Shaker amp and when caps get too big

[Magnus Danielson]

From: Toby Paddock <tpaddock at seanet.com>
Subject: Shaker amp and when caps get too big
Date: Fri, 10 Dec 1999 22:55:44 -0800

> Some more info on the shaker amplifier I mentioned.  
> Sorry, this got kind of long.
> 
> There are 92 19,000uF caps (1.75F total).  On startup, there are 
> resistors limiting the inrush current for a few seconds.  Then relays 
> short the resistors.  The rails are at +/-32VDC.  There is nothing 
> modern about it, it's just a big linear amp (the new ones are switchers).  
> The transistors are screwed into square tubing with cooling water 
> flowing through.  It's a little weird because they are not insulated.  
> Unscrew a transistor and water runs out.  A de-ionizer cartridge 
> keeps the conductivity of the water low.  Each transistor has 
> 0.5 ohm series for load sharing.  Each group of 6 transistors goes 
> through a 30A fuse.  If a transistor fails (shorts) the fuse blows 
> and takes that group off-line. 

I would precharge the caps in groups, those, first group precharge, then
when it has charged up the next is in for precharge etc. This way you can
spread it over time. This is even better than having just one precharge step
for all of them.

> It drives an electromagnetic shaker.  Pretty much a speaker with 
> an 18 inch dia voice coil. Instead of a cone, it pushes a metal surface 
> that the test article is bolted to.  In fact, it's a little bit like a *vintage* 
> speaker in that instead of a permanent magnet, it has a DC coil to 
> provide the DC magnetic field.  The field coil is made out of copper 
> tube with cooling water flowing through it and runs about 250-400 amps DC.

Hmm... you don't want to short that amp with your favorite screwdriver, it
will not remanin to be your favorite screwdriver...

> Is it true that old tube amps ran the B+ supply through the DC field coil of 
> the speaker to give the magnetic field AND the coil's inductance 
> smoothed out the B+ ripple?  

Some amps used this, the neat thing is that you get a inherent protection
since when the tubes starts to pull too much current will the weak power supply
lower the voltage and the current in the DC coil will be reduced.

> This one doesn't actually use a moving voice coil.  The AC drive coil 
> is stationary and the moving armature is a big fat aluminum ring.  
> The AC field induces some oh-my-God big currents in the ring and 
> then the ring wants to move in the DC field.  
> Advantages: no electrical connection to the moving 
> armature and no lightweight moving voice coil to rip itself apart.  
> Disadvantages: will not work down to DC and I don't think it's very 
> efficient.  Drive coil and armature are cooled by a large blower on the 
> roof pulling air through it.  I think Advent made a car speaker with a 
> coaxial tweeter driven suspiciously like this.

Cool solution.

> There is no air enclosure to work against so it doesn't put out very 
> good bass audio.  It can get painfully loud, but the bass response 
> is not that great.  The spec'ed range is 5-2000Hz.  Max stroke is 
> 1 inch.

I guess that the stroke-length used for maximum sinusoidal responce may not
be 1 inch all the way up to 2k, right?

How much load may this baby be loaded with?

> An accelerometer is mounted on the shaker to provide feedback to 
> the controller.  Some can be mounted on the test article to 
> see resonances.
> 
> 250,000 watts is the full load INPUT power.  Actual amp power is 
> probably, let's see... about 60V x 750A... only about 45kW pk maybe. 
> The rest goes to the DC field, pumps, blowers, and 
> steam belching out of the cooling tower.  
> 
> Wait a minute, I just thought of something.  If all the transistors are 
> the same, how can you have + and - drive?  I think I've got 
> something wrong here.  One *problem* is that it's so reliable I 
> haven't had to poke around in it's guts often enough to 
> remember the details.

The use of NPN and PNP transistors is due to simplicity in design, but you
can spend time to do a drive curcuit that uses only NPN transistors for the
last power stage.

How many steps of current amplifying does it have?

I mean, you must drive a lot of current into the bases so you need a bunch
of transistors for that, and then those have to be driven... this is what
happends at the edge of a modern CMOS chip (basicallly just a bunch of
inverters interconnected).

Also, do I assume right that is uses an H bridge or?

Anyway, thanks for sharing this with us, now we all run of building 45kW amps
for our center subwoofer of our homebuilt Dolby Suround sound room. Wonder when
they make a floor shaker subchannel so that you get the right feeling in the
disaster movies ;)

Cheers,
Magnus