How amplifiers affect how speakers sound?

[oal]

“In fact, it will be shown that loudspeakers are the single most important element in sound reproduction. Electronic devices, analog and digital, are also in signal paths, but it is not difficult to demonstrate that in competently designed products, and effects they may have are small if they are not driven into gross distortion or clipping.”
(Dr. Floyd E. Toole: “SOUND REPRODUCTION”, p.16)

This sentence matches my experience. I can tell the difference between two different types of speakers, but it is rare that I notice a difference when I change the DAC or amplifier, as long as I keep the sound pressure the same. So, I tried a simple objective experiment - comparing the frequency response and distortion of the same speaker with the three amplifiers. Let me share the results.

1. Measurement Equipment and Software

• ADC: Cosmos ADC (for the amps)
• Dummy Load: 4Ω/200W x 2 (for each channel of the amps)
• LPF: DIY Low Pass Filter (LPF used for the digital amps TOPPING MX3s and Sylph-D100 P02)
• Microphone: miniDSP UMIK-1 (for the speaker)
• Software: Room EQ Wizard (REW) Pro

2. Equipment under test

DAC: TOPPING D30Pro. Here is the review of the speaker by Amir.
https://www.audiosciencereview.com/forum/index.php?threads/topping-d30pro-review-balanced-dac.20259/
This is an excellent DAC with SINAD = 120dB, used for all measurements.

Speaker: Revel M105, a flat response and low distortion speaker. Here is the review of the speaker by Amir.
https://www.audiosciencereview.com/...ds/revel-m105-bookshelf-speaker-review.14745/
General Specifications of Revel M105 are:
https://www.revelspeakers.com/products/types/bookshelf/M105-.html

Revel M105’s Nominal Impedance is 8Ω (by specs), measured minimum impedance is 3.8Ω and measured impedance at 1kHz is 10Ω (by Amir’s measurements). Its Sensitivity (2.83V/1m) is 86dB (by specs).

Amplifier 1: Marantz NR1200. Here is the review of the amp by me.
https://www.audiosciencereview.com/.../marantz-nr1200-stereo-receiver-review.44685/

Amplifier 2: TOPPING MX3s. Here are the specs.
https://www.toppingaudio.com/product-item/mx3s
TOPPING MX3s has Low and High Gain modes. I used the Low Gain mode for the test.

Amplifier 3: Sylph-D100 P02. Here is the review of the amp by me.
https://www.audiosciencereview.com/...sylph-d100-p02-amplifier-module-review.46613/

Let’s compare the FFT Spectrum (5W into 4Ω) of the three amps. As you see, Marantz NR1200 has ‘Fair’ level SINAD and a mains power leak. TOPPING MX3s has ‘Good’ level SINAD but its noise floor looks higher than NR1200’s. Sylph-D100 P02 has ‘Excellent’ level SINAD, it is clean in both noise and distortion.

I don't usually listen to music at a very loud volume. It's less than 80dB SPL, about 2 meters away from the speakers. I think the amplifier's output power is less than 2W per channel. Here are the curves of THD+N vs Power, the cursor is at 1W.

Then, THD vs Power, again, the cursor is at 1W.

As you can see, there is a difference in the performance of the three amplifiers, ranging from Fair to Excellent.

3. Measurement
To measure frequency response and distortion of Revel M105, I used “Make a measurement” in REW.
https://www.roomeqwizard.com/help/help_en-GB/html/makingmeasurements.html
With the “Check levels” button and volume knob of the pre-amp, I set the Level to 80dB SPL at 1 meter from the speaker.

Three measurements were taken (Take1, Take2 and Take3) for each amplifier so that measurement errors (or run to run variations) and differences between amplifiers could be distinguished. Here are the curves of Frequency Response, the cursor is at 1KHz, 80dB SPL. The measurements were taken in my living room; of course there are standing waves below 200 Hz, but ignore them and focus on the differences between the speakers. I think the differences are small.

Next, THD. Here are the curves of THD vs Frequency (%). Distortions of less than 1% will be barely audible. There seems to be a somewhat large gap only at 220 Hz. The Shlph-D100 has less than 1% distortion (-42dB or 0.8%), while the NR1200 and MX3s have more than 1% (-37dB or 1.4%), a difference of about 5dB.

Let's change the vertical axis to absolute sound pressure (dB SPL) so that we can see the volume of the distortion. The cursor is at 1kHz, 40dB SPL (average signal level 80dB SPL minus 40dB). THD at 220 Hz is less than 40 dB SPL. Not only will you not be able to tell the difference in distortion, but distortion itself is not audible. Overall, the NR1200 distortion is larger than the other amplifiers’, but the differences are small.

4. Conclusion

I listened to the speaker subjectively and could not tell the differences by amplifiers, but when I measured them objectively, I could see the differences. However, the differences between Fair and Excellent amps are small and would not be audible to most people.

Obviously, amplifiers with serious defects (such as significant noise or distortion, non-flat FR, or channel mismatch) should be avoided, but a Fair level amp works well.

I would have just re-examined what many people already know. However, I hope that the results of this experiment are helpful.

 

[Old_School_Brad]

Sometimes, conducting these tests yourself serves as a good sanity check -I’ve done it too. Even when I know the outcome, I still occasionally perform blind tests on DACs.
It’s a helpful way to stay grounded, in my opinion.

 

[Matt_Holland]

Thanks for putting this together.
I think we all agree that when competent amplifiers are used within their limits, they sound the same. However, I’d like to see comparisons of amplifiers’ behaviour close to and into their non-linear operating regions. This is where I suspect people hear differences, even if subtle. I also think that using amplifiers at their limits (even only momentarily) is a typical use case and so we should consider their behaviour at their limits.

 

[Kolllak]

Can you also check multitone with Umik?
I compared Arcam SA30 with Topping PA7 Plus (KEF R3 Meta) and got the same THD via sweeps but some deltas for multitone.
Something like the pic below.

THD is mostly the same for both.

 

[Mart68]

Let's see a load that drops to 2 ohm, I think we will see audible issues then.

 

[Matt_Holland]

Can you also check multitone with Umik?
I compared Arcam SA30 with Topping PA7 Plus (KEF R3 Meta) and got the same THD via sweeps but some deltas for multitone.
Something like the pic below.

THD is mostly the same for both.

Is it possible that some people can hear this difference?

 

[Blumlein 88]

Thanks for putting this together.
I think we all agree that when competent amplifiers are used within their limits, they sound the same. However, I’d like to see comparisons of amplifiers’ behaviour close to and into their non-linear operating regions. This is where I suspect people hear differences, even if subtle. I also think that using amplifiers at their limits (even only momentarily) is a typical use case and so we should consider their behaviour at their limits.

With amplifier power so cheap these days why use an amp pushed to or over its limits? If you are hearing your amp when pushed too far, don't, get a properly sized amp. We've had this idea for at least 50 years that we hear differences when amps get into the non-linear region. Yet why waste time on that, the fix is easy. Get a bigger amp.

 

[Matt_Holland]

With amplifier power so cheap these days why use an amp pushed to or over its limits? If you are hearing your amp when pushed too far, don't, get a properly sized amp. We've had this idea for at least 50 years that we hear differences when amps get into the non-linear region. Yet why waste time on that, the fix is easy. Get a bigger amp.

I agree. But in reality, a lot of people love their underpowered amplifiers and claim their sonic virtues even though these are due to characteristic distortion, etc.

 

[solderdude]

Is it possible that some people can hear this difference?

There is something called 'masking effect' which prevents softer sounds near a stimulus not to be heard. Lossy codecs make use of this effect.
Based on that a 40dB difference in noise floor (unknown what causes that) between loud tones flanking that noise floor is extremely unlikely to be audible.

 

[Old_School_Brad]

I agree. But in reality, a lot of people love their underpowered amplifiers and claim their sonic virtues even though these are due to characteristic distortion, etc.

It's difficult to determine what's happening without examining each case individually, as there are numerous variables and potential explanations for it. The most common explanation is likely some form of perception bias, even if it is unrecognized and unacknowledged by the individual.

 

[Speedskater]

Using a microphone to measure power amplifiers, just adds overlooked uncontrolled variables.
After setting your volume levels, measure the response at the loudspeaker's cable terminals.

 

[Petrushka]

Thanks for putting this together.
I think we all agree that when competent amplifiers are used within their limits, they sound the same. However, I’d like to see comparisons of amplifiers’ behaviour close to and into their non-linear operating regions. This is where I suspect people hear differences, even if subtle. I also think that using amplifiers at their limits (even only momentarily) is a typical use case and so we should consider their behaviour at their limits.

I seem to recall at least one amp, the Wiim Pro, won’t clip, because the digital circuitry limits the output. I don’t know how this sounds, but it seem like a good idea, if only for protection.

 

[Matt_Holland]

I seem to recall at least one amp, the Wiim Pro, won’t clip, because the digital circuitry limits the output. I don’t know how this sounds, but it seem like a good idea, if only for protection.

I have a Wiim amp driving Revel M16, listening position is about 3m from them. For fun playback levels I’m at about 80% volume but i wouldn’t say it’s crazy loud. After that the amp starts to sound a little strained. Once it accidentally played at 100% and it wasn’t badly distorted so maybe there is so limiting.

 

[pogo]

Reading this, I don't think the Umik-1 is a suitable microphone: Link

 

[milosz]

At levels below the onset of clipping, I've never seen any competently done blind testing to show that ANY listeners can hear differences between properly operating mid-fi or high-end solid state amps on speakers whose impedance magnitude/phase isn't "difficult."

In my experience there are some speakers with what I consider "wacky" impedance characteristics that bring out audibly bad characteristics in some amps, which is why I'm glad that Amir has been testing amps into loads of varying reactance.

In addition amps with very low output source impedance can exhibit audible differences in bass quality / response compared to other amps with some speakers, this is related to damping factor. But really most if not all solid state amps have very low output source impedance / high damping factor these days so again there is very little difference - if any - to be heard in terms of bass quality, absent consideration power delivery as a more powerful amp generally makes bassheads happy.

Now, with tube amplifiers things are quite different. Their generally high output source impedance typically interacts with the speaker/crossover impedance to create a frequency response that is NOT flat, even though the amp measures flat into a dummy load. I think that this is largely responsible for "tube sound," along with how tube amps distort near their clipping level- they have fewer higher order harmonics in their output as clipping nears compare to solid state amps which produce essentially square waves at clipping with harmonics up into the AM broadcast band.

In this graph, the black line shows the frequency response of this tube amplifier when driving the impedance presented by a typical speaker system - you can see it's NOT flat. This kind of departure from transparency is audible to most people capable of critical listening. Most any solid state amp would show much lower deviation from flat response.

This is an Audio Research I/50 tube amp as reviewed by John Atkinson of Stereophile, who says "...The 4 ohm transformer tap's source impedance at low and middle frequencies was low for a tube amplifier, at 0.61 ohms, but rose to 2.3 ohms at the top of the audioband. As expected, the 8 ohm tap's source impedance was higher, at 0.97 ohms at 20Hz and 1kHz, rising to 5 ohms at 20kHz..."

Note that a decent solid-state amp will typically have an output source impedance of 0.1 ohms or less- for example Benchmark AHB2 0.09 ohms at 20 Hz.

A lot these speaker-vs-amplifier considerations change when using active speakers in which there is no crossover filter used on the amplifier's output, and although most drivers / enclosures themselves have somewhat complex impedance-frequency characteristics they are nowhere near as broad as the combination of drivers plus crossovers inside a passive speaker. And some drivers have almost flat impedance curves, similar to a length of wire or a plain resistor- planar drivers come to mind.

 

[oal]

Can you also check multitone with Umik?
I compared Arcam SA30 with Topping PA7 Plus (KEF R3 Meta) and got the same THD via sweeps but some deltas for multitone.

Yes, there were also small differences in my multitone measurements. However, they are not signals but noise floor between the tones (35-40dB smaller than the tones). Are the differences not audible as Matt_Holland and solderdude pointed out? Just like the 5dB difference of distortion in the “THD (H2...9) (%) vs Frequency (Hz)” chart.

 

[HaveMeterWillTravel]

“In fact, it will be shown that loudspeakers are the single most important element in sound reproduction. Electronic devices, analog and digital, are also in signal paths, but it is not difficult to demonstrate that in competently designed products, and effects they may have are small if they are not driven into gross distortion or clipping.”
(Dr. Floyd E. Toole: “SOUND REPRODUCTION”, p.16)

This sentence matches my experience. I can tell the difference between two different types of speakers, but it is rare that I notice a difference when I change the DAC or amplifier, as long as I keep the sound pressure the same. So, I tried a simple objective experiment - comparing the frequency response and distortion of the same speaker with the three amplifiers. Let me share the results.

1. Measurement Equipment and Software

• ADC: Cosmos ADC (for the amps)
• Dummy Load: 4Ω/200W x 2 (for each channel of the amps)
• LPF: DIY Low Pass Filter (LPF used for the digital amps TOPPING MX3s and Sylph-D100 P02)
• Microphone: miniDSP UMIK-1 (for the speaker)
• Software: Room EQ Wizard (REW) Pro

2. Equipment under test

DAC: TOPPING D30Pro. Here is the review of the speaker by Amir.
https://www.audiosciencereview.com/forum/index.php?threads/topping-d30pro-review-balanced-dac.20259/
This is an excellent DAC with SINAD = 120dB, used for all measurements.

Speaker: Revel M105, a flat response and low distortion speaker. Here is the review of the speaker by Amir.
https://www.audiosciencereview.com/...ds/revel-m105-bookshelf-speaker-review.14745/
General Specifications of Revel M105 are:
https://www.revelspeakers.com/products/types/bookshelf/M105-.html

Revel M105’s Nominal Impedance is 8Ω (by specs), measured minimum impedance is 3.8Ω and measured impedance at 1kHz is 10Ω (by Amir’s measurements). Its Sensitivity (2.83V/1m) is 86dB (by specs).

Amplifier 1: Marantz NR1200. Here is the review of the amp by me.
https://www.audiosciencereview.com/.../marantz-nr1200-stereo-receiver-review.44685/

Amplifier 2: TOPPING MX3s. Here are the specs.
https://www.toppingaudio.com/product-item/mx3s
TOPPING MX3s has Low and High Gain modes. I used the Low Gain mode for the test.

Amplifier 3: Sylph-D100 P02. Here is the review of the amp by me.
https://www.audiosciencereview.com/...sylph-d100-p02-amplifier-module-review.46613/

Let’s compare the FFT Spectrum (5W into 4Ω) of the three amps. As you see, Marantz NR1200 has ‘Fair’ level SINAD and a mains power leak. TOPPING MX3s has ‘Good’ level SINAD but its noise floor looks higher than NR1200’s. Sylph-D100 P02 has ‘Excellent’ level SINAD, it is clean in both noise and distortion.
View attachment 409067

I don't usually listen to music at a very loud volume. It's less than 80dB SPL, about 2 meters away from the speakers. I think the amplifier's output power is less than 2W per channel. Here are the curves of THD+N vs Power, the cursor is at 1W.
View attachment 409068

Then, THD vs Power, again, the cursor is at 1W.
View attachment 409069

As you can see, there is a difference in the performance of the three amplifiers, ranging from Fair to Excellent.

3. Measurement
To measure frequency response and distortion of Revel M105, I used “Make a measurement” in REW.
https://www.roomeqwizard.com/help/help_en-GB/html/makingmeasurements.html
With the “Check levels” button and volume knob of the pre-amp, I set the Level to 80dB SPL at 1 meter from the speaker.
View attachment 409070

Three measurements were taken (Take1, Take2 and Take3) for each amplifier so that measurement errors (or run to run variations) and differences between amplifiers could be distinguished. Here are the curves of Frequency Response, the cursor is at 1KHz, 80dB SPL. The measurements were taken in my living room; of course there are standing waves below 200 Hz, but ignore them and focus on the differences between the speakers. I think the differences are small.
View attachment 409071

Next, THD. Here are the curves of THD vs Frequency (%). Distortions of less than 1% will be barely audible. There seems to be a somewhat large gap only at 220 Hz. The Shlph-D100 has less than 1% distortion (-42dB or 0.8%), while the NR1200 and MX3s have more than 1% (-37dB or 1.4%), a difference of about 5dB.
View attachment 409072

Let's change the vertical axis to absolute sound pressure (dB SPL) so that we can see the volume of the distortion. The cursor is at 1kHz, 40dB SPL (average signal level 80dB SPL minus 40dB). THD at 220 Hz is less than 40 dB SPL. Not only will you not be able to tell the difference in distortion, but distortion itself is not audible. Overall, the NR1200 distortion is larger than the other amplifiers’, but the differences are small.
View attachment 409073

4. Conclusion

I listened to the speaker subjectively and could not tell the differences by amplifiers, but when I measured them objectively, I could see the differences. However, the differences between Fair and Excellent amps are small and would not be audible to most people.

Obviously, amplifiers with serious defects (such as significant noise or distortion, non-flat FR, or channel mismatch) should be avoided, but a Fair level amp works well.

I would have just re-examined what many people already know. However, I hope that the results of this experiment are helpful.

Do you have any time response data? For example, settling time toggling between maximum V and minimum V?

 

[HaveMeterWillTravel]

Thanks for putting this together.
I think we all agree that when competent amplifiers are used within their limits, they sound the same. However, I’d like to see comparisons of amplifiers’ behaviour close to and into their non-linear operating regions. This is where I suspect people hear differences, even if subtle. I also think that using amplifiers at their limits (even only momentarily) is a typical use case and so we should consider their behaviour at their limits.

Ah, but how do you define "competent"?

 

[HaveMeterWillTravel]

I seem to recall at least one amp, the Wiim Pro, won’t clip, because the digital circuitry limits the output. I don’t know how this sounds, but it seem like a good idea, if only for protection.

That is the premise of McIntosh's power guard. How well it operates, I do not now.

 

[HaveMeterWillTravel]

At levels below the onset of clipping, I've never seen any competently done blind testing to show that ANY listeners can hear differences between properly operating mid-fi or high-end solid state amps on speakers whose impedance magnitude/phase isn't "difficult."

In my experience there are some speakers with what I consider "wacky" impedance characteristics that bring out audibly bad characteristics in some amps, which is why I'm glad that Amir has been testing amps into loads of varying reactance.

In addition amps with very low output source impedance can exhibit audible differences in bass quality / response compared to other amps with some speakers, this is related to damping factor. But really most if not all solid state amps have very low output source impedance / high damping factor these days so again there is very little difference - if any - to be heard in terms of bass quality, absent consideration power delivery as a more powerful amp generally makes bassheads happy.

Now, with tube amplifiers things are quite different. Their generally high output source impedance typically interacts with the speaker/crossover impedance to create a frequency response that is NOT flat, even though the amp measures flat into a dummy load. I think that this is largely responsible for "tube sound," along with how tube amps distort near their clipping level- they have fewer higher order harmonics in their output as clipping nears compare to solid state amps which produce essentially square waves at clipping with harmonics up into the AM broadcast band.

In this graph, the black line shows the frequency response of this tube amplifier when driving the impedance presented by a typical speaker system - you can see it's NOT flat. This kind of departure from transparency is audible to most people capable of critical listening. Most any solid state amp would show much lower deviation from flat response.

This is an Audio Research I/50 tube amp as reviewed by John Atkinson of Stereophile, who says "...The 4 ohm transformer tap's source impedance at low and middle frequencies was low for a tube amplifier, at 0.61 ohms, but rose to 2.3 ohms at the top of the audioband. As expected, the 8 ohm tap's source impedance was higher, at 0.97 ohms at 20Hz and 1kHz, rising to 5 ohms at 20kHz..."

Note that a decent solid-state amp will typically have an output source impedance of 0.1 ohms or less- for example Benchmark AHB2 0.09 ohms at 20 Hz.

View attachment 409349

A lot these speaker-vs-amplifier considerations change when using active speakers in which there is no crossover filter used on the amplifier's output, and although most drivers / enclosures themselves have somewhat complex impedance-frequency characteristics they are nowhere near as broad as the combination of drivers plus crossovers inside a passive speaker. And some drivers have almost flat impedance curves, similar to a length of wire or a plain resistor- planar drivers come to mind.

35 years ago, I had multiple "gimme" conditions of blatantly obvious conditions (not including clipping) which manifested in audible differences.
I enjoyed "breaking" the test.