Skip to main content

Hi guys,

Could you suggest something here? I have to make a CD from tape for commercial release. It's classical piano. The original tape was recorded using dbx noise reduction :roll: , so after decoding all quiet passages have kinda poooofs--you know what I mean. Is there any way to get of rid of that?


Ethan Winer Sun, 02/01/2004 - 06:56


> Sounds like you might have used the wrong decoder for what you are decoding or the level match was not correct. <

Or maybe he doesn't have a decoder? I once had to master a tape that was encoded, but the customer didn't even know if his tape was Dolby or dbx! I eventually got it to sound pretty decent by using a multi-band expander. But it took a lot of fussing around with!

I'll also mention that dbx is not level dependant like Dolby. It's a continuous 2:1 curve with high frequency emphasis / de-emphasis. As I recall, all that differed between Type I and Type II was the emphasis curve, not the actual ratio.


KurtFoster Sun, 02/01/2004 - 13:20

DBX I is broadband. It was intended for use with reel to reel recorders. If you are working with a tape recorded on any Teac or Tascam RTR recorder and it is encoded, the chances are it is DBX I..

DBX II only works on the higher portions of the audio leaving the lows alone.. I am not sure where it starts working but I seem to recall everything above 500 Hz gets compressed.. DBX II was intended for use with cassette based recorders.. all the TASCAM potty studios used it.. Fostex used Dolby C and later Dolby S (an abreviated version of SR).

All noise reduction schems are very level dependent. You have to present the decoder with exactly the same level signal that the tape was encoded with. If you play a DBX'ed tape back through a play back system that is biased differently or that has been calibrated for a different tape formulation, you will encounter decoding errors.. I have a feeling this is what you are encountering..

anonymous Sun, 02/01/2004 - 16:44

Thank you Gentlemen,

I am confused, as Ethan says that DBX is not level dependant, and Kurt says that it is yes, level dependant. The last time I used DBX was many years ago in Tascam porta-studio, and it seemed to have the same problem.
Ethan, as I wrote in my original message, the tape WAS decoded. Some studio transfered it on CD for this person. However, he still has original tape.
Will it be worth trying an external DBX NR unit and play with different levels until finding the best result, and if yes, which model would you recommend. I would think it is Type I, as the tape was recorded professionaly.

realdynamix Sun, 02/01/2004 - 17:13

:) DBX is very level dependant, or you get tracking errors. Anything recorded with DBX on open reel will have a calibration tone. This is Type1 I am referencing to, type II was used in home cassette recorders. DBX does not work all the time, only when the level drops below a certain level. If you can get the master, and a type1 unit, chances are the standard playback level setting on the decode will match, it will either be set for -10 semi pro gear, or +4db, pro gear. DBX1 has high frequency pre-emphasis circuits in it. The same type unit is a must, and tape algnment is a must to have it track correctly. Otherwise, you are going to hear the errors, piano would be very noticeable. DBX1 is complex, but not as much as the multi-band Dolby of the period.


anonymous Wed, 02/04/2004 - 17:17

Today I checked the tape. It is 1/4" 2 track, +4db, and it has calibrating tones. The tape was recorded in 1984 and played four times, so I think it might lost some HF. Anybody knows parameters of HF emphasis, so I could use EQ to compensate for this loss during decoding? I checked 'New audio enciclopedia', they mention that compression ratio can be varied from 1:1 to 1:2, so should I play with different ratios, as well?


P.S. On ebay DBX 150 type I is going pretty cheap. How 'good' is this unit? It seems I saw this one a lot in different studios.

KurtFoster Wed, 02/04/2004 - 18:17

The DBX 150 is a very good version of DBX.. it will probably be cheap because there is not much demand for that type of thing any longer.

On another note, we need to straighten out a couple of things... first, if a tape is recorded on a machine that runs at +4, it really makes no difference if you play it back on a -10 machine or even one that runs at -20dB nominal level. These levels (+4, -10, -20) are only the operating levels of the machines electronics and have nothing to do with the levels recorded to tape.

To confuse things a bit, you can calibrate a machine to record at +4 dB level ... which would be 4 dB louder / hotter than was originally intended. This means that the machine will record and playback a level from tape 4dB hotter but see it as 0VU. So if there is somthing on the tape box that says +4, it probably means the machine was set (calibrated) 4dB hotter than normal.. and if you playback machine is not calibarated the same, you will encounter decoding errors.. with a loss of highs being one of them.

Also DBX is broadband and affects things recorded at all levels. Its dolby that only works on low level recording shutting off when higher levels are encountered and where noise is masked by signal. With DBX, the recordist should attempt to keep levels below 0dB when encoded, to avoid crappy sounding side effects..

realdynamix Thu, 02/05/2004 - 02:40

:) What Kurt say's about the level on tape is correct. There was 6 levels to consider with a DBX 150.
(1) The level driving the encoder
(2) The level driving the input of the recorder
(3) The level (and EQ) to tape
(4) The level (and EQ) off tape
(5) Level at the out of the recorder to decoder
(6) Level out of the DBX back into the chain

The test tones are only to line the play system up. All the other levels are set and forget. Items 3,and 4 are alignment settings in the record/play electronics of the recorder.

So, If you play this tape's test tones, typical, 1khz, 100hz, 1khz, 10khz, and they ALL read 0 on the play meter, just set the level of the output of the player to +4db (with decoder in bypass) then while switched in, set the decoder level output to +4db. (If this is -10db gear, use -10db as the decks output reference level) You should hear no artifacts, unless it is a bass solo, where asperity noise may need to be masked.

I have decoded 20+ year old DBX tapes with no problems. The tracking errors I mentioned above will only happen if the play level and EQ (off tape) are out of alignment. If you hear puffy sounds, or sprites, and uneven high frequency levels, then the machine's play electronics are out of alignment.

Properly aligned, the piano should take advantage of the extended dynamic range, with no annoying effects.
Get the unit if you can, or take the tape and your 2 track in somewhere to decode.


Ethan Winer Thu, 02/05/2004 - 07:25


> I am confused, as Ethan says that DBX is not level dependant, and Kurt says that it is yes, level dependant. <

I admit it's been more than 20 years since I owned and used a dbx NR system. It was a Type I pro unit used with an MCI 16 track 2-inch deck. Kurt may be right, but I'm 99% sure the system uses a continuous curve. As I recall that was one of the big advantages dbx had over Dolby A, which was not only level dependant but also had to be calibrated for each of its four bands.

I agree with Rick that the tape recorder's alignment is important, and will skew the decoding if it's incorrect.


Ethan Winer Thu, 02/05/2004 - 10:55


> I am certain that DBX is level dependent. <

It's been so long since I saw a dbx NR manual I can't be sure of anything. :D But since I like to argue :eek: here's a link that supports what I said about continuous 2:1 compression expansion:


Here's another quote I found at http://audiotools.c… that agrees with both of us :tu:

DBX type 2 noise reduction

A simplified version of the DBX noise reduction system intended for use with systems with limited bandwidth such as cassette recorders, 8 track recorders and carts. The type I system did not work very well with those types of systems since it works on a fairly linear basis across the spectrum so if the recorder had problems with bandwidth, phase shifts, dropouts or any sort of notches or bumps along the frequency spectrum the type I decoder would have problems. The type II system solves these problems by ...

So it looks like the basic mechanism is indeed linear 2:1 compression / expansion which should be level-independent, but as you suggested errors in the medium can harm the accuracy.


anonymous Thu, 02/05/2004 - 11:57

Thank you Rick, Kurt and Ethan.

I opened books and found... let's say, 'very little' information on DBX NR. There is one thing, which still confuses me. The 'Handbook of Recording Engineering' says: "Essentially, the input signal is pre-emphasized, and is fed to a voltage-controlled amplifier, whose gain is controlled by a level-sensing circuit that provides a 2 to 1 compression ratio…. The differences between Type1 and Type2 are in detection threshold…."
'The New Audio Cyclopedia' explains DBX as: "…DBX differs from the Dolby system because it lacks a threshold…. .The action is not frequency nor dynamic range sensitive". The provided graphs look 2:1 continuous type-- -80db->-40db, -60db->-30db, etc.
I feel that I am still missing something, is it different wording?

Ethan Winer Thu, 02/05/2004 - 13:06


> how the right calibrating settings supposed to affect the flaws of the medium? <

Beats the heck out of me! I can see how a flawed frequency response could mess things up. So at least that aspect of calibration - as opposed to absolute level - will have an effect. Since the compression / expansion is 2:1, I believe any errors in the recording medium will be doubled. So if the tape recorder's native response is 2 dB down at 15 KHz, then it will be 4 dB down after dbx decoding.


anonymous Thu, 02/05/2004 - 14:50

Now this part makes perfect sense. Thank you Ethan.

Still two sentences from the books I mentioned above confuse me:
"The differences between Type1 and Type2 are in detection thresholds as well as in the shape of pre- and de-emphasis curves…."

"DBX differs from the Dolby system because it lacks a threshold…. "

realdynamix Thu, 02/05/2004 - 15:26

:) Some more info to digest.
Reduction of noise of a DBX typeI

@ +4dbm signal level, hiss and high frequency modulation noise are reduced by 10db on signals with dominant energy below 500Hz.

@ -16dbm signal level, there is an aditional 10dbm noise reduction.

@ -36dbm signal there is an additional 20dbm noise reduction.

@ -55dbm level, there is an additional 30dbm
noise reduction.

@ -76dbm there is a total of 40dbm noise reduction.

IMO, as long as you can get the test tones to read 0vu on the output meters from 100Hz to 10Khz, you should get a pretty good decode.

There is a level sensor in the pre-emphasis, which will cause greater compression with significant HF components (prevents tape saturation) and the potential for tracking errors. The expansion post-emphasis will track this as long as alignment is good.

SSSsssshhh. Remember the ads? :D Just a cute chick with her finger to her lips.


anonymous Fri, 02/06/2004 - 04:20

dbx = "decilinear" expansion

I worked for David Blackmer (dbx / Earthworks founder) for 14 years and know a little bit about compression / expansion NR systems.

All dbx systems where designed to be level-independent. This was a key differentiator when marketed against the level-dependend Dolby systems. One of David's many breakthroughs was the design of an RMS level detector which provided a linear DC voltage output corresponding to AC input signals over an input range of 60dB. This meant that both the compressor and expander in a companding circuit could be presented with a input signal anywhere within a 60dB dynamic range window and still provide a decibel-linear output. (Compressed or expanded).

However, if the recording or transmission media have frequency or phase anamolies or introduce distortions between the output of the compressor and the input of the expander these anamolies will be exacerbated by the 2:1 expansion ratio expander.

Most likely the sound Marik describes as "poooff" is what we used to "out of band noise modulation artifact" and is not an encode/decode level match problem. This problem occured most frequently in solo piano recordings where slow, single note bass register passages where played. During playback, the compressed bass notes would be expanded correctly at a 2:1 ratio BUT these same notes did not have enough MF and HF spectral energy to mask the accompanying "shadow" of tape hiss that would be modulated upward as the expander restored the dynamic range of the bass notes at a 2:1 ratio.

For this reason, the attack and decay time constants in a dbx NR system used a program-dependent "active capacitor" circuit to adjust RMS detector tracking to provide the most accurate signal tracking and lowest ripple voltage in order to track signal envelopes most accurately.

The out-of-band noise modulation problem is an inherent design limitation of all dbx systems. We attack this problem some years later when David developed a multiband deciliner NR system. I did the reduction to engineering practice design work on this system which separated compression / expansion action into LF, MF and HF bands. I demonstrated a 7-channel matrix surround sound the system to the Swedish Film Institute in March 1986 which achieved 75dB signal to noise ration from 16mm film.

The system was being considered by AMC theater for possible use in the distribution of feature films on 16mm format in order to save on shipping charges. AMC cancelled the project and the multi-band decilinear system was never commercialized.

best to all, MJ

anonymous Fri, 02/06/2004 - 04:44

Further Memories of Decilinear companding

Another aspect of decilinear companding David hoped to commercialize was the concept of "user adjustable program dynamic range".

A decilinear expander with an adjustable expansion ratio from 1:1 to 2:1 could be used by the listener to taylor decilinear encoded program material to the listening environment - reduced dynamics range for a noisy automobile interior or full dynamic range for a quite listening room at home.

A level-dependent companding system with transfer curve "knees" could not be used to achieve this goal.

The decilinear concept took the entire notion of "level match" out of the picture and also offered the possibility of independent control of playback dynamics.

KurtFoster Fri, 02/06/2004 - 07:00

Well we have it from the "horses mouth" so to speak and I have to admit I was mistaken on this one. Surprising to me as I have used tape with DBX many times and I actually owned both four and eight track recorders along with accompaning DBX NR systems. It always seemed with the four track (where you could adjust the input and output levels) that the machine needed to be set at 0VU input and output for the DBX to track correctly. I guess I just sniffed too much fresh tape fumes once too often ....

anonymous Fri, 02/06/2004 - 07:10


I don't doubt you noticed the level-sensitivity you describe.

Many analog tape recording engineers continued to drive high levels to tape even when using dbx in order to get that nice tape saturation effect. Of course, that introduced a non-linearity into the encode / decode chain (put a "knee" in the transfer curve so to speak) that would turn a non-level dependent NR system into a level dependent one.

In the very early days of dbx Inc. (say '72 or so) David would accompany the first NR systems to high profile studios customers in NYC and teach them how to use the dbx system and stay within the linear operating range of tape. but, boys will be boys and smack the tape...

As a side note, the very earliest dbx systems installed in NYC where plauged by encode / decode errors that took a long time to figure out - everything bench tested fine. Turned out strong extremely low freq rumble from subway trains was making its way through the signal chain to tape and drive the expander nuts upon playback. The solution was to keep wide band audio path response but tighten-up the control signal pass band to eliminate ELF sensitivity.

anonymous Fri, 02/06/2004 - 11:58

Thank you Michael. The problem with this recording is exactly how you describe it. Not very uplifting.
Huh, now.... what can be done?
My initial idea was to run and mix a tape hiss through the whole track to mask these artifacts, and then use NR plugin by Waves, to remove it. However, this (as any other softaware NR) exhibits other artifacts, which I don't like, esp. on classical piano. Any other ideas?

Ethan Winer Fri, 02/06/2004 - 13:50


Thanks very much for the excellent posts. I was pretty sure dbx NR systems are intentionally level-independent, but I learned long ago not to engage Kurt in an argument so I didn't push it. :D

BTW, I might have been in the minority back in the early 1980s, but I never pushed our 16-track MCI very hard. Yeah, you can jam analog tape to "11," but distortion rises quickly once you get much above 0 VU. And that distortion is cumulative - a couple of percent when recording, a few more when you mix, yet more in the copy of the mix, etc.

Thanks again for the great info. Do us all a big favor and hang around, okay? :tu:


realdynamix Fri, 02/06/2004 - 15:19

Originally posted by Kurt Foster:
It always seemed with the four track (where you could adjust the input and output levels) that the machine needed to be set at 0VU input and output for the DBX to track correctly. I guess I just sniffed too much fresh tape fumes once too often ....

:D Hey, speaking of fumes, do you remember when Scotch Tape out of the box smelled like Christmas, and another product called CAPITOL II tape, excellent stuff.

RE: Your comment on alignment, this is normal to make this assumption, given firstly, through-put system alignment. +4db @ the input, on the tape and +4db on the tape return back to the console channel. The DIFFERENCE when using DBX is that the meters on the recorder NO LONGER represent what is displayed at the console.

So, in order to keep with a standard levels through the DBX process, once setup, the meters on the recorder respond to DBX processed signals.

If the inputs and outputs are NOT aligned, and you are attempting to drive the tape with a hotter level with it's inputs influenced by DBX, you will most likely throw the whole system alignment off as well. This is the reason the input and output level controls are on the unit.

Later versions used standard set levels with-in the units, there were control cables to enable switching from play to record/monitor etc. DX4 etc.

But, once dbx was enabled, the only true way to see what was going on with levels were the console meters. I have marks on my 2 track that are set for record and playback of DBX encoded tapes. This is where My (notion of level dependancy) came from, if I put +4db in, I expect +4db out.
Would we expect it any other way?

Tell us more MJ, I think the dbx system was great, and affordable as well. I still have 2 150's, you never know where or when you may need them. I also have an outboard Dolby B unit, an MXR compander (no compare to dbx) and the original Burwin dynamic noise reducer. I also have a Roland SN-700, multi-band noise reducer.

Thank you for your contribution to a great NR system.


anonymous Fri, 02/06/2004 - 19:54

Thanks for the warm welcome. My areas of expertise are limited but I'll try to contribute where I have some experience.

re Marik's noise modulation problem. Two solutions suggest themselves to me. The first would be to use a variable decilinear expander like a dbx 119 and run it at a ratio lower than 2:1. The result will be slightly compressed dynamics but the upward noise modulation will be reduced. And yes, mixing some nice tape hiss with the expanded signal output will help to mask the dynamically varying noise modulation.

The second approach would be to split the compressed signal into three bands, each driving it's own expander, and then sum the output of all three expanders. I'll have to admit I've never tried this and there's bound to be some mistracking, but I do know that by separting the HF expander from the LF expander it is possible to reduce HF noise modulation by LF program material to inaudibility.

If possible Marik, you might try a 200Hz low pass with 18dB/oct slope, a 200Hz - 2kHz bandpass and a 2kHz highpass with 18dB/oct slope. Of course, this assumes you've got three dbx expanders lying around to follow the band split filters.

Some clever person could probably do this in software - not me though ;)


User login