NE570-based NR compander (P-522) by JR

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

NE570-based NR compander (P-522) by JR

Post by JR. »

I decided to start a new thread rather than hijack the RMS thread.

Here is a schematic from my '80s tape NR.. This is based on a medium preformance gain cell (NE572), but I have long been of the opinion that gain modulation artifacts can dominate the sound quality differences between VCAs

Image

As usual I am image challenged so here is a link to the schmatic hosted on one of my sites.

http://www.johnhroberts.com/p-522.jpg

I will try to give a brief overview touching all bases, but there are a lot of bases here so I will not go into a lot of detail unless folks are interested in specific aspects.

First in general you will notice the compressor is far more complex than the exapander... The compressor must accept the worst case transient signals and squeeze them down to a more manageable dynamic range and whitened frequency spectra for tape media.

I will describe the comp features as the expander is a subset of these features applied symmetrically for playback.

- starting from top left.. simple unity gain input buffer. Inverting since compader is inverting, so non-inverting overall.

- wow filter, .1uF C2 and 2Meg resistor to output of inverting gain stage forms a level adaptive sliding HPF. At very low level (high inverting gain), pole shifts up to suppress wow when it could dominate envelope and cause playback mistracking from warped or out of round records. At higher signal levels when wow does not dominate envelope HPF slides down for full response. Note: this dynamic HPF filter might be of use without the NR.

- audio path receives +10dB of HF pre-emphasis (above 500 Hz or so). IIRC this is slightly less than +12dB used by DBX.

- gain cell has some HF boost in it's audio path to take back some of the HF boost to deal with tape media's limited very HF response.

-rectifier side chain has extensive filtering to make it primarily responsive to mid band signal content that will be accurately reproduced by tape media (think cassettes).

- Ok here is where the secret sauce comes in.. The ne572 has two mirrored rectifier outputs designed to provide a fast attack and slow release. I am tricking the fast attack section by connecting it to an inverting opamp virtual earth input instead of a small cap. Since it is held approximately at ground the internal follower between this fast port and slow port is effectively defeated.

----- first order effect of inverting opamp and connection is to make real .1 uF cap at slow port, look more like a 1 uF+ slow cap, for modest signal level changes.

------the second 1uF cap connected through a zener configured transistor serves as a fast attack charge circuit. One unique characteristic of this specific topology is that the fast attack current is "borrowed" from the second 1uf cap and paid back over time so long term (ave) control voltage tracks very accurately independently of small transient or level related mistracking. A conventional fast att circuit that pulls fast charging current from rails, would cause significant and audible control voltage tracking errors from simple level differences between encode/decode, and even from things like phase response errors that impact transient peak amplitude expression (think cassette phase response).

------in addition to the primary fast attack circuit, there is a secondary, fast attack mechanism, when the opamp clipps on the - supply rail. Since it can no longer subtract current from the .1uF on slow port, it reverts to it's real .1uF (fast) value.

----- the transistor at the - input provides a Overload LED indication to tell you you are hitting it too hot, and serves the secondary function of clamping that node at a diode above ground so internal fast attack path remains defeated.

------ In addition to all the fast attack speed up circuitry, the leading edge of severe transients coming from dead silence can always pass through before the finite rise time of control voltage can respond so the back to back diodes and resistor divider across compressor stage opamp forms a soft clipper at 20 dB gain and several Vp-p. (Note: I could have allowed even faster attack control path by not clamping the opamp minus input at a diode drop, but listening tests determined that allowing this internal fast path to operate on transient overloads, caused much more audible mistracking artifacts. )

===========
-- The playback expander provides an identical version of fast/slow attack and frequency shaping networks so this will dynamically track typical recorded signals very well..

While this design may alter the first mSecs of the cannon fire transient on that classic Telarc 1812 overture recording, it does so with no audible artifacts while retaining most of the impact an feel. I am tempted to debate the absolute accuracy of the recording wrt that actual acoustic event anyhow. I invested many hours with my custom tone burst generator, to make this NR handle the most severe signals I could throw at it with freedom from audible artifacts.

I am very proud of this effort even though it fell flat in the marketplace. I was not trumped by technology or a better design but by low cost offshore (Japanese) manufacturing and a stronger brand image than I enjoyed, selling for less assembled than my kit unassembled. :oops:

JR
Cancel the "cancel culture", do not support mob hatred.
Crusty
Posts: 69
Joined: Mon Aug 20, 2007 1:42 am

Re: NR compander (P-522)

Post by Crusty »

Thanks for posting this, John! I only wish I could design something this elegant with so few parts! Altering the attack circuit like that is really wild - I've never seen that done before (and I've stared at a lot of 570 circuits LOL). I want to give some thought to how I could apply some of those techniques to the That chips - the zenered transitor/cap circuit in particular.
JR. wrote:------in addition to the primary fast attack circuit, there is a secondary, fast attack mechanism, when the opamp clipps on the - supply rail. Since it can no longer subtract current from the .1uF on slow port, it reverts to it's real .1uF (fast) value.

----- the transistor at the - input provides a Overload LED indication to tell you you are hitting it too hot, and serves the secondary function of clamping that node at a diode above ground so internal fast attack path remains defeated.
I love neat details like these...
JR. wrote:Note: this dynamic HPF filter might be of use without the NR.
Yes, in my preamp! LOL


Thanks again for posting, many questions to follow...
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: NR compander (P-522)

Post by JR. »

In fact the node on the THAT RMS chip where the cap connects is not unlike the node on the 572.. While when you play games with att/release the merit to using RMS is diminished somewhat in my view. In one studio compressor I did back in the '80s I used a NE572 just as a dual output rectifier since it was cheaper than several opamps and assorted parts.. I just left the gain cells unused.

Keep in mind.. in designing a companding NR a great deal of effort goes into managing overload aspects that could lead to mistracking, after the record media tramples the signal, so this is a slightly different focus than dynamics processing. Which only has to sound good from one pass, and doesn't have to be symmetrically unwound.

JR

PS: If you want clever you should look at my TS-1 circuits... I designed them and I have trouble going back decades later and figuring some of what I did out.
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: NR compander (P-522)

Post by mediatechnology »

Very nice John! I printed out the schematic so I could follow it along with your text. Elegantly simple.

I like what you did in the detector and yes the 570 and the rms seem to be similar in being able to use this trick. The diode limiter for overshoot protection with it's soft-limiting would seem to be better than to allow hard clipping even for short durations.

Unless you object John here's a copy of the TS-1 schematics. I may put these over in "Document" too. During my travels with SSL it was very unusual to go to a studio that didn't have a TS-1 in their shop. The TS-1 is another example of making a few parts do as much as possible.

http://www.ka-electronics.com/Images/jpg/Loftec_TS1.jpg
http://www.ka-electronics.com/Images/jp ... tector.jpg
http://www.ka-electronics.com/Images/jp ... _TS1_2.jpg

While we're on the subject of the 570 I thought it would be worthwhile to post this link found over in "Document."

http://www.ka-electronics.com/images/pd ... ations.pdf
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: NR compander (P-522)

Post by JR. »

For any interested in discussing details of the TS-1 I'll answer specific questions, but there is too much going on for me to do a point by point overview of everything that is unusual in that design... it wold be easier to list what isn't unusual. I'll also need a schematic I can actually read (Hint- Wayne if those were saved as PDF I could zoom in, but that probably takes more memory.) I should have a schematic around, but don't have one handy. I'm not big on dwelling on the past, despite how this all sounds.

edit- I found a schematic- so I can answer specific questions- /edit

My boss (me) is silently screaming at me for even thinking this, but I am damn tempted to revisit my old TS-1 using modern technology. I can now do everything the TS-1 did, plus a bunch more for a few dollars worth of digital silicon... I think Gold Line is still selling the old TS-1 design for some obscene amount.

I'm thinking not only could I cover the the sine wave gen, freq counter, DB meter, but i could add a cheap mic and do crude SPL, speaker polarity, I could even do a FFT with $5 DSP chip... I could measure THD. IMD, reactive impedance , etc ad nauseum... I am very tempted to KISS,,, not aim for silly high bench performance, but target something similar to old TS-1 level performance but with tons of features...

But like i said I need to work on my day job,, right now I just figured out to make a dual mirror image PWM output so I can cancel out the PWM fundamental at AC zero (at zero AC both PWM outputs are opposite polarity square wave so sum to zero PWM period fundamental before filtering, instead of 100%. Still needs to be filtered but interesting.

JR
Last edited by JR. on Tue Apr 07, 2009 10:00 am, edited 1 time in total.
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: NR compander (P-522)

Post by mediatechnology »

John - I think I have a copy of the scan you sent that I could turn into a pdf for the TS-1.

I also want to re-size your NR image. It may take me a couple of days 'cause there's some stuff goin' on here.

Are you familiar with the NTI line? http://www.nti-audio.com/
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: NR compander (P-522)

Post by JR. »

mediatechnology wrote:John - I think I have a copy of the scan you sent that I could turn into a pdf for the TS-1.

I also want to re-size your NR image. It may take me a couple of days 'cause there's some stuff goin' on here.

Are you familiar with the NTI line? http://www.nti-audio.com/
I forgot.. or haven't been paying attention.. There's also the Phonic for less money, and they are all on second or third generations by now.. No need to go into a market with competitors when I own the one I'm in, for now.. New to be making hay while that sun shines..

JR

PS: no worries, that's why I also posted a link to image,..
Cancel the "cancel culture", do not support mob hatred.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: NR compander (P-522)

Post by JR. »

raf wrote:Cool John

I have been curious about this since you referred to it recently... I did the simple version in 81 IIRC and it worked pretty darn well if you were careful with levels.
I coupled it with a variable filter NR project that I built and recorded a lot of FM off air stuff with that combo back then.

Anyway, thanks for posting this.

Ps, always wanted a TS-1 but never had the cash back then.
Yup the original NR was OK, but not in the same league with later one, but I missed my market window... cest la vie.

JR
Cancel the "cancel culture", do not support mob hatred.
Crusty
Posts: 69
Joined: Mon Aug 20, 2007 1:42 am

Re: NR compander (P-522)

Post by Crusty »

Dammit, now I'm mulling mirrored pwm signals...man, what's that for? Building synthesizers now? LOL

As far as test gear, that basic stuff is covered. More interesting would maybe be something in between that and a full-blown AP rig. Maybe the next step between it and an AP is a high quality sound card and software, but I wouldn't like having a pc running in the shop all the time. Multiple channels and automated testing would be cool though...

I have a box made by Terrasonde http://www.terrasonde.com/ The basic package is geared toward acoustics mostly. I have a downloaded upgrade called Test Bench that seems to be mostly an afterthought. It has level, phase, amp. sweep and distortion measurements, though nothing particularly refined about it. My old Flukes are more versatile measuring level. The signal generator of the Terrasonde puts out up to 25dB which is why I bought the color version - but the sine has a hideous noise on it that is visable but considered "out of band" (as they kept telling me). Apparently normal for these things. Definitely many things I would change if building a real test bench box but where do you draw the line?
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: NR compander (P-522)

Post by JR. »

That Terrasonde is now made and sold under Sencore Brand.. as far as I can tell $3000 + more for specialized software. The NTI box sold by Neutrik that Wayne mentioned is more reasonable at $500 or so, and then there's a hand held from Phonic for maybe $400.

Admittedly the TS-1 was designed in a simpler time when you couldn't do 99% of everything with a decent computer sound card. I should not waste time on this while I still feel there is a place for something that covers the TS-1 functions + a little something extra for <Phonic price..

It appears there are some nice codecs with 2x 16b d/a with 1x 12b A/D with some variable gain and interface for inexpensive microphone. Apparently targeted for cellphone apps but I could combine this with cheap DSP chip and make an interesting box.. I recall my old school bench spectrum analyser with only 50dB dynamic range and it was useful. This could do 60dB+.

Oh well.. I need to work on my work... regarding the mirrored PWM, that is probably not the real name for it.. when self taught and figuring stuff out for yourself, this happens. The basic problem when using a digital PWM to make audio is the trade off for dynamic range or resolution which is limited by the number of clock ticks in PWM base period. But as you increase the number of ticks to get resolution you push the base period lower and down into audibility. What I am trying to address is the energy that must be filtered out at AC zero or 50% duty cycle. In this case you effectively have a full amplitude square wave at the PWM period that must be completely filtered out. My approach is to make a second PWM running at the same clock period but opposite phase (or polarity) and combine them before filtering. At AC zero there are two identical but opposite square waves that cancel each other out, as the signal increases or decreases from there the filter sees just the number of ticks positive or negative..

In addition to canceling the base square wave, the second PWM stream increase effective dynamic range 1 bit (6 dB). It appears that the PWM period will still be there in full force at signal maxima and minima, so I need to build this and listen to see if it actually makes an audible improvement. My gut feeling is that it will, but I have been wrong before.

I am mainly trying to save the cost of adding a D/A since the PWMs are built into the basic micro so effectively free. Since I'll still only be getting 8b or so dynamic range (at best) I could drop in a cheap digital pot and be done... beyond 8b the next step cost effectively is those cellphone codecs. So if I have to start communicating with the digital pot, for less than a dollar more I could get 16B.. tempting. And now my tuner platform could be repurposed as a test set... somebody stop me... :lol: My boss is yelling at me (in my brain).

JR
Cancel the "cancel culture", do not support mob hatred.
Post Reply