Building A Reactive Load for Amplifier Testing

[amirm]

So interest has been expressed in more comprehensive power testing of amplifiers going beyond just resistive load. Audiograph makes such a box for automated testing but the retail is north of $25000. Way too much money for the limited use we will get out of it. But you do get cool graphs out of it:

First question is whether what they are testing is the right set. I personally don't see a reason to go down to 1 ohm.

Is +-30 and +-60 degree phase the right targets? What is typical in speakers?

I don't mind spending the money to build one if others chime in with the design and questions like above.

 

[March Audio]

I think the worst load I have seen measured was 1.8 ohms, a magico speaker from memory. I will do some research. Without further evidence to the contrary my view would be that 2 ohms is adequate to represent 99.9% of real world speaker loads.

Need to do more research regarding phase angle limits.

 

[Hugo9000]

The Apogee Scintilla was infamously difficult, wasn't it? A long time ago, however, and the company is long dead.

 

[RayDunzl]

Is +-30 and +-60 degree phase the right targets? What is typical in speakers?

A model immediately preceding but likely similar to mine...

 

[March Audio]

A model immediately preceding but likely similar to mine...

View attachment 28525

Holy moly that's pathological!

A bit OT but my initial reaction when I see stuff like that is that it's bad design that should be avoided rather than expecting an amplifier to cope. Luckily at 20 kHz there won't be any significant content so perhaps not such a huge problem.

 

[GrimSurfer]

Perhaps the issue of going down to 1 ohm isn't to find a load comparable to a tough to drive speaker (even though Ray found an example of how low things can go).

Maybe it is a way to ascertain the ultimate capacity of an amp's power supply.

Dunno...

My thoughts are that being able to test the limits of an amp's power supply would be a worthwhile undertaking. There is a lot of stuff out there over rated (on the output side) and under supplied (on the input side). The ability to quantify power limitations would likely tie in to output limitations, regardless of the topology of the amp.

 

[MZKM]

Phase almost never goes past +/-45°, don’t know what’s typical, just that almost no speaker I’ve ever seen has gone past that (45° requires 2x the heat dissipation).

Some electrostatics and some regular loudspeakers go below 2ohm in the bass, but it is rare.

My question is on impedance >8ohm. Some speakers go to 16ohm, 32ohm, or even higher. I know that from 8ohm to 4ohm the wattage should double and most amps do at least 1.5x, but how do most amps handle 32ohm, do most all come close to putting out 1/4x the wattage relative to 8ohm?

Also, going off Stereophile’s simulated loudspeaker, when including 6ft of speaker cables, an output impedance of 0.2ohm (common) results in about +/-2dB variation in frequency response (and 0.25ohm results in about +/-2.5dB variation).

 

[Wombat]

FYI: Stereophile circuit

Duncan's

 

[March Audio]

Perhaps the issue of going down to 1 ohm isn't to find a load comparable to a tough to drive speaker (even though Ray found an example of how low things can go).

Maybe it is a way to ascertain the ultimate capacity of an amp's power supply.

Dunno...

My thoughts are that being able to test the limits of an amp's power supply would be a worthwhile undertaking. There is a lot of stuff out there over rated (on the output side) and under supplied (on the input side). The ability to quantify power limitations would likely tie in to output limitations, regardless of the topology of the amp.

Even if it would never be exercised in the real world?

I think there is a trade off been cost, complexity and functionality. Perhaps a better initial question would be do we want to try and cover all conceivable situations, or just the 99%.

 

[March Audio]

Phase almost never goes past +/-45°, don’t know what’s typical, just that almost no speaker I’ve ever seen has gone past that (45° requires 2x the heat dissipation).

Some electrostatics and some regular loudspeakers go below 2ohm in the bass, but it is rare.

My question is on impedance >8ohm. Some speakers go to 16ohm, 32ohm, or even higher. I know that from 8ohm to 4ohm the wattage should double and most amps do at least 1.5x, but how do most amps handle 32ohm, do most all come close to putting out 1/4x the wattage relative to 8ohm?

Also, going off Stereophile’s simulated loudspeaker, when including 6ft of speaker cables, an output impedance of 0.2ohm (common) results in about +/-2dB variation in frequency response (and 0.25ohm results in about +/-2.5dB variation).

With the higher impedance you do run into voltage output limitations

 

[March Audio]

FYI: Stereophile circuit

Duncan's

This will be more complex as we need to switch in different combinations of LCR.

 

[DDF]

Woofers show minimum impedance about 1 to 2 octaves above box fb, before inductive rise. I thinks it reasonable to use a model approximating a 2.5 design, dipping to worst case 2 to 3 ohms around 100 Hz. This also happens to be where average music shows peak power spectral density, long term averaged, so its a legitimate stress test.

When I design crossovers its easy to reach 60 deg impedance phase angles through crossover but I always endevour to cap it around 30 deg if possible, though it may lead to some small sacrifices in response or a larger component count. I think it would be easy to find designs achieving highly reactive impedances but Im not sure its reasonable to expect an amp to shine under such a load. I agree with @March Audio , its worth combing through Stereophile measures and nailing it down.

I wouldnt worry about 2 ohms at 20 kHz, thats just a silly design.

 

[Blumlein 88]

I think you'll find it very rare for loads to be past 60 degrees. We don't need to go to worst ever load to get a very good read on how an amp will handle tough loads. And to see if lots of amps fold when the going gets tough. I do find it not terribly uncommon for a few loads to get near 2 ohms. I don't see a need to test lower though I own a speaker that gets to that level. Maybe even 4 ohms is fine with the reactance included.

Using an example might simplify it some. Find a fairly terrible ported speaker and use it to make one tough load. Pick one fairly tough ESL-like capacitive load and use it. Then do pure resistive testing to 2 ohms.

 

[amirm]

Woofers show minimum impedance about 1 to 2 octaves above box fb, before inductive rise. I thinks it reasonable to use a model approximating a 2.5 design, dipping to worst case 2 to 3 ohms around 100 Hz. This also happens to be where average music shows peak power spectral density, long term averaged, so its a legitimate stress test.

To be clear, this will be a single frequency load. So we need to pick a frequency (1 kHz?) and then design the load around it.

 

[restorer-john]

I've been thinking about this for a while myself.

Some thoughts:

• The cube tests power (CEA burst) at 1KHz at 1% which is well into clipping. As such, the loads and their associated wiring, connections and terminals are not as critical as they would be if you were using the same loads for low level THD testing. Costs will be lower.
• Continuous testing means Rs of at least 300W, CEA burst is way lower requirements. For people wanting just one load box for both continuous and burst, higher power parts will be needed, plus much higher quality connections to the main R loads.
• +/-30 and +/-60 L and C plus 8/4/2 ohms should be sufficient. I see no problem extending it to 1 ohm R only.
• R loads can be series/paralleled to reduce parts count and have more accurate loads
• 2 channels, required for single channel and both channels driven- that is essential. That doubles the parts.
• Various control for the unit- automated perhaps stepping through each combination and using the logging and analyzing functions of the AP, then extract the data points and create the 3D plot of 20 points. (I know nothing about the interface or application environment for controlling/interfacing the AP as I don't own one)
• Manual control via rotary selection driving relays for spot manual testing for us guys without APs. Perhaps a 'next' step button to move to the next combination.
• temperature controlled fan/s to keep the unit/loads cool
• Logic driving high powered low contact resistance relays for all combinations.
• Having a second surface plot layered in the main 3D plot that shows single vs both channels driven- it would highlight inadequate PSUs.

 

[amirm]

The cube tests power (CEA burst) at 1KHz at 1% which is well into clipping. As such, the loads and their associated wiring, connections and terminals are not as critical as they would be if you were using the same loads for low level THD testing. Costs will be lower.

That's a good point. Solves a lot of issues I had with building a high-performance load.

Should this be our target though?

 

[restorer-john]

Should this be our target though?

Good question.

It will allow like for like comparisons with actual power cube tests already out there and it will add value to the tests you do.
You can still do all your precision THD testing on your existing high accuracy loads.

The power cube tests are about raw power into difficult loads and a relatively low resolution 3D representation of that. Personally, I'm not a big fan of burst numbers taking all the limelight, but they are one of several valid tests. They are also much kinder on loan gear. In other words, you won't be blowing other people's gear up as often.

 

[Blumlein 88]

Well if you are going to limit frequencies I think you still need more than one. Maybe 50 hz, 1 khz and 10 khz. Among other things I think you'll find reactive loads will alter the FR of amps. I think this contributes to a different sound even if the amp isn't fully over-driven or overwhelmed.

 

[Wombat]

KISS - Component swapping:

Good enough for most electrical Labs.

 

[restorer-john]

Well if you are going to limit frequencies I think you still need more than one. Maybe 50 hz, 1 khz and 10 khz

It gets crazy with all the combinations and the component count will go through the roof.