Cute new schiit speaker amps

[Hatto]

Instead of using a single frequency sine wave, here is what you'll get with a signal that is representative of the characteristics of music -- M-Noise, AES Standard AES75-2022.

AES Standard » AES75-2023: AES standard for acoustics — Measuring loudspeaker maximum linear sound levels using noise

This standard details a procedure for measuring maximum linear sound levels of a loudspeaker system or driver using a test signal called Music-Noise. In order to measure maximum linear sound levels meaningfully and repeatably, a signal is required whose RMS and peak levels as functions of...

www.aes.org

M-Noise

M-Noise is a new test signal that promotes standardized measurement of a loudspeaker system’s maximum linear output.

m-noise.org

M-Noise crest factor = 7.832 (= 17.9 dB). See below (reference):
View attachment 273580

Assuming the short term transient power of the Schiit Rekkr is 1.5x its rated output, its clipping voltage is 6.928 V.

Rated output power (Schiit Rekkr) = 2 W @ 8 ohm​

V_rms @ rated power = sqrt(2 * 8) = 4 V​

V_pk @ rated power = sqrt(2) * V_rms = 5.657 V​

​

Assuming a short term power headroom of 1.5x rated power​

Clipping voltage = V_pk * sqrt(1.5) = 6.928 V​

With M-Noise, for a speaker with 88 dB SPL sensitivity, that means a maximum of 77.9 dB SPL, 1 m, free field, before the on-set of clipping.

M-Noise V_rms = V_pk / Crest Factor = 6.928 / 7.832 = 0.8846 V​

M-Noise SPL at clipping = 20 log10(0.8846/2.83) + 88 = 77.9 dB​

See also Charles Sprinkle's comments on amplifier power rating:

Kali IN-5 Studio Monitor Review

Finally saw it. Looks like we finally got a good and "cheap" monitor. H2 is a bit high, but who cares, H3 stays under 0.5% at 96 dB and under 0.2~0.3% after 500 Hz; not even changing with SPL! And as already said, the diffraction is moved in a much more benign area and with a lower level. Will...

www.audiosciencereview.com

Excerpt:

...​

So which way is the right way to measure power? I have something to say about that. The traditional way of rating "power" as V^2/R with sine waves is in my opinion complete nonsense. Power amplifiers don't amplify power. They amplify voltage. People don't hear watts. They hear SPL. We provide output "power" ratings as a reference, and yes, they are honestly measured and specified as described above. But I hope we all understand that actual power produced into a reactive load with complex impedance using real music or program material is going to be substantially less than headline "power" ratings.​

​

Thank you for the detailed explanations, much appreciated. As long as we have numbers, we can discuss what they actually mean.

I'm not so sure aboutyour last line of calculations:

M-Noise SPL at clipping = 20 log10(0.8846/2.83) + 88 = 77.9 dB

This formula, calculating peak SPL using Vrms is valid for a uniform signal, it shouldn't be applicable to such a non-uniform signal.

Just as you quoted it yourself:

Power amplifiers don't amplify power. They amplify voltage. People don't hear watts. They hear SPL.

For any given Vrms, V_pk (thus peak SPL) can very wildly from one audio signal to the other. Therefore it's not accurate to calculate peak SPL with Vrms alone for coplex audio signals (as opposed to a uniform signal).

So let's look at those voltages more closely:
First of all, I measured Rekkr's clipping voltage as 9V using a digital oscilloscope, so I'll refer to that value rather than 6.928V you assumed. Now, as long as the amplifier provides V_pk=9V, the peak SPL (which is at the point of clipping) should be at that power level, regardless of the shape of the signal. In other words, peak SPL at any given V_pk is a fixed value, becuse they both represent the same ceiling (I don't know how to calculate peak SPL from V_pk, so if anyone knows, any help would be much appreciated).

The only difference between signals with high and low crest factors at equal peak SPL (thus equal V_pk) should be the mean SPL (that of high crest factor signal being lower).
So for a signal with such high crest factor, instead of trying to calculate the peak SPL from Vrms, we should be doing it the other way around: Calculate the mean SPL when V_pk is at clipping point.

Example given:

Above is an audio signal from Rekkr output with a V_pk/Vrms value of 3 (7.4V/2.48V). I don't know what value it corresponds to in terms of crest factor (help, anyone?), but it's certainly lower than M-noise, which has a V_pk/Vrms value of 4.5 at roughly the same peak voltage. This is what M-noise looks like with a very close V_pk:

Vrms = 1.60V compared to the Vrms = 2.48 for the track above for roughly the same V_pk values.
(Actually it was 1.62V excluding the second of silence at the end, but let's discard that error).

I took quick and dirty SPL measurements (at around 1.2m distance) for both tracks using an Android app with signal levels (with almost matching V_pk levels) as shown above.
Music file: 93dB SPL peak and 81dB SPL average
M-noise: 94dB SPL peak and 77dB SPL average
Now I'm not claiming these values are perfectly matching with calculated numbers, but this little experiment shows SPL peak is one and the same for a given V_pk value. It also reflects the difference between average SPL values differing with differing Vrms. The fact that difference between peak and average SPL for M-noise (17dB) matches almost exactly with 17.9dB crest value gives me even more confidence.

Therefore, I believe we can confidently say that peak SPL for a given V_pk is the same regardless of the source (signal). What differs from one signal to the other (one crest factor to the other) is the average SPL for a given peak SPL. Since max V_pk (hence peak SPL max) is determined by the clipping point of the amp, the only thing different between tracks with different dynamic range would be the average SPL at clipping point (again, peak SPL being the same).

So, at what point does "loss of fidelity" occur? It occurs when peak of the audio signal reaches the clipping voltage of the amplifier and you keep on incerasing the volume, at which point you have reached (maybe even exceeded) the peak SPL of your system already.

Here is a high dynamic range music track with even higher crest value than M-noise, with V_pk/Vrms = 5
When there's no clipping, Rekkr can reproduce the entire audio signal under V_pk = 8.6V

Again taking a quick and dirty SPL measurement:
High dynamic range music track: 101dB SPL peak and 81dB SPL average (notice the measured difference is 20dB)
Here is what happens when I push the volume further up:

V_pk doesn't go any higher than 9V but Vrms goes on rising (up to 4.84V in this example), first compressing the dynamic range, then turning the audio signal into utter garbage past the red line.

But none of this ever happens before you hit 101dB SPL peak (or whatever the actual max SPL peak level of the amp is), which is already annoyingly loud in my opinion, and that's the only opinion I have ithroughout this write-up.

I hope this helps. Now, I'm no audio engineer, so it would certainly help me if anyone can show me the error of my ways as they catch them.

 

[NoxMorbis]

This looks like something that should cost 50% of what they are asking for it

They don't even state the power on the Amazon page. Just going on about how much power do you "really" need, and simplicity is beautiful., etc. They make nice stuff, but it's too expensive for what you get.

 

[NTK]

Thank you for the detailed explanations, much appreciated. As long as we have numbers, we can discuss what they actually mean.

I'm not so sure aboutyour last line of calculations:

This formula, calculating peak SPL using Vrms is valid for a uniform signal, it shouldn't be applicable to such a non-uniform signal.

Just as you quoted it yourself:


For any given Vrms, V_pk (thus peak SPL) can very wildly from one audio signal to the other. Therefore it's not accurate to calculate peak SPL with Vrms alone for coplex audio signals (as opposed to a uniform signal).

So let's look at those voltages more closely:
First of all, I measured Rekkr's clipping voltage as 9V using a digital oscilloscope, so I'll refer to that value rather than 6.928V you assumed. Now, as long as the amplifier provides V_pk=9V, the peak SPL (which is at the point of clipping) should be at that power level, regardless of the shape of the signal. In other words, peak SPL at any given V_pk is a fixed value, becuse they both represent the same ceiling (I don't know how to calculate peak SPL from V_pk, so if anyone knows, any help would be much appreciated).

The only difference between signals with high and low crest factors at equal peak SPL (thus equal V_pk) should be the mean SPL (that of high crest factor signal being lower).
So for a signal with such high crest factor, instead of trying to calculate the peak SPL from Vrms, we should be doing it the other way around: Calculate the mean SPL when V_pk is at clipping point.

Example given:
View attachment 273651
Above is an audio signal from Rekkr output with a V_pk/Vrms value of 3 (7.4V/2.48V). I don't know what value it corresponds to in terms of crest factor (help, anyone?), but it's certainly lower than M-noise, which has a V_pk/Vrms value of 4.5 at roughly the same peak voltage. This is what M-noise looks like with a very close V_pk:
View attachment 273650
Vrms = 1.60V compared to the Vrms = 2.48 for the track above for roughly the same V_pk values.
(Actually it was 1.62V excluding the second of silence at the end, but let's discard that error).

I took quick and dirty SPL measurements (at around 1.2m distance) for both tracks using an Android app with signal levels (with almost matching V_pk levels) as shown above.
Music file: 93dB SPL peak and 81dB SPL average
M-noise: 94dB SPL peak and 77dB SPL average
Now I'm not claiming these values are perfectly matching with calculated numbers, but this little experiment shows SPL peak is one and the same for a given V_pk value. It also reflects the difference between average SPL values differing with differing Vrms. The fact that difference between peak and average SPL for M-noise (17dB) matches almost exactly with 17.9dB crest value gives me even more confidence.

Therefore, I believe we can confidently say that peak SPL for a given V_pk is the same regardless of the source (signal). What differs from one signal to the other (one crest factor to the other) is the average SPL for a given peak SPL. Since max V_pk (hence peak SPL max) is determined by the clipping point of the amp, the only thing different between tracks with different dynamic range would be the average SPL at clipping point (again, peak SPL being the same).

So, at what point does "loss of fidelity" occur? It occurs when peak of the audio signal reaches the clipping voltage of the amplifier and you keep on incerasing the volume, at which point you have reached (maybe even exceeded) the peak SPL of your system already.

Here is a high dynamic range music track with even higher crest value than M-noise, with V_pk/Vrms = 5
When there's no clipping, Rekkr can reproduce the entire audio signal under V_pk = 8.6V
View attachment 273660
Again taking a quick and dirty SPL measurement:
High dynamic range music track: 101dB SPL peak and 81dB SPL average (notice the measured difference is 20dB)
Here is what happens when I push the volume further up:
View attachment 273662
V_pk doesn't go any higher than 9V but Vrms goes on rising (up to 4.84V in this example), first compressing the dynamic range, then turning the audio signal into utter garbage past the red line.

But none of this ever happens before you hit 101dB SPL peak (or whatever the actual max SPL peak level of the amp is), which is already annoyingly loud in my opinion, and that's the only opinion I have ithroughout this write-up.

I hope this helps. Now, I'm no audio engineer, so it would certainly help me if anyone can show me the error of my ways as they catch them.

I'm answering your first part first regarding the use of RMS voltage to calculate the SPL.

M-Noise is a steady signal. Being noise, there is randomness, but the signal is "statistically" steady. To show that the RMS voltage of the signal gives its SPL, I'll start with the addition of 2 tones. For reference see:

Total dB level adding of incoherent or noncoherent uncorrelated sound sources signals combine sound levels two sources resultant level audio logarithmic decibel scale decibels or SPL sound pressure level add signal noise levels noncoherent incoherent

Total dB level adding of uncorrelated incoherent noncoherent sound sources signals combine sound levels two sources resultant level audio logarithmic decibel scale decibels or SPL sound pressure level add signal noise levels incoherent random phase acoustic - Eberhard Sengpiel sengpielaudio

www.sengpielaudio.com

The Fourier theorem tells use that any signal can be decomposed into a summation of a series of sinusoidal signals. Therefore, the total SPL is the "complex sum" of each individual component signal. Below show why. (This forum does not support mathematical equations, so I'm using screen captures.)

Notes:

• I only showed sine waves without phase offsets. The above proof can be readily generalized to signal components with phase offsets.
• The integration time T, in the above derivations, is an integer multiple of the periods of the components. However, if T is not an exact multiple but is much larger than the periods of the signals (i.e. the sample include a large number of periods of the signals of interest), the result will approach that from an exact multiple T with small errors. You can look at the signal sample as one that includes many complete periods plus one incomplete period. If the number of complete periods is large, the error from the incomplete period will be small relative to the total.

 

[Hatto]

I'm answering your first part first regarding the use of RMS voltage to calculate the SPL.

M-Noise is a steady signal. Being noise, there is randomness, but the signal is "statistically" steady. To show that the RMS voltage of the signal gives its SPL, I'll start with the addition of 2 tones. For reference see:

Total dB level adding of incoherent or noncoherent uncorrelated sound sources signals combine sound levels two sources resultant level audio logarithmic decibel scale decibels or SPL sound pressure level add signal noise levels noncoherent incoherent

Total dB level adding of uncorrelated incoherent noncoherent sound sources signals combine sound levels two sources resultant level audio logarithmic decibel scale decibels or SPL sound pressure level add signal noise levels incoherent random phase acoustic - Eberhard Sengpiel sengpielaudio

www.sengpielaudio.com

The Fourier theorem tells use that any signal can be decomposed into a summation of a series of sinusoidal signals. Therefore, the total SPL is the "complex sum" of each individual component signal. Below show why. (This forum does not support mathematical equations, so I'm using screen captures.)

View attachment 273815
View attachment 273816
Notes:

• I only showed sine waves without phase offsets. The above proof can be readily generalized to signal components with phase offsets.
• The integration time T, in the above derivations, is an integer multiple of the periods of the components. However, if T is not an exact multiple but is much larger than the periods of the signals (i.e. the sample include a large number of periods of the signals of interest), the result will approach that from an exact multiple T with small errors. You can look at the signal sample as one that includes many complete periods plus one incomplete period. If the number of complete periods is large, the error from the incomplete period will be small relative to the total.

My understanding is that, the SPL referenced in those integrals you shared refer to the average SPL of the signal, rather than the peak values. Otherwise there are (at least) 2 inconsistencies with reality:
1. I have demonstrated above that one can drive the M-noise signal above 77.9dB SPL using even less than 2W(rms), without any indication of clipping.
2. I have demonstrated with actual measurements that 2 audio signals having the same Vrms value can have 2 different peak SPL values which directly correlate with V_pk rather than Vrms.

 

[Trudius]

@Matias In case you want to add these to your spreadsheet, the Gjallahorn has 0.005822% THD+N at 4Ω 5W, or 84.7dB SINAD AVGed across both channels:
View attachment 266959

Power at 1% is 18.04W:
View attachment 266961

The Rekkr on the other hand, doesn't even reach 5W so not sure if/how you'd want to add it.
It can output 3.78W at 1%:
View attachment 266962

Both graphs are from Schiit's measurement reports (also attached below).

I noticed that the company is rating these 2 amps quite conservatively, perhaps they want to overstate the "less better" whatever that means. However if you walk the monoblock level sweep at the end of the pdf for the Rekker, you can see the breaking point at almost exactly 5W @~94dB, so it could get a rating on Matias's chart. At 1% THD+N it does reach ~6.5 W.

 

[staticV3]

However if you walk the monoblock level sweep at the end of the pdf for the Rekker, you can see the breaking point at almost exactly 5W @~94dB, so it could get a rating on Matias's chart.

It can't, because that sweep is at 8Ω and Matias' list only tracks 4Ω data.

 

[NTK]

My understanding is that, the SPL referenced in those integrals you shared refer to the average SPL of the signal, rather than the peak values. Otherwise there are (at least) 2 inconsistencies with reality:
1. I have demonstrated above that one can drive the M-noise signal above 77.9dB SPL using even less than 2W(rms), without any indication of clipping.
2. I have demonstrated with actual measurements that 2 audio signals having the same Vrms value can have 2 different peak SPL values which directly correlate with V_pk rather than Vrms.

I'm answering the remaining part of post #261.

Oscilloscope screen 1.

V_pk = 7.4 V, V_rms = 2.48 V, Fs = 5 kHz. This gave a crest factor (CF) of 7.4/2.48 ≈ 3x. The crest factor for M-Noise, per item 15, is at least 17.5 dB = 7.5x (for every 10 seconds interval in the provided test signal). That is a huge difference, and cannot be totally explained by phase shifts of the signal by the Schiit amp. The only plausible reason, IMHO, was that the amp was clipping.

BTW, the sampling rate in the test was 5 kSa/s (kilo-samples/second). That was way too low and would have caused errors in the measurements (e.g. from aliasing). As the M-Noise WAV file has Fs = 96 kHz, it is only appropriate to use a measurement sampling frequency at least that. Oscilloscopes usually change their sampling frequency on you when you change the horizontal time scale of the display. Something to watch out for.

Oscilloscope screen 2.

V_pk = 7 V (same as screen 1), V_rms = 1.6 V, Fs = 20 kHz. More evidence that the amp was clipping. In this case CF = 7/1.6 = 4.4x. CF must be the same if the amp is operating in its linear operating range -- V_pk and V_rms must scale up or down together.

The SPL measurements with microphone in a room cannot be compared to the calculated SPL from the speaker sensitivity. First, there is an unknown amount of room gain. Second, the speaker sensitivity rating is not verified. The non-linear frequency response of the speaker will also affect the mic SPL measurements.

Oscilloscope screen 3.
V_pk = 8.6 V, V_rms = 1.72 V, Fs = 10 kHz, CF = 5x.

Oscilloscope screen 4.
V_pk = 9.0 V, V_rms = 4.8x V, Fs = 10 kHz, CF = 1.9x.

This, IMHO, is an example of operating the amp well into clipping without noticing it. So, let me bust another audio myth here. Contrary to what most people say, when an amp clips nicely (clean, sharp, and symmetric, no funny oscillations, immediate recovery) and not severely, it is quite benign (well, at least until it is not). See Dr. Alex Voishvillo's paper (Figure 17).

(PDF) Assessment of nonlinearity in transducers and sound systems - From THD to perceptual models

PDF | Research of the audibility of loudspeaker nonlinear distortion has not shown good correlation between traditionally used metrics (harmonics and... | Find, read and cite all the research you need on ResearchGate

www.researchgate.net

 

[Deleted member 48726]

I'm answering the remaining part of post #261.

Oscilloscope screen 1.

V_pk = 7.4 V, V_rms = 2.48 V, Fs = 5 kHz. This gave a crest factor (CF) of 7.4/2.48 ≈ 3x. The crest factor for M-Noise, per item 15, is at least 17.5 dB = 7.5x (for every 10 seconds interval in the provided test signal). That is a huge difference, and cannot be totally explained by phase shifts of the signal by the Schiit amp. The only plausible reason, IMHO, was that the amp was clipping.

BTW, the sampling rate in the test was 5 kSa/s (kilo-samples/second). That was way too low and would have caused errors in the measurements (e.g. from aliasing). As the M-Noise WAV file has Fs = 96 kHz, it is only appropriate to use a measurement sampling frequency at least that. Oscilloscopes usually change their sampling frequency on you when you change the horizontal time scale of the display. Something to watch out for.

Oscilloscope screen 2.

V_pk = 7 V (same as screen 1), V_rms = 1.6 V, Fs = 20 kHz. More evidence that the amp was clipping. In this case CF = 7/1.6 = 4.4x. CF must be the same if the amp is operating in its linear operating range -- V_pk and V_rms must scale up or down together.

The SPL measurements with microphone in a room cannot be compared to the calculated SPL from the speaker sensitivity. First, there is an unknown amount of room gain. Second, the speaker sensitivity rating is not verified. The non-linear frequency response of the speaker will also affect the mic SPL measurements.

Oscilloscope screen 3.
V_pk = 8.6 V, V_rms = 1.72 V, Fs = 10 kHz, CF = 5x.

Oscilloscope screen 4.
V_pk = 9.0 V, V_rms = 4.8x V, Fs = 10 kHz, CF = 1.9x.

This, IMHO, is an example of operating the amp well into clipping without noticing it. So, let me bust another audio myth here. Contrary to what most people say, when an amp clips nicely (clean, sharp, and symmetric, no funny oscillations, immediate recovery) and not severely, it is quite benign (well, at least until it is not). See Dr. Alex Voishvillo's paper (Figure 17).

(PDF) Assessment of nonlinearity in transducers and sound systems - From THD to perceptual models

PDF | Research of the audibility of loudspeaker nonlinear distortion has not shown good correlation between traditionally used metrics (harmonics and... | Find, read and cite all the research you need on ResearchGate

www.researchgate.net

Wouldn't it be obvious if you zoom in on the peaks enough?

 

[NTK]

Wouldn't it be obvious if you zoom in on the peaks enough?

If you know what you are looking for, but may not be so obvious for a noise signal that looks like a bunch of random spikes.

However, it should be easy enough to look at the measured V_rms and V_pk from the 'scope. If you turn up the volume, and the V_pk to V_rms ratio (= crest factor) doesn't change, you don't have clipping. If the crest factor decreases, you do.

Alternatively, you show the waveform of the source signal with another channel of you 'scope. With appropriate vertical scaling, they should look the same. If the highest peaks of the output are lower than the source, you have clipping. However, if the amp causes significant phase in the output, the output waveform shape may look different from the source.

 

[Gorgonzola]

View attachment 267020

Balanced Audio Technology VK-56SE power amplifier Measurements

Sidebar 3: Measurements

www.stereophile.com

$8495

The sad thing is I really didn't have to look too hard.

Ok, Ok, this BAT offering has disturbingly high noise and distortion, (especially for $9k). But look at the harmonic distortion spectrum JA came up with ...

Balanced Audio Technology VK-56SE, Low output tap, spectrum of 1kHz sinewave, DC–10kHz, at 1W into 8 ohms (left channel blue, right red; linear frequency scale).

LOTS, and I do mean lots of 2nd and 3rd order harmonics. These harmonics, (particularly so much), caramel coats the sound -- so many audiophiles just love that. To heck with accurate reproduction.

 

[Hatto]

I'm answering the remaining part of post #261.

Oscilloscope screen 1.

V_pk = 7.4 V, V_rms = 2.48 V, Fs = 5 kHz. This gave a crest factor (CF) of 7.4/2.48 ≈ 3x. The crest factor for M-Noise, per item 15, is at least 17.5 dB = 7.5x (for every 10 seconds interval in the provided test signal). That is a huge difference, and cannot be totally explained by phase shifts of the signal by the Schiit amp. The only plausible reason, IMHO, was that the amp was clipping.

BTW, the sampling rate in the test was 5 kSa/s (kilo-samples/second). That was way too low and would have caused errors in the measurements (e.g. from aliasing). As the M-Noise WAV file has Fs = 96 kHz, it is only appropriate to use a measurement sampling frequency at least that. Oscilloscopes usually change their sampling frequency on you when you change the horizontal time scale of the display. Something to watch out for.

Oscilloscope screen 2.

V_pk = 7 V (same as screen 1), V_rms = 1.6 V, Fs = 20 kHz. More evidence that the amp was clipping. In this case CF = 7/1.6 = 4.4x. CF must be the same if the amp is operating in its linear operating range -- V_pk and V_rms must scale up or down together.

The SPL measurements with microphone in a room cannot be compared to the calculated SPL from the speaker sensitivity. First, there is an unknown amount of room gain. Second, the speaker sensitivity rating is not verified. The non-linear frequency response of the speaker will also affect the mic SPL measurements.

Oscilloscope screen 3.
V_pk = 8.6 V, V_rms = 1.72 V, Fs = 10 kHz, CF = 5x.

Oscilloscope screen 4.
V_pk = 9.0 V, V_rms = 4.8x V, Fs = 10 kHz, CF = 1.9x.

This, IMHO, is an example of operating the amp well into clipping without noticing it. So, let me bust another audio myth here. Contrary to what most people say, when an amp clips nicely (clean, sharp, and symmetric, no funny oscillations, immediate recovery) and not severely, it is quite benign (well, at least until it is not). See Dr. Alex Voishvillo's paper (Figure 17).

(PDF) Assessment of nonlinearity in transducers and sound systems - From THD to perceptual models

PDF | Research of the audibility of loudspeaker nonlinear distortion has not shown good correlation between traditionally used metrics (harmonics and... | Find, read and cite all the research you need on ResearchGate

www.researchgate.net

I'm going to call voltage crest factor as CF(V) as that's the only thing I can measure with the oscilloscope at the end of the amplifier. And let's forget about the different signals that I used for demonstrating different things and continue with the M-Noise signal alone.

Oscilloscope screen 1.
...
V_pk = 7.4 V, V_rms = 2.48 V, Fs = 5 kHz. This gave a crest factor (CF) of 7.4/2.48 ≈ 3x. The crest factor for M-Noise, per item 15, is at least 17.5 dB = 7.5x (for every 10 seconds interval in the provided test signal). That is a huge difference, and cannot be totally explained by phase shifts of the signal by the Schiit amp. The only plausible reason, IMHO, was that the amp was clipping.

Oscilloscope screen 1 was NOT for the M-noise, as I clearly explained underneath. While it's better to have a higher sampling freq, it doesn't have a dramatic effect on the end result. If we take M-Noise file at different sampling rates:

Oscilloscope Screen 2.1 - 5kSa/s sampling rate -> CF(V) = 416/93 = 4.491

Oscilloscope Screen 2.2 - 50kSa/s sampling rate -> CF(V) = 680/147 = 4.627

As you can see from the screenshots, raising the sampling rate to 50kSa/s (which is over twice the max freq, of the sampled M-Noise audio file) reveals there's only 3% difference for the CF(V) compared to a sampling rate of 5kSa/s for the same signal. I think the difference is well within the error range.

V_pk = 7 V (same as screen 1), V_rms = 1.6 V, Fs = 20 kHz. More evidence that the amp was clipping. In this case CF = 7/1.6 = 4.4x. CF must be the same if the amp is operating in its linear operating range -- V_pk and V_rms must scale up or down together.

AND

If you turn up the volume, and the V_pk to V_rms ratio (= crest factor) doesn't change, you don't have clipping. If the crest factor decreases, you do.

EXACTLY!

Oscilloscope screen 2 was for M-Noise output. Copying here again:

Oscilloscope Screen 2.3 - V_pk = 7.20V, Vrms = 1.60V => CF(V) = 4.500 which is only 3% different than the previous signal (Oscilloscope Screen 2.2) at CF(V) = 4.627 (with much less amplification).
As you can see V_pk and Vrms scale proportionately until V_pk = 7.20V, so we can clearly say that there is no sign of clipping until there

The SPL measurements with microphone in a room cannot be compared to the calculated SPL from the speaker sensitivity. First, there is an unknown amount of room gain. Second, the speaker sensitivity rating is not verified. The non-linear frequency response of the speaker will also affect the mic SPL measurements.

Well, then disregard ant SPL calculations based on any fomulae altogether because at this point (before clipping) an uncalibrated sound level measurement yields 94 dB SPL peak at 1m for M-Noise and it was disturbingly loud regardless of SPL reading on my phone, nothing even near 77.9dB SPL.

Oscilloscope screen 3 was again for another audio file, just to demonstrate (compared to M-Noise) that when it comes to music with high dynamic range, audio signals can have different Vrms at same V_pk values and vice-versa.

Oscilloscope screen 4 was a visualized example of actual clipping where CF(V) started dropping below 4.5

Here is what we have demonstrated so far:
1. Two audio signals having different Vrms values can have the same V_pk
Corollary to 1: Two audio signals having same Vrms values can have different V_pk
2. Peak SPL scales with V_pk, for HDR, non-uniform music.
3. Average SPL scales with Vrms, for any type of audio signal, regardless of shape.
4. Clipping occurs when V_pk is at max of what the amplifier can provide.

It's pure logic from here on:
Combining corollary to1 & 2: Vrms alone is not enough to calculate the peak SPL level for HDR, non-uniform music. Therefore any formula which calculates peak SPL using Vrms alone is unapplicable to complex audio tracks.
Combining 2 & 4: Clipping occurs after at least one instance of the audio signal reaches peak SPL. To find the point of clipping, one has to determine the first point where V_pk reaches the max that the amplifier can provide and measure amp gain at that point.

As a side note: 9V max output into 8ohms puts Rekkr's instantaneous max power at around 10W per channel.

 

[darmok]

Thank you @Hatto for actually pulling out your scope and taking some measurements. A couple of points:

1. The experiment I performed is significantly more pessimistic than using a signal like M-noise and looking at crest factor, since I effectively used a crest factor of 1 and set the peak voltage to a level corresponding to rated output at THD, which is much lower than clipping.
2. I think you should be using EPDR (or a guess at such) when computing power from voltage, not a nominal 8 ohms. When in doubt 4 ohms is probably a good guess at minimum EPDR for a nominal 8 ohm speaker unless your speakers lie about their impedance (*cough* Klipsch).
3. With a clipping voltage of 9V and a gain of 4, the good news is that you can feed Rekkr from a 2V DAC through a passive preamp like SYS or an active preamp like Magni in low (unity) gain without worrying about ever clipping. Whether that’s enough power for you is still an exercise for the reader.

 

[l0f3y3]

Subjective impressions of Schiit Gjallarhorn in Monoblock

Been listening to the Gjallarhorns via monoblock in my small home office.

Spotify & Tidal
WiiM Mini
SMSL SU-8s DAC - Slow Minimum PCM Filter, Rich 2 sound color
SMSL HO200 as Pre-Amp Low and Med Gain toggle
Fluance RT-85 Ortofon Blue
Fosi Box X2 Phono pre with GE tubes

Very satisfied with the performance of the Gjallarhorn. It has been a few weeks now. Plenty of power to drive 4 ohm, 85 DB sensitivity JBL L52s, Triangle Borea BR03 and Denton 80th Anniversary speakers in my rotation. I listen below 75db, rarely above. Mostly jazz off tidal, some electronica, and a variety of vinyl, some classic and modern rock, vintage and remastered classical & jazz, ambient and house electronica.

Overall the detail and sound stage is the best I have owned. The amps are fast and have a lot of detail and refinement across the spectrum, they certainly do not hinder the upstream gear. The SMSL DAC and Preamp work really well with the amp monoblock pair. I have zero complaints about the amps. When I flip the power switch of the APC in the morning the amps respond with delightful clicks a few seconds later and are ready to go after a few minutes of warm up.

I learned a lot about the other amps I own, a 6.5 watt/channel class A tube amp (nobsound 6p1) and 40 watt/channel Class AB Fosi HD-A1. I A-B switch between the Gjallarhorns and the other two amps on occasion. The FOSI has ended up surprising me. It lacks the detail of the GJallarhorns in the upper and mid midrange but has a bit more punch on the low end, More of a smooth punch though. There is some music when played through the JBL's that I find the FOSI's extra watts open the L52s up just a bit more. When comparing the two amps through the Triangle's with 90 db sensitivity the difference is more evident and the GJallarhorns show more sophistication in presenting the music and how it renders the individual instruments. I know the FOSI HD-A1 tested poorly here but subjectively, is quite a remarkable sounding amp and quite a bargain for a Class AB amp. It will still have a place in my listening setup for some situations, Old vinyl, classic jazz and acoustic music, are smoothed out by the FOSI. But for modern recordings and remastered vinyl and Tidal MQA sources the Gjallarhorns and very pleasurable to listen to. I am considering swapping out the SMSL SU-8s DAC for the DO200 mrkII for the dual chip config, upgraded processor and form factor that matches the HO200 amp. Plus it is on sale right now.

I have taken to the Triangle speakers and as such am planning up stepping up to the Triangle Comete EZ at some point in the future (based on various reviews in the wild).

Concluding: the Gjallarhorns have met and exceeded my expectations, they run cool, drive my various speakers well, sound excellent and allow me to explore and enjoy the music more. They also provide me a serious reference for evaluating and understanding the performance of my other amps that are all in the same price range. Literally these Schiit amps are the perfect fit for my listening habits and my small home office space. (12'x12'x10' ceiling). I literally cannot think of any other balanced mono-block amp that sounds this good available at this $600 price. The nearest is no probably the Topping LA90 or Schiit's VIDAR 2 or Aegir. Which in mono-block will be around $1800.

 

[JiiPee]

I think we should acknowledge that all audio products are not meant to be allrounders, but they have been designed for a specific use case. Smaller Genelec "ones" - 8331 and to some extent 8341 - for example, are sublime near field monitors, but they don't meet all needs of the amirs of this world who want to listen their music loud and in bigger spaces. The same applies to these small pieces of Schiit (sorry about that). They work fine for many of us, who just don't need much power, but of course, they are not the right choice for everyone.

Having extra power in reserve never hurts as such, and who knows - I might need it in future, but keeping it down has it's advantages too: like possibility to design a high quality amplifier with a small physical size and reasonably low price.

Having said that, I like this thread. Lots of well reasoned posts showing that there are knowledgeable people participating on this forum.

 

[Trudius]

Subjective impressions of Schiit Gjallarhorn in Monoblock

Been listening to the Gjallarhorns via monoblock in my small home office.

Spotify & Tidal
WiiM Mini
SMSL SU-8s DAC - Slow Minimum PCM Filter, Rich 2 sound color
SMSL HO200 as Pre-Amp Low and Med Gain toggle
Fluance RT-85 Ortofon Blue
Fosi Box X2 Phono pre with GE tubes

Very satisfied with the performance of the Gjallarhorn. It has been a few weeks now. Plenty of power to drive 4 ohm, 85 DB sensitivity JBL L52s, Triangle Borea BR03 and Denton 80th Anniversary speakers in my rotation. I listen below 75db, rarely above. Mostly jazz off tidal, some electronica, and a variety of vinyl, some classic and modern rock, vintage and remastered classical & jazz, ambient and house electronica.

Overall the detail and sound stage is the best I have owned. The amps are fast and have a lot of detail and refinement across the spectrum, they certainly do not hinder the upstream gear. The SMSL DAC and Preamp work really well with the amp monoblock pair. I have zero complaints about the amps. When I flip the power switch of the APC in the morning the amps respond with delightful clicks a few seconds later and are ready to go after a few minutes of warm up.

I learned a lot about the other amps I own, a 6.5 watt/channel class A tube amp (nobsound 6p1) and 40 watt/channel Class AB Fosi HD-A1. I A-B switch between the Gjallarhorns and the other two amps on occasion. The FOSI has ended up surprising me. It lacks the detail of the GJallarhorns in the upper and mid midrange but has a bit more punch on the low end, More of a smooth punch though. There is some music when played through the JBL's that I find the FOSI's extra watts open the L52s up just a bit more. When comparing the two amps through the Triangle's with 90 db sensitivity the difference is more evident and the GJallarhorns show more sophistication in presenting the music and how it renders the individual instruments. I know the FOSI HD-A1 tested poorly here but subjectively, is quite a remarkable sounding amp and quite a bargain for a Class AB amp. It will still have a place in my listening setup for some situations, Old vinyl, classic jazz and acoustic music, are smoothed out by the FOSI. But for modern recordings and remastered vinyl and Tidal MQA sources the Gjallarhorns and very pleasurable to listen to. I am considering swapping out the SMSL SU-8s DAC for the DO200 mrkII for the dual chip config, upgraded processor and form factor that matches the HO200 amp. Plus it is on sale right now.

I have taken to the Triangle speakers and as such am planning up stepping up to the Triangle Comete EZ at some point in the future (based on various reviews in the wild).

Concluding: the Gjallarhorns have met and exceeded my expectations, they run cool, drive my various speakers well, sound excellent and allow me to explore and enjoy the music more. They also provide me a serious reference for evaluating and understanding the performance of my other amps that are all in the same price range. Literally these Schiit amps are the perfect fit for my listening habits and my small home office space. (12'x12'x10' ceiling). I literally cannot think of any other balanced mono-block amp that sounds this good available at this $600 price. The nearest is no probably the Topping LA90 or Schiit's VIDAR 2 or Aegir. Which in mono-block will be around $1800.

The power of the 2 Gjallarhorns is roughly equivalent to 1 Aegir. The single Aegir is more costly but it is also rated at 4 ohms, the monos are not.

 

[foreigner4]

If someone with amplifiers has high-frequency noise from a computer in the tweeters or 50/60 Hz hum, then just use a power cable without a third ground prong. This will eliminate the effect of ground loops.
Do not rush to argue, do it, then write the result.
Good luck everyone, enjoy.

 

[Hatto]

...just use a network cable without a third ground prong.

You mean an ungrounded power cable?

This will eliminate the effect of ground loops.

Would connecting multiple chases with bare wire work as well?

 

[foreigner4]

The speaker that has a larger cone will need to move more air because it has a larger air piston, moving more air = more watts.

In reality, loudspeakers with large bass cones have greater sensitivity, including bass. The higher the sensitivity, the higher the efficiency and the lower the power consumption.

 

[foreigner4]

You mean an ungrounded power cable?

Quite right.

 

[foreigner4]

It helped me with Vidar 2. Interestingly, Denon PMA-1500mk2 and Cambridge MA10 did not make noise in my system with a PC and DAC, and for Vidar 2 I had to disconnect the ground so that the RF noise from the PC disappeared.