Small audio interface distortion spec's

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

iampoor1

Well-known member
Joined
May 11, 2013
Messages
899
Location
California
Hello

Looking to setup a TrueRTA based audio test setup on my PC.
Of course, the next question is "what interface"!?  ;D

I am leaning twards either the focusright scarlett solo or 2i2. Has anyone used these with that setup? Whats the distortion measure? First or second generation? I have owned the saffire pro40 ionterface for about 8 years now and it has been pretty dandy!

I was originally looking at the quantasylum qa401 but cannot swing the 400$ right now.

Any other recommendations would be nice, I am running a windows 7 PC, so drivers should not be an issue.
 
I use a 2i2 that is now several years old. Distortion is orders of magnitude below that which I want to measure.

Cheers

Ian
 
ruffrecords said:
I use a 2i2 that is now several years old. Distortion is orders of magnitude below that which I want to measure.

Cheers

Ian

Great!
Do you by any chance know the figure offf the top of your head? I have heard its in the .003% range. Not that it really matters at that level, just curious.  ;D
I have also read your post about the interface box you built for it, seems nice.  8)
 
I haven't used any of the cheaper Focusrite interfaces for measurement but one thing to note is that a THD+N spec is a single number, or rather it is but only at a given level.  Usually cheaper interfaces will have a sweet spot somewhere around -10dBFS, which translates to something around +4 or +8 dBu.  As you get closer to 0dBFS the distortion will climb, sometimes quite alarmingly :)

That is one thing that separates a high quality AD/DA from cheaper units, a better unit will have a wider area of low distortion.

That said, it doesn't have to be an issue in practice, but it can limit your ability to drive DUT hard or accept a high level back from a DUT.



 
iampoor1 said:
Great!
Do you by any chance know the figure offf the top of your head? I have heard its in the .003% range. Not that it really matters at that level, just curious.  ;D
I have also read your post about the interface box you built for it, seems nice.  8)

As I recall it was below -100dBFS in a loop back test with REW, In my test set up, 0dBu is at about -9dBFS so this represents about -90dBu or about 0.003%.

Cheers

Ian
 
ruairioflaherty said:
I haven't used any of the cheaper Focusrite interfaces for measurement but one thing to note is that a THD+N spec is a single number, or rather it is but only at a given level.  Usually cheaper interfaces will have a sweet spot somewhere around -10dBFS, which translates to something around +4 or +8 dBu.  As you get closer to 0dBFS the distortion will climb, sometimes quite alarmingly :)

That is one thing that separates a high quality AD/DA from cheaper units, a better unit will have a wider area of low distortion.

That said, it doesn't have to be an issue in practice, but it can limit your ability to drive DUT hard or accept a high level back from a DUT.

Gotcha, thanks, that is very good information to know! I think I will add an active gain stage with a ne5532 and a THAT1646 active line driver for pushing harder loads up to +20-22dbm. Might have to add a passive attenuator in the front if distortion becomes an issue. I really just want something to get me buy that I can also use for field work.  8)

ruffrecords said:
As I recall it was below -100dBFS in a loop back test with REW, In my test set up, 0dBu is at about -9dBFS so this represents about -90dBu or about 0.003%.

Cheers

Ian

Thats pretty fantastic for such an affordable interface!
 
You  may want to check out Room EQ Wizard (REW) audio measurment software.

It's free and has some major improvements over TrueRTA - particularly in that it has real-time THD measures and flawless handling of asio multi-channel interfaces.

I was a TrueRTA user and I changed over to REW a few years back.  I do miss the 'dBu' scale from TrueRTA however  :)

...

Also, to get THD lower than the  'low double zero' range ...  ie. 0.002%  at -3dB FS means a more expensive audio interface - think RME or Motu or if you are able to use thunderbolt then Focusrite Scarlet etc.

That said, the M-Audio ProFire has an analog 'loopback' THD at -3dB FS  (corresponding to something like +15dBu balanced) at around 0.002% at 1KHz sine.

That's around 100usd  or so.  My old Motu 828Mk2 did quite a bit better, at 0.00023%.

Getting better than that is RME territory and the more up market interfaces.

Here's a snap of the M-Audio Profire 610  loopback spectra with THD measures ...  -10dB FS nominal  ... corresponding to +4dBu or so.  (+18dBu max output is 0dBFS)
 

Attachments

  • M-AUdio Profire610 Analog Loopback 96KHz 64K FFT 50%Hahn_nominal spl_baseline1_1.jpg
    M-AUdio Profire610 Analog Loopback 96KHz 64K FFT 50%Hahn_nominal spl_baseline1_1.jpg
    90 KB · Views: 23
Measuring the outputs/inputs with the CRO and using REW with the M-Audio Profire 610 analog loopback  ...

->  1KHz sine at max output rew -3 dB FS  (analog bal out to analog bal in)
      4.7Vrms  or +15.7 dBu    at  0.0032% thd+n

SO to convert from REW's dBFS readouts to dBu (600ohm load)  is approx 'add 18.7 dB'.
that relationship holds true all the way up and most the way down ...  of the measurement range.

It has a -131dBFS 'loopback' noise floor which is -112.3 dBu.

There is a 0.4dB 'error due to loading' in the measured output when operating into a 10K-ish load.

A further 0.1dB difference in level across  the M-Audio Profire 610 Analog Chn 3 and 4.

So - a loopback test of -20.0- dBFS in REW outputs -20.4dBFS at the CRO, and returns to REW  reading out  -20.6dBFS and -20.7dBFS on chns 3 and 4 respectively.

So the basic 'error' on readings here is 0.6dB with a further 0.1dB diff across the channels.

My old Motu 828 Mk2 was more like 0.4dB at it's worst.

True RTA had a neat feature  allowing 'calibration' of the output levels and input levels (pseudo independently).

I would specify an 'output calibration factor' of 0.4dB and a 'input calibration factor' of 0.2dB and the True RTA sig-gen and RTA dBu readouts would line up with the CRO (reality based community).

REW as we know displays mostly in dBFS and as yet doesn't 'do dBu'  and also not 'display calibration'.

REW does do 'mic SPL calibrations' and 'house curve' kind of thing, but those are other stories entirely.

SO - the upshot is even a cheap interface with  'balanced analog connections' can of quite high quality and very usable for measurement with that top shelf measurement application 'Room EQ Wizard'.

Knowing the 'dBFS to dBu' conversion factor is useful to 'reality checks'.

I use for a 'nominal test signal level'  +4 dBu, which on the M-Audio Profire 610 and REW is -14.7 dBFS.

Here's a 'baseline' snap annotated to show what's what ..

 

Attachments

  • M-Audio_Profire610_Measurement Baseline 01a.jpg
    M-Audio_Profire610_Measurement Baseline 01a.jpg
    120.2 KB · Views: 17
And once that is done, the same test snap with annotation becomes a pretty standard template which I use for further tests ...

Now I have characterised my new M-AUdio Profire 610 (around 100usd eby) I can get on and do measurements of distortion, freq resp, phase resp and a zillion other things besides.

----

The limit of my 'large signal' measurement resolution is my CRO .... it has measurement cursors to 0.01V. My DMM goes down to 0.001V. 

For the really low stuff, such as whats reported by REW at the 'noise floor' range, it is low uV range.

To measure reasonably outside of the computer, one needs to have super low noise amplifiers to bring the signal where I can measure with CRO/DMM.

 

Attachments

  • M-Audio_Profire610_nominal+4dBu_baseline 01a_small.jpg
    M-Audio_Profire610_nominal+4dBu_baseline 01a_small.jpg
    125.3 KB · Views: 12
Also to note, is that I have a 48KHz measurement bandwidth in the above snaps.

That requires of course a sampling rate of 96KHz.

The noise figures and 'peak error' figures are a little lower if the BW is restricted to 22KHz  (44.1 KHz sampling rate).

The measurements in the spectral display also vary a little depending on the settings ...  window type, overlap, fft points, averaging etc.

In addition, displays that are showing only low level components measure a little  higher than the same components in the presence of a large amplitude components. I call this a  'low level measurement  crest factor' and it is uaually 0.5dB or so.

ie. measuring 'hum floor', the reported figures are approx 0.5dB 'noisier'  on 'silence' than they are with a high amplitude test signal present.

All good fun and knowing where the errors creep in means making more valid characterisations of equipment under test :)
 
FInally, one notes that the noise performance of the interface balanced analog ins and outs dictates the minimum amount of THD you can measure in  the equipment-under-test.

OPamps are typically in the 'low triple 0' range for +4dB levels ....  the lowest I can measure on the Motu was around 0.00026% THD+N. 

Here it is more like  'medium double 0'  .... 0.0062% or so ...  this interface isn't really suited to do the really low level stuff.

Discrete stuff is usually a little higher noise than opamps ....  so this interface is not too bad for that.

Tube is usually well higher  in THD+N  ...  like 0.1% and higher ...  all the way to 4% or even more  :) 

This is where I spend most of my measurement time .. plates, cathodes, ripples and lols.
'must be robust' is the quality I admire most at this measurement baseline!

....

So the upshot is if you are mostly concerned with the really low THD stuff, like less than 0.006% at +4dBu, one needs to go to the next level of interface performance. .... in my world, that would be an RME box or a Motu box.

I like the M-Audio boxes because they are cheap enough to use at the bench, which  can be a little of a  hazardous environment.

Plus, I have several of hem, they work great with Windows 7 64bit, which is my preferred OS. The Motu  828Mk2 was always whining and *****ing and carrying on in Windows 7_64, until it's untimely demise and subsequent dumpstering.



 
Hi Alexc

Thank you so much for all of this information. I am going to mull over it for the next few days. I actually just purchased a profire 610 on your recommendation. All of my PC's have a firewire port, and the build quality looks pretty fantastic. I bought one on Ebay for 50$, so cannot beat the price!
 
Way cool :)

One day I *will* get an RME and measure confidently down to the 'low low'.

FOr now, I have another tube box to get thru.

The THD measures one can do on tube stages .. with a simple (but safety conscious) probe is very satisfying.

8)
 
Oe thing to rememeber is that if you want to measure mic pres for example, you will need an analogue domain attenuator to create low level test signals like -60dBu. The reason is you will get triple 0 distortion figures from your sound card at 0dBFS but if you use it to directly create a -60dBu signal, the distortion will be at least 60dB higher by definition.

Cheers

Ian
 
Ian, could you explain a bit further ...

I've surely noticed the thd figures get higher as the output level reduces, up to a point  ...

Thanks!
 
alexc said:
Ian, could you explain a bit further ...

I've surely noticed the thd figures get higher as the output level reduces, up to a point  ...

Thanks!

Simple example. A 16 bit system can represent plus or minus 32000 different voltage levels with 3200 being the highest. A -60dbFS signal is one thousandth of this so it only has plus or minus 32 of the levels available to represent it - basically about 6 bits. Resolution and hence distortion is therefore 1 in 64 or just over 1%.

Cheers

Ian
 
Back
Top