Here's an interesting set of products: these are high-speed power amplifiers which take an input signal over coax (optionally summing two inputs), boost it by a variable amount depending on some gain switches, and then drive a coax output with a good deal of power behind it. The 4005 can handle a good combo of speed and high voltage, with up to 50VA and 130Vpp into 50 ohms at 1 Mhz, which is pretty impressive - definitely not something well-known like an audio amp, or the kind of design that you can doodle on the back of an envelope and make it work the first time. I can't find any specs on the 4020, but judging from the front panel it's a higher-power more-featureful version of the 4005: the gain range goes higher, it shows output voltage/current, and it advertises itself as also being a bipolar power supply (by using the "bias" knob to adjust the DC level, I assume):
Let's look inside the 4005 first:
Main visible features are a big-ass transformer (left), a pair of large aluminum capacitors holding up a rectifier board (bottom-left), a PCB with a bunch of power transistors (right), and a pair of fans which pull air over the transformer and then send it over the power transistors' heatsink.
Following the flow of power...
The AC input goes through a fuse to an input voltage range selection switch, which sends it to either the 100V or 220V taps on the transformer's primary:
The secondaries then supply +/- 17V and +/- 90V:
Isn't it great when things are well-labeled? The +/- 90V goes to the rectifier and large pair of cans for some peak-charged goodness:
From there, it's down to the part which took me a lot of continuity checking to figure out. Turns out if you pull off the main PCB, there's a whole 'nother board underneath, with the power supplies for everything:
The two pairs of massive transistors at left are paralleled bipolar emitter-followers which regulate the raw +/- 90V down to +/- 87V. The smaller transistors right next to them are the pre-drivers which provide their base current:
I haven't figured out what the right-most pair of transistors are for. Their emitters are connected to the +/- 90V (NPN emitter to -90V, and PNP emitter to +90V), and their collectors end up running through some cabling up to the top amplifier board, but it's not obvious from there what the signals do, and the traces don't look particularly high-current. My best guess is that they either provide switched (unregulated) +/- 90V power to the top amplifier board (waiting until the other rails come up, for example), or because of the emitter resistors, that they provide some kind of current-source bias to the top amplifier board, with an extra few volts of overhead.
There's also a bridge rectifier (at far right) which rectifies the +/- 17V transformer taps, and a couple linear regulators in the middle which drop it down to +/- 15V for the analog controls (op-amps, etc.):
Back up the top amplifier board then: this now has nice regulated +/- 87V supplies from the bottom linear regulator board, and +/- 15V auxiliary power as well. The input signal comes in on either back or front connectors, both of which take the coax a few turns around some toroids (for noise suppression and common-mode rejection, I'd assume):
The input signals then run to a board attached to the front panel, where they go through some kind of bandwidth shaping I think, and the gain selection. There's a few op-amps here, and the gain-selection buttons either switch out their feedback resistors or select/bypass certain fixed-gain amplifier stages.
After this, the signal ends up down on the amplifier board. Those power devices turn out to be MOSFETs, mounted in sockets (can't see it here, but they conveniently have D S G markings):
The basic idea, which took more probing and swearing, is that the power amp is (unsurprisingly) a class-B amplifier, with all these MOSFETs paralleled together. The N- and P-channel devices alternate down both sides of the heatsink, and have source resistors for sharing. The N-channel gates are all bussed together, and the P-channel gates are all bussed together as well (through some series resistors, in both cases). There's some complex biasing which I couldn't quite figure out (may finish tracing the circuit one day), and the MOSFETs on the end sections of heatsink act as pre-drivers for the paralleled MOSFET gates:
Let's take a look at one part of the board (it's pretty symmetric):
You can see the two common gate connections and the common output trace all winding their way down the row of MOSFETs here. Each MOSFET also has its own local ceramic-disc decoupling cap for either the +87V or -87V rail. There's also a couple op-amps over here, with some adjustment pots nearby. I'm about 99% sure that these are responsible for wrapping a feedback loop around the big paralleled MOSFET array to keep everything nice and linear.
A schematic is probably called for at this point to keep track of everything so far:
The N- and P-channel gates have a pair of resistors and parallel caps between them, which probably is meant to have a fixed voltage across it for biasing. The low-power transistors, through the inexplicably long chains of diodes, probably are current sources which set up a constant current through the resistors to keep a constant voltage drop - this is similar to using a couple diodes between the base of the NPN and PNP transistor in a BJT-based class-AB or class-B amplifier. The curious part though is the lack of temperature compensation that I can see. As power transistors heat up, their DC characteristics shift and the biasing changes. In an audio amp, a big deal is made of doing the biasing voltage with transistors that share the power transistors' heatsink, so that the temperatures of the devices track pretty well and the biasing can remain fairly constant over temperature. Here, there seems to be nothing except the power MOSFETs on the heatsink. Maybe the biasing for these particular devices changes little enough over temperature that it doesn't matter, or the feedback loop with the op-amp(s) has enough gain that even noticeable variations in biasing don't actually hurt the performance. Or maybe the THD specs for this kind of product are loose enough compared to an audio amp that you can get away with a lot more.
The spec sheet also mentioned an output protection circuit: in audio amps, this can get pretty sophisticated with output current and voltage sensing creating a boundary with multiple slopes to stay inside the power transistors' safe operating areas, while still covering as wide of an operating range as possible. Here, I'm not actually sure how the current sensing is done, because I didn't find any resistors between the paralleled sources and the output (except the inductor-bypassed output resistor, which doesn't have sense traces and doesn't work at DC). The +/- 87V regulators do have emitter resistors though, so maybe the linear regulator board actually does the current sensing?
One thing that I noticed about this amplifier board is the layout. The NF engineers really knew what they were doing - anyone can do a good-enough layout for a non-HDI design like this if you have 4 or 6 layers available for putting your power and ground planes everywhere, but it takes way more skill to do a good 2-layer layout like this one for a high-speed high-power amplifier. Some features are:
1. The local decoupling at each transistor that I already mentioned.
2. The aluminum 33uF bulk capacitors buffering the +/- 87V supplies at all 4 corners of the MOSFET array, to minimize the distance to any individual transistor.
3. The N-gate to P-gate RC network, duplicated once on each side of the heatsink to keep the gate signals from starting to get away from each other with distance as the stray inductance and resistance adds up.
4. The well-placed ground pours on the bottom side:
Anyways, the output of the massively-parallel MOSFET class-B stage goes to a disconnect relay at the back, and then a source resistor bypassed by an inductor (as seen on the schematic) for some kind of frequency compensation, and a couple clamping diodes to keep the output from exceeding the +/- 87V rails:
There's a chassis-mount device on the bottom of the case, next to the linear regulator board, with coax coming in and going out of it: my guess is that it's a 50-ohm power resistor to supply the 50-ohm-source-impedance outputs:
Let's now look at the
4020, which as discussed is the 4005's older brother:
Everything looks the same, but, well, bigger. The amplifier board appears to have more pre-drivers, as well as more solid power distribution with metal busbars and way more buffering caps:
The power transistors now hang down vertically from the bottom. You can see them here, along with the big TO-3 packages on the bottom heatsink that are probably the linear regulators (like the bottom board in the 4005) in this design:
The rectifier and DC bulk capacitor board has grown:
...and there's now a separate board with some colorful resistors for driving the front-panel meter and indicators:
...but the front-panel input and gain-scaling board looks pretty much the same:
That's pretty much it, so have some scenic views of what seems like the protection circuit, biasing, and pre-drivers:
Home
Help
Search
Login
Register
