Note: I performed extensive frequency, level, and actual studio tests on the 864 clone today, and several interesting details were revealed.  Text has been edited to reflect that. 

From 1954 through at least 1963, the Federal Television Corporation built an audio limiter called the AM 864/u for the US Air Force and US Army.  The 864 is a simple, rugged device that accepts 600 ohm balanced or unbalanced line-level signal, offers a single front-panel input-attenuator control, and compresses the output level at a 10-to-1 ratio once the threshold point is reached.  The output is also 600-ohm balanced or unbalanced, and it offers a maximum 36db of gain.  The rear panel of the unit displayed the threshold and ratio controls, although these are confusingly referred to as (respectively) CURRENT and THRESHOLD in the manual and schematic.  Attack and release times are fixed, and the manual indicates them at .05″ and 2″ respectively.

As you can see in the diagram above, the 864 was intended to be used as what we call a ‘broadcast limiter’ – the final step in the signal chain before the broadcast transmitter.

Download the original 1963 manual for the AM864 (apologies to whomever did the epic work of scanning this 55pp document; I have long forgotten where I got this file from)

DOWNLOAD: Federal-AM-864-U-Manual copy

After having scratch-built an Altec 436 compressor years ago, I wanted to try building an 864.  The circuits are very similar, although the 864 uses the older 1940s-era octal tubes and uses a feedback circuit from the plates of the input tubes (rather than the output tubes, as in the 436) for its compression control signal.  More importantly, though, the 436 remains a bit of an oddball underdog in the vintage-compressor market while the 864 enjoys a very strong reputation.  Anyhow, like the 436, the parts cost to build one of these things is negligible, so I figured what the hell.

This is going to be a very long + detailed +technical article, so I’m going to ask y’all to please click the link below if you dare to READ-ON,,,,

(L to R: 6SK7 + 6SK7 inputs, 6SQ7 compression-control, and 6SN7 output tubes)

After 16 hours of fabrication and assembly, my Federal AM846 clone was finished, and I am really impressed with the sound.  Far more than I ever have been with my Altec 436 (or any original 436/8 that I have ever heard, either….).   Now, there are about a million threads and article on the internet written by other dudes who have scratch-built these things, so there is no need for me to offer a full account of the proceedings.  If you wanna try building one yrself, I suggest you start where I did: google “Federal AM864” and start reading+ downloading.  However: as is often the case, I made many slight modifications to the unit that you might find helpful.  So consider this article as yet another addendum to the already-substantial body of online information regarding this circuit.

Above is the schematic that I used for my build.  It proved to be accurate and easy-to-follow.  Although the device uses very few components, there are a lot of multiple-interconnections between components, so there is surprising amount of wire inside the box.  Also – since this is a fully balanced device – there is some additional potential for error in wiring.  To make the assembly go fast, I used many colors of wiring:  The usual green/yellow for filaments, blue and brown for plates, red for B+, green for grounds, striped for center-taps, and I even invented my own new coloring convention: Purple/Grey for metering circuits.

As is also my standard practice, I used a single heavy-copper buss bar for all ground connections.  In classic style, the buss bar contacts the chassis at one point and one point only: the input connector.  In the photo above, you can see the buss bar starting in the lower left corner.  If you look closely, you can see that it is mechanically secured at pin one of the XLR output jack, and the chassis-connection point is immediately to the left; after passing the input XLR, it heads north and then east towards the power supply, where it is ultimately secured at the ground lug of the IEC socket (not visible here).  Other points of note in the photo above:  note that the input transformer (the cylindrical grey item center-north) is mounted as closely to the input-attentuator pot (visible in the extreme upper-right) as possible.  This is because the signal becomes high-impedance once it exits the input transformer, and good layout dictates that we keep out high-impedance signal runs as short as possible.  Also: you can see my choice of signal capacitors here.  As always, I use Solen caps.  I feel that they are the best price/quality signal cap on the market, plus they are extremely small physically, which makes things much easier to work with.  The one exception is the single Sprague orange-drop, chosen simply because I lacked a Solen in that value.  One more note regarding the signal caps: the manual specs .1uf coupling caps for the interstage coupling; I used .22.  Now, for all I know, .1 would have still allowed for full 20-2ok bandpass, but .22 is a more typical value in amplifiers that are expected to pass full low-end (IE., the class Fender Bassman).   Anyhow, not sure if this circuit alteration was necessary, but since this component is not incorporated in the feedback loop, it certainly can’t hurt.

For the resistors, I used 2-watt metal oxide power resistors for anything that dropped a lot of voltage, and carbon-comp or carbon film for the rest.  The BIG resistor at the right is a Dale 15k 10w wirewound.  I have no idea why the schematic specs a 10w resistor here – seems WAY overkill – but again, anytime you are going to drop a lot of voltage, use a wirewound or metal oxide resistor.  IMO, Carbon comps are just too noisy for this application in pro audio gear (but – they are FINE in guitar amps, where the overall system noise floor is necessarily much higher owing to inherent pickup noise ETC).

I naturally moved the ‘Current’ and ‘Threshold’ pots to the front panel as well.  One note regarding the pots:  the schem does not spec audio or linear taper.  Now, the input atten pot obvs needs to be audio taper.  The other two pots would seem to want to be linear taper. I used a 500k linear for the ‘threshold'(aka ratio) pot, but I was all out of 1/2watt 1K linear pots, so I used an audio taper here; it works, but not very well – all the action happens at one end of the swing.  Gotta change that out.

The release time of the 864 is very slow, and I wanted to speed it up – it’s pretty rare that I want a 2″ release time in the studio.  The release timing is apparently dictated by C1 and R1.   I tried changing C1 to a .1, a .2, and a .5, but they all caused some degree of motorboating weirdness.  So i changed R1 instead – to 1M- and this shortened the release time considerable with no ill effect.

OK so on to the power supply.  The 864 needs 30ma 350v CT and 2A 6.3 volt heater supply.  If we omit the 5v rectifier tube (as I did), we can get away with using the cheapest plate/filament transformer known to man, the Edcor XPWR024.  It is spec’d at precisely those voltages, and it retails for a ludicrous $20.01  This is about 1/3rd the price of any Hammond unit that you would need to get the job done.  I have bought dozens of these things, and I’ve yet to have one fail.  Highly recommended for any low-current application where you don’t need more than 200v plate voltage or you are willing to use a voltage-doubler in the power supply.

Since the Edcor does not have a 5V winding, and there was not enough current available from the filament winding to use a 6×5 (6-volt) rectifier tube, I just used a pair of $0.02 IN4007 diodes instead.  If you do this: remember: rectifier tubes allow for a gradual build-up of B+ voltage in a circuit.  So if you chose to use a solid-state rectifier (EG., a pair of diodes), it’s important that you incorporate a standby switch in the circuit.  Otherwise everytime you power the thing on, the tube plates will get hit with full B+ voltage when they are still cold.  This will adversely affect tube life.  In this case, I implemented a standby switch in the standard manner: it acts to switch the 350v center-tap to ground.

To the right of the Edcor you can see the filter cap can.  This is my favorite guitar-amp muti cap, the JJ 500v 20/20/20/40uf.  They sell for a not-cheap $17 with mounting clamp, but I have bought dozens if not hundreds of these things and they are a great price/quality/size value.

If you look at the original schem, you will see that the B+ filtering consists of simply two stages:  12uf, followed by a 15h choke, followed by 24uf (achieved by paralleling two 12uf sections).  Now, I may be a simple cave man who fell into some ice and later got un-thawed by your scientists, but just two stages of power-supply filtering in any audio device seems like a terrible idea (even in a fully differential device where any common ripple should in principle be cancelled out in the output transformer).   Clean that shit up man!  Jesus.  So i omitted the choke, paralleled two of the 20uf sections of the JJ, and ended up with a 20uf/40uf/40uf three-stage filter.  I isolated the stages with 470 ohm 5-watt resistors; these resistor values gave me 191 volts on the 6SN7 plates, which is a mere 6v deviation from the reading spec’d in the manual.  Done and done.  Now, in general, I am in favor of chokes, but I was trying to keep the cost down on this thing, and I am satisfied that I ended up with a much cleaner power supply than the original unit had.

OK bored yet?  The final power supply note:  heater wiring.  The original schem indicates a really janky heater wiring scheme in which one side of each heater (and one side of the transformer filament winding) simply go to ground.  Now, who fkkn knows, MAYBE due to the fully balanced nature of this unit, this has no adverse effect in the end, but in 99.99% of cases this type of heater wiring scheme will make for a noisier unit.  Since the Edcor power trans has a center tap for the heater wiring, I simply ran that to ground and then wired the heaters each with their own heater-secondary lead.  This is the standard way that most AC heaters are wired in pro audio gear.  Oh and a note on that: ordinarily, i would never build a pro audio unit with AC heaters; instead, I build a simple DC supply like this:

(the lead at the junction of R26 and R25 goes to ground BTW).   But, in this case, I knew that hum would likely not be a problem, so I skipped this step and ran the heaters balanced-AC as I described above.  In the end, this was the right move, as the completed unit is dead-fkkn-quiet.

OK so on to the audio transformers.  Let’s start with the input.  The schem specs a 600:10K audio transformer.  The 10k winding needs to be center-tapped (and WELL), as the variable bias voltage that creates the compression-effect is introduced to the input-tube grids via this center tap.  So what does this mean for us in terms of picking a transformer?  Well, we can use basically any 600 (or 150): >5K, <30K audio transformer, so long as:

*it is magnetically shielded (otherwise it will pickup and amplify hum) – so no Edcors here!

*the high-impedance winding is truly center-tapped, with good balance.

*it has a wide-enough frequency response

*it can handle enough level (dbm) for line-level applications

*we understand that a secondary winding of less than 10K will give us less-than-spec’d total-system-gain, and a secondary winding of more than 10K will give us more-than-spec’d total-system gain.  Practically speaking, and value between 5k and 30k should be fine.

Based on what I had in my transformer-case that day, I chose a UTC O-10.  This is a 30K c/t: 600 ohm transformer intended for push-pull-plates to line application.  Note that I have ‘flipped’ it and I am running it in reverse, but this has never been a problem with audio transformers in my experience.  AFAIK, a 30K c/t : 600 is a 600: 30K c/t.  The O-10 is full-frequency, but it is only rated for +8 operation.  After extensive testing, I have found that it indeed gets distorted below 100hz at approx +10 level.  Plus, its just got too much voltage gain.  I am going to order a Sowter #8540 and try that out,,, stay tuned.

Now on to the output transformer.  The manual specs a 10K center-tapped to 600 ohm transformer.  I feel like this is quite low of a load impedance for the plates of a 6SN7, so I went with a 15K:600 Edcor WSM600/15K.  This impedance variation also has the additional benefit of slightly reducing the overall system gain, which is beneficial in my instance, since I used an input trans that added more gain than spec’d.  Also, I keep this model Edcor in stock at all time – I have used dozens of these in projects for clients, and no one seems to mind that they cost $10.23!  They really do sound good, and the frequency response is there.  Plus they handle .5watts (26.5 dbm).  True, they are not magnetically shielded, but since they are a voltage step down device, it’s not such an issue as with a line-to-grid transformer, which is a voltage step-UP device.

Nonetheless, chassis placement of unshielded chokes and audio-transformers is critical relative to power transformers, so use good practice and scope out your placement using the classic headphone technique.  If you don’t know this technique, you can read about it any of the old military signal corp course books or just check it out at this link. 

Now, full disclosure: I have always and only used these Edcor transformers single-ended, cathode-follower style.  Never with DC current on the center tap. After running my 864 clone for a couple of hours in a mix session today, I did not notice any issues.  But will it eventually melt?  Only time will tell I suppose.  I will probably just swap it out for a nice 10-watt Motorola that I pulled from a PP 6K6 broadcast line amp,,, stay tuned for that as well.

Last but not least: the meter.  The manual never actually says exactly what sort of meter it is, but somewhere online I read that it is a 2ma DCma meter.  I have a box full of DCma meters, but 2ma is an unusual scaling – 1ma is much more common.  So, that’s what I used.  In keeping with the military theme of this device, I chose a 2″ panel-mount General Electric U.S.N (united states navy) CG-22421 from my pile of ye old round meters.

It works pretty well.  It shows output level moving, it shows compression happening, no, it’s not set to any sort of objective scale but the needle hangs out in the expected area and it provides good visual feedback of what my ears are telling me.  Testing shows that +15db correlates ‘.8’ on the scale.   I did notice that the diode bridge introduces noticeable distortion in the output audio, as well as a slight loss of level, so I would not recommend actually using the meter for output level indication while tracking or mixing.

Above: the unit prior to wiring

My final comments will concern the chassis fabrication.  It was important to me that the thing look right; I mean, christ, it’s gonna live in a recording studio where clients have reasonably high expectations about the quality of the kit.  I don’t think that you can buy a complete chassis these days that in any way resembles the original AM864 chassis, but we can get to a pretty reasonable approximation by making a system from various sundry components.  In my case, I used an old Hammond 1444-series 17x8x2 aluminum box that I had in inventory (no longer made – try this instead).  I made the bottom plate out of aluminum sheet that I keep around, and the faceplate is a $15 Mid-Atlantic flanged-aluminum four-space blank panel that I picked up at RedCo.  The steel side brackets are some ancient dead-stock Bud units that are long out-of-manufacture, but Bud now offers these instead, which should perform a similar task.  The handles on the front of the unit are my favorite Bud H-9160b 3.5″ chrome handles.  Put it all together and you get something like this:

Above: L to R: meter switch, “Current Control” (aka threshold), input attenuator, “Threshold” (aka compression ratio), power indicator lamp, standby switch, mains switch

Now, the 4-space panel turned out to be excessive – I could have gotten away with a 3-space – but until I knew exactly which meter I was going to used, I could not be sure.  In the future, I’m going to use the 2″ high chassis, a 2″ round meter, and a 3-space panel.

Further reading:

Here’s a dude who manufactures a Federal AM864-inspired unit for $1499 street

Here’s a nice-looking DIY build of this unit from GroupDIY