Theory of Load Isolators?

Hi,

I am looking at the datasheet for the Jensen JT-OLI-3. It shows a 3.7uH inductor parallel to a 39 Ohm resistor.

The text says:

The load isolator is designed to isolate op-amp outputs from capacitive loads
such as a long cables. Its series impedance is about 0.3 S across the audio
spectrum, but increases to about 40 ohms at RF frequencies to prevent amplifier
instability or oscillation while preserving a low audio output impedance.

Click to expand...

Then I see the specs and charts. It does not explain how the device works. If I were an engineer I'm sure it would make perfect sense. Does anybody care to explain to me exactly how this thing works?

Some math would be nice to see also if it doesn't go much beyond basic algebra... :?

Thanks!

its a low pass filter, looks like the resistance of the inductors wire to low (audio bandwidth) frequencies and 39 ohms to higher frequencies, helps to prevent HF oscillation

I think of it more as a raising impedance at high frequencies so that from the output device's perspective, load capacitance essentially has an R in series with it.

mike p

> It does not explain how the device works.

????

"The load isolator is designed to isolate op-amp outputs from capacitive loads such as a long cables. Its series impedance is about 0.3 S across the audio spectrum, but increases to about 40 ohms at RF frequencies to prevent amplifier instability or oscillation while preserving a low audio output impedance."

That's How It Works!

Maybe you are not understanding the problem?

Pro Audio sometimes drives loooong lines. Just for an example (not worst-case) say 300 feet or 100 meters. That has about 10,000pFd of capacitance.

What is the impedance of 10,000pFd?

> Some math would be nice to see also if it doesn't go much beyond basic algebra...

You can use simple algebra, or just use a Reactance Chart.

I happen to recall that 10,000pFd is 1K ohms at 17KHz. Capacitive impedance falls with frequency. So it must be 10K at 1.7KHz, 100 ohms at 170KHz, 10 ohms at 1.7MHz.

The 990 amp can drive low impedances, but not 10 ohms.

We don't put much 1.7MHz into our amps, but amps like the 990 -will- amplify past 10MHz. And we apply feedback. We must be sure the amp is not overloaded at MHz or it will oscillate. (A full explanation of feedback stability is too much algebra to get into here.)

OK, the 990 will drive 75 ohms, put a 75 ohm load between its output and whatever strange loads we need it to drive. Then, worst-case, it "sees" 75 ohms.

Ah, but some loads want "zero impedance" drive. Transformers give best bass response and lowest bass distortion with very low impedance sources. Splitters to drive an unknown number of loads should have very low source impedance so levels do not vary with number of (possibly shorted) loads.

Jensen's approach (also found on many many loudspeaker amps) is a parallel L-R combination. 4uH at 20KHz is just 0.5 ohms, which is "nearly zero" for our purposes. Of course it is 0.05 ohms at 2KHz. At 200KHz it is 5 ohms, still reasonably low. At 2MHz it would be 50 ohms; however it has 39 ohms resistance in parallel, so it actually tapers slowly to 39 ohms at several MHz.

So for the common audio loads, resistance in parallel with capacitance, the "isolator" allows great drive over the audio band, but decouples the capacitance so it does not strain the amplifier in the ultrasonic range.

A different way to look at it: the combination of 39 ohms, 4uH, 150-600 ohm loads, and whatever capacitance, forms a "crossover network" as used in speakers. Everything below about 500KHz is "woofer" and lands in the 150-600 ohm audio load, with "good damping". Anything higher lands in the 39 ohms, which Jensen has figured is safe for their amplifiers.

While perhaps not the "how" but "why" for output buildout impedances.

When ever we use negative feedback around a gain stage there are stability constraints associated with phase shift (lag) in that feedback path. If a large capacitance is hung directly from the output of a typical opamp, it will form a pole with the opamps output impedance that cause the feedback to lag and can cause the opamp to oscillate.

A small resistor is typically adequate to prevent oscillation but a resistor in parallel with an inductor will deliver lower source impedance at audio frequency, with similar protection at HF.

JR

PRR, you forgot to mention inductance of cables. Actually, they represent both distributed capacitance and inductance, and have a transfer impedance that depends both on capacitance per length and on inductance per length of the cable. On frequencies where length of cable is comparable to length of the wave we can't say about which reactance is dominant without measuring the cable length; rather we have to load the cable on both ends with certain resistances, otherwise reflections will occur from points of an impedance mismatch causing comb-filter effect.

Here is a good article that describes the theory of long lines:

http://www.epanorama.net/documents/wiring/cable_impedance.html

However, for short cables of 100 meters of length or less on audio frequencies capacitive reactance will dominate, so we may assume a capacitive load of opamps driving cables.

Opamp has output resistance, that with capacitance of a cable will form a R-C network (so called integrating network), or low-pass-filter. What is worse, output resistaqnce of an opamp is non-linear, it means that the lower is load resistance, the more distorted will be output signal.
Also, such a network consisting of a resistance and capacitance shifts phase up to -90 degrees. Now, assume that every transistor stage shifts phase on high frequencies, sum all phase shifts, and get 180 degrees starting from some frequency. If on that frequency an amplifying factor in the feedback loop (you remember, that opamps in audio are always used with a negative feedback) is greater than 1 AND phase shift is 180 degrees it means that on such a frequency a negative feedback becomes positive and your amp starts to oscillate! I.e. instead of just amplifying an audio signal it also generates a radio frequency signal!
Now, since output resistance of opamp is non-linear, as I mentioned before, an amplitude of oscillations will be changed by an audio signal, so as the result you will hear DISTORTIONS!

To prevent such oscillations a cable may be connected through an inductance that has the higher impedance the higher is frequency, right?

Not so simple... Inductance and capacitance in series form a L-C contour that on certain frequency has almost zero impedance, i.e. output of the opamp will be almost shorted on certain frequency. Below this "tuning frequency" a capacitive part will dominate... Now, you will have a radio transmitter that works on more stable frequency than before, firing distorted audio signal around.

What to do, how to solve the problem?

Add a resistor that will absorb an energy from the contour making it's selectivity less sharper. Either in series, or in parallel. As the result, a phase-angle Q of the L-C contour will be lowered, it is exactly what that guys are doing adding what you call "an isolator". Connecting in series you will increase output resistance of your amp. Connecting in parallel with capacitance of the cable you load your amp on audio frequencies. Connecting in parallel with an inductance you lower impedance on very high frequency that is inaudible. Such a simple theory. :grin:

Wow, this is hugely helpful. It sounds to me like the isolator works in a very similar manner to a ferrite bead.

PRR
You can use simple algebra, or just use a Reactance Chart.

Click to expand...

I'll do a google search for this...

PRR
A different way to look at it: the combination of 39 ohms, 4uH, 150-600 ohm loads, and whatever capacitance, forms a "crossover network" as used in speakers. Everything below about 500KHz is "woofer" and lands in the 150-600 ohm audio load, with "good damping".

Click to expand...

This paints an excellent picture for me. I think all I needed to do was look at this from a more basic point of view than I was.

JR
When ever we use negative feedback around a gain stage there are stability constraints associated with phase shift (lag) in that feedback path. If a large capacitance is hung directly from the output of a typical opamp, it will form a pole with the opamps output impedance that cause the feedback to lag and can cause the opamp to oscillate.

Click to expand...

Now I see where the "isolation factor" comes into play. I suppose with something like the 990, use of an isolator allows for a smaller value compensation cap in the feedback to give a higher stable bandwidth.

WB
Now, you will have a radio transmitter that works on more stable frequency than before, firing distorted audio signal around.

What to do, how to solve the problem?

Click to expand...

I've never thought of that as being a problem :wink: :wink:

WB
Connecting in parallel with an inductance you lower impedance on very high frequency that is inaudible. Such a simple theory.

Click to expand...

Yes it is, and very much like the ferrite bead.
Transformers and inductance in general aren't things I am used to working with. They are rarely seen in analog synthesizers which is what I tinker on mostly. Anytime I see anything coil-like it tends to be somewhat mysterious to me.

This has all been very educational to me. Thank you all for taking the time to enlighten me.

[quote author="somnium7"]

WB
Now, you will have a radio transmitter that works on more stable frequency than before, firing distorted audio signal around.

What to do, how to solve the problem?

Click to expand...

I've never thought of that as being a problem :wink: :wink:

[/quote]

The problem is not in radio waves, the problem is in distortions: you may assume a shape of the transfer function of opamp close to letter S. It means, that oscillations will have less amplitude (or even stopped) close to zero volt and close to power rails. They impact on the transfer function making it more curved (distortions increased!) also, demodulated they will add more garbage to your sound.

I found out (by experimentation) that cheapo tl071 / 061 often used in guitar-related onboard electronics, which have inbuilt output resistors (iirc 128ohm for 071) require a simple but effective output circuitry to "not suck too much tone": 10-22uF elco in circa 15-22k shunt-to-ground, 560-680ohm in series to output jack, but shunted with a 100-220pF to ground. Problem free, cheap, stage/studio, goes well into DIboxes, consoles, preamps...
Don't know why "pro gear" designers often use diff. configuration which to me sounds worse.

but that's OT.

[quote author="tv"]
Don't know why "pro gear" designers often use diff. configuration which to me sounds worse.
[/quote]

Because it:
1. measures better using common methods
2. simpler to design using cut-n-paste
3. inertia of thinking
4. lack of understanding