I have become very confused about how to ground AF power amps. I have an old 69 Marshall "69" 100W plexi with 2/4X12 speaker cabinets and it is getting old and has been repaired once (about 1980). I had to replace the power cord and switch myself and used a three lead power cord with the center (green) lead cut off (it has a polarity switch) and a cheap RS mini SPST on/off switch in about 1980. An amp tech in Texas replaced the output transformer at his time (I allowed a group in Colorado called Aphrodite to use the amp for a concert with Buddy Miles and they blew it up! Lead player was Snuffy Walden who did all the guitar in "30 something" ). Anyone have a suggestion as to how I should ground this beast or should I just leave it alone? I get zapped all the time if I'm not careful.

I have read a lot about a "star ground" but find little agreement on exactly how to implement it. My general impression is that the power ground (e.g. green wire of the power cord) should be grounded as close to the power outlet as possible and the signal ground should be grounded as far from the power ground as possible. All signal grounds go to the single star signal ground and all power grounds go to the power star ground.

Is this correct?

LouTheBlueGuru

 

I won't claim to be an expert. But I THINK that, typically, in audio amplifiers, the safety earth ground and the signal ground are never connected directly to each other. I think that if they are connected at all, it is usually through a "safety disconnect" network, which should consist of two diodes in anti-parallel, and a small-value resistor (10 to 100 ohms seems typical) in parallel with the diodes, and a 0.1uF capacitor also in parallel with the diodes, which would all be connected from the signal's star ground to the safety earth ground. But I'm not sure what ratings you'd need, on those components.

I believe that the safety earth ground (the ground wire from the three-prong AC mains connector) usually is only connected to the metal chassis and/or outer metal case, and should not be directly connected to any of the other grounds in the system, i.e. the various power supply grounds and signal grounds. (But I've also seen it done differently, sometimes.)

Hopefully, someone with more knowledge about this than I have will jump in to provide more and better information.

- Tom Gootee

http://www.fullnet.com/~tomg/index.html

-

 

You can go to most any hardware store and pickup an 18AWG 2-wire polarized-plug lamp/appliance cable. This is the type of plug where the neutral blade is wider than the hot blade, and will only fit into the outlet in one direction. You can also get a nice long cord of 12ft. or more (but if you go beyond this, look for 16 AWG wire).

After swapping out the AC cord, learn which way to consistantly set the polarity switch so you don't get zapped. This is not a guarantee, but better than what you have currently.

A 3-prong is not a guarantee either, because there is no way to know if the ground is carried by whatever outlet you plug into, and you may have to use an extension that defeats its purpose anyways.

 

Nomurphy, your suggestion on using a two pronged plug with a defined ground pin (the fat one) is excellent. This seems like a much better solution than my three wire plug with the earth ground cut off. I think an important point here is that these old amps were not designed with the more modern three wire grounding system.

 

Gootee, I am still confused about the so called "star ground." I am building an LM3886 chip amp and in most pictures I can find everything, and I mean everything, is routed to the star ground directly. This seems to be wrong from what I have read, but it seems to work for a lot of people building these things.

 

Gootee, I am still confused about the so called "star ground." I am building an LM3886 chip amp and in most pictures I can find everything, and I mean everything, is routed to the star ground directly. This seems to be wrong from what I have read, but it seems to work for a lot of people building these things.

I'm not an expert at this stuff, yet. But I'm working on it.

I think that I would have separate star grounds for "signal" and "power" grounds, if I couldn't route ALL grounds separately, ALL the way back to the "ground" side of the main PSU filter cap(s), which is usually my preferred star ground point. And then I might want to see what would happen if I put a small resistor between the signal star ground and wherever I connected the two star grounds together, especially if they had to run together for a while, after that. (Instead of a resistor, a smaller wire diameter might work, too.)

But, as I said, I'd really rather run every ground separately, as much as possible. However, things can get more complicated, for certain designs, depending on how the power supply is configured, for example, and which audio channels share which power rails, etc etc etc.

I guess a lot depends on how good the layout is. The shorter the current paths are, the less it would matter.

Then again, there might also be some number of grounds that CAN be run together, with no ill effects. And that could save board space or wire or labor, etc. So it would be good to know which grounds can and can't be run together, for both performance reasons and "cost" reasons.

One main thing, for understanding how to best do grounding, is to figure out where the return currents must flow (and which ones are "nasty"). Then realize that PCB traces and wires are best thought of as components, not as pure conductors. Return currents always induce a voltage between the two ends of a trace or wire. So you don't want a large, spiky return current sharing a ground trace that also goes to, say, an opamp or chipamp input pin.

You can usually see what's happening, fairly well, if you include the parasitic components, when doing a simulation. I use the excellent, free LTSpice program, from http://www.linear.com . (And there's an excellent support and discussion group, called LT-SPice, at http://www.yahoogroups.com .)

You can find estimates, or on-line calculators, for the resistance and inductance of different types and sizes of wires and pcb traces, on the web.

If you model all (or just some, at first) of your ground returns, and even your power supply wires or traces, with an inductor and resistor in series, you can plainly SEE (at least some semblance of) what happens when you run things together or separately.

As far as I know, there are literally NO decent spice models for any of National's chipamps, available on the web. So, to at least get something I could use to simulate chipamp circuits, I went to ti.com and got their OPA541 model; the OPA541E version, actually, since it models some extra stuff that might be important.

For doing something like this, I usually just stick an (almost-)squarewave voltage source in, for the input, sized to give an output that's at least, say, 3/4ths of the chip's maximum voltage swing (so the effects will be pretty easy to see), with something like an 8-Ohm load from the chip's output to "ground" (I actually use a spice model of a speaker system and its cables). [Later, you can also use a sinewave input and let spice calculate the THD (total harmonic distortion), for various combinations of grounding layouts, etc.]

For a simple example, with a non-inverting configuration, try connecting your output Zobel network's ground, and your ground for the resistor from the inverting input, to each other, and then run that through something like 50 nH and .001 Ohms in series, and THEN to the spice ground symbol (to simulate running the grounds together). And then try it with TWO 50nH and .001 Ohms combos, one for each , BEFORE they connect to each other and then connect to ground symbol, or to their own separate ground symbols (to simulate running separate grounds). You'll definitely see the difference!

You can get as detailed as you like, modeling the parasitic effects, and do things like model every power pin decoupling cap's ground return, for example, and power supply lines, signal lines, etc etc etc. (Side note: You can also model the parasitics WITHIN components, too, such as, to begin with, by paralleling each resistor with a .3pF cap, and adding a small resistance in series with each cap and inductor.)

But you can usually get away without trying all kinds of random combinations of return paths, etc, by first looking for where the "sensitive" voltages or currents are, and where the "obnoxious" ones are.

I sometimes start out by inserting some approximate L-R model into each ground, separately, i.e wherever I have a ground symbol dangling. Then I run the "transient" (time-domain) simulation with something like the squarewave I mentioned above (make sure the rise and fall times aren't too fast for the chip's slewrate, though), and just click LTSpice's "current probe" on each ground's parasitic resistance, to see the plot of the current. It's then fairly obvious which ones have big fluctuations. (Hint: You can also do something like this to see the effects of using different sizes of decoupling caps on power pins, for example.) And you can click the LTSpice "voltage probe" just upstream of the parasitic L-R, and see what voltage the current is inducing, at that end of the ground trace or wire.

I just ran one of my chipamp simulations, and did that, and it's REALLY obvious that I don't want anything sensitive to share a ground run with the Zobel network that's on the chipamp's output, because, for a 20v peak-to-peak squarewave output, and the Zobel's ground returning to the power supply through 35nH in series with .00056 Ohms, I see current spikes of +/- 0.8 Amps, which causes, AT THE ZOBEL'S END of the ground-trace, VOLTAGE spikes of about +/-50mV!

Meanwhile, the voltage on the inverting input resistor's end of its ground trace is only doing spikes of about +/- 3.5 mV. And the amp's output is a very-nicely-shaped squarewave.

BUT, when I connect those two "grounds" together, and run them both through the SAME L-R, to get back to the power supply, then there are spikes of about +/-70mV on their end of the ground trace, and the amp's output squarewave now has a nasty 1.5V overshoot on the leading edges of the squarewaves.

Sorry to have gone on for so long, about all of that. But it's relatively easy to set it all up, especially for simple circuits like monoblock chipamp audio amplifiers, and might be pretty helpful for getting a handle on grounding issues.

Regards,

- Tom Gootee

http://www.fullnet.com/~tomg/index.html

-

 

Just to be clear, the "fat" prong is NOT ground, it is neutral (although usually/normally/often equivalent to ground).

Only the round prong on the 3-wire is earth ground, and ONLY if 3-wires were properly run to the outlet.

I had a friend that once rewired his home before I knew him; he not only used 3-prong outlets and DIDN'T connect the ground because he used "standard" 2-wire Romex, he also unwittingly wired the black/white in reverse. It was years before I understood why we were always getting zapped and constantly having to play around with polarity switches, and why the 3-prong plugs on some equipment I built seemed so ineffective -- until one day in passing converstion he mentioned that white was "hot" ...oops.

 

Just to be clear, the "fat" prong is NOT ground, it is neutral (although usually/normally/often equivalent to ground).

As long as we're nitpicking (always good where safety is concerned) I woud use the phrase "at ground potential" rather than "equivalent to ground." They serve two different purposes and, as Nomurphy cautions, should never be considered interchangeable.

 

The general method is to only connect signal and earth ground through a capacitance to give any potential surges/EMI a way out. This is generally a smaller cap to allow noise out. Please keep in mind EMI won't travel down regular wire due to skin effects and rather needs surface area to travel, so it may surve no purpose without proper wiring.

The chassis is generally then connected also to earth ground for safety as well as EMI shielding, however if the chassis has no metal components this also serves no purpose and you don't really need to worry about earth ground, save its' potential EMI filtering outlet as mentioned above if connected via some capacitance.

 

Please note the above mentioned is generally only a method used to reduce noise if noise is a concern. In some instances it may actually feed noise in and as such the capacitive coupling should be removed and earth ground only used for safety.