LM13600/13700 compressor

[tv]

Hi

I'm new here and I would like to ask some questions regarding construction of a compressor/limiter based on LM13700 (I have two chips) and easy obtainable low-cost components (NE5532, TL084, BF245/256, BC550...), my finances are limited and would need the circuit as a DAW complement, and make is as fine as possible. I've made up my mind about input/output stages, which seem a bit too much, but I'll use it with computer-card ADAC which is suboptimal (jack-type, balanced), so I read jensen papers on servo I/O circuits and they don't seem cost-prohibitive (won't use transformers).

What I need is a gain-control cell and sidechain circuit with attack/release (maybe autorelease). I suspect the best way to get best performance would be to use 13700 as a "variable resistor" in NE5534(2) feedback path.

I know LM13600/13700 aren't the greatest but 13600 was/is used in "pro" gear (e.g. Tl Audio C1 compressor), and I have seen some limiter schematics (for PA) which use it as a gain controling element in low-noise opamp (NE5534) feedback loop.

I have seen some sidechain circuits here, and there's one below for 13700 (in german - I think the sidechain HPF filter could be easily moded to variable SK topology with additional C5 and dual 1M pot), but I don't know if they could be adapted for use with LM13700 as "variable resistor".

I'd also like to add a "key listen" option, but don't know if CMOS switches are any good for such application - would be LM604 better for audio use?

I have prior discrete audio-building experience, but haven't done anything in last 8+ years.

Please, help !
(some advice, maybe somebody has a good, working schematic to take a look at?)


what I looked at before posting here:
--------------------------------------------------------------------
limiter with lm13700:
http://www.elektronik-kompendium.de/public/schaerer/otalim.htm
http://www.elektronik-kompendium.de/public/schaerer/bilder/otalim03.gif

that 4301 control circuits:
http://www.thatcorp.com/datashts/dn115.pdf
http://www.thatcorp.com/datashts/dn118.pdf
http://www.thatcorp.com/datashts/dn125.pdf
(I don't really understand how exactly these sidechains work, but descriptions seem cool to me - I like the looks of dn118 fet circuit)

tl audio c1 (lm13600)
http://www.soundonsound.com/sos/mar99/articles/tlaudio.htm

limiters in power amplifiers:
http://www.yorkville.com/downloads/servman/smcr5.pdf
http://www.ofgb.org/reference/Music/Qsc/Service%20Bulletins/RMX%20service%20manual.pdf

jensen balanced input/output:
http://www.jensen-transformers.com/an/an003.pdf
http://www.jensen-transformers.com/as/as030.pdf
(I would replace LM11 with TL08x and capacitors on output - i.e. w/o transformer)
http://www.jensen-transformers.com/an/ingenaes.pdf
(figure 10 - God knows what comes out of computer card)

 

[jhaible]

Honestly, forget the 13600 / 13700 for any serious studio application.

I have used it, quite often, but only to build a compander around circuits which are noisy anyway (BBD delays, or frequency shifters, for instance). But even in such an application, you want *at least* put the OTA into the feedback loop of an opamp, to keep its (the OTA's) input level fairly constant. But even then, the CV feedthru of these chip is not nice.
Better go with a FET, a Vactrol, a dBx VCA, or - if you're determined to use an OTA, the CA3280 or better build your own around a good transistor pair.

IMO, OTAs are great for music synthesizers, but no good for pro audio.

JH.
(Who once, in the blindness of youth, and misguided by a one-eyed publication, built a single ended "noise reduction" around a 13600, which created more noise than it filtered ... )

 

[tv]

Thanks

Yes, I wanted to put the otas into a feedback loop. I just don't know how to get the right sidechain for it.

My budget is prohibitive to use more dedicated components atm. and I'd like to make a good circuit with 13700s which I already have. TLAudio seem like they did just that.

Fets would be nicer, but I suppose I'd have to have a measurement equipment available to calibrate the circuits - which I don't. Otoh, otas are supposed to have very good stereo tracking, that's why I'd like to use them instead.

Any help?

 

[bcarso]

I concur with JH---they are fine parts for really rough work. They have however lousy noise and offsets and really have to be coaxed to get even 90 dB range out (I suspect a record at that). I have used them extensively in cheap multimedia speakers and even as forward path gain control elements. But much better for compression/limiting is as a feedback element, since at zero compression they are effectively out of the circuit.

If you drive them differentially they can have greatly reduced distortion over any of the datasheet configurations. But they will still be noisy. There 's a reason why THAT parts cost some money---they are big suckers and you don't get all that many off a wafer, for one thing.

 

[tv]

Thanks

I had a look at some measurements (googled) for otas, i.e.:
http://hem.bredband.net/bersyn/vca.htm
with corresponding schematics.

Seems like they improved on a datasheet applications, I suppose 13700 could be a little better in figures.

You suggested otas being fed differentially - did you mean they should be fed via a balanced buffer? Like 1968 fet compressor stage in the other thread? A totally cheap circuit would be if I fed the balanced input straight into ota and used it as a balanced-to-unbalanced stage. Would that work?

But I'm sure there must be a way to get much higher performance from these otas - If someone could look into pdf-s I've posted above and the TLAudio C1 review (it uses LM13600 as gain-control element). In PA limiters (see pdfs) otas are paralleled as "variable resistor" in opamp FEEDBACK path. I suppose this must lower the THD and noise figures very much, but havent done anything on a protoboard yet.

edit: how could I use the "true RMS converter" (circuit 27) from datasheet in a functional sidechain (said I have 2 chips)?

Any further insight?

 

[bcarso]

The key to performance out of the 13700 is proper use of the linearizing diodes (really diode-connected transistors of the same type as the input pair).

I like to tie the diode bias about as high as possible while still leaving room for the current mirrors to work---say about two diode drops below Vcc.

Bias the inputs with an identical pulldown current on each. At these impedances a simple pair of matched resistors to the negative rail is sufficient.

The run differential signal currents into each input base/bias diode cathode. Again, not much is gained by making a current drive per se--- a pair of resistors will work, unless you are trying to work at very low supply rails. You can a.c. couple to the input resistors if you don't need response to d.c., which is usually the case.

So this is really a differential current in, single-ended current out configuration. Keep any and all currents, peak, below a mA---the transistors in these parts get unhappy with more.

Use as a feedback element with the output current going to the inverting input of an op amp. A fixed feedback R for the op amp limits the maximum gain when the OTA is shut off (no "Iabc" current). Throw in another unity-gain inverter to generate the feed for the other OTA input's resistor. You will need a little feedback C for the main op amp to keep things from screaming.

 

[tv]

Thanks. Really appreciated.

I understand the part of fbloop inclusion with unity inverter. But I don't know how to deal with the linearizing diodes. Do you have any picture? Or is there something I overlooked in datasheet? I'm little lost here.

-------------
Another Q: In theory, I could use ONE 13700 with minimum parts (ONE 5532 to drive heaphones and what's needed for a sidechain - I suspect a single TL072 plus both darlingtons in OTA would be available) to construct a simplistic comp/limiter for headphone amp which would be driven directly from BALANCED jack/xlr? I.e. signal directly to +/- inputs of LM13700 via resitor network? And ota output directly into inverting opamp stage to drive headphones? Sort of like here, but with balanced input:
http://hem.bredband.net/bersyn/VCA/lm13600%20vca%202.htm

except, to have correct phase on output, I'd need to route + from XLR to - input of ota and - from XLR to + input of ota. Am I right?

Furthermore, can be ota inputs shunted with caps to ground? Or will that yield oscillations? Would it be better to include a zobel between + and - inputs of ota?

 

[bcarso]

[quote author="tv"]Thanks. Really appreciated.

I understand the part of fbloop inclusion with unity inverter. But I don't know how to deal with the linearizing diodes. Do you have any picture? Or is there something I overlooked in datasheet? I'm little lost here.

-------------
?[/quote]

I haven't seen the precise approach I outlined in anyone's datasheets. I think the app guys at National ( and before that, at RCA) despaired of anything requiring much in the way of external components, which is odd.

PM me here with an email addy and I'll sketch something and transmit.

 

[PRR]

Bild 3 on this page appears to be a reasonably refined limiter. Input at Ue, output at Ua. The input could be differential if you duplicated C1 R1 R2 into the other input pin of the OTA. The output is reasonably low impedance unbalanced; if you gotta drive headphones or 600 ohms, you add a 5532; if you really need balanced output, add two opamps.

In general: the 13700 is a dandy one-dollar chip, good for many things. But the noise is limited by small current in the 1-cent transistors, and the distortion soars as you approach 20mV. As a forward-path gain cell, 76dB dynamic range is fabulous for such a topology, 60dB (1uV-1mV) is more typical. As a feedback path, you may dodge input overload and it may be "dead" at zero limiting, but there is always some point in the range where noise rises. It works for PA-amp protection limiters; I can't imagine using it in a critical listening path without a TON of lily-gilding. (It appears bcarso has gilded that lily, perhaps for some commercial work where a penny saved is a kilo-buck of profit and worth some brain-strain.)

What dBx/THAT gain cells add, beside a dime's worth of transistor area (lower noise), is a current-steering topology that can either increase its own output or cancel its own input. This push-me/pull-you scheme is an order of magnitude better. Moreover, unlike National, THAT corp is an audio company, and has published good working plans for high-performance studio limiters. Some of National's audio application notes would, pardon me, be good for a singing toaster or other kitchen appliance. If only for the convenience of stealing a known-good design, I'd say the few bucks it costs to get a THAT chip is money well spent.

 

[bcarso]

quote: (It appears bcarso has gilded that lily, perhaps for some commercial work where a penny saved is a kilo-buck of profit and worth some brain-strain.)

The application was a multimedia-oriented 4 satellite + subwoofer ("4.1") powered speaker system, that required a volume control on the right front satellite. To avoid both a microcontroller/shaft encoder setup along with digitally-controlled step attentuators, or a big fat unreliable cable/connector combo and a four-gang pot in the satellite, I made a control voltage out of the d.c. level on the single-ended amp output to the satellite, putting the speaker coupling cap inside the sat. It also powered an LED in the sat (this was done with three wires total IIRC). Then the a.c. riding on the d.c. from the potentiometer wiper was lowpassed out at the central subwoofer, and the scaled and offset voltage controlled 4 VCA's made from LM13700's. To get more range I controlled both the normal bias current that sets OTA transconductance, and as well the linearizing diode currents, one giving way to the other at a breakpoint. Differential current drive of the OTA inputs was essential for acceptable distortion.

The product didn't do well although those who bought it praised it profusely. The main problem was that at the time it was introduced, many were migrating to 5.1 systems.

 

[SSLtech]

[quote author="bcarso"]The application was a multimedia-oriented 4 satellite + subwoofer ("4.1") powered speaker system... blah, blah blah...[/quote]

Damn!

-A singing toaster which harmonises to the housewife making breakfast sounds easy by comparison!

:wink:

Keef

 

[bcarso]

The singing toaster might have sold better too....

 

[Rob Flinn]

20 years ago I built a stereo compressor from a design published in the UK magazine Home & Studio recording. This used the NE 572 (I think) which I think was still quite cheap & an improved version of the NE571. As far as I`m aware the NE572 is much better than the 13600/700.


I also built compressors using the 13700 with a design from the same magazine. This was part of a series of modules that plugged into a rack & was marketed undr the Tantek brand name. I still have the boards somewhere, but have pillaged them for parts. I do seem to remember them being rather noisy.

 

[Rob Flinn]

The singing toaster might have sold better too....

Yes, but a toasted singer would have been more funny to watch !!

 

[tv]

Thanks to all

@bcarso: PM sent.

 

[SonsOfThunder]

Analog Devices SSM2164. Four nice VCAs in one package. Nuff said.

HTH!
Charlie

 

[jdiamantis]

If you use the LM13700 as a "variable resistor" (check the app notes) you can build a processor similar in dynamic response to a fet type limiter, with repeatable results. CRL (now part of Orban) used similar circuits with these chips with decent results. I built (and sold) a couple of processors using the RCA CA3082A chip, which was a better part both noise-wise and distortion wise, but to the best of my knowledge, is no longer available.

 

[henk]

In the Harrison K series (2000,3000,4000,5000 and 10000) amps the all had an lm13600 in front to use it as an limiter/compressor.
And the worked great , never had trouble with these big amps.