Zobel networks

[Biggles]

Hi all, I was just wondering exactly why Zobel networks were (are?) used on the output of audio amplifiers. Over the years I had kind of understood they were in place to improve stability, as a sort of top cut filter, using a series connected resistor and capacitor, but it crossed my mind that you don't seem to see the configuration used so much nowadays. Is this due to more stable amplifier design, germanium/silicon technology changes or have they a more subtle purpose that I have missed? I remember every amplifier in the seventies and eighties seemed to have a Zobel network on the speaker output.
Alan.

[G6Tanuki]

I've always assumed the Zobel-network was there to prevent the amp going unstable/into oscillation at higher frequencies where the inductance of the speaaker-coil becomes significant. Some early transistor amplifiers would readily 'take off' when feeding speakers whose crossover-networks weren't well-mannered.

When using old-style audio amp chips like the TBA800/810 series and the LM380/383/384 I _always_ fitted a Zobel. Many more modern audio chips integrated the low-value-resistor part of the network into the chip itself, meaning you only needed to connect the 100nF or so capacitor from the relevant pin to ground - in which case it's not immediately obvious that there's a Zobel network there unless you look at the chip's architecture - the capacitor-to-ground would appear at first glance to be just another decoupling component.

[Craig Sawyers]

A Zobel is essential to ensure amplifier stability under even mildly inductive loads. Leaving one off is exceptionally hazardous and inviting VHF oscillation.

Similarly a small value (5uH or so) but high current air cored inductor with a damping resistor in parallel of 10 ohms, in series with the output. Again that is essential for stability into practical complex impedance mainly capacitive loads (like a loudspeaker with a crossover). Without that network invites incinerated output transistors.

What makes you think that a Zobel is no longer used?

Craig

[G6Tanuki]

Originally Posted by Craig Sawyers

a small value (5uH or so) but high current air cored inductor with a damping resistor in parallel of 10 ohms, in series with the output.

Yes: I never came across any "professionally-made" single-component version of this but in times-past we scramble-wound 30-or-so turns of 22-gauge enamelled wire over the body of a 'fat' low-value noninductive resistor.

[kalee20]

I'd assumed that it was to safeguard the output transistors.

A loudspeaker is predominantly inductive at higher frequencies. If a high-frequency sinewave is applied, at the end of a positive half cycle the current is a maximum. The next half-cycle, the voltage will go negative, but as the current is still positive it will flow via the upper transistor still - which will have an increasing voltage across it. Second breakdown is all the more likely to happen.

Connecting a series C-R network across the output means (if the sums are right) that the load will look purely resistive. So at any frequency, at the end of a voltage half-cycle, the current will also be zero and during the next half-cycle the current will flow in the appropriate transistor. And max current flow will then occur when voltage across the transistor is a minimum, so safe-operating area violation is less likely to occur.

(I'm not saying it won't help stability too).

[Radio Wrangler]

The full 'zobel' network has two parts. In series with the output is a small value inductor with a small value resistor across it. The second part is a series resistor/capacitor shunting the amplifier output.

The stability margins of feedback amplifiers diminish with increasing frequency. If a load looks inductive, or even worse, capacitive at frequencies above the audible range, RF oscillation and destruction can ensue.

So the series inductor starts to disconnect the amplifier from the cable to the speaker. And if there is some capacitance which could resonate with our added inductor, the resistor is there to spoil the Q of the resonance and keep things damped. In this frequency region, the shunt R/C comes into play and gives the amplifier a nice resistive load, damping things even further.

In the audible frequency range, the amplifier sees the speaker as its load, as you go higher, the amplifier segues into seeing the resistors in the zobel elements.

A designer no longer with us eschewed Zobel networks in his rather expensive amplifiers. Along came trendy speaker cables with zillions of fine strands all interwoven, half carrying current out, half carrying it back. The capacitance was remarkable. The trendy amps went down in a blaze of RF.

The moral of the story is to not let religious creeds and trends design either your amplifiers or your wires. Mr Zobel was a careful scientific type at Bell Labs if my memory serves.

David

[Biggles]

Thanks for the replies. I didn't fully appreciate the complexities of the simple two component network, accepting that they were on most amplifier circuits and that was that. I just assumed that they weren't as common as they used to be, particularly with modern IC designs, but as G6 states, they may well be fabricated internally. I seem to remember the component values would be typically 10 ohm in series with something like a 0.047 microfarad?
Alan.

[mhennessy]

Originally Posted by Biggles

but as G6 states, they may well be fabricated internally.

Can't say I've met this - any examples?

[Radio Wrangler]

Originally Posted by Biggles

the simple two component network,
Alan.

They should be a four component network, for the full thing. The bit that does the critical work is the series inductor and its damping resistor. It's usually made from several turns of wire wound on a large-bodied resistor looking like it shorts it.

All told, it looks like a diplexing filter. At low frequencies, the amplifier is connected to the speaker cable and the speaker. At high frequencies, the amplifier is connected to the resistance part of the series R-C that goes to ground.

The tansition region is designed to be well damped and to not have a significant resonant peak.

The design of the Zobel is based on the design of the time-constants and gains of the feedback system and hence knowing what range of frequencies it would most likely oscillate at if badly loaded.

Some amplifier designs omit the series element and just have the shunt R-C, so they fail to isolate the bad load, but just add damping resistance shunting it. Better than nothing, but it misses the boat.

I think that in most amplifier designs, the values in the Zobel network were just stolen from a previous design and rarely calculated from scratch.

Look at the value of those few turns over the series resistor and compare it to the size of the inductances that speakers and crossover networks present. They look inductive to the amplifier at far lower frequencies than the Zobel series inductor. There isn't much isolation going on. It's capacitance loading on a feedback amplifier ar RF frequencies which is the chief problem being solved.

Everyone says the zobel network is there to handle inductive loads. It's become folklore, but the maths says otherwise.

Zobel was a major player in the evolution of network theory and filter design, and did a fair chunk of the work which got organised into what's now called 'Modern' filter design. We owe a lot to the guy, so it's a shame that his name is now only associated with a simple stabiliser tagged on the end of audio amplifiers.

David

[Diabolical Artificer]

Take a look at the Boucherot cell too, which is often called a Zobel network by mistake. See - https://www.youtube.com/watch?v=-Pn1odN7DNA&t=958s for a pretty good explanation.

"We owe a lot to the guy, so it's a shame that his name is now only associated with a simple stabiliser " True, his cousin Mr Nobel is far more famous for blowing things up and giving out prizes.

Andy.

[Radio Wrangler]

Originally Posted by mhennessy

Can't say I've met this - any examples?

Unlikely... they involve component values which would be amazingly difficult to fabricate on an IC.

There are techniques for modifying feedback networks so they better tolerate driving capacitive loads. These are features of some opamps and are needed in three terminal regulators.

David

[kalee20]

Originally Posted by Radio Wrangler

Zobel was a major player in the evolution of network theory and filter design, and did a fair chunk of the work which got organised into what's now called 'Modern' filter design. We owe a lot to the guy

Let's hope he was given a Zobel prize

[G6Tanuki]

The chip-with-the-built-in-Zobel-resistrsr - I don't recall the part number but it was a stereo SMD one fitted to a pair of 868MHz wireless-headphones I got secondhand around 15 years back (the base-station transmitter-bit was dead so I converted the headphones to Bluetooth for use with my phone).

At the time I was intrigued at the space-efficient inclusion of the Zobel-resistors into the chip: from memory there were only another 2 or 3 resistors on the headset-PCB.

[Radio Wrangler]

Ah, it would have just been purely resistive, and they'd have tolerated the inefficiency in a low-power amp.

David