timing network in variable mu - hi impedance vs. low impedance

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

hop.sing

Well-known member
Joined
Jun 21, 2005
Messages
160
Hi

I have a question concerning the timing network in a variable mu compressor. I posted it in the "Another poor mans fairchild" thread, but got no response, maybe it is better suited here.
What is the difference, soundwise, from a network like in the fairchild, or the network rotheu posted, and a high impedance network like in prrs "very poor mans" sidechain. I am sure there is some distortion happening in a fairchild, but are there other reasons for different sounding compression?
I did read everything I found from prr, but he mostly covers the technical side of the problem (I learned alot from him!). But what about the sound, if technically the hi impedance sidechain is no problem.
I built the prr sidechain and  can get 15dB (ca -10V on the grids in my design) compression on two parallel pcc189, and it can be really fast with a 0.2 mS attack. The only modification on my design is a 5532 buffer after the timing network to keep the grid resistance low enough for the 4 triodes. But maybe the buffer helps also with the ability to get fast gainreduction, I did not try without. But the 5532 is defintly able to drive 4 grids into -11V very fast.
The compressor sounds great.
The design consists of an input transformer into a stepped atty into two pcc189 in parallel (four tubes for stereo) into a lundhal 1660 configured as 2.25:1 output, no buffer. The comp can drive loads down to at least 1k without a problem.
If I do not need more than 15 dB of compression, is there still a reason for a power amp sidechain, solid state or tube?
Is there a difference in compression characteristics and/or sound?
The only comparison I have is the 670 emulation from IK multimedia and I get very similar compression curves with my sidechain, but I like my comp much better overall than the software (I should, shouldn't I?).
Thanks
Tobias

Tobias
 
> a network like in the fairchild, ....., and a high impedance network like in prrs

I do not understand.

Fairchild used 600 ohm winding. The TL072 can make 10V at 20mA, or a 500 ohm load. There are other differences, but they are NOT very far apart, except that all voltages (signal and control and power) are scaled lower.

Absolute impedance never matters. It is all about impedance relationships. Your explicit impedances, and the ones you can't do anthing about (tube impedance, stray capacitance). A fast attack slow release simple network requires a low driving impedance and a high load impedances. Tube grids are exquisitely high impedance. Chips give lo-Z cheaply. The general impedance relationships can be easily met.

> the 5532 is defintly able to drive 4 grids into -11V very fast.

? ? ?

A grid is a small bit of metal in an empty bottle. It is EASY to "drive very fast". In an audio limiter we "must" slow it down.

The grid is >200Megs||50pFd. (Or >50Meg||200pFd for four.)

The timing network is typically 100K-1Meg and 0.5uFd (500,000 pFd).

If you think the grids are slowing you down, change to 980K and 0.499,8uFd. (No difference).

> the 5532...

.... has a large DC bias current. With typical 1Meg timing resistor, this gives a large DC offset voltage.

> I did not try without.

Maybe you should? It seems a simple experiment.

> What is the difference, soundwise, ... the fairchild

The Fairchild has EIGHT fat hot triodes which can directly deliver +20dBm straight and +16dBm while limiting without gross strain. If you use any less, you must have a post-amplifier. Once you have a post-amplifier, there is no reason to strain your variable stage or to use a basket of tubes; fix it in the post-amp. No, it won't be quite the same.

> I get very similar compression curves with my sidechain

If the curve isn't similar, it isn't a limiter.

> The only comparison I have is the 670 emulation from IK multimedia

How do we know the emulation is anything like the original?

And don't forget: it is the music not the machinery.
 
Hi PRR

Thank you for the response. As you see, I am a beginner in electronics but willing to learn.

Absolute impedance never matters. It is all about impedance relationships. Your explicit impedances, and the ones you can't do anthing about (tube impedance, stray capacitance). A fast attack slow release simple network requires a low driving impedance and a high load impedances. Tube grids are exquisitely high impedance. Chips give lo-Z cheaply. The general impedance relationships can be easily met.
I understand that tube grids are very high imedance and that a tl072 has no problem to drive them within the chips voltage limits and if you need more voltage, you could use a opa2604 or another +-24 voltage chip.
But that is the reason I always wondered about the need for a beefy sidechain, if you avoid release resistors smaller than 1000 ohm and do not need more the -18v, you can always use a chip to drive it.
I guess the original fairchild is the only compressor that needs more negative control voltage.
That is the timing network that made me wondering:


rotheu said:
Timing network with independent attack and release controls and true stereo linking (two linked networks are isolated and don't affect each other)

Fairchild660ModTiming.jpg

Why so complicated? Is there any benefit from a design like this? Just raising the series restance before the timing cap seems to be much easier than to change the value of the cap and than having to drive a 150 ohm release resistor with a slow attack/fast release.

The 5532...
.... has a large DC bias current. With typical 1Meg timing resistor, this gives a large DC offset voltage.
The offset voltage was not that bad in my application  (0.05V), but I tried without and it works just great, thanks!
But now the 4 grids have to deal with a max impedance of 2.2 meg with a long release, but the tubes  seem to be happy.
The Fairchild has EIGHT fat hot triodes which can directly deliver +20dBm straight and +16dBm while limiting without gross strain. If you use any less, you must have a post-amplifier. Once you have a post-amplifier, there is no reason to strain your variable stage or to use a basket of tubes; fix it in the post-amp. No, it won't be quite the same.

I tried without a post-amp during trouble shooting and it worked just fine.
I use two pcc189 paralleled and use the lundahl 2,25:1 highZ line transformer (used 1:1 as interstage before) and now I go straight to the output, without a buffer amp. Sounds somehow better to my ears.
I use the tubes well within their limits 105V/9 mA.
I did some math and the plate resistance in heavy limiting gets really high: 3k no limiting, which is fine, but something like 15k or more with -5V on the grid, which should not work so well any more.
But still... the four triodes work well even under those conditions, probably because they do not have to drive 600 ohms, but rather 10k in my studio and my levels are usualy pretty moderate.

Thanks for help and a happy new year!
Tobias


 
> Why so complicated?

The drawing is complicated. But at any one time the switches are in a cerain position, so the circuit is always simple:
vpjgx2.jpg


> having to drive a 150 ohm release resistor

I have no idea why he did it this way. I had to poke it to even accept that it would work, but it does.

I note that rotheu explains his thinking in the other thread; "in case of Fairchild, I think having resistance in series with CV (attack potentiometer) will change the character of the compression and the way it behaves."

You need series and shunt resistors (or other way to control time). The ratio of resistors is the ratio of attack to decay. In many audio situations, this needs to be a large ratio.

You have two limits on resistors. How big and low-impedance an amplifier you want to use, and how large your Mu-tube grid resistors may be.

The 660 runs its multiple Mu-tubes at very high power, so must use a fairly small maximum grid resistor, OTOO 200K? When we work the 12AU7 at 100V it does not run real hot, and even if it drifted it would be in no danger. While book-spec is 1Meg per grid (500K for two tubes), I'm quite sure 1Meg/2tubes is safe, and we could go higher (but don't need to).

This leads to the classic design with one part-uFd cap, a 50K-1Meg variable decay, and 1K-50K variable attack.

In the 660, the budget was tighter, so there's at least two main caps to keep the resistor values reasonable.

Instead of basing the design on the maximum allowed impedance, rothew has based design on the minimum impedance, the ~~150 ohms out of the sidechain power amplifier. At 330uFd, which is a heavy audio load against 150 ohms, it is not even an R-C network any more: the sidechain amplifier sees a near-short. We must ask what the maximum current is, and use Slew Rate formula to find the rate of change of voltage on the cap. I estimated a couple 6V6 at 150 ohms could drop 0.25A, which will slew 330uFd to 50V in 66mS-- near-enough the 33mS marked on the knob.

So it "works". It seems like tube-abuse, but I don't see how the 6V6 can possibly be harmed. While slewing, the sidechain amplifier NFB is ineffective, but I don't see a problem there. As the cap voltage approaches final value, the slew-limit changes to R-C action, and the NFB begins to take hold. Usually an amp wobbles as it comes out of saturation, enough to mess-up a 8-bit ADC conversion, but probably not enough to matter for slamming 50V grids.

I would not do it that way, but every designer is different.

The ramp is linear, not exponential. That's different (and affects how you interpret "time contact", but I don't think it is inherentlay better or worse. For "transparent" limiting you will be in attack-mode for VERY short times, and you could not hear a difference. In transmitter protection where time of overmodulated signal influences complaints from other stations and license authorities, an exponential may get you out of trouble quicker.

My only "solid" objection is that instead of depending on the well defined impedance of a resistor to set the attack, we now depend on the maximum saturated tube current to set the attack. Through the 1950s when tube production was "rational", you could be sure you would not get a lot more than spec current, and good fatories would reject low-current tubes. Today's tubes may be much more variable. In the 660 plan, you could trim this with screen voltage.

 
> I tried without a post-amp ... use the lundahl 2,25:1 highZ line transformer

The 660 was aimed at matched 600/150 impedances, old-time studio-gear.

10K loads are 16 times easier to drive, you can use 16 times less post-amp. Indeed you can eliminate the post-amp if your mu-stage is fairly strong and your loads are always light.

I jjust have this thing about always offering a <1K output even when not needed. I have had to drive a dozen 10K loads, and it is nice to not worry about loading.
 

Latest posts

Back
Top