Fender Mustang Micro Frequency Response Curves

Charles Eric LaForest, PhD., 2026

This page contains frequency response curves of amp models measured from a Fender Mustang Micro, a self-contained headphone practice amp which plugs directly into a guitar/bass/keyboard, etc... For reference, here is the English User Manual and the Firmware Update page.

The manual, like so many manuals of musical gear, is vague on the specs and full of "inspired by" and "based on" language which offers access to the sound of famous amps while disavowing accuracy. I was hoping that by directly measuring the behaviour of the device, I'd learn something actually useful about modelling amp tone stacks and their speaker cabinets. So, I won't refer the amp models by the amps they claim to model, as there's no reason to believe the claims, and I'll instead refer to models by the LED colour they were assigned in the user interface and manual.

I'm not the first to try doing this, and in 2016 an unknown person at GUITMOD did a very similar set of measurements on the Mustang v2 modelling amp. Go see their charts for an interesting comparison, particularly since the Mustang v2 separates speaker cab modelling from amp tone stack modelling.

Measurement Methodology

The input test signal is a collection of sine waves, one for each semitone (in 12-TET, 440 Hz concert tuning) existing between 50 Hz and 10 kHz. They are generated with a script (generate_all_notes.py) and played over my computer's headphone output. The output level is monitored with an oscilloscope and set to approximately 250 mVpp (peak-to-peak), which registers as about 31 mVrms for this signal. The resulting waveform looks and sounds almost like noise, as if you pressed all the keys on a pipe organ at the same time.

A collection of chromatic semitone sinewaves is much better to use than pink noise:

• it doesn't need filtering since it's inherently bandwidth-limited when generated
• it has 12 notes per octave, so equal energy per octave like pink noise, and the sample points are evenly spread out on a log plot
• the noise floor and any distortion products are visible in the gaps between notes
• there is little FFT leakage between notes, except at the lowest frequencies, so Flat Top windowing is usable for accurate level measurement

The input signal is brought to the device with its EQ is set to White ("Flat") and amp model set to Cyan (a "studio preamp" with no frequency effect), and the output volume of the device is adjusted (monitored on the same scope as the input signal) until the total gain is unity.

The input and output are also monitored with a Red Pitaya STEMlab 125-14 set as a dual spectrum analyzer (see example view on the left), with Flat Top windowing, a range of 50 Hz to 10 kHz, and the FFT sample count max'ed out at 16k. After some experimentation, the zoom and position of the display are left constant and set to place the input signal at the bottom and the output using as much of the display as possible for best resolution. We can now see the entire spectral response of the device at once, and each spectrum is captured as a JPG image and as a CSV dump of values for post-processing.

Post-processing and plotting is done with a Python script (fmm_plot.py) driven by a Bash script to generate all cases (fmm_generate_all_plots.sh). The CSV file is read in and contains the frequencies (in Hz), input levels, and output levels (in dBm). The absolute dBm values are not meaningful, since the input impedance isn't set right. We find the peaks of the input signal, and filter the whole dataset to the location of those peaks. We then subtract the input signal from the output signal to give a correct relative dB measure. We can then plot this data on a log scale to show a detailed frequency response chart for a given device setting. Each chart point is labelled with the frequency and the dB value relative to input. (It's a bit crowded, sorry.)

General Observations

• The response curves show broad humps, which look like the typical EQ curves of guitar amp tone stacks, and multiple notches which look like the responses of speaker cabinets. There are also very steep roll-offs (-24 dB per octave or more) at the extremes, which is also a characteristic of speakers. Thus, I think the amp models here combine amp models and speaker/cabinet models, and I'll analyze them as those two parts.
• There's little use for perfect numerical measurements here, as too many variables in real life affect the resulting timbre, and I'm not trying to replicate here, but extract general principles. The overall system behaviours are linear (think RLC filters, phase-cancellation notches, etc...). So I just trace the chart with my finger and mind's eye to where a curve should be, and measure from there. This gave suprisingly consistent results across charts.
• I also swept the input with an oscillator, and noticed the output phase rotates dozens of times over the entire range. This makes showing any distortion waveform meaningless, as we don't know the relative phase of the generated harmonics, so I'm not addressing that.
• There are input protection diodes, as any input signal past 1 Vpp starts to show odd-order distortion products, even on the Cyan "Studio Preamp" amp model, indicating symmetric clipping.
• The volume knob is likely linear, not logarithmic, as the second half of the rotation barely increases the perceived volume. (This is my general beef with the Fender Mustang Micro: without an amp model in place, it can't get loud enough for practical use.)

EQ Settings

There are five equalizer settings applied globally, each denoted by a LED colour. All these EQ settings are measured with the Cyan "Studio Preamp" amp model, which is claimed to be flat in frequency. There's no charts here as the EQ effects are simple and better described in text.

• The "Brighter" Red EQ has a flat +1.7 dB boost overall with an extra +1.5 dB boost curve from 200 Hz to 2 kHz, peaking at 700 Hz.
• The "Bright" Yellow EQ is a flat +1.7 dB boost, returning to 0 dB at 50 Hz. (I forgot to note the upper limit, likely still the 18 kHz cutoff). This is the true "flat" setting.
• The "Flat" White EQ setting is not actually flat, but a low shelf boost starting at about 1 KHz, reaching +1.6 dB at 400 Hz, then staying constant until about 50 Hz, where it returns to 0 dB. There is a sharp cutoff at 18 kHz. It's a small boost, but over a large chunk of spectrum, so it might be audible and meant to account for the lower overall bass response of most headphones.
• The "Dark" Green EQ has a low shelf boost of +1.5 dB starting at about 700 Hz, then transitions to a -1.5 dB high shelf cut starting at about 2 kHz.
• The "Darker" Blue EQ has a low shelf boost of +1.4 dB starting at about 625 Hz, where it's at 0 dB, then transitions to a -3 dB dB high shelf cut starting at about 2 kHz.

Clean Amps

The Clean Amps have little to no distortion, and apply an overall boost of +10 dB to +12 dB to the mVrms signal value. Keep this in mind when you see rather large gains in the amp models: they do add a lot of loudness as well as frequency shaping.

• White (Raw data: fmm_eq_white_amp_white.csv)
• Red (Raw data: fmm_eq_white_amp_red.csv)
• Green (Raw data: fmm_eq_white_amp_green.csv)
• Blue (Raw data: fmm_eq_white_amp_blue.csv)

The humps are as-expected from general amp models: a bass boost centered around 200 Hz, and a treble boost centered around 3 to 4 KHz, which together create a mid-scoop around 500 Hz to 700 Hz. This is pretty typical across guitar and bass amps and pre-amps.

Amp Model	Humps
White	+14 dB at 190 Hz	+25 dB at 3900 Hz
Red	+10.5 dB at 195 Hz	+20.6 at 3136 Hz
Green	flat +9 dB down to +7 dB at 1500 Hz	+25 dB at 3951 Hz
Blue	+10 dB at 270 Hz	+17.3 at 3323 Hz

The notches vary, as one would expect from different modelled speakers, cabinets, multiple drivers, mic placement, etc... but show up most frequently around 150 Hz to 200 Hz, 830 Hz, 1500 Hz, and 3 kHz. The other notches alternate their presence, which likely affects the timbre of each model. Notice how White and Green have effectively identical notches, and Red and Blue also (with an extra notch at 784 Hz).

Amp Model	Notches
White	-5 dB at 190 Hz	-3 dB at 830 Hz		-7 dB at 1570 Hz	-8 dB at 3 kHz		-12 dB at 5 kHz
Red	-3 dB at 155 Hz		-9 dB at 1046 Hz	-10 dB at 1479 Hz	-6 dB at 2489 Hz	-5 dB at 3729 Hz		-2 dB at 6646 Hz
Green	-6 dB at 186 Hz (broad and pointy)	-3.5 dB at 830 Hz		-7 db at 1569 Hz	-9 dB at 3 kHz		-10 dB at 4978 Hz
Blue	-5 dB at 165 Hz	-3 dB at 784 Hz	-3 dB at 1108 Hz	-4 dB at 1475 Hz	-3 dB at 2794 Hz	-2 dB at 3735 Hz		-6 dB at 7903 Hz

Here's a speculation: If you round-off the various frequencies a little, it's easy to imagine one or more harmonic series (of notches) with missing entries, suggesting one or more comb filters arising from the signal interacting with delayed versions of itself, with some non-linearities skewing the notch frequencies. Most of the signal energy would be around 200 Hz, and if there was one or more delays in the 2.5 ms range, then that would generate significant notches at multiples of that frequency range. Something to experiment with later...

White Model Compressor

The White amp model includes a compressor "with a low compression setting", but there's no truth to that. I first checked that the frequency response of the model is not affected by varying the input level (it's not). Then I chose a 660 Hz test tone as that frequency has nearly unity gain in this amp model. Then, by looking at the spectrum analyzer and zooming in on the 660 Hz peak, I increased the input level until the input level exceeded the output level, indicating compression. How much the input has to increase (in dB) to produce a 1 dB increase in output measures the compression ratio (ignoring attack/release times). The spectrum analyzer cursors are quite useful to mark levels.

• The output tracks the input 1:1 until 156 mVpp to 172 mVpp. That's the "knee" of the compressor, which spans an input range of about 0.84 dB.
• After the knee, increasing the output by 1 dB to 194 mVpp requires increasing the input by 11 dB to 640 mVpp.
• Increasing the output another 1 dB to 216 mVpp requires increasing the input by another 16 dB to 1.22 Vpp.

So this compressor has a sharp knee (less than 1 dB) with a ratio of 11:1 minimum. That's not low compression, that's a limiter. That said, it does work very well and sounds completely transparent, as evidenced by the unchanging frequency response and likely fast attack/release times.

Crunch Amps

The Crunch Amps start to have significant distortion, and except for the Yellow amp model, there isn't much to see here. I did not estimate the humps/notches as it wouldn't add anything useful, and the distortion makes the measurements suspect.

• Yellow (Raw data: fmm_eq_white_amp_yellow.csv)
• Magenta (Raw data: fmm_eq_white_amp_magenta.csv)
• Turquoise (Raw data: fmm_eq_white_amp_turquoise.csv)

High-Gain Amps

The High-Gain Amps are exactly that. I had to measure them a second time at 5.7 mVrms input, about as low a level my computer could output with any quality, to make sure the spectrum contained any actual information. Both measures are on the same chart, with the 5.7 mVrms measurement marked as "low", showing the compression from clipping when measured at 31 mVrms as above. There seem to be some fairly consistent humps and notches, but I'm not estimating them, as I am not confident about their meaning. What notches that seem constant don't really add anything new.

• Orange (Raw data: fmm_eq_white_amp_orange.csv, fmm_eq_white_amp_orange-low.csv)
• Raspberry (Raw data: fmm_eq_white_amp_raspberry.csv, fmm_eq_white_amp_raspberry-low.csv)
• Spring Green (Raw data: fmm_eq_white_amp_spring_green.csv, fmm_eq_white_amp_spring_green-low.csv)
• Violet (Raw data: fmm_eq_white_amp_violet.csv, fmm_eq_white_amp_violet-low.csv)

Final Thoughts

It looks like a broad and useful range of amp models could be interactively created with well-understood, all-analog hardware. The main idea is to allow control of the relative heights of the low and high humps, the depth of the mid-scoop, and their centre frequencies. And then add fixed steep cutoffs at the lowest and highest frequencies, and a few variable notches, to model the speaker cabinet response.

• A pair of 4th order Butterworth filters: one LPF around 5 to 7 kHz, and one HPF around 50 to 70 Hz, to model the edges of speaker response. Each of these needs a single TL072, which is cheap.
• A pair of semi-parametric (fixed Q, variable gain and frequency) boost filters to generate the low and high humps as well as the implicit mid-scoop. (And maybe a third cut-only filter to allow for deeper mid-scoops.) We can't use shelving EQ here as each hump must be a broad resonnance peak to alter timbre effectively, and the filter slopes at the edges will add-up with the above fixed filters and produce the even steeper cutoffs measured from real speakers.
• Alternately: a pair of fixed filters for the high and low humps, paired with a high/low shelving EQ to move the humps up and down. This is the approach taken by the SansAmp BDDI and its clones like the Behringer BDI21. Both also include an third, mid-scoop filter.
• To create the notches: one or more feed-forward comb filters composed of adjustable delay lines in the sub-5ms range passed through tunable band-pass filters before being mixed back with the original signal.