dbx compression

[wavfile]

Does anyone have a suggestion for using DCart32 for decoding a tape recording made with dbx type 1 or 2?

Thanks!

[Craig Maier]

Re: dbx compression

The most that I can tell you is that one would use the Dynamics Processor to do so. Although I know the settings to use, I, as a representative of Diamond Cut Productions, Inc. would place the company in some jeopardy if I revealed those time constant and compression ratios. That would possibly be considered a copyright infringement by dBx since we pay them no royalties. Perhaps someone not associated with our company may be able to give you the information that you seek.

[wavfile]

Re: dbx compression

I understand. I've been able to get decent results by trial and error.

Wav

[wavfile]

Re: dbx compression

By the way, Thanks for posting the Dcart32 Patch to 3.12. It's been a great help!

Wav

[rfc]

Re: dbx compression

I have the manual for both the dbx Model 124 and 224. The 124 has a lot of detail. I suspect this was their first consumer model and they were trying to reassure users by explaining how it worked in detail.

On page 35 of that manual they give freq resp. curves for various points in the dbx II record/playback process. So let me just throw out some of the data here and see if this is useful.

First, on the surface, recorded signals are generally compressed with a 2:1 ratio. That is stated in the manual in a number of places.

Second, there are two dbx systems the Pro and the dbx II (consumer) version. I only have information on dbx II. The consumer version alters the record and play back curves in a non-linear way to improve tracking in the consumer environment. There is a different curve for both LP Discs and Tape - primaily below about 100Hz.

OVERALL PROCESS:
There is a simple bandpass filter ahead of the entire dbx II record/play back process which has a 3db roll off at 27Hz and 27kHz. Those numbers are stated in the text.

PLAYBACK RESPONSE
The reference level of the playback curve is +5dB which is confusing to me (I thought it would be 0dB or 3dB). So I will just verbally describe the dbx II overall record playback response curve as I see it on page 35.

It is essentially a flat +5dB between 100Hz and 1kHz. At 1kHz there is a linear rise to +10dB at 7kHz. Above 7kHz, there is what looks like an exponential roll-off from +10dB(7Hz)to 0dB at 20kHz. Below 100 Hz, it appears to exponentially roll -off to -10dB at 20Hz.

The Tape version is the same, except at 100Hz it rolls-off to -3db at 20Hz.

I'm sure there are better specs available somewhere, but hopefully this could get you started.

I was considering doing the dbx Encoding on the PC last year. After calling around, I think I ended up at a CBS division asking for the playback info (Sony ate CBS who ate dbx,if I recall corectly). They had no information and didn't even know where to look - according to the gentleman I talked to. In any civil case you have to show damages, and my impresion is that the dbx II system will never generate a penny for those who own the intellectual property. The dbx Pro system may be a different story.

[Craig Maier]

Re: dbx compression

rjc - The processes that you describe are "do-able" with the Diamond Cut algorithms, with the exception of the 27 KHz rolloff (unless you use 96 KHz sampling). The filters that you will be using will be the bandpass filter, the Dynamic Noise Filter and the Dynamics processor.

[rfc]

Re: dbx compression

Thanks Craig, that's good to know. I'm not too worried about the 27kHz filter as I think its main job is eliminate high frequecies that may get boosted by the recording process. The playback side of the process appears to have a 20kHz roll-off. I only mentioned the 27kHz filter for completeness.

FREQ RESPONSE CORRECTOR
I have a Stereo Review (& Shure) test record with frequecy sweeps. I could run these through my dbx unit and get the dbx playback curve for the PC. Are there any Diamond Cut tools that would let me graphically "draw" over the freq response waveform, the inverse ampltude v freq. response envelope to get a flat sweep?

TAPE DROPOUTS
I thought a variant of this same tool would be handy for tape drop outs (where you can see the erroneous amplitude envelope and you just want to boost it back to a flat envelope). Incidentally, in most of the tape drop outs I've observed, it is only one channel that drops outs. I think another usefull variant of the tool would use the rms value of the good channel to correct the bad channel.

Cheers - rfc

[Craig Maier]

Re: dbx compression

The Diamondcut software has better tools for characterizing the frequency response of a system. The "make waves" has the ability to create either a swept sinewave or to create white noise of various amplitudes. That, in conjunction with the spectrum analyzer should give provide you with the information that you seek. My preferred method is using white noise and the analyzer. One way to do this is to create a five minute wavefile of the white noise with the software program. Open up a second version of DC Art and use one to play the wavefile and the other to read the system response via the spectrum anzlyzer. That will represent the transfer function of the system under test. As for tape dropout, it is often only the high frequency component of the signal that drops out, except in extreme circumstances. You could use the channel blender on a selective basis to correct for this anamoly, since your ear will not detect the momentary reduction in stereo seperation.

[cwzell]

Re: dbx compression

I'm also very interested in dbx decoding - lots of cassettes I'd like to archive. I'm looking into this through some other sources I have.

Although I think you realize it, the 2:1 encoding compression is logrithmic, that is 2dB in is 1dB out - don't use linear amplitude!

One potential problem: As I understand it, the dbx encoder control channel is bandpass filtered between 100Hz and 10kHz. So I think what we want to do is expand the FULL signal based on the logrothmic content of a FILTERED version. How can we do this with DCArt. Ignoring this may be problematic if there are sections with considerable content < 100Hz - it may sound jumpy, a standard problem when using expanders. Not too worried about the stuff above 10kHz - seems to me that it is rare to have lots of content that high.

The rise from 1kHz to 7k, and drop back down to 0 at 20kHz rfc mentioned must be the pre-emphasis that others warned me about - nice to have the details. I think this was to get rid of some pumping artifacts.

Hey - I may have a manual from my dbx128 - maybe they have some details there also - I'll look tonight (my dbx224 has no manual - got it used).

Another interesting part is how to choose a reference level for the decoding expansion (ie. above which you expand up in level below which expand down). Maybe normalizing first to near 100%, then everything will be expanded down.



[This message has been edited by cwzell (edited 10-10-2000).]

[Craig Maier]

Re: dbx compression

You might want to consider using the top three bands of the Punch and Crunch algorithm found under the effects menu of the latest release of our software. Those three bands come close to doing what you want to accomplish. (Punch and Crunch is a multi-banded dynamics processor - - - including Punch (expansion) and Crunch (compression)). Compression and expansion ratio's are adjustable. Thresholds are also adjustable along with attack and release times.

[rfc]

Re: dbx compression

Craig, I like the idea of using white noise instead of a Freq Sweep, though its been fun watching the peak walk across the freq spectrum.

I was actually attacking two problems: dbx Records and dbx Tapes - each with a different transfer function.

It is difficult for me to use a real time setup like you suggest (if I understand you correctly), because the A/D recorder and turn table are 30ft away in another room. But I can still accomplish what you suggest for the dbx Tape transfer function (which is going to be the easiest to characterize).

Its the dbx Records that are tricky. I can't easily use the technique you describe due to the RIAA and possibly Phono Cartridge elements that are eliminated from the transfer functon.

The LP recorded Freq sweeps that I have start at 500Hz. I did verify that in general, the curve is as I described in the earlier post - above 500Hz. I used Cool Edit Pros FFT filter to "draw" a dbx playback response curve and process some record recordings. The results were poor - with too much treble.

So I will focus on the White Noise/Tape recorder approach.

[rfc]

Re: dbx compression

FYI - The Sound Well in Berkley California claims to have purchased the "pre-1990" dbx company and sells copies of the manuals.

I have emailed them asking if they own the dbx intellectual property and if they are willing to license it.
http://www.thesoundwell.com/

[cwzell]

Re: dbx compression

thanks for the info rfc. I also emailed them - but I just told them about all the the dbx LPs and tapes I have that I really want to archive on CDR, and did they know how I could get detailed encoding information? We'll see what happens. Please post back if you find out anything at all. I have quite a bit of music that I'd like to decode.

Cheers

[rfc]

Re: dbx compression

Yes I had some email from Rich yesterday. I think when he states on the Web that he bought the old pre 1990 dbx, he means that he bought their inventory. He states he does not beleive he owns the designs but that dbx pro at http://www.dbxpro.com probably owns the design.

I agree. The dbx Pro groups Web site claims to sell a dbx 140x that looks like a rack mount dbx II encoder/decoder. Several distributers said that the item is no longer on the dbx Price list but was last listed at about $319. I have a call into dbx Pro's engineering manager to see who I need to talk to about dbx II licensing and if they still sell dbx II encoder/decoders.