REW Beta Release V5.20 beta 13

[John Mulcahy]

A few minor updates for beta 13 and one bug fix.

The sweep file version number (saved within the WAV when a measurement sweep is exported) has been increment and I have added a check that the sweep file loaded for a File playback measurement is the current version before it is used. The sweep file has undergone a few minor changes through the V5.20 beta releases, using a file from an earlier release may produce incorrect responses at the lowest and highest frequencies.

The input and output scaling controls now use "FS sine" instead of "0 dBFS" so that the figure is independent of the "Full scale sine rms is 0 dBFS" setting.

The signal generator voltage level display now uses V, mV, uV or nV according to the value. Values change in steps of 0.1 dB when using the arrow buttons.

I have added an option to capture the graph image in monochrome, using 4 different line styles instead of different colours, handy if an image is intended for print.

A Linux option has been added to force stereo access to audio interfaces with Java drivers as some interface drivers (E-MU 0404, for example) may report they support multichannel at the requested sample rate but may not operate properly.

The SPL meter now starts automatically when it is opened.

Pressing Cancel while a measurement is in progress stops the measurement but now doesn't close the measurement dialog.

The bug fix: The stepped sine dialog did not allow 0 dBFS as a test level level when the "Full scale sine rms is 0 dBFS" option was selected.

 

[mvs0]

I have just installed rew beta13 on my new pc. I noticed that this version can not read my old measurement files (.mdat).
I looked at the headers.
My old version mdat files have "REW Measurement Data File V2w" in the header.
The beta 13 version has "TMreq Imported Measurement Data File"

Can anyone confirm my findings?

I have a lot of old measuremens and would like to keep them readable

 

[tankm]

No issue with 5.20 beta 13 reading mdat file from 5.19 and 5.20 beta 12. I just started using REW, and don't have anything older than 5.19.

 

[John Mulcahy]

I have just installed rew beta13 on my new pc. I noticed that this version can not read my old measurement files (.mdat).

Please attach an example of a file that won't load. Everything back to V3 is loading OK for me. "TMreq Imported Measurement Data File" is the old file header.

 

[tankm]

Hi John,

Not sure if it's too late for any feature request or if this is a right way to request for feature. Pardon me if it's not.

1. In the EQ Filter, when doing "Delete this filter set" to reset all the filters, will be handy to have a select-able option to "Exclude Manual Filter". ie a way to reset AUTO filters only.

2. In "Target Settings", current slope is limited to "12 dB/octave" and "24 dB/octave". Possible to include "6 dB/octave" and "18 dB/octave" ?

Thanks John.

 

[John Mulcahy]

For item 2 probably better selecting the Speaker driver target, than you can choose the order of the low and high pass responses. I'll add item 1 to the feature request list.

 

[tankm]

For item 2 probably better selecting the Speaker driver target, than you can choose the order of the low and high pass responses. I'll add item 1 to the feature request list.

Thanks John.

 

[KSTR]

Hi John,
Inspired by this post showing a potential bug I found a strange behaviour when measuring my soundcard in loopback, the Impulse Response (and derived data like Step Response), that is, the phase portion if it, doesn't seem to computed correctly sometimes.... it depends on the frequency setting of the sweep.

Settings:
- Soundcard : RME Adi2 Pro FS, 192kHz, analog loopback via cable, DAC and ADC set to same reference levels, all processing off, etc
- REW general : ASIO interface, 192kHz
- Measurement dialog : Start freq 20Hz (wrong result) or 19Hz (correct display), Stop freq 20kHz, 256k sweep length, no timing reference, t=0 at IR start, level -6dBFS

The measurement with 19Hz start freq shows the correct and expected step response display (minimum phase high-pass behaviour), but with 20Hz choosen it's getting really funny and shows a linear phase response (symmetric step)!

Interestingly, all other things derived from the IR seem to be correct, notably phase display is identical, the only thing that is off is the distortion display where an incorrect rising H2 at low frequencies is displayed (H2 dominant, but all other are wrong, too). Could be explained by the wrong time domain rendering where the left side tail of the IR seems to reach into the region of the Hx "pre-echos" (as per Farina method of distortion extraction from log-sweep) and that causes a leakage component there.

EDIT: updated plot to show non-normalized data.

PS. Another glitch I got is that the measurement process hangs forever at "checking noise floor..." when I use 128k sweep length at these settings, all other lengths seem to work.

 

[John Mulcahy]

I've fixed the 128k sweep issue for the next build, sweeps that short shouldn't be allowed at 192 kHz.

For the sweep range, to ensure valid frequency response data over the requested range REW starts its sweep at half the desired start frequency and ends it at twice the desired end frequency (or half the sample rate if lower). If the start is below 10 Hz (so a requested start below 20 Hz) the sweep uses a combination of a linear sweep from 0 to 10 Hz and a log sweep from 10 Hz to the sweep end. That is to avoid spending excessive time at the lowest frequencies, since a log sweep spends equal time in each octave of the span. If the start is 20 Hz or higher the sweep is purely logarithmic starting at half the requested start. In calculating the impulse response the frequency domain data is windowed to the range of the sweep to avoid introducing noise from regions where the sweep has little or no energy. That frequency domain windowing produces time domain ringing, since the impulse response is band limited to the range of the sweep. With short sweeps and high ratios of end frequency to start frequency that can mean the time domain ringing overlaps the distortion images.

The tl;dr version is to use long sweeps if you are interested in distortion and start sweeps from 0 unless there is a specific reason not to, such as measuring a tweeter.

 

[KSTR]

Ah, I understand, thank you for the profound answer. Clever trick, this usage of linear sweep for ULF.
From the hard borderline of the effect at 20Hz I had already imagined that there must be a sound engineering reason.
Thanks and best regards, Klaus

 

[Breeman]

Hi John,

I was playing around with the measurement file from FIR Designer, available here https://eclipseaudio.com.au/downloads/12in_2way_ir_44p1k.wav I imported this file into REW as audio data since importing it as an impulse response removes the excess phase. The issue that puzzles me is that when trying to remove the time of flight by inserting -12,1458; a value calculated by another program, in the "t=0 offset (ms)" field. The impulse gets shifted to the right instead of toward zero as I would have expected it, see screenshot below. A positive number shift the impulse to the left. Is this how this feature suppose to work or am I using it incorrectly?

 

[John Mulcahy]

Yes, that's how it is supposed to work. Moving t=0 12 ms earlier means features will be 12 ms later than they were.