Behringer Super Fuzz SF-300 on the bench

One Internet reviewer says of the Behringer Super Fuzz that “the primary purpose of this pedal [...] is to bludgeon the listener.” I thought we'd better get it on the bench and see if it can bludgeon the dynamic signal analyzer.

A Behringer Super Fuzz
guitar pedal sits on an electronics work bench, connected to a dynamic signal
analyzer. The pedal is made of orange plastic and has four tiny black knobs
near the top. The stomp switch is the full width of the pedal and about two
thirds of its height. It hinges at the top and has a thin rubber pad with the
Behringer logo covering its bottom half. The input and output jacks are on the
right and left sides while the power jack is on the right side. Also visible
are a notebook, a keyboard, and an open can of Miller Lite beer.

My initial impression from handling the pedal is that it has been inexpensively made: the plastic is plasticky and the knobs are kind of sad. On closer inspection, I was actually kind of delighted by what appears to have been a comprehensive exercise in cost-engineering. I guess I'm that kind of nerd.

Methodology

I was able to use essentially the same instrument setup for these measurements as I did for the Tube Screamer measurements. The DSA configurations for frequency response and distortion measurements are available as FREQRESP.STA and DISTORTN.STA.

Findings

The pedal was connected to a bench power supply set to 9 Volts. With the input jack disconnected, the pedal drew no measurable current. With the input jack connected to the DSA, it drew 8.7 mA and with the effect turned on it drew 10.4 mA.

I started out with a frequency response measurement with the pedal turned off. Since the stomp switch seems to be momentary, I supposed that the pedal might use buffered bypass. I was surprised to find a flat frequency response with the effect disabled. FBYPASS.DAT

The effect has three different modes as well as some adjustments, so I settled in for some comprehensive frequency response measurements. In both the Fuzz 1 and Fuzz 2 modes, the bass control seems to boost or cut the response around 115 Hz while the treble control seems mostly to increase or decrease the cutoff frequency on the high end. In Fuzz 2 mode there's also a second (much lower) treble peak that becomes more prominent and moves up in frequency as the treble control is increased.

Fuzz 1 setting

Treble Bass Low scoop/peak High cutoff File
Minimum Minimum -9.6 dBVRMS at 112 Hz 690 Hz F1MINMIN.DAT
Minimum Middle 590 F1MINMID.DAT
Minimum Maximum +7.9 at 114 532 F1MINMAX.DAT
Middle Minimum -9.8 at 114 819 F1MIDMIN.DAT
Middle Middle 765 F1MIDMID.DAT
Middle Maximum +7.7 at 116 622 F1MIDMAX.DAT
Maximum Minimum -10 at 114 1,400 F1MAXMIN.DAT
Maximum Middle 1,200 F1MAXMID.DAT
Maximum Maximum +7.6 at 114 941 F1MAXMAX.DAT

Fuzz 2 setting

Treble Bass Low scoop/peak High cutoff Second treble peak
Minimum Minimum -6.5 dBVRMS at 116 Hz 433 Hz -24 dbVRMS at 2,000 Hz
Minimum Middle 317 -28 at 2,000
Minimum Maximum +11 at 116 397 -26 at 2,700
Middle Minimum -6.5 at 116 456 -16 at 2,700
Middle Middle 323 -20 at 3,000
Middle Maximum +11 at 112 397 -18 at 2,900
Maximum Minimum -6.5 at 116 488 -7 at 3.8
Maximum Middle 340 -11 at 3,900
Maximum Maximum +11 at 112 418 -9.4 at 4,000

Fuzz 2 setting data files: F2MINMIN.DAT, F2MINMID.DAT, F2MINMAX.DAT, F2MIDMIN.DAT, F2MIDMID.DAT, F2MIDMAX.DAT, F2MAXMIN.DAT, F2MAXMID.DAT, F2MAXMAX.DAT

The frequency response plot below shows the bass scoop that can be achieved by setting the bass control to its minimum setting in Fuzz 1.

A frequency response
plot labeled to indicate Fuzz 1 setting with treble set in the middle and 
bass set to minimum shows a double-peak with the bass peak at about 60 Hertz
and five decibels below the treble peak at about 400 Hertz and 12 decibels.
A relative marker shows the dip between the peaks 9.8 decibels below the
treble peak. The bass peak has a bandwidth of about 50 Hertz and the treble
peak has a bandwidth of about 500 Hertz.

And this plot shows the peak that can be produced by setting it to its maximum setting in Fuzz 1.

A frequency response
plot labeled to indicate Fuzz 1 setting with treble set in the middle and
bass set to maximum shows a single bass peak whose roll-off proceeds through
an inflection point on its way to an eventual treble roll-off. The bass peak
is at about 100 Hertz with a bandwidth of about 50 Hertz. The inflection point
is at about 300 Hertz and eight decibels down from the bass peak. The treble
corner frequency appears to be about 500 Hertz.

The frequency response plot below shows the second treble peak characteristic of the Fuzz 2 setting.

A frequency response
plot labeled to indicate Fuzz 2 setting with treble and bass both set to the
middle shows a broad peak at around 100 Hertz and a narrower peak about 20
decibels down at about 3,000 Hertz. A marker indicates that the trough between 
the peaks at about 1,300 Hertz is 24.4 decibels down from the main peak.

In the Boost setting, both the treble and bass controls each seem to adjust a peak or scoop at about 6,200 Hz and 116 Hz respectively.

A pair of
frequency response plots labeled to indicate Boost setting with treble and
bass at minimum versus treble and bass at maximum shows the maximum/maximum
plot with two almost-equal-height peaks at about 120 Hertz and about 6,000
Hertz in front of the minimum/minimum plot with scoops at the same frequencies.
The two plots show about the same response at 20 Hertz, about 15 decibels
difference at the bass peak/scoop, about 3 decibels difference at around 400
Hertz, about 23 decibels difference at the treble peak/scoop, and about 20
decibels difference at 20,000 Hertz.

As for distortion, both the Fuzz 1 and Fuzz 2 settings show prominent second harmonics in particular and high even-order content in the distortion products more generally. At high input levels, both settings produce second harmonics that are a bit below the fundamental. As the input level falls off, the second harmonic overtakes the fundamental. At very low input levels, the fundamental once again dominates. There are some subtle differences between the two settings and at different gain levels, but this basic pattern holds across all of the settings I measured and seems to be characteristic to this type of fuzz effect. Overall, the most salient difference was that the Fuzz 1 setting had higher output than the Fuzz 2 setting at equal gain settings.

Fuzz 1 setting

Gain Source Fundamental 2nd harmonic Harmonic distortion
Even order Odd order Total
Minimum -60 dBVRMS -52 dbVRMS -72 dbVRMS 11% 0% 11%
-40 -33 -34 93 6 93
-20 -40 -22 759 141 772
0 -31 -24 228 144 269
9 o'clock -60 -48 -63 17 0 17
-40 -31 -27 147 12 148
-20 -37 -22 528 139 546
0 -30 -26 153 143 210
Middle -60 -44 -55 29 1 29
-40 -34 -24 298 31 299
-20 -35 -22 454 140 475
0 -29 -27 139 132 192
Maximum -60 -31 -30 115 8 115
-40 -38 -22 644 139 659
-20 -30 -95 185 145 235
0 -28 -27 121 118 168

Fuzz 1 setting data files: D1A60.DAT, D1A40.DAT, D1A20.DAT, D1A0.DAT, D1B60.DAT, D1B40.DAT, D1B20.DAT, D1B0.DAT, D1C60.DAT, D1C40.DAT, D1C20.DAT, D1C0.DAT, D1D60.DAT, D1D40.DAT, D1D20.DAT, D1D0.DAT

Fuzz 2 setting

Gain Source Fundamental 2nd harmonic Harmonic distortion
Even order Odd order Total
Minimum -60 dBVRMS -61 dbVRMS -90 dbVRMS 4% 2% 5%
-40 -41 -50 38 6 38
-20 -48 -38 568 487 748
0 -40 -40 246 266 363
9 o'clock -60 -56 -79 7 1 7
-40 -39 -43 63 23 68
-20 -45 -38 386 436 583
0 -38 -42 216 229 315
Middle -60 -52 -71 11 1 11
-40 -42 -40 161 89 184
-20 -43 -35 337 407 529
0 -38 -42 198 223 298
Maximum -60 -40 -46 49 11 50
-40 -46 -38 450 457 641
-20 -39 -41 231 237 331
0 -37 -43 174 211 274

Fuzz 2 setting data files: D2A60.DAT, D2A40.DAT, D2A20.DAT, D2A0.DAT, D2B60.DAT, D2B40.DAT, D2B20.DAT, D2B0.DAT, D2C60.DAT, D2C40.DAT, D2C20.DAT, D2C0.DAT, D2D60.DAT, D2D40.DAT, D2D20.DAT, D2D0.DAT

As gain is increased in the Fuzz 1 setting with moderate input signal strength, the fundamental comes up in power faster than the second harmonic. The net result is lower total harmonic distortion and a shift from mostly even-order distortion products to more balance between even- and odd-order products. This effect is present but less prominent also in the Fuzz 2 setting.

These two plots show the effect in Fuzz 1 setting with a source signal of -20 dBVRMS at minimum and maximum gain.

A frequency domain
plot labeled to indicate Fuzz 1 setting with minimum gain and a -20 decibel
source signal shows a second harmonic about 20 decibels higher than the
fundamental and third harmonic. By 25.6 kilohertz, the distortion products
drop off to more than 60 decibels below the second harmonic. A marker indicates
that the second harmonic is 17.5 decibels above the fundamental. A label
indicates total harmonic distortion of 772 percent.

A frequency domain
plot labeled to indicate Fuzz 1 setting with maximum gain and a -20 decibel
source signal shows a second harmonic at about the same level as the
fundamental and third harmonic. By 25.6 kilohertz, the distortion products
drop off to more than 40 decibels below the second harmonic. A marker indicates
that the second harmonic is 5 decibels above the fundamental. A label
indicates total harmonic distortion of 234 percent.

Switching from Fuzz 1 to Fuzz 2 has a similar (but not exactly the same, I guess) effect to increasing the gain setting, as these two plots show. Both were taken with the same gain setting and source signal. You can see a similar effect to the prior two plots where the fundamental comes up faster than the second harmonic, giving lower total harmonic distortion and more balance between even- and odd-order harmonic content.

A frequency domain
plot labeled to indicate Fuzz 1 setting with gain set to the middle and a
-40 decibel source signal shows a second harmonic about 10 decibels above the
fundamental and distortion products below -90 decibels relative to one volt
RMS by about 12 kilohertz. A marker indicates the second harmonic is 9.4
decibels above the fundamental. A label indicates total harmonic distortion
of 299 percent.

A frequency domain
plot labeled to indicate Fuzz 2 setting with gain set to the middle and
a -40 decibel source signal shows a second harmonic about the same as the
fundamental and distortion products below -90 decibels relative to one volt
RMS by 25.6 kilohertz. A marker indicates the second harmonic is 1.8
decibels above the fundamental. A label indicates total harmonic distortion
of 184 percent.

The boost mode is — as one might expect — completely different from the two fuzz modes. It produces essentially no distortion at reasonable source signal levels. At unreasonable source levels, it can be coaxed into distortion.

Boost setting

Gain Source Fundamental 2nd harmonic Harmonic distortion
Even order Odd order Total
Minimum -60 dbVRMS -51 dbVRMS 1% 1% 2%
-40 -32 0 0 0
-20 -12 0 0 0
0 7 -27 dbVRMS 2 2 3
Middle -60 -43 0 0 1
-40 -24 0 0 0
-20 -4 -63 0 0 0
0 9 -16 8 30 31
Maximum -60 -30 0 0 0
-40 -11 -71 0 0 0
-20 8 -23 3 9 9
0 9 -19 11 43 44

Boost setting data files: DBA60.DAT, DBA40.DAT, DBA20.DAT, DBA0.DAT, DBC60.DAT, DBC40.DAT, DBC20.DAT, DBC0.DAT, DBD60.DAT, DBD40.DAT, DBD20.DAT, DBD0.DAT

Here's a plot showing the negligible distortion at reasonable source signal levels in the boost setting.

A frequency domain plot
labeled to indicate Boost setting, gain in the middle position, and a source
level of -20 decibels shows a second harmonic about 60 decibels
below the fundamental and no other visible distortion products. A marker
indicates the second harmonic is 59 decibels below the harmonic. A label
indicates the total harmonic distortion is 0.11 percent.

Contrast the preceding plot with this one showing distortion coming in on the boost setting as we get to unreasonably-high source signal levels.

A frequency domain plot
labeled to indicate Boost setting, gain in the middle position, and a source
level of 0 decibels shows predominatly odd-order distortion products tapering
off to about 60 decibels below the fundamental by 25.6 kilohertz. A marker
indicates the second harmonic is 25 decibels below the fundamental. A label
indicates total harmonic distortion of 31 percent.

While the balance of distortion products and the way they change with input level in the Fuzz 1 and Fuzz 2 settings is measurably different, the overall pattern is comparable: these do not seem to be two wildly different distortion techniques. I suspect that the difference in character noted by ear is due in part to the difference in distortion products but perhaps moreso to the relatively more pronounced differences in tone shaping between the two modes.

The boost is clean with what seem like thoughtful tone-shaping options. I suspect it would be handy for musicians who decide to park this pedal permanently at the front of their signal chain.

I hope you found this write-up interesting and helpful. If you have any questions or comments, please let me know! If this is the kind of thing you're into, you may enjoy our other work.

Do our ideas intrigue you?

Subscribe to our newsletter and be the first kid on your block to get our latest pedal write-ups and tear-downs and to hear about Hot Schmitt, the fuzz pedal we're working on. We can drop you an email when we post new stuff if you sign up.

Aaron D. Parks
Parks Digital LLC
4784 Pine Hill Drive, Potterville, Michigan
support@parksdigital.com