Getting different results with Realtek analog output set to external speaker vs headphones with UMIK-1

[jimpsu]

I spent all day yesterday doing a lot of measurements on REW trying to find out why I am seeing differences depending on whether I use the headphone output selection or the external speaker output selection. The soundcard software is Realtek which is a pretty common one. I even looked at the tones coming out with an oscilloscope and was not happy with what I am seeing. The sound levels vary quite a bit over frequency especially at the low frequencies. There is a big difference whether you select headphones or speaker. There is also a difference with the actual UMIK-1 microphone measurement of the speaker too as you can see in the attached document. The REW document says that no soundcard calibration is needed if you are using a USB microphone like the UMIK-1. I assume that the software knows the output level that it can measure and then also the level from the USB mic all probably done digitally so it can compute the response by a direct comparison of each. If the output level is lower at some frequency, then the software knows that and will take it into account with the mic input digitally. I am just guessing this since there is no real info that I can find on exactly how it works inside. I tried this on a Dell desktop and laptop both on Windows 10 Pro and both I7 6 cores both with Realtek sound.

It is not good that the tones are not constant with frequency. This means you could not tune it by ear like you do with the tones you have on CD which will be constant since they are created and played back digitally.

I have no idea why there would be so much variation in amplitude from the soundcards since they should be very constant with everything being done digitally. This makes no sense.

REW is of no use at all for putting out tones for tuning by ear if it is this bad.

It looks like the speaker output is more clean and better signal to noise ratio than the earbud. I am not sure why since the levels are set the same to the mic. I do not hear a lot more noise with headphones for one vs the other so not sure what is going on there either.

 

[John Mulcahy]

REW uses a log sine sweep. If you would like to study the waveform you can save it to a WAV file from the REW generator by selecting the Meas Sweep signal type. Details of log swept sine measurements and comparisons with other transfer function measurement methods are in the paper linked on the REW website:

• Swen Müller, Paulo Massarani, "Transfer Function Measurement Using Sweeps", JAES Volume 49 Issue 6 pp. 443-471, June 2001. An extended version of the paper is available at CiteSeerX

If you want to generate tones, use the Sine signal on the generator. There is an option to control the generator frequency using the graph cursor.

Recent Windows 10 builds commonly include default noise cancellation processing on stereo mic inputs, so make sure enhancements are off for the UMIK.

 

[jimpsu]

Hello John,

Thanks so much for your informative reply. I am familiar with everything you mentioned and made sure that all audio enhancements were off on both the output and the input of the computers. I am using just Realtek Audio and not Realtek HD Audio. Maybe there are some hidden settings that I cannot find and I have really tried to look. I decided to connect a good DAC up to the laptop thru its USB output and looked at both Audacity and REW in generating tones and sweeps. In both cases using the USB output and looking at the DAC analog output on an oscilloscope, the levels were exactly constant from 1 Hz to about 15 kHz where there started to be a roll-off where I was probably starting to see the filter for the sampling rate I was using. As mentioned before the response was not constant at all with either the headphone or speaker selected output from the 3.5mm output connector to the scope. I am using the MiniDSP SHD as the 1st stage in my triamped system and it has both analog inputs and a USB digital input. I then used the USB output from the laptop and fed that into the SHD and ran REW again with the 6.5 inch Purifi speaker driver into my custom enclosure and it was much different and what I feel is the correct response compared to using analog outputs to the analog inputs of the SHD. I also connected the scope up to the speaker and the sweeps were perfectly constant in level as they should be. This was not true with the analog connections.

I would highly advise a recommendation be to always use digital connections between a computer and the system if possible since the analog sections of the computer can have enhancements and behavior like I have experienced and completely giving garbage results. I would still like to find out and will keep investigating what is going on with the analog outputs in both of my high end computers and will then let the forum know if I find out.

I have also redone my Dirac Live calibration of my total system at the listening position and it is clearly better than what happened when I did it with the analog outputs. I did notice that the results above about 300 Hz were pretty much the same with the analog speaker output vs the USB output but a lot different below 300 Hz. Also, the headphone output was more noisy compared to speaker output at the higher frequencies.

This is all very curious but I am very happy with REW now that I am using everything digital from the computer to the stereo system.

I have attached some plots to show the different results and the new result of the Purifi speaker that I believe is correct using the USB output. If one doesn't have a USB input on their first stage, then they could get an inexpensive USB to analog adapter (AES-EBU XLR, Coaxial S/PDIF, or Optical Toslink) that one can find online. I tried one of those too instead of the more expensive DAC and it also had perfectly level sweeps, etc. to my oscilloscope.

I hope this will help others who might also be having some strange results using the analog outputs in their computers.

Thanks again and I am happy now for sure.

Jim
Professor in EE at Penn State and Audio Hobbyist and Author for many years

 

[John Mulcahy]

Dell tend to install Waves MaxxAudio on their computers, there should be some link to it on the Realtek Audio control panel. It can have some odd effects. I have a Dell, though I've never used the built-in audio on it.

 

[jimpsu]

Hello John again.

Thanks so much for pointing out Waves MaxxAudio. I reinstalled it and noticed that it is the EQ is activated as a default. When I deactivated it, the tones and sweeps are pefectly flat like they should be. I also noticed that it would do some automatic adjustments of volume too. Wow, that is really important to keep this turned off for all options if one is using a Dell computer. I tried to shut it down and uninstall it completely but then I could not get it to acknowledge that I plugged headphones or another device into the 3.5 mm jack. I kept it installed and on but made sure no audio enhancements are turned on.

Dell users - please be aware of this since it can really screw up the REW results.

I still like using the USB output the best and not use any of the analog sections of the Dell.

Thanks again John and this is now solved and a closed post for me.

Jim at Penn State