any love for a new THAT comp discussion?

[JohnRoberts]

The 4301 seems to have so many different things that it can do.  And there are a ton of compressors around here.  And I think there are still Pico-type schematics floating around online.

Sorry to try to steer the thread differently, but why not design some kind of gate or dynamic noise reducer into a nice little package (maybe 500 series, maybe not - I don't think anything like that exists for that format).  I've got an ancient Burwen VCA-based noise reduction unit and would love to have more. 

I'm just saying there are a lot more possibilities.

I recall the Burwen, (and dynafex?), and sundry other single ended NR approaches over the years.

I designed a (dbx) VCA based single ended NR that was a combination of sliding LPF and downward expander, so the both of them working together gave a lot more apparent NR with less audible artifacts from each technique alone. It took 2 VCAs per channel, one to make the sliding filter and one for the downward expander, but it sounded good (to me)... Unfortunately it was a little too cool for the room (Peavey"s AMR distribution) that wasn't a good channel to push +4 studio gear through, so it didn't sell very well.  :'(

FWIW this particular product wouldn't work with my generic two PCB approach, since the standard VCA path could only handle the downward expander half, not the sliding LPF.

There are other interesting single ended NR technologies, but they are too much of a veer for simple VCA projects.

JR

 

[gemini86]

Re the question about posting Pico schematics this may help
http://waynekirkwood.net/forum/php/viewtopic.php?f=7&t=300&sid=18ce537c9ddc1d3350c334e7b849971a

That's exactly why I said that. Somebody put a lot of time into the pico, gotta respect that.

 

[Biasrocks]

You'd think it was the holy grail, or is that sacred cow?

While I respect the fact that there's good reasons to protect a commercial product - "$$". Why he's worried about a previous diy version, which is completely different from the now commercial version is beyond me.

We're talking about a 3 chip THAT Corp based compressor, Using the 4301p analog engine (VCA/RMS detector) along with a few resistors, capacitors and controls added, culled from the design notes with a few small changes. Anyone with access to the schematics can clearly see that's the case. There's nothing innovative or ground shaking going on in that circuit.

I say have at it.

Mark

 

[gemini86]

I started drawing up a scheme last night. Pretty cut n paste but it's only purpose is to start routing a board. I think johnroberts two board idea is great. I don't want to keep repopulating a whole compressor every time we make a sidechain revision. It would be cool to fit this into 500 format if possible, but one step at a time.

 

[StephenGiles]

I would like to see the trick used in the Aphex Compellor (in circuit form to use with a THAT compressor chip) to hold the amount of compression until the input varies more than a preset amount, in order to minimise noise and pumping - I'm probably over simplifying though!!

 

[living sounds]

Legality aside, taking something that a person doesn't want to give away is not very gracious.

I guess it comes down to personal philosophy, my participation here is mostly to help people who want to learn, not trying to get something for nothing.

JR

But he originally put it there, didn't he? I'm not talking about any current commercial product, but what was formerly a public domain DIY project. That's a different issue IMO.

 

[buschfsu]

Id like to see this for mastering so a stereo version would be prefered. As for the pico didnt he forego attack and release controls in favor of auto/rms/peak. Id really like to have manual sidechain control even if its a bit more trouble

 

[JohnRoberts]

I would like to see the trick used in the Aphex Compellor (in circuit form to use with a THAT compressor chip) to hold the amount of compression until the input varies more than a preset amount, in order to minimise noise and pumping - I'm probably over simplifying though!!

While it would be easy to literally do that with a digital side chain, the common technique to get that in analog involves a ripple filter that is made from an opamp capacitance multiplier that is clamped, so for small ripple currents the opamp is active and can make the capacitance look large for small ripple. When the amount of current exceeds the preset threshold, the opamps can no longer correct, and the capacitance looks like it's real, much smaller value. So in use, this sounds like a very long slow time constant for small changes, but fast small time constant for large changes. You can even get fancy and limit the ripple filter thresholds for different amounts of attack or release current.

I published a pretty complicated variant on this level dependent cap multiplier to a different thread here a few years ago. I used in a tape NR kit back in the '80s, but there is already a level dependent ripple filter in the schematic in the OPs first post. Look at U3B near the middle of the cut and paste schematic.

The - input of the opamp is connected to the bottom of the RMS detector smoothing cap. Normally the bottom of the cap would just connect to ground, but now current flowing into the top of the cap, also goes into the opamp inverting input which uses negative feedback to cancel that current out. In effect making the cap look much larger than it really is. All good, but the two back-to-back LEDS across the feedback path of the opamp means at about 2V in either direction the opamp clamps and can no longer make the cap look big, and it reverts to it's nominal value.  Note: the current threshold is roughly 2V/2M  or +/- 1 uA. This circuit is just a little extra trick smoothing on the rms detector, but it demonstrates the concept of level dependent  time constants.

JR

 

[StephenGiles]

Thanks for that John, we are flying from London to Buenos Aires overnight this Saturday - this will help to pass the time (14 hour flight!)

 

[JohnRoberts]

Legality aside, taking something that a person doesn't want to give away is not very gracious.

I guess it comes down to personal philosophy, my participation here is mostly to help people who want to learn, not trying to get something for nothing.

JR

But he originally put it there, didn't he? I'm not talking about any current commercial product, but what was formerly a public domain DIY project. That's a different issue IMO.

I repeat this isn't about the legal status of Roger's IP, or his intent at any point in time. I suspect if folks are really compelled to, they can find footprints scattered around the WWW...  I see no reason to open up that can of worms for this thread, here and now.

Any folks who just want to clone a pico, or any other existing compressor, don't expect me to play along, or support that. My understanding was that this thread is about a fresh design based on THAT app notes and shared input. If you want something just like an XYZ, why not just get an XYZ?

Back in the day I had to roll my own VCAs and log conversions... What's left to design after using the THAT chips is the relatively easy part, but still hard enough to be interesting.

The main reason I see to make a compressor from scratch these days, is to make something different you can't buy for any price. Hopefully it won't suck.

JR

 

 

[gemini86]

This Pico thing has been beaten to death before. The op was intending to start a new thread on a project based on the app notes compressor. (I also expressed interest in rekindling the thatcomp flame in another thread.) Can we just leave the Pico alone now? If we must debate the nature of internet public domain designs, lets start a thread in the brewery.

JR, your experience in this area is invaluable to all of us. We really do appreciate your time and hope you chime in as often as you're willing/able to. What, if anything, would you change on the basic app note compressor as it sits now?

 

[JohnRoberts]

What, if anything, would you change on the basic app note compressor as it sits now?

I don't know? Nothing, everything, it depends... I already mentioned I'm not building my ideal compressor because I don't have all the tools I need (PC/MAC interface, and motivation.. I don't actually use a compressor).

I would suggest that you guys, use that app note as a starting point, build it up and mess with it. Listen to it a lot and figure out what it does well and what it doesn't.

Maybe tweak values in all the different circuit blocks to see what they do and understand where they should end up.

One thing I used when designing dynamics circuitry was a tone burst generator.. actually a glorified gate that took a clean sine wave input and diced it up as desired (pan between burst and dry, with variable on/off times). Tone bursts are IMO extremely useful when designing dynamics processors, because you can put in a perfectly chopped tone burst, and listen to how mangled it gets by the compressor. With music as a source you could listen for a long time and not hear the weaknesses you can parse out in a few minutes with a difficult tone burst.  (I also published a circuit for that here too some time ago. A few cmos parts and JFET gate.)

In my experience there is a lot of bench time involved getting a dynamics processor from OK to really good. The good news is you are pretty much starting out with OK, using a good chip set and app notes. But you will need to decide for yourselves what really good means to you.

JR

 

[myker]

Ok, I found the gerbers and associated info for the THAT design PCB's that I had made if you guys want to take a look at them. Its 3 or 4 separate pcbs on one pcb. There were a few things that I didn't really like about the PCB's, some holes were too small, etc. If you guys give me your e-mail i will send em' your way for a look.
-mike

 

[gemini86]

my email is listed on my account, just click the mail button under my screen name.

 

[buschfsu]

My intention was NOT to clone a pico. As a matter of fact the DIY version I know has only auto att/rel. I'd like to use these great parts for something new that maybe meets my mastering needs for smooth transparent stereo but I'm open to other ideas. Really want to understand and explore side chain cv possibilities

 

[ruairioflaherty]

Ok, I found the gerbers and associated info for the THAT design PCB's that I had made if you guys want to take a look at them.
-mike

Hey Mike,

Shoot me a mail at ruairioflaherty (at) eircom (dot) net

As I say I'd like to use it as a base for some experimentation on another idea I have.

Cheers,
Ruairi

 

[JohnRoberts]

Here is a schematic I scratched out for my tone burst gate... caveat... this is not the actual schematic of my old units so not proved out, but close enough, it should work with minor tweaks. Another nice test technique, you can run audio through the burst gate, to hype up the transient or dynamic content of music for testing how the compressor reacts to transients.

There have been several interesting threads over on Wayne's forum about dynamics processing design considerations. You might discover that I have mixed feelings about the merit of RMS.  The image above is something I scratched out to make the smoothing cap used on a THAT RMS detector "variable"... So you could literally adjust the apparent cap size, and therefore speed of both attack and release with one knob. This is a variation on the cap multiplier.  This is also something I never checked out so while I believe it will work, I don't know how desirable the result.

http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=61&start=30

One of several threads that could be researched to learn more about dynamics processing.

JR

 

[myker]

it would be cool to come up with a compressor using that chips exclusively, that could be used for mastering.

 

[buschfsu]

That's where my head is at. But if it sounds good it wouldn't be hard to put into a 500series mono also.

 

[ruairioflaherty]

John,

The burst gate looks to be a very useful tool for this kind of development work, I recall you mentioning it before and I hope to get time to construct one.  I love the idea of the cap multiplier circuit to vary the apparent cap size and RMS timings.  One other notion I like is the Massenburg/MacNeill concept of multiple RMS timing paths combined with a peak sensing path and all combining to provide a more accurate read of the dynamic information in the signal.  I need to get back to the 4301 datasheets and get a refresh but I'm trying desperately to stay focused on one project at a time, for a beginner like me it takes a long time to absorb stuff..

Cheers,
Ruairi