Limiter/Compressor based on Beam Deflection Tube

Hi,

Since the subject was discussed here some time ago, I thought I would share my ideas on building a compressor/limiter based on a beam deflection tube.

I looked at the characteristics of a few BDTs, and the 6ME8 stood out as allowing a pretty large linear deflection voltage swing, which is desirable in a limiter input circuit. The idea is to drive the audio voltage on the deflection electrodes.

I could have audio signals coming in at +30 dbu or more, which is about 70 volts peak-to-peak. Looking at the attached curves, the deflection voltage could swing 40 volts each way, in a diffamp giving > +30 dbu, before going into the nonlinear part of the transfer function.

The 6EM8 being a beam deflection tube, the control grid G1 sets the anode current and the deflection "plates" steer the beam to one anode or the other, the total current of the 2 anodes remaining constant.



Looking at the ratio of deflection voltage to plate current differential, using the deflection electrodes as a "differential signal grid" gives a Gm of approximately 175 at unity to 35 uMhos at max GR over the range of -6 to -9V on grid1. The operating range is shown in red and green for the respective 2 anodes

Grid 1 will be used to control the gain over about a 14db ratio, or a maximum 14db gain reduction. Not exactly a crusher but should be useful for a lot of tasks.



With respect to the G1, it the tube has a pentode characteristic and within the chosen operating range (in red) is quite linear over anode voltage. This implies that as the anode voltages change wrt one another as signal is output, the total current will remain constant.

It looks to me like a purely resistive load would give a fairly linear output, especially if the topology is kept balanced differential throughout.

One problem this limiter will have in common with so-called "variable mu" limiters is that as the control voltage increases there is a common mode output current step that can cause an annoying audible "thump" when the limiter is used at fast attack times.

In this operating range the common mode current over the range of CV is about 6 mA, and the signal current is about 7 mA p-p, so I conclude a simple resistor load on each anode is going to be trouble for "CV isolation". I think I'll start with a common mode choke like a LL1667, with a resistor load anode-anode chosen to give a 1:1 signal ratio from deflection plates to anodes when the DC anode current is about 8 mA per anode.

The sidechain amplifier only needs to swing 3V into the control grid of a pentode! Is this too good to be true?


Here's a simplified schematic. The input is transformer coupled with a signal balance network to null out the f2 distortion due to unequal signal swing through the signal path top vs bottom.

Then cap coupled to a DC balancing network to set the quiescent current balance between the 2 anodes by tweaking the deflection voltages.

The gain control voltage from the sidechain drives the control grid of the 6ME8 through a cathode current servo. Remembering that the signal doesn't change the cathode current, just steers the beam back and forth, this trick should help linearize the gain response to a 2-8V control voltage input.

Back to the signal path, the BDT anodes are loaded by a common mode choke that should provide the first rejection of the control voltage induced common mode thump signal while passing the audio at a level determined by the BDT current swing and the load resistance.

This then DC couples to the output stage which can provide controllable gain through the "plate feedback". The load resistor also has an adjustable center tap so that the CV thump can be balanced out and rejected by the diffamp output stage. The cathodes of the output stage are AC coupled but DC separate, allowing perfect DC balance and AC current sharing.

The output stage gain should be adjustable between 1x and 3x (0-10db) which is in the loop and will set the compressor's small-signal gain.

The output signal is taken back to the sidechain amp where it is rectified and processed for threshold, attack, release settings and applied to the 6ME8 current servo.

All of the balance adjustments make it look complicated, but I think they'll be necessary for good performance. I don't know how good the dynamic balance can be. One characteristic of this limiter may be an increase in f2 distortion as the Gain Reduction (GR) kicks in.

Cheers,

Michael J Koster
http://redwoodcoastmusic.com

Thanks Darius,

Glad to hear the basic circuit sounds good.

Thanks for the links. I had researched the idea on AA, TubeDIY, and this forum also. Don's scans are interesting and helpful and I would like to see the compressor circuit and figure 8...

The 6AR8 looks good if you want audio gain, as in a preamp etc, but I chose the 6ME8 for it's large voltage swing and what looks to me a better characteristic for a compressor/limiter in a recording or mastering chain. I guess I'm trying to trade off gain (and gain reduction potential) for linearity.

I still need to design my sidechain. The circuit is just a place holder for the time being. There's no rectification either ;-)

I'm thinking, for the sidechain, of using opamp stages and something simple like emulating the Fairchild 660/670 time constants and program switch to start with. (thanx for the schemos of these BTW!)

Michael

> One problem this limiter will have in common with so-called "variable mu" limiters is that as the control voltage increases there is a common mode output current step that can cause an annoying audible "thump"

It is in fact just another vari-Gm machine, but with a single cathode and electron stream, and audio injected high up the tube.

Not only does it want to thump, maximum level goes down as gain is reduced, which is great for a volume "pot" but not in a Limiter.

> an increase in f2 distortion as the Gain Reduction (GR) kicks in.

That's always on our plate; but push-pull amps can be pretty "off" and still give low 2nd.

I think your plan makes sense.

I wonder if 47uFd is enough between the ECC99 cathodes, but I'm too lazy to think and you'll know when you run freq response. It probably only has to be 10V. If you plan to use Film, of course they don't come that low-V and 47uFd is already a big cost. (And perhaps your 0.47u caps compensate the 47u's loss?)

I keep thinking "you don't need all those trims and 400V!!!", but every one has a valid reason. We can hope you will discover that several of those trims are "no effect", but without heating-up a 6ME8 we can't know which ones are unnecessary.

> a maximum 14db gain reduction

Yeah, in that range.

Somebody is gonna want "more, more!", but if you are happy, that's all that matters.

An unanswerable question: what is the Noise referred to the deflection electrodes? Total dynamic range at input is noise-to-overload. We know we can stuff 30V at them, far more than any vari-Gm input. But if the deflection geometry means small random charge distribution gives large "hiss", then dynamic range may not be huge. On a vari-Gm we can estimate hiss from Gm, and fortuitously we get least hiss at max gain. OTOH, that Gm is "hot" and has 2.5 times the hiss of a 25C part. In the beam deflection tube, we just don't(?) know the noise sources and magnitudes. Is the deflection space equivalent Temperature high or low or meaningless? Is partition noise stirring things up? The contraption is strange, and the market application (NTSC "color") was high-level lo-fi so it was never characterized.