Any ETF experts here? (1 Viewer)

[RichardHOS]

I have an odd problem with setting up ETF with my hardware. I've posted the problem at AVS and AA forums... if there is anyone here that really knows their ETF stuff that hasn't/won't see the post at the other foums, speak up and I'll describe what's going on.

 

[ThomasW]

What does Doug say when you call his 1-800-302-1423 support number?

 

[RichardHOS]

To be honest, and I explained this in one of the posts on another forum as well, I wanted to explore the online community's resources first. Often it's a simple thing I'm overlooking, and though the company may have customer support just for such questions, for some reason I feel like I'm wasting their time. When a forum member helps out, they are donating their time. Odd, yeah, but that's me.

More than that though, I've often found that the various responses I get to such questions online ultimately lead me to a deeper understanding of the problem. It might take longer to solve, with plenty of erroneous suggestions and more work on my part, but all of that usually helps me to understand what is really happening. If I called customer support right off, and they said "oh, we've seen that before with that sound card, open this option panel and check this box here..." I might get a quick and correct answer but have no increased comprehension of what was causing that anomaly.

Not like I wait around forever though... if there aren't any suggestions that seem to be heading the right direction over the weekend, then I'll call Acousti Soft directly see what they have to say.

lol... probably a lot more answer than you cared about, and I know it might be a bit of a strange outlook on things, but hey... you asked! I suppose I'll post the question here as well, in case someone here has some insight.

 

[RichardHOS]

From another forum:

In preparation for a serious calibration attempt in the soon to be new theater, I downloaded ETF (5.983 I believe) and began the process of setup. I assume that the setup part should function fine before they send my license file.

Hardware:

dbx 386 mic preamp and ADC. It has a tube gain section, but for this purpose I have the tube drive essentially turned off, with all the gain in the SS section.

The dbx 386 is sending S/PDIF into an M-Audio Delta Dio 2496 soundcard, using coaxial connection.

For the soundcard setup in ETF, I have the Delta Dio analog outs connected to the line inputs on the dbx 386, which then sends the signal back to the Dio after ADC. I have a Behringer ECM8000 to use for measurements.


Current settings:

No dithering or noise shaping currently on the 386, and at the moment it is set to 48kHz sampling and 24bit word. Delta Dio is set to sync to the incoming S/PDIF signal, and expects a 48kHz sampling rate. Sound card buffering in ETF is set to 48kHz.

I have adjusted gain on the 386 to reach a -2dB signal using the ETF level check.


Problem: I get a strong impulse at 0ms where it should be with a very flat response elsewhere, except for a series of weaker and decaying impulses spaced roughly every 1.6ms or so. On the linear frequency response chart I see a horrid 10db frequency response variation with a very regular sine shaped frequency dependancy for the raw soundcard device channel, but the compensated system response looks quite flat still.

When I try other sampling rates (other than matched 44.1/44.1) or try using the internal Delta Dio clock for incoming S/PDIF sync (instead of the clock in the dbx), I get the resampling variation expected.

Latency settings don't change anything. I get the same results on both channels of the dbx, and noise shaping and dithering don't change things either.

My hunch is that this regular error is caused by the slight mismatch in the Delta and dbx clocks (with the Delta generating ETF's test tone and the dbx doing the A/D conversion back into the software).



Any suggestions or ideas for how to "correct" this? I know that ETF claims the soundcard's inherent frequency response won't matter due to using one channel as a feedback correction (which does seem to work, as evidenced by the still flat system freq. response), but I'm uneasy about this for three reasons: (1) the Delta/dbx combination together have a nearly ruler flat frequency response normally, and (2) the "horrible" frequency response soundcard example they show in the help files has only about a +/- 2dB swing, and (3) the variation is large and regular enough to manifest itself as the quite noticeable series of ~1.6ms spaced measured impulses. Relatively speaking, what I'm measuring would probably be considered about the worst of the worst situations for a soundcard.

Anything I can do to correct this without purchasing still more hardware? I do have an SBLive! 5.1 card with the requisite line in/out, which would require me to purchase a cheap mic pre for the ECM8000. The Delta Dio card doesn't have analog inputs at all (it doesn't have an ADC internally at all), so an outboard ADC is required for that card. The dbx has a wordclock input, so in theory it could sync to the Delta's output, except that the Delta doesn't have a dedicated clock output. Any ideas on making use of one of the unused digital outputs on the Delta to eliminate what I can only assume is a clock sync problem?

And I suppose the most important question is will this repeating impulse response series cause errors in my measurement attempts (which will eventually include raw driver, enclosure, and room measurements)? I measure a muted and/or open loop input level of -75dB in ETF, which according to them would be fabulous (they quote "better than -50dB" for "high end" cards). If I lose the ability to make useable distortion measurements because of the described problem, I would be most disappointed.

And last but not least, if there is a better place to ask these questions, by all means point me that direction. I'd rather exhaust the online community help before going directly to ETF with the problem... generally because I learn more in the process this way.

 

[ThomasW]

I talked with my resident guru and he suggested using the analog sections of the dbx .

He says that computer audio products are usually happiest operating at 48kHz, but that frequently outboard devices like the dbx aren't, they prefer 44.1kHz. As a result one can to get sampling rate problems when the devices linked with an all digital path. So he suggests having only one digital clock in the circuit that of the M-Audio unit

 

[RichardHOS]

That would probably have been my choice, but unfortuntately my specific model of soundcard (the Dio 2496) has no analog inputs whatsoever. I have no choice, so long as I stick with this hardware, but to use an outboard A/D converter (and this particular dbx unit has selectable 44.1/48/88.2/96 kHz sampling with 16/20/24 bit wordlength, so I would guess it shouldn't have a problem with either my soundcard or ETF on those grounds).

I'm also not sure how to use the Delta Dio to control the clock in the dbx... the dbx has an AES/BEU wordclock input, and the Delta has no such output. I'm sure there is some device that will take a digital output from the Delta and generate a synced wordclock for the dbx, but the money might as well be spent on a cheap consumer soundcard in that case that has an analog input.

Maybe I can rig a way to send the dbx analog outputs into my SBLive. I'll work on that this weekend. Seems a shame though to have this nice soundcard with exceptional SNR (from my recording hobby) and not make further use of it.

Thanks for trying TW... I'll give Doug a call Monday if nothing promising develops over the weekend.

 

[RichardHOS]

Well, ETF seems to be happy enough with the SBLive! for now. Haven't received any promising suggestions on any of the forums yet, so I'll probably contact ETF when I get a chance during lunch next week. Someone mentioned an ETF discussion group, and google pulls up a few links back to ETFacoustic's site, but I can't find mention of a discussion group anywhere there.

As for ETF itself, man... this software is powerful and easy to use. Screw the manual RatShack SPL meter and test CD method, which is what I was trying to do for the past few days since my BFD arrived. Hitting a button to instantly generate a new frequency response graph is much better if you're just using a trial and error method for generating filters (which seems to work as well as anything else - my experience thus far is that trying to generate an inverse filter mathematically for freq/gain/bandwith settings just doesn't produce the results you expect... just as easy to dial in the correct filter manually).

 

[ThomasW]

Good news Richard

BTW the ETF group is basically dead but here's the url
http://forums.delphiforums.com/ETF5