Woody 1x12 - SM57 Steps IRs

James Freeman

Rock Star
Messages
3,155
Note: Please don't click the 'Reply' button on this post to not quote the entire post, I may change or add stuff.

This time no tricks, just normal IRs captured with sweeps, deconvolved and normalized, the IRs are raw, no eq or processing whatsoever.
Each speaker captured individually with a single SM57 in 2.5mm (0.1") steps from Center to 8cm (3") left and about 2cm (0.75") away from the metal mesh, it is up to you to find the sweetspot like you would with a real speaker.

Some background;
I wanted to capture finer distance steps so I can find the sweetspot without going back and re-capturing the cab if I was not satisfied with an older capture. Now I can "move the mic" digitally with enough resolution to find the perfect point of each speaker.
I was of course inspired by the Quad Cortex and Line 6 Metallurgy where I can move the mics freely and wanted to try the same with my own cab and speakers. It was a tedious work to capture so many IRs with just one SM57 so I can definitely appreciate what IR vendors are doing.

Current Speakers:
Alnico Blue
Classic Lead 80
G12H Anniversary
G12H Redback
G12H 75Hz Heritage
G12H-75 Creamback
G12M EVH
G12M Greenback
G12H-65 Creamback
Vintage 30

Download Link:
Code:
https://drive.google.com/file/d/1laMXRM65_WnmTTZKgc7ofmZkhIYdj2nB/view?usp=sharing




V30+GB.jpg




EDIT:
Oct 27th; Added more speakers, download link changed.

EDIT2:
Oct 30th; Added more speakers and reshot all old speakers, download link changed.
 
Last edited:
Added more speakers, link changed.

Alnico Blue
Classic Lead 80
G12H 75Hz Heritage
G12H Anniversary
G12M EVH
G12M Greenback
Vintage 30

I may add the 65 and 75 Creambacks and Redback later, but in my personal opinion you can get better sounds by blending the "classic" speakers.
 
Each speaker captured individually with a single SM57 in 0.25mm (3/32") steps from Center to 8cm (3") left and about 2cm (0.75") away from the metal mesh, it is up to you to find the sweetspot like you would with a real speaker.
I'll be offering some remarks and analysis of the results of this project as time allows, but first I must point out that your mic-position increments are just under 2.5, not .25, millimeters. (3/32)*25.4 = 2.38. .25mm = .010". That's ten thousandths of an inch. It would take lots and lots of IRs to cover any meaningful distance if you only move the mic .25mm each time.

Edit: There's either an inconsistency in your file-naming protocol or some IRs are missing. You go from 0.00 to 1.0, then 1.25, 1.5, etc. That implies that there should also be 0.25, 0.5, and 0.75 IRs. What am I missing?
 
Last edited:
I'm proceding on the assumption that 0.00 represents the axial measurement and the IRs that follow were taken at 2.5mm steps from center. If I am mistaken in that please correct me. I'm also focusing on the Greenback set due to my familiarity with those speakers. The same observations will apply to the other IRs.

Here's on-axis (".00") overlaid with the next IR ("1.00"):
1and2magvsfreq.jpg

Edit: Two prominent response features in the above are due to the microphone rather than to the speaker: the large peaks at ca. 115Hz and 3.5kHz. These are both due to the proximity effect that all directional mics have. Both peaks are broadband and easily audible.

This is the response of the first 21.3ms of each IR, which is all you need for the cab sound and all that most modelers can convolve, with no smoothing. The differences are well within the repeatibility of the measurement: IOW, you could acquire two IRs from the same mic position and see differences of this magnitude. Hopefully nobody thinks they hear differences between these two....

Here's an unsmoothed plot of the magnitude differences between the above two IRs, zoomed in to make the differences more obvious:
1and2diff.jpg

Note that, up to ca. 6kHz, the differences are +/- 1dB and all very narrowband. Also note that the apparent huge difference at ca. 7.2kHz is due to the difference in the depth of a notch that is extremely deep in both IRs. These are the kinds of small difference that your mesh grille can cause. Without smoothing of any kind, many differences that are visible in a magnitude plot like the above are not audible.

More to follow.
 
Last edited:
Thanks Jay.

Fixed, yeah 2.5mm so about 0.1", I used a metric ruler and tape markings on the grille mesh.
0.00 (dead center) and 1.00cm are so similar in sound I skipped 0.25, 0.5 and 0.75.
Initially I shot in 5mm steps but I noticed that it wasn't enough resolution around the sweet spot so I halved it to 2.5mm.

I'm proceding on the assumption that 0.00 represents the axial measurement and the IRs that follow were taken at 2.5mm steps from center.
Yep.

More to follow.
:nails


EDIT:
I also thought that 8cm (3") is enough because it started to change the sound in a bad way (imo), slightly further away than what is "mid-cone".

EDIT2:
@jay mitchell , note that this Greenback is from June 2021 (date code FG28) UK made from Thomann, and in my opinion it sounds honky by itself but just right to blend with other speakers.
 
Last edited:
Here's another comparison, this time between the 0.00 file and one of my IRs. Caveats: my IR was taken from a Marshall 1960A clone (4x12) loaded with Greenbacks of ca. 2015 production. Detailed response differences are not the point of this comparison. The responses have been smoothed to 1/6 octave, which still shows some details that are not audible.

NFandFF.jpg

The two overwhelmingly audible differences are the low-frequency hump at ca. 120Hz and the rising response from 1kHz to just below 6kHz in the 0.00 IR. Those are both entirely due to the response of the microphone and are present to some extent in almost all of James's IRs.

Having taken many, many IRs of the same cab on multiple occasions going back 35+ years, I'm well aware of the magnitude of a project such as this. My observations are not intended as criticism; they are simply to illustrate the influence of various elements in response measurements.
 
Last edited:
Those are both entirely due to the response of the microphone and are present to some extent in almost all of James's IRs.

The Shure SM57 and Celestion speakers were all bought from Thomann between 2020 and 2022 except the Vintage 30 which is from 2010.
The Poweramp is TPA3255 300W Class D driving the cab to a loud bedroom volume, enough to rattle a few things in the room.
The sine sweep was sent from my DAW via a Focusritre audio interface into the Poweramp and recorded back with the SM57.
The recording chain was clean, distortion free and boring, no extra coloration or 'character' from the electronics as far as I'm aware.

I'm not sure the anomalies you see are specific to my IRs, I did nothing extraordinary, just the tried and tested SM57 in front of Celestions.
Here is a graph of YA Friedman Blend Mix 01 and my raw mix of Greenback 5.50 + Vintage 30 4.50 you can hear in the first post.
Shockingly similar despite the completely different cabs and other unknown variables, not to mention the Celestion lottery.


Woody vs YA.png
 
The Shure SM57 and Celestion speakers were all bought from Thomann between 2020 and 2022
Irrelevant. You're misunderstanding what I'm saying.
except the Vintage 30 which is from 2010.
The Poweramp is TPA3255 300W Class D driving the cab to a loud bedroom volume, enough to rattle a few things in the room.
The sine sweep was sent from my DAW via a Focusritre audio interface into the Poweramp and recorded back with the SM57.
The recording chain was clean, distortion free and boring, no extra coloration or 'character' from the electronics as far as I'm aware.
Again, irrelevant.

I'm not sure the anomalies you see are specific to my IRs,
Nor did I ever say that they are. They are, in fact, common to all IRs that are captured with a directional mic placed close to the speaker. In case this isn't already clear to you, SM57s are directional mics (cardioid) and have prominent proximity effects.

Here is a graph of YA Friedman Blend Mix 01 and my raw mix of Greenback 5.50 + Vintage 30 4.50 you can hear in the first post.
If you get far enough off axis, the rising HF response of the mic is partially offset by the falling HF response of the speaker. Those IRs still both extend to just shy of 6kHz before they begin to fall off. No Greenback ever does that. See my plot above. Guitar speakers' actual HF responses all look a lot like that plot. A 12" transducer can't hold up to 6kHz, even on axis, without major EQ assistance. In the case of close-mic'ed IRs, that assistance comes from the mic.
 
Last edited:
Say Jay, why don't we discuss other types of IRs in another thread?
Although what I've posted here is purely educational and is specifically addressed to your IRs, I'll take your suggestion to heart. FYI, it's always a good idea when you're employing testing and measurement processes - as you're doing here - to maximize your understanding of the contributions that each element in the signal chain makes to the results. For example, an IR that sounds "honky" or "scooped" may sound that way because of the response of the mic, not the speaker.
 
I've added my other speakers and reshot old ones with louder volume including 0.25 0.50 0.75, took two full days of eardrum piercing sine sweeps (I used earmuffs).

First post and download link updated.

Welp, that's all the speakers I've got worth capturing.
I got more microphones like e906, M201, M88, Beta 57A, SM58. sE8, V7X, Large Condensers and Measurement Mics that may be useful, but I think blending two different speakers and distances with only two SM57's can get excellent results worthy being on records.


speakers.jpg
 
Hi KHAN!

18mm (3/4") Baltic Birch Plywood
W: 61cm (24")
H: 47cm (18.5")
D: 36cm (14") ← big factor.

Closed back oversized 1x12 cab similar to Zilla Fatbaby 1x12 in spec.
The tuning of this cab is very low, electrical impedance resonance peak is lower than of a 4x12 which makes it sound massive for a 1x12.
The ratio of internal air volume to cone area has a lot to do with how big the cab sounds.

Not the prettiest cab, but very functional, and it was way cheaper to DIY, you get one cab from a single 244cm x 122cm (8' x 4') plywood sheet.

1x12 cab.jpg


two-1x12.jpg
 
Last edited:
I've done some more experimenting with micing and noticed that concentric position matters a lot with the same axial distance, sounds like a slightly different speaker of the same type, I think the reason is the paper cone is not perfectly symmetrical and does not have the same thickness all around.
I did try to rotate the cab on its side and got consistent results so it is the cone itself not the distance to the floor.
Yet another factor to take into consideration when close micing a guitar speaker.

EDIT:
I'm trying to think of a more efficient method than capturing 2.5mm steps in 8 directions resulting in hundreds of IRs per speaker.
 
Last edited:
I've done some more experimenting with micing and noticed that concentric position matters a lot with the same axial distance, sounds like a slightly different speaker of the same type, I think the reason is the paper cone is not perfectly symmetrical and does not have the same thickness all around.
That is not the reason. It's one thing to observe an effect, quite another to correctly diagnose the cause.

I did try to rotate the cab on its side and got consistent results so it is the cone itself not the distance to the floor.
Consider that the cab is not cylindrically symmetrical. That makes a substantial contribution to the sound of the speaker in ways you apparently haven't realized. Think about relative arrival times of sound that is reflected and diffracted back to the mic from the cab edges. Differences in this time smear cause differences in which frequencies get reinforced and cancelled.

Edit: There is also the effect of the grille openings and their relationships to the location of mic diaphragm. I've had occasion to test speakers with that exact perforation pattern. I say from hands-on experience that you'd be much better off acquiring IRs with the grille removed. Obviously that would require a different methodology for mic placement, but you'd then find that there are far fewer unexpected response variations from one IR to the next.
 
Last edited:
Back
Top