Dynamics before or after RiAA?

[karl hungus]

Hello all

I’m redoing my mastering chain to be mostly analog, except for the deessing, and am curious about other people’s signal paths and reasoning.

If you had an outboard EQ, and/or RiAA box, to do your RiAA curve and cutterhead resonance duties, would you have that as your last piece before your amps, or would you put a limiter after that to protect your head?
I’m on the fence as my initial inclination is to use the limiter as a final safety, but it also seems odd to squash or impact that high top end arc of the custom RiAA/anti-resonance curve I’m using. In some ways it seems more logical that curve be the final bit of processing, but have any of you come to regret that decision with a weird transient blowing your head or anything?

If anyone feels like elborating on their whole chain, particularly the order of things, that would also be very informative.

I tend to prefer:
Elliptical EQ plugin for mono bass-> surgical EQ plugin if needed-> deessing plugin-> analog compressor -> analog EQ(s) -> analog limiter -> final RiAA curve (or perhaps not)

Cheers

[Jccc]

For my signal chain i usually go:

Music source - mixing board / Eq - RIAA box - Cutting Head Amplifier - Cutting head.

i have a fuse inline with the cutting amplifier to cutting head that protects the head incase a loud signal ever hits it.

But my set up is all mono and im not cutting masters or anything.
It seems to work for me

[markrob]

My two cents is that the limiter should be at the end of the chain. The RIAA adds more treble boost and you want to kill the short duration low energy fast spikes that result in the need for tons of extra power reserve just to let those peaks through without clipping your power amp. By this point in the chain, you have a ton of high frequency boost applied to flatten the head response along with the RIAA. For a moving coil head, maybe as much as +30-40db at 15-20 Khz. I've measured crest factors of 25db (average program RMS to peak ratio) or more at the end of the chain depending on the source material. Having a fast brick wall limiter at the end of the chain allows you to increase the average power to the head for a given peak power reserve. You do have to be careful not to overdo the amount of limiting or it will cause audible results. Depending on the source material, you may not be able to apply much.

Mark

[karl hungus]

Interesting replies, thanks.

It sounds like its a good idea to control that energy despite the risk of squash. Markrob, but one concern is the potential risk of changing the whole curve by limiting it. If the treble curve is being limited, it seems that one might be driving up the lower register too, and changing the ratio due to the squash.
Have you run into issues flattening the curve, or do you mean using the limiter where its not effecting the material with any compression at all?

[markrob]

On the program material I've been playing with, you are normally only lopping off really short duration high frequency transients (e.g. a snare hit). No compression is used at all. If you pre-process the audio with all of the filtering in place, you can see by looking at the wav file (or on a scope if you are all analog), that there are many short spikes that are well above the normal program level (that 25db of crest factor at work). If you tame them via a limiter with a fast attack and decay, your not going to affect the sound too much unless you set the threshold too low. A good limiter (look ahead is nice for a digital version) should be pretty transparent as long as you don't push it too hard. Depending on your DAC and/or amplifier, you might get away with letting it clip these peaks and in effect creating a crude fast attack limiter. I find that a good limiter setup such that the amp never clips is preferable. Since my path is all digital up to the amp, my limiter prevents DAC clipping and via proper gain staging amp clipping as well. I'm hitting the head with 1-3 watts of average power, but the peaks would hit 500 watts or more if not limited a bit. Every 3 db of limiting I can do, allows me to increase the average power by a factor 2 as compared with the no limiting case.

Mark

[dmills]

My take is that everything post IRIAA is velocity not groove amplitude, so place processing depending on what exactly you wish to do dynamics on.
Post IRIAA limiting is essentially limiting on velocity (which you do need due to the cutter head geometric limits), but you also want this limiter to respond to acceleration (Which is a playback limitation due to the effective replay stylus diameter), so ideally you need to limit on both flat velocity and velocity with a 6dB/octave slope.

You can of course do this limiting pre IRIAA, it just needs an extra differentiator to get a velocity control signal (Mostly the problem is above a few kHz so a single zero gets it done near enough). Hell you do the limiting pre IRIAA, but take the control signal from the output of the IRIAA stage....

Pre IRIAA without the differentiator is limiting on groove amplitude, which should not be too much of an issue providing your HPF is set somewhat reasonably.
What can make sense to do post IRIAA is things like spring back equalisation as the roughly +6dB/octave of the IRIAA is effectively preemphasis and reduces the effective noise contribution of this stage, there is however a trap in that you want a design that contributes little low frequency noise as that is significantly amplified on playback, decisions, decisions...

Limiting in the broadband sense pre IRIAA usually makes little sense as the transfer function thru the cutting lathe to the disk is anything but linear phase, and indeed should NOT be linear phase (It should be the inverse of the replay curve, including the replay phase response), so limiting and clipping going in will NOT produce anything that looks particularly limited on the groove amplitude, and may increase the HF present making the de-esser more of a player.

[karl hungus]

This is very informative, thank you for the detailed replies.

I’m going to take your example and put the limiter last. It sounds like the best way to get a cleaner and slightly generally louder cut, loudness war aside.

[markrob]

I suspect that the voltage signal right at the power amp input is closer to an acceleration command rather than velocity in the case of a moving coil head. As a DC voice coil motor, head current is directly proportional to its generated force or acceleration. The head back EMF and the presence of its L/R time constant will make it deviate a true current mode if fed from a voltage source. But for much of the audio band, I think it will be close. If you setup a local current feedback loop around the power amp and head, both factors would be eliminated. I think this true for both open loop and closed loop heads. If the head is a closed velocity loop, the error signal fed to the power amp will have to do the same effective EQ that would be done open loop to flatten the head.

Mark

[dmills]

Yea, I was kind of assuming that the doings to make the IRIAA output effectively velocity were post all of this (either feedback or a mess of post EQ if running open loop).

Do not that any limiting MUST be placed outside the feedback loop as anything in the loop will just be automatically equalised out by the feedback (And may screw up the loop phase response sufficiently to make it oscillate).

Current mode drive, while it removes a lot of issues with the L/R time constant and the back emf, also implies zero electrical damping, so may be a poor fit for an open loop head, makes sense in a feedback design however where the damping is done by the error amplifier.

If experimenting with current mode drive you will likely want an R/C snubber across the load so that the load becomes resistive above the audio band but before the amplifier gain intercept as otherwise the rising impedance of the inductive load will likely stuff up the stability.

[karl hungus]

Sorry, but is Souri’s style of head an open or closed loop? I’m not as technically informed to follow the last two posts, unfortunately.
Should I be reconsidering placing the EQ last with this type of head?

[markrob]

Its an open loop head. He does have a feedback version, but I'm not sure its very common and there is debate as to how effective it is. I don't think the EQ order is that important. You just have to make sure your gain staging is well thought out such that you don't cause clipping or noise buildup anywhere in the chain. You really only need one stage of EQ that encompasses the RIAA and corrective EQ. However, you may find it better to spread the EQ over two or more sections to avoid the need for really large boosts. One thing to note. The standard RIAA recording curve matches the theoretical open loop head response between 500 and the head resonance. If you do the full RIAA curve as a block, you will be adding corrective EQ to undo that portion of the curve. I created a version of the RIAA that has no 500hz turnover point to avoid this. Just a small boost below 50 hz and a +6db/oct boost above 2122 Hz. I also add to this a 1st order shelving EQ at 1Khz with about 20 db of gain above 10 Khz to compensate for the -6db/oct roll-off of the head above resonance. This is a crude first approximation of the EQ needed to flatten the head. I do my test cuts with this in place and then use a a 3rd EQ (graphic in my case) to do the final correction. This includes the cancelation of the main system resonance. Hope that makes sense.

[karl hungus]

Thanks Markrob, makes sense.

I’m fortunate enough to be running the Bettermaker EQ so it’s like Fabfilter controlling an analog section. My RiAA curve will be generated by a sweep/noise for the actual head and dialed in digitally. I think it might be the best of both worlds.