How is amplifier frequency response measured?

[escape2]

Sorry if this was addressed somewhere earlier - not sure how to search for this - maybe there is a thread that describes this that someone can point me to?

How is frequency response of an amplifier measured? I understand it's done with a frequency sweep generator, but is it done by subjecting the amp to a specific load/impedance, such as 4 Ohm, 8 Ohm, etc? And if so, is that load kept constant across entire frequency spectrum? We all know that various speakers have greatly varying impedance across their frequency response, and so wouldn't this varying impedance affect how much power an amp can deliver at each one of those frequencies and thus altering its frequency response?

 

[DonH56]

Usually with a resistive load (e.g. 4 or 8 ohms) and at 1 W (full-power frequency response is also usually tested and specified).

Yes it can vary with load (speaker) impedance depending upon the amplifier's output impedance.

This is discussed quite often, probably nearly every amplifier and many speaker threads, and there is a short article here: https://www.audiosciencereview.com/...amping-factor-and-speakers.23968/#post-807327

 

[escape2]

Yes it can vary with load (speaker) impedance depending upon the amplifier's output impedance.

Great write up and work. Exactly what I was looking for. Thank you.

So another words, these flat-looking amplifier frequency response graphs that Amir publishes are of limited value, but then it wouldn't be practical for him to test how an amp behaves with every possible speaker out there.

 

[DVDdoug]

, but is it done by subjecting the amp to a specific load/impedance, such as 4 Ohm, 8 Ohm, etc? And if so, is that load kept constant across entire frequency spectrum?

Generally, yes.

We all know that various speakers have greatly varying impedance across their frequency response, and so wouldn't this varying impedance affect how much power an amp can deliver at each one of those frequencies and thus altering its frequency response?

The simple answer is that amplifiers are designed for "constant voltage" (voltage that's independent of the load) and they are usually very-good at that. If there is a published gain spec, that's the voltage gain (often expressed in dB).

Speakers (and headphones) are also tested/specified with a constant voltage source so the actual power (wattage) varies across the frequency range. This is not a concern unless the speaker impedance drops to the point of over-stressing the amplifier by "pulling" too much current and power.

Any variation with load is a function of the amplifier's output impedance (or source impedance) which is in series with the load and effectively makes a Voltage Divider where the voltage is proportional to the impedance. There is sometimes a published spec called Damping Factor which is the ratio of the load impedance to source impedance. (They never directly give you the output impedance, only the recommended load impedance.) A damping factor of around 1000 is common which means you are loosing 1/1000th of the signal internally.

Source impedance is often higher (worse damping factor) with tube amps so it's more common for tube amps to interact with different speakers in unpredictable ways. And sometimes it's an issue with headphone amplifiers.

 

[DonH56]

Great write up and work. Exactly what I was looking for. Thank you.

So another words, these flat-looking amplifier frequency response graphs that Amir publishes are of limited value, but then it wouldn't be practical for him to test how an amp behaves with every possible speaker out there.

Most SS amps have low enough output impedance that variation with typical speaker loads is minimal. There are always cases of speakers with widely varying and very load impedances, natch. As mentioned above, tube amplifiers tend to have much higher output impedance and thus greater load (speaker) sensitivity.

Amir looked into getting a specialized load that would generate performance over a ranger of impedances (the Power Cube) but it is very expensive and somewhat limited in power handling capability.

 

[pma]

Most SS amps have low enough output impedance that variation with typical speaker loads is minimal.

There is a general exception - class D amplifiers without post filter feedback. In fact most of the class D amps on the market. Due to output LC their FR is grossly modulated at higher frequencies, depending on the speaker complex impedance. And yes, there is a proof in ABX DBT. Many ASR members use these cheaper class D amps.

 

[JWAmerica]

Great write up and work. Exactly what I was looking for. Thank you.

So another words, these flat-looking amplifier frequency response graphs that Amir publishes are of limited value, but then it wouldn't be practical for him to test how an amp behaves with every possible speaker out there.

Do you see a lot of 2ohm or 32ohm speakers? Testing 4ohm and 8ohm covers the vast majority.

 

[JWAmerica]

There is a general exception - class D amplifiers without post filter feedback. In fact most of the class D amps on the market. Due to output LC their FR is grossly modulated at higher frequencies, depending on the speaker complex impedance. And yes, there is a proof in ABX DBT. Many ASR members use these cheaper class D amps.

View attachment 148245

Would be nice if manufacturers would do something about this as it's been a known issue for a while.
https://www.aes.org/e-lib/browse.cfm?elib=15210

But implementation has only begun recently
https://www.eeweb.com/post-filter-feedback/

And my tweeters have a strong peak at 27khz.

 

[escape2]

Do you see a lot of 2ohm or 32ohm speakers? Testing 4ohm and 8ohm covers the vast majority.

That was not my point. My point was that impedance for one speaker greatly varies across its frequency response spectrum, from as high as 20+ Ohm to as low as 3 Ohm. So why are we using constant resistive load for testing when no speaker actually poses such a constant load on the amp?

 

[JWAmerica]

That was not my point. My point was that impedance for one speaker greatly varies across its frequency response spectrum, from as high as 20+ Ohm to as low as 3 Ohm. So why are we using constant resistive load for testing when no speaker actually poses such a constant load on the amp?

It would be an interesting experiment, but I think we'll find only fundamentally flawed designs will produce the aberrations you're looking for, which we can likely see symptoms of in other measurements.

 

[kukocz]

Do you see a lot of 2ohm or 32ohm speakers? Testing 4ohm and 8ohm covers the vast majority.

That was not my point. My point was that impedance for one speaker greatly varies across its frequency response spectrum, from as high as 20+ Ohm to as low as 3 Ohm. So why are we using constant resistive load for testing when no speaker actually poses such a constant load on the amp?

A good example is Tannoy XT8F:

There’s almost 40 ohm difference across the frequency range.

 

[Annoyed82]

If you're comparing receivers or external amps with different frequency response ranges , say one is 20Hz-20Khz and another is 10Hz-25KHz-is there going to be a noticeable difference in sound between the two, all other things being equal?

 

[escape2]

If you're comparing receivers or external amps with different frequency response ranges , say one is 20Hz-20Khz and another is 10Hz-25KHz-is there going to be a noticeable difference in sound between the two, all other things being equal?

All other things being equal, no. Both of these already cover the full frequency range that a human ear can hear.

 

[mdsimon2]

To the original point of this thread, at some point @amirm was testing with a complex load but I have not seen those measurements in a while. Any reason why this test is no longer included? @pma's testing of the AIYIMA 07 has been very interesting to me as it definitely points to audible frequency deviations.

Michael

 

[RayDunzl]

My soon-to-be-antique not-green Krell FPB 350mcx into a MartinLogan reQuest.

"The usual REW measurements at an calculated estimated measured 4.72V across the output for 5W into the 4 ohm speaker load."

Original post here...



Stereophile Impedance measurements from the most recent prior version of the speaker.

 

[BoredErica]

There is a general exception - class D amplifiers without post filter feedback. In fact most of the class D amps on the market. Due to output LC their FR is grossly modulated at higher frequencies, depending on the speaker complex impedance. And yes, there is a proof in ABX DBT. Many ASR members use these cheaper class D amps.

Saw your post here: https://www.audiosciencereview.com/...07-tpa3255-review-amplifier.18984/post-634199

Is FR performance at a steady state 4 or 8ohm at all indicative of what performance they would have in a complex speaker load?


edit:
oh you seem to say here (https://www.audiosciencereview.com/...pping-pa5-review-amplifier.28512/post-1010512) that yes, 'with some difficulties'.