with cheap deals and hope that everything will work out for a couple of weeks.
As you said above: you get what you paid for it.
If you buy off of ebay for cheap you are going to experience the hassle of going through the AAALLL service manual of your cheap machine than tried to fix it...try to find the right parts...then deal with calibration.......Have i forgotten to mention something?

Well.....time consuming.

I had to deal with an oscilloscope once and with a signal generator then.
I have gotten the last for free since it was shipped as in working condition...instead was dead....

i took several hours along with a friend of mine that kindly helped me out which wihtout i wouldn't have fixed it (thanks Conrad!).

So i decided that in the future...igf i want to purchase a good piece of equipment...will only be for calibrated and guaranteed items.
This won't be cheap....you understand?

yeah...i can also be lucky and try to play on ebay....but i just don't have the energy/time to do so.....if i buy an instrument is because i am supposed to use it from that moment on for my projects and not to buy another projects

Said that: yes, i have noticed that measurements depend upon input level...that stinks...but oh well.

BTW i can sample over 48KHz......why are you saying that EMU won't work over that frequency?
You mean the output? if you uncheck the option on the control panel of the "control", it will work.

What is Patch mix? does it come along with the CDs of the EMU? do I have o install it?

 

[fotios]

Hi amico Stefano
Under a first thought, i had no intention to post this reply to not worry any one, because also me i am in the same economics like you and worried after the following facts.
Syn88 wrote:
"The EMU won't work at anything over 48KHz sampling. Everybody is waiting for new Windows drivers for quite some time."
And by some way he is right. I can't understand how this Spectra Plus can gives so good results. I fear maybe uses some trick in background.
The fact is, if i make the same measurement with the "Virtins M.I. Pro V3.1" - which is a very serious software - it gives worse results. This soft, it is very modern and continuously upgraded. In the last version 3.1, included drivers except for sound cards also for two DACs of N.I. as well for its recently presented dual channel ADC device of 8bit/100MHz which costs 250USD (included the software that i bought before one year as well the shipping). Take a look in Virtins site. This device seems to be a professional interface, and i am feeling idiot because i paid the same ammount for the EMU0404. I have paid also 160 euro before 2 years for an X-Fi Platinum, as well another 150 euros for a :bs: ADC of 44,1KHz (Audio Codec stuff). I have spent total 510 euros for toys.
Except all others, the v3.1 it detects also the ASIO of sound card if exists. In WinXP it founds automatically the ASIO of EMU, instead in x64 not! I am forced each time to get into settings to configure the soft for the ASIO drivers while in the same time in the control panel of EMU the sampling rate changed from 192K to 44,1K! Amazing because is internally locked in EMU.
One and very serious thing, it is that M.I. Pro3.1 does not presents the initial continuous improvement of values and after the gradual deterioration such as Spectra Plus. The values presented from M.I. Pro3.1 are very stable from the begining of measurement and have an inclination of +/-2% only which caused of course from the EMU. I have tried my function generator, and the indication of spectrum is very stable and accurate.
SpectraPlus is a famous PC based analyzer, used from Sound Technology also in its proffesional SpectraRTA132 (if i remember well).
I have the sense that the measurements taken with Virtins M.I. are more reliable and very close to those given from EMU in its technical specs. brochure.
See in the spectrogram below:

An externally hosted image should be here but it was not working when we last tested it.

With friendly greetings and big worry
Fotios

 

[BFNY]

It may help to be consistent (in how you are making the measurements ) if you want to directly compare them. Same window, same sample block size, and same measurement - i.e. power spectrum with say 50 averages. I would recommend flattop window for amplitude accuracy.

As you increase the block size, the apparent noise floor is going to appear to drop. Also, as you increase the number of power spectrum averages, you'll characterize the noise and it will also drop, to a point.

Years ago people paid $25k for an FFT analyzer with less than 80db dynamic range. Now you have well over 100-105 dB for $150.

But there are some things to be careful about. What's the input impedance? Low frequency rolloff? It may change with a change in source impedance. Are there filters limiting the HF? What frequency, what type, and how much alias protection do they provide versus sample rate used. Those are some things to make sure you know, instead of whether it does 105 vs. 107 dB.