Bob Cordell's Power amplifier book

I wouldn't recommend that if for no other reason than the rails are pretty dirty.

But the rails and ground are shorted together by the decoupling capacitors.

If you draw the +ve current loop that flows in the zobel in a standard configuration, it’s out of the positivie terminal of the decoupling cap, through the output transistor, into the zobel, out of the zobel and returning to the negative terminal of the decoupling cap. If you connect the zobel emitter to collector instead, all you’ve done is remove the deoupling caps from the loop. Never tried it though. A real-life comparison of the two approaches would be worth doing.
 
But the rails and ground are shorted together by the decoupling capacitors.

If you draw the +ve current loop that flows in the zobel in a standard configuration, it’s out of the positivie terminal of the decoupling cap, through the output transistor, into the zobel, out of the zobel and returning to the negative terminal of the decoupling cap. If you connect the zobel emitter to collector instead, all you’ve done is remove the deoupling caps from the loop. Never tried it though. A real-life comparison of the two approaches would be worth doing.

Hi Harry,

It would be great if the decoupling truly provided an ac short between the rails and ground. They don't. I connect the ground ends of the rail decoupling capacitors together, then connect that point to the ground. In some amplifiers I have also used an X capacitor from rail-to-rail.

Cheers,
Bob
 
Hi Harry,

It would be great if the decoupling truly provided an ac short between the rails and ground. They don't. I connect the ground ends of the rail decoupling capacitors together, then connect that point to the ground. In some amplifiers I have also used an X capacitor from rail-to-rail.

Cheers,
Bob

Relative to the size of the resistance in a typical zobel, the decoupling capacitance is a short, assuming that electrolytics are augmented by surface-mount ceramics or even small-pitch (2.5 mm or less) plastic caps.

A 1 uF 0805 cap is self-resonant at about 5 MHz, after which it is inductive. The reactance will be about 125 mOhm at 20 MHz.

A 1 uF plastic cap with 2.5 mm lead pitch will be self-resonant at about 3.2 MHz, after which it is inductive. The reactance will be about 315 mOhm at 20 MHz.
 
One could also place the zobal(s) right at the output device(s) itself... minimum path lengths. But also the goal for me would to not dump RFI to common ground unless you can be quite sure none of it will couple to the circuit via common z paths.... field coupled or conducted.
Note: separate leads back to common point isnt RFI proof as the lead length can (and will, if given a chance) couple to other parts/wires/leads even other separate parallel ground leads.

There may be benefit to building the amplifier circuitry in a radius around the PS.... PS central on the amp PCB.... PS central with circuit going out towards the OPS. Rather than running wires back and forth to a more remotely located power supply (though the XFMR could be further). I'm just saying.....


THx-RNMarsh
 
Last edited:
The HF ground return currents automatically follow the traces above the ground plane if allowed to, with no breaks in their paths.
This is much better in general than routing traces. You do need to take special care with high impedance junctions like op amp inputs, sometimes removing the planes in those areas.
Yes to removing the plane under any nets connected to the OPA inputs.

For return currents, Grounding of Mixes Signal Systems has the clearest accurate explanation.

In particular, he points out that if you can’t route single track earth returns exactly where you want them on the ground plane layer because of breaks, adding the ground plane doesn’t help. For stuff without digital, earth planes are usually a bad idea.
 
Relative to the size of the resistance in a typical zobel, the decoupling capacitance is a short, assuming that electrolytics are augmented by surface-mount ceramics or even small-pitch (2.5 mm or less) plastic caps.

A 1 uF 0805 cap is self-resonant at about 5 MHz, after which it is inductive. The reactance will be about 125 mOhm at 20 MHz.

A 1 uF plastic cap with 2.5 mm lead pitch will be self-resonant at about 3.2 MHz, after which it is inductive. The reactance will be about 315 mOhm at 20 MHz.

Hi Harry,

I think this misses the point. There is dirt on the rails that is not on the ground. Why inject dirt, even through a 4.7 ohm Zobel resistor, into the output of the amplifier?

Cheers,
Bob
 
more cancellation methods

Cancellation techniques are a fun place to look for solutions. A cap from PS rail to rail cancels differential-mode signals, for example; A much needed solution IMO for unregulated PS in audio PAmps.
In regards to ground planes.... you can cancel or reduce signal strength from radiating by the use of a ground plane above and below the signal trace.
For a mention and description of this cancellation method,you can look up pg 474 in Electromagnetic Compatibility Engineering (H.W.Ott). It may be applicable to wideband audio amplifiers to use multiple ground planes as well as separate power ground and signal ground planes.

If you go this route, you might as well go all the way to image planes.
See EMC and the Printed Circuit Board. IEEE press. M.I. Montrose. or Wiley Inter-science. It says .... layout made simple. But dont believe him.




THx-RNMarsh
 
Last edited:
if you can’t route single track earth returns exactly where you want them on the ground plane layer because of breaks, adding the ground plane doesn’t help. For stuff without digital, earth planes are usually a bad idea.

This is an important point as well. One factor nobody ever mentions is how large you make the annular space between the ground plane and the isolated nodes on the board. The smaller you make this space, the better the plane looks - as if it covers almost the whole PCB - but this also increases any parasitic capacitance from nodes to that ground plane. Depending on topology etc this can be a nice recipe for instability.

I've had cases where a prototype amp with a ground plane added actually ruined the stability that was OK without it.

You can't look at a ground plane and automagically assume it makes your amp 'better' in any way. And if you start adding speaker returns, supply decoupling and yes zobel returns to that plane, all bets are off.

Jan
 
Here is my take on the issue of output stage grounding. I have some experience in switchmode power electronics, so in my last build I tried laying out the output stage for minimum inductance like I would a halfbridge inverter. The board is 4 layer and the two inner planes are used for the supply rails, the bottom layer is the output and the top ground. The Zobel network is R7 and C3.

I measured better than 0.01% THD at 20kHz and 100W/4 ohms.

Is this overkill? Yes. But the 4 layer boards were a lot cheaper than silver foil in beeswax capacitors. 🙂

blamelesstest21.jpg

blamelesstest22.jpg
 
Last edited:
Here is my take on the issue of output stage grounding. I have some experience in switchmode power electronics, so in my last build I tried laying out the output stage for minimum inductance like I would a halfbridge inverter. The board is 4 layer and the two inner planes are used for the supply rails, the bottom layer is the output and the top ground. The Zobel network is R7 and C3.

I measured better than 0.01% THD at 20kHz and 100W/4 ohms.

Is this overkill? Yes. But the 4 layer boards were a lot cheaper than silver foil in beeswax capacitors. 🙂

Nice build.
I like the idea to screen the feedback cable for obvious reasons; though the downside is that the extra lead capacitance *may* cause instability....
Pick your devils.

jan
 
Nice build.
I like the idea to screen the feedback cable for obvious reasons; though the downside is that the extra lead capacitance *may* cause instability....
Pick your devils.

jan

Perhaps not much of a down side IMHO as the capacitance brought in by the screen is usually no more than a dozen pF as long as the coax isn't insanely long. And any half-decent amp design shouldn't be destabilize by an nF cap at the output.

An added benefit by using a screened feed back wire is that the screen conductor can double as a dedicate GND connection between the speaker return terminal, if placed on the PCB by design, and the IPS signal GND. This will provide with a very low inductance path for the feedback sampling current loop, which seems especially valuable in CFA designs where low value feedback resistances are employed.
 
I was thinking more that the cable cap could act as lead cap *across* the feedback resistor. Probably depends on how it is implemented, where the screen is grounded, and where the fb resistor is (closer to the output or closer to the input).

Call me suspicious 😉

Edit: from the PCB pic above it looks like the fb R is at the input side. Probably safest.

Jan
 
Edit: from the PCB pic above it looks like the fb R is at the input side. Probably safest.

The screened cable is driven directly by the low impedance speaker output. All of the feedback network is on the driver board at the other end of the cable. Putting the feedback resistor on the power board, so that the coax was driven by a high impedance signal, would probably have turned out nasty. 😱

The capacitance of the cable is probably an order of magnitude less than the junction capacitances of the output transistors.

that the screen conductor can double as a dedicate GND connection between the speaker return terminal, if placed on the PCB by design, and the IPS signal GND. This will provide with a very low inductance path for the feedback sampling current loop
Yes, I used the cable screen as the ground reference for the input stage in exactly this way.
 
Last edited:
Here is my take on the issue of output stage grounding. I have some experience in switchmode power electronics, so in my last build I tried laying out the output stage for minimum inductance like I would a halfbridge inverter. The board is 4 layer and the two inner planes are used for the supply rails, the bottom layer is the output and the top ground. The Zobel network is R7 and C3.

I measured better than 0.01% THD at 20kHz and 100W/4 ohms.

Is this overkill? Yes. But the 4 layer boards were a lot cheaper than silver foil in beeswax capacitors. 🙂

Nice build.
I like the idea to screen the feedback cable for obvious reasons; though the downside is that the extra lead capacitance *may* cause instability....
Pick your devils.

jan

Hi Jan,

I always bring the feeback signal at full output amplitude right back to the series resistor of the feedback network, located right at the input stage. In this case, screening may still be a good idea, but <10pF of the screening will be insignificant.

Cheers,
Bob