HOLMImpulse: Measuring Frequency- & Impulse-Response

[KSTR]

I would find file-I/O for everthing very useful (especially for offline processing) :
- stimulus (and the inverse, too, while you're at it)
- DUT's response
- impulse (that is, additional to the text export we already have)

My preferred formats for all three would be
- plain array of raw double, .DBL (+ dialog for the samplerate, then), since that is as "lossless" as it gets
- 32bit float .WAV
- 24 oder 16bit int .WAV (w/ normalizing + dither, maybe)
And for the pulse an estimate (or the precise value) of the sample offset between harmonics.

Background: I'd like to have access to a pure Farina Core as such, with no frills. For both (self-)educational and practical reasons.

Acourate's excellent and free LogSweepRecorder (http://www.acourate.de/) has most of this functionality (plus more params for the Farina processing) -- please take a look at it (note eg the cos envelopes of start and end of sweep and arbitrary freq range, and the HF-emphasis/deemphasis). Alas the author has chosen to scramble the doubles in his output files for pulse and inverse.

Personally I'dont need "yet another program" to display the info that's "buried" in the impulse in various way, but then again it would be nice to have that (and some quality/calibration controls as mentioned) too, like step response derived from the main pulse, etc.

The frontend, besides some minor quirks, looks perfect to me (is that due to those 63MBs of that required .NET package?)

- Klaus

[gedlee]

Originally posted by KSTR

Personally I'dont need "yet another program" to display the info that's "buried" in the impulse in various way,
- Klaus

I agree. Whats needed is something that will export the raw data from the log sweep convolution (thats is there now as I understand, but I have not used it) and allow us to see the input and output signals for checking on levels, clipping, and such. The ability use a two channel measurement for things like impedance would be nice as well.

[KSTR]

Now for the strange stuff (besides I managed to crash it)

ARTA FR (click to see full size):


HOLMI FR+Pulse :

EDIT (ingore 0.4dB normalization "shift" -- pls default to 0dB)

Same conditions, of course.

That's for sure NOT the looback LF response (the impulse looks ok ---> wrong graphic interpolation for the sparse data points at LF, it seems -- don't use any, please)

[gedlee]

Agreed, I got the same thing and wondered what was going on.

[critofur]

Excellent! So simple to use, and the export function looks great!

All that's missing is distortion analysis?

Thank you for sharing.

[askbojesen]

Thank you KSTR!

Originally posted by KSTR
Now for the strange stuff (besides I managed to crash it)
HOLMImpulse :

EDIT (ingore 0.4dB normalization "shift" -- pls default to 0dB)
That's for sure NOT the looback LF response (the impulse looks ok ---> wrong graphic interpolation for the sparse data points at LF, it seems -- don't use any, please)

1. I don't use interpolation
(Expect linear drawing from point to point)

2. I cannot reproduce these low-freq artifacts
What is the phase response ?
- The phase response reveals clock-mismatch

What are the settings?
(Screenshot of the "Settings for device and signal" tab)
- Samplerate =
- LogSweep start frequency =
- Signal Length =
- Measurement Save Length =
- Keep in/out stream alive = yes/no
- Microphone calibration = yes/no
- DAC-ADC calibration = yes/no
- If you save the measurement (file.hlm) and sent it to my email address (With a private message I'll investigate)

[askbojesen]

Hi, gedlee + KSTR!

Originally posted by KSTR
I would find file-I/O for everthing very useful (especially for offline processing) :
- stimulus (and the inverse, too, while you're at it)
- DUT's response
- impulse (that is, additional to the text export we already have)

My preferred formats for all three would be
- plain array of raw double, .DBL (+ dialog for the samplerate, then), since that is as "lossless" as it gets
- 32bit float .WAV
- 24 oder 16bit int .WAV (w/ normalizing + dither, maybe)- Klaus

As I understand you, then you would like:[list=1][*]"Do not normalize" option[*]"Do not non-hole-number time-shift option[*]Save measurement signal as wave, ascii text[*]Save The Inverse measurement signal as wave, ascii text[*]Save The impulse as wave, ascii text[*]Save The recorded response as wave, ascii text[/list=1]
(HOLMImpulse already has the "Save The impulse as ascii text"
in lossless double precision format)

or what?

[KSTR]

Hi Ask,

I remeasured today with my Marc8 8Ch.Stereo-I/O Soundcard, the result stays the same. I'm using ASIO-drivers under XP

The settings:



And the response, again


The phase matches the mag in perfect minimum phase fashion, from what I can see. And the shape of that ripple (which is linear spaced, with a linear f-axis) is what puzzles me and Earl (looks akin to a VERY high order chebysheff response or something). This is definitely not the true LF response (I "know" this soundcard very well, and checked for correct responses many times with various means).

This was the "natural" system response. With calibration it is all flat (some minor ripple left)... which is a no-brainer since we then refererence the loopback onto itself. The problem is the basic response to start with and it seems to be independent of the card (@other users: pls confirm this).

BTW, do you actually calculate the "measurement signal" response with the same process or do you just draw flat lines on the screen, assuming it is flat per se? I think this might be the core of the problem, I mean if you did an internal "loopback calibration" of the calculation process then the symptom would be accounted for, and the DAC-->ADC loopback would then only show the effective loopback error of the hardware and not the systematic behaviour. Still better would be to address the root cause, not the symptom, of course.

>> 2. I cannot reproduce these low-freq artifacts
>> What is the phase response ?
>> - The phase response reveals clock-mismatch
The signal is running loopback (1m cable) into the same specimen of the converter chip (AK4524), so I see no chances for clock mismatches as there is only one clock line in action. There might be some low-freq jitter, though. And constant latency etc. If I had used unsynced multiple soundcards this would be a problem.


>> 1. "Do not normalize" option
>> 2. "Do not non-hole-number time-shift option
Just default the shift to 0dB and provide a "normalize" button. "Shift" as a label also is a bit misleading, though I know what you meant (linear offset in a dB mag plot). "Normalization Gain" might be more readily understandable.

>> 3. Save measurement signal as wave, ascii text
>> 4. Save The Inverse measurement signal as wave, ascii text
>> 5. Save The impulse as wave, ascii text
>> 6. Save The recorded response as wave, ascii text
Yep, that would be fine (in some .WAV format. The ASCIIs for signal, inverse and response will get pretty large)

The ASCII export is already very good and flexible.... as is the whole tool in general.

Thanks for your effort (see these points as a wish list -- I'm not entitled to force anything, only your boss is )
- Klaus

[John Sheerin]

I tried measuring with the program this weekend. I could not get a reasonable level signal out of the program. Windows mixer was turned up all the way. Measuring with Speaker Workshop produced normal levels with the same settings. I could hear the signal being output - it was just at a very low level. I'm using a Behringer UCA202 USB sound card on a laptop running Windows XP SP2.

[KSTR]

>> If you save the measurement (file.hlm) and sent it to my email
>> address (With a private message I'll investigate)

I'm unclear about what to save exatly.

The impulse of the "natural response" as a .CAL-File (what is .HLM??) or the FR? Zipped I could attach it here (only 38k)