808 Snare - Mutations
This post originally appeared on the SS-30M blog on 21/10/21
tl:dr
I tried to set my Hexinverter Electronique Mutant Snare (MS) module to match the TR-808, on which it is based. I found that the tuning of the MS 'shells' is quite high and not like most 808s. The original 808 schematic values give a frequency for the high snare oscillator of 499 Hz
and later component change brought it right down to about 335Hz. My
research, which included checking more examples' frequencies, shows that most 808s and its modern clones use the revised component
values and thus a lower pitched snare 'shell'. That's including the Roland ACB digital products. The MS is using the original component values, hence is higher in pitch. Noting the variation in tuning ratio between the two oscillators and also bearing in mind that the Q value changes decay time, it should be possible to modify the MS to match the sound of most 808 snares and clones.
Ensnared
The Roland TR-808's snare drum is the topic of this post, but more specifically I was trying to tune my Hexinverter Electronique Mutant Snare (MS) module to match the 808. It is based on the 808, so something close should be (or ought(?) to be) achievable. I was trying to work out how to do this and found myself caught up in a tangle of questions regarding 808 snare tuning. I will also add that I have not come to bury the Mutant Snare. Hexinverter do not claim the MS as a clone and merely say that the design is based on the 808. This is not a negative critique of the choices they made. I love my Mutant Snare! Anyway, I'm starting with a deep look into the 808 snare pitches.
808 Snare Analysis
Going Into Service
Here's what the 808 Service Notes have to say about the snare drum (SD) shell, or pitched element.
Two "bridged T-Networks", meaning two filters that are set to feedback and self-oscillate. These oscillators, because that's what they are used for here, provide "Fundamental wave-forms and harmonic wave-forms" - I will come back to that. And you can tailor the "output ratio" to get a different "characteristic" with VR8. VR8 is the Tone control.
Here is the block diagram:
Each oscillator is labelled 'IC14', indicating the schematic reference for which chip the op-amps are housed in. Note the Snappy control sets the amplitude of the input pulse to the oscillators and the noise envelope. The noise goes through a VCA under envelope control and a high-pass filter before being mixed with the shells. I won't be referring to the noise element again, but I will say that this element of the MS isn't particularly simialr to the 808 either.
T-Time
The oscillators are the famous bridged-T network (or Bridged T-network) type which, with a simple op-amp cicuit, produce sine waves that decay when the input is hit with a short pulse. This is a pinged resonant filter, which can be done with any such filter, but in a handily, cheap design.
Portion Out
Before doing that calculation you have to know what the values of the components are, and there is more than one answer to that question.
Firstly, towards the end of the Service Notes there is a section on 'Design Changes & Improvements'.
That section doesn't include anything about the SD though, because these changes are all recommended to get the 808 working correctly and the SD is working faultlessly on the "early" units. Note that this is dated June 1981, and the 808 was in production between late 1980 and 1982.
Instead, going to the schematic there is another note off to one side:
R191 relates to the trigger and, although interesting to see that
there is a different trigger level to each oscillator, it will not bother us, for now. It's a minor change in value.
Let's look at the schematic.
C58 and C61 are the 'C2' of each oscillator's bridged T. Now it's ime to do some maths!
Safety In Numbers
Fortunately, someone has already done the number crunching on the 808 for me. It's always nice to have someone else's answers in case you are fooling youself. This calculation came from Kurt James Werner's post here: https://kurtjameswerner.tumblr.com/post/51352144814/chuck-tr-808-emulator-snare-drum-sd-emulation
f_left = 173.334 Hz (running the calculation with the original C58 value gives f_left = 249.629 Hz ).
f_right = 335.976 Hz (running the calculation with the original C61 value gives f_right = 498.999 Hz )
Thanks, Kurt!
Side (Drum) Note
Before I saw Kurt's page I read Gordon Reid (hallowed be thy name) on the subject of snare synthesis in his Synth Secrets series for Sound On Sound. I was left scratching my head though because his analysis on the 808 snare was, well, wrong. Kurt confirmed my doubts and and gently points out where Gordon Reid got confused. Yes, yes, wrong! He just gets the block diagram muddled and appaently misunderstands the arrows in a potentiometer symbol with signal flow. Ah well. Feet of clay and all that.
Just Typical!
As Kurt points out, the Service Notes also provide a table with values "typical and variable" for each drum voice on the 808.
Here, the High oscillator is 476 Hz and the Low 238 Hz and so Kurt reckons that this table was not updated with the 'portion changed ' of the SD schematic. His calculation shows that the table values are close-ish to the original values for C58 and C61. Close; not exact.
Super Hiro
This conclusion makes sense also as that table is nicely formatted and typeset, whereas the amendment to the schematic is just typed (type-writer'd) in. Kurt thinks that they forgot to update the table, and I agree. In fact I would go further. I think this table is the specification that the sound designer gave to the circuit designer, Hiro Nakamura. It is mentioned in Roland's Story of the TR-808 that...
"...the engineers used analog synthesis to recreate the sounds—initially on Roland’s System-700 modular system - before recreating the parameters on the TR-808’s analog circuitry."
It also states that Hiro was asked to design the voice circuits, so I am guessing that Tadao Kikumoto drew up a spec based on experiments on the modular. In the photo below you can see Tadao, posing with prototype 808s, and a System 100m lurking above, but maybe they used the System 700 as reported.
 |
| Tadao Kikumoto (chief designer?) with prototype 808s - sadly I can't quite read the component values |
An Uncertain Ratio
Now let's come back to that comment in the description of the SD:
"...two bridged T-networks for fundamental wave-forms and harmonic wave-forms".
The first harmonic is double the frequency of the fundamental, a ratio of 2:1. For example 238 Hz x 2 = 476 Hz, just as in the table of typical values above. This table seems to be a kind of idealised specification though.
The actual harmonic, when using the rev B 'C2' value is 173.334Hz x 2 = 346.668 Hz, which is not what you get from the other oscillator. That is 335.976 Hz, which is 10.692 Hz lower. The actual ratio is 1.93:1.
What about the original values?
Fundamental would have been 249.629 Hz giving a desired harmonic of 499.258 Hz. The actual was 498.999 Hz, just 0.259 Hz lower and a ratio of 1.96:1.
What they gained in lowering the pitch they lost in accuracy of the harmonic, but not by much. Why did they allow that though? It might have just sounded better that way or ended up that being slightly out when real component values were used was all they could do. There is another (ahem) factor* involved though, which may have forced the change in ratio. I will return to that in due course.
(* Q, I'm talking about the Q factor).
Playing The Field
What about real 808s though? Using Audacity I took some examples of 808 snares and analysed their spectrum. It was quite easy with an isolated sample to find two peaks for the two oscillators, and work out how the snare was tuned.
I listened to the first known use of the TR-808 by Yellow Magic Orchestra on tour in 1980, but they only use the clap and not the snare.
High There!
Here's a video from someone with two 808s (I know, right!?). One with the earlier 'high' snare and a later one with a 'normal' tuning, presumably after the changes noted in the schematic, although I found that both were very high. Almost so high I wonder if the digital audio has been pitched up.
The turnings are like this:
Low (newer) 808 - 220 & 433 - 1.989 ratio.
High (older) 808 - 247 & 526 - 2.13 ratio.
With the older unit the high oscillator is particularly high, being 25 Hz above the components calculation. The newer unit is not as low as the calculation would suggest it should be. It's sort of in-between, but closer to the older, high pitch component selection. In fact this is the only example I found with such a high oscillator frequency. It just goes to show how much variation there is in 808 tuning.
What about other examples?
Sampling Sample Samples
My CS1x (what?) has a sample with 202 and 374 Hz ratio 1.85. That is closer to the revised values but still quite far off.
A random sample from my library, probably of a free CD of drum samples from the 90s - 173/338 ratio 1.95. This is tuned very close to the revised components values and as it really is supposed to be a typical example of an 808 I trust this to be a 'good' example of an 808 snare.
Synthmania has all the 808 sounds up and even plays the snare with snappy off and eiether end of the tone control to make measuring the frequency really clear and easy. This one is 177/340 Hz. What's also cool about this is that they provide the serial number of the 808 - 209265. This can be input to the Roland Serial Number Decoder and we learn that it was made around June 1982. Quite a late one.
Electro 808 Beats
Magic Juan
In the film All Hail the Beat by Nelson George, Detroit Techno Originator Juan Atkins says that he got an 808 very early on. He says this again in this electronic beats interview :
"We must have had one of the first 808 machines."
And he reckons he had "the first 808 in Michigan" too. Alleys of Your Mind/Cosmic Rain Dance came out in 1981 so there's no doubting that he was there at the very start. What's interesting is that the sound on those early tracks is the revised, lower pitched snare 159/332 Hz.
Bam & Baker
"Anyway, since we were basically trying to emulate the Kraftwerk sound, when we went into the studio we decided we needed a drum machine. What's more, after hearing an 808, we knew that was what we wanted, so I looked in the Village Voice and saw an ad stating, 'Man with drum machine, $30 a session.' His name was Joe and he had an 808, so we paid him to come in and do the beats."
https://www.soundonsound.com/people/afrika-bambaataa-soulsonic-force-planet-rock
Note that on Confusion by New Order (August 1983), which was produced by Baker, the 808 is 189/334 Hz, so, by now, Arthur had his own 808 with a different snare tuning.
In another twist on this part of the story I found in a Gearslutz thread that someone made their own Planet Rock cover version because they thought their 808 sounded so much like the one on the recording.
https://youtu.be/kll-f69zfb8
His serial number is 010731 which puts it as being built December 1980 and the tuning is 254/477 Hz. He also says that an example an from e-Bay auction with s/n 062399 is high, and that is built in May 1981 and another with s/n 093553 is lower, and that was made in August 1981. Recall that the Service Notes with the portion changed was dated June 1981, in-between these two examples. I found 062541 on YT, and that is from May 1981 and it's high - 235/506 Hz.
I think I've made my point there. It does indicate that Juan Atkins got his hot off the press in the second half of 1981 and had Alleys of Your Mind out in double quick time though. I can't find an exact release date, but that's the only thing that's making sense, unless he had it serviced and updated really fast.
Tilt your head back let's finish the cup
Now, how about the Beastie Boys', in fact Ad Rock's, 808? This is a machine with some serious history as not only was it used by the Beasties, but it was an in-house machine for Def Jam and the Wang Chung studio. It's Your's! T La Rock and Jazzy Jeff, LL Cool J and Run DMC. Taking a look at the sound of this machine from Brass Monkey, it is 175/326 Hz. Not a particularly good harmonic ratio of 1.86 and although the low oscillator is close to factory the high has wandered down a bit.
Hi, tolerance!
As we look at some real 808s it's becoming clear that poor tolerance in components must be causing what I'm seeing as some pretty wide variation in the tuning. Hence, if you're making a clone, you might want to factor that in and to some extent the MS let's you do that.
First though, some other clones.
A Rack Of The Clones
Now we move to clones and how they approach this change to the SD tone.
There are several types of clones. Digitally modelled by Roland using their Analogue Circuit Behaviour (ACB) technology in the TR8/8s 08 and cloud 808. Analogue voice clones in Eurorack format, full sized analogue recreations of the whole 808 machine with sequencers and so on.
Full clones are now led by the Behringer RD-8, joining the DIY Acidlab Miami and E-Licktronic Yocto.
Straddling both full recreation and Eurorack is the System 80 880 and then we have the 8raw8 which is just the voices with no sequencer in desktop format.
TR Rated
The Roland digital TR models are all more or less identical in terms of the snare. They all go with a fairly low fundamental of 167 and a 326 harmonic giving a ratio of 1.94. This is just slightly lower than the revised values. I should point our here that I'm not certain if any 808s were really as per the calculation. Kurt (see above) also notes that the rest of the non-ideal circuit could produce a different frequency than the equations would have you believe.
In any case it is possible to tune the snare on the ACB 808s. Although the ratio is likely to be fixed.
Euro-hacks
Eurorack is blessed with a few options for those in need of an 808 sound. The most well known being the Tip-Top SD808, but there was also an Analogue Solutions SD88, there's a Sound Force 808 snare (DIY), Falafular Snare8 (DIY), Acidlab 8SD (discontinued).
I could go on with more but they start to edge in 808 inspired territory instead of cloning.
The story behind these clones is often that the designer had access to a few actual 808s and tuned their design to sound 'right'. As we've seen though, it's only 'right' to one particular 808 unless you can adjust the tuning of both oscillators and match it up to your own ideal. Which brings us back to the Mutant Snare.
Mutant Snare
The Mutant range has a philosophy
There are already some pretty awesome clones of vintage gear out there in the modular world. The Mutant Drums were not made to try and fill that role. Although each Mutant began as a favourable classic drum topology, we wanted to create something modern and different, while still maintaining an analog nature. Features like CV inputs and signal routing not found on classic analog drum machines were incorporated to make these circuits powerful creative tools in your modular synthesizer
So, is it a clone or not? In may ways yes. The core of the sound generations is the same and the noise generator and oscillators are faithful. The mutations really come in the form of additional controls but the core of it is an 808 snare and you might expect to be able to obtain that sound from it.
 |
| My Mutant Snare |
Mutant Snare has a fixed 'high' oscillator set at 491 Hz and the 'low' oscillator is variable between 176 Hz and 1141 Hz. The fixed oscillator is tuned to a similar frequency as the initial run of 808s, before the revised components, and the second can be tuned to the same frequency as that too. That is the 'Shell Pitch' control. The noise is transistor based and the same controls are provided in terms of Tone and Snappy. The additional stuff takes it beyond what an 808 or most other clones can do, but I'm interested in the shell tuning only here.
The fixed, 'high' oscillator (although it switches to being the 'low' if you push the 'Shell Pitch' high enough) places the Mutant Snare firmly in the higher end of 808 snare tuning. Recall that the original schematic had the higher oscillator tuning at 499 Hz, the specification was 476 Hz and the revised schematic brought it down to 336 Hz. My survey of the 808s and the clones, including Roland's own, finds that the 'high' oscillator is typically in the range of 325 Hz - 345 Hz, or around +/- 3% of the revised design.
It seems that the Mutant Snare was designed with component values close to the original schematic, which is not what I would have expected. When I came to tuning mine to a standard 808 snare sound I found it was not sounding quite 'right'. It was 808-ish, but only after checking out all the other examples and doing some measuring did I realise why. Some commentators have said they weren't happy with the punch of the Mutant Snare. Maybe this is why, except the noise has to be part of this too. As there is no way to adjust the higher oscillator to match more typical sounding 808s I'm left wondering if I can modify it instead.
Mutated Mutant Snare
The Mutant Snare was initially offered as a DIY build. There are still instructions around for the build here: http://data.mutantbrainsurgery.hexinverter.net/hexinverter_MUTANT%20SNARE_assembly%20manual_v105.pdf
However, there is no calibration of the snare tuning and the passive components, which set the tuning, are SMD and were provided pre-assembled on the board. I make this clear to observe that there was no optional tuning of the snares.
Mutant Analysis
Fortunately, the BoM is still available and although there's no schematic or detailed layout provided, the pictures in the assembly guide are sufficient to trace tracks as well as locate and identify components.
Going from the (Shell) Tone control it's quite simple to find the decoupling cap from each oscillator and from there back to the rest of the bridged-T. I'm interested in the fixed 'high' oscillator, to start with.
Mutant Fixed Osc
Decoupling - C27 (0.47uF)
Right-hand or 'high' Bridged-T oscillator using the component references from the 808 SM
R1 - R109 2K + R76 220R (not 2K2)
R2 - R82 1M
C1 - C30 6n8
C2 - C31 6n8
This is almost exactly as in the Service Notes schematic and what Kurt called f_right (above) except for R1 being made of two resistors and a very slightly higher value. Now I did my own calculation for frequency to check and I obtained 496.7 Hz, which is just a couple of Hz lower, but you can see how that small extra value on R1 made an impact.
I measured my actual MS at 491 Hz, which is lower than that, but I'm not going to get in to guessing why. Anyway, it confirms that the MS is using the original schematic values.
Great, so how could I modify this? There are two options. Change the fixed values (easy-ish) or somehow add a variable control.
808 Status Quo
Changing the fixed values to decrease the frequency maintains the basic functionality and is the easiest option. Even if I do go beyond that I need to think about doing this anyway, so I'll start here.
This change would mean increasing the value of one of the four components. Roland opted to up the value of C1 (C58) on the low oscillator and C2 (C61) on the high oscillator. So, not the same cap in each case. Hmm!
Right on Q
I wonder if there's any significance in that decision, and then I recall that there's also a Q calculation. In filters we think of Q as something to do with the resonant frequency, but Q is actually about damping of oscillations and when we ping a resonant filter the oscillations decay at a rate determined by the Q factor. This means that Roland may have been tweaking both the the frequency and the Q in a deliberate way when they changed values. Any change to the components could affect the Q factor if the ratios are altered. As they only changed one of the capacitors this is what actually happened.
Q-808
Getting into the maths of Q a bit further, when C1 and C2 are the same value, the denominator in the Q equation is always 2 and Q is then only determined by the ratio of R1 and R2. The capacitor values are irrelevant in this case. Note that for resonance to occur the system's Q must be greater than 1/2, so where the capacitors are equal and the denominator is 2 the numerator must greater than 1. As this is derived from the square root of the ratio between R2 and R1, R2 must always be higher than R1.
In this case Q = 10.66. The spec sheet in the 808 Service Notes gives a decay time of 60ms, but not how that is derived. It may just be a target for the designer to achieve, but I will come back to that later on.
For the revised values used by Roland on the 808 they changed different capacitors in each case. Let's see what the changes did to the Q factor.
|
Revision |
Low Oscillator Q |
High Oscillator Q |
|
A |
17.4 (680R, 820K, 27n, 27n) |
10.7 (2.2K, 1M, 6n8, 6n8) |
|
B |
16.3 (680R, 820K, 56n, 27n) |
9.9 (2.2K, 1M, 6n8, 15n) |
In both cases the Q factor decreased, in-line with the decrease in frequency. This is because although different caps are changed one is increased and the other decreased. It's still not really clear why they selected the changes in this manner though.
As revision A had a higher Q it would be more damped and the decay would have been shorter. Does this matter? If I'm going to alter just one fixed component, then probably not that much. It must be considered though, because if my change alters the frequency to match the revised units but not the Q then it won't be (and therefore sound) the same.
Simplest Mod
If I changed C31 on the MS to 15nF it would bring it in line with most 808s and I might be quite happy with that. I might have to look at the variable shell too though because the Q factor is worth considering and in any case the lowest frequency of the variable/low oscillator still isn't quite low enough - 176Hz vs 167Hz - and might be more interesting if it went lower instead of higher.
Shell 2 Pitch Control
To add a variable control to the pitch of the other shell of the MS it could be as easy as replacing a fixed resistor with a variable resistor, or a variable resistor and some other resistors. Before I do that though, I wanted to check and see how Hexinverter did it.
Shell 1 Pitch Control
The other side of the Tone control goes to C29 (0.47 uF. Again, this is the decoupling cap for the output of the oscillator). From there, the track winds down to the 'A' op-amp in IC8 (TL072). Another track goes off to the jumper for selecting cut-off CV source, so this CV is taken from Shell 1.
Meanwhile around IC8 we see the bridged-T made from R59 (R1) 680R, C21 and C23 (C1 & C2) both 27nF. These are all exactly as per the original 808 schematic.
OTA Control
R2 is more complicated and this is where the control comes in. Initially it seems that R108 and R54 are in the place of R2. These are both 1M, which more than doubles the value in the schematic. R54 also connects to R62 220K and that tracks over to a pin on IC1, an LM13700.
This is more or less what I expected to find. The R2 resistance isn't simply connected to a variable resistor; there's an OTA in the circuit and a voltage from the Shell Pitch control sets the resistance. In theory this means that a CV could be added to control the Shell Pitch, although that feature isn't provided on the MS. I expected this is because, in general, it's better to give you more control. Hexinverter didn't actually make full use of this, but I wonder if he planned to originally. Manually twisting the shell Pitch control throws up some cool tabla type effects, so this would be good to put under CV control.
Continuing the tracing of the circuit, the track from R62 connects to pin 12 on IC1 which is, unsurprisingly, the output of one of its OTAs. At this point I'm going to stop the analysis of this side of the circuit. Suffice to say that there's a significant amount more.
Shell Pitch Plus?
It's also worth pointing out that the rest of this OTA and the other, a V2164, are fully utilised with no spare capacity. And finally, it would not be at all difficult to add a CV control to Shell Pitch. The control is a voltage divider between ground and rail (+12V, I think).
Adding a CV pitch control to shell 2 would be a lot more work, if I were to simply replicate that design and it's not really required.
March Of The Mods
Adding a simple variable resistor would be relatively easy to do, so surely someone has already done this to an 808. Searching for 808 mods, and R198 in particular, I came across this comment on synthforum.nl by Arnoid, back in 2005.
The next four modifications are based upon "The Mix" article on
modifying the 606, namely the snare modifications. The 606 snare
is quite similar to the 808 snare, though the 808 snare goes a little further circuitry wise. I compared the schematics for the 808 snare with that of the 606's, and was able to determine how to adapt the 606 snare mods with the 808's.
----------------------------------------------------------------------
Snare drum tuning (both oscillators)::
The 808 snare, unlike the 606, has two drum tone oscillators. The pitch of these are controlled by "R197" and "R198".
Check out the Snare Drum Pitch Mod in "part 2" of the 606
modification article (diagram 3).
It involves removing "R112" in the 606 and replacing it with a 1=
meg pot and a 470 k resistor.
In the 808, you will need to add this modification for each of
the two drum oscillators.
Instead of removing "R112" (as done for the 606 snare pitch mod) you will need to remove "R197" and "R198" inside the 808, and replace each with a 1 meg pot and 470k resistor, as they show in Diagram 3.
https://www.synthforum.nl/forum/synth-hardware/vintage/10849-voor-iedereen-met-een-tr-808/page7
So, the 808 mod was copied over from the 606. Fine, let's look at the 606 (yeah, this used to be the SS-30M blog).
606 In The Mix
This is the 606 SD schematic
 |
| Geoff Waterson 606 snare mod |
"This will allow you to tune the snare up and down about two octaves."
Give The Drum (err) Sums
Let's run some numbers using that idea.
Starting with the MS values for R1, C1 & C2 and adjusting the R2 value. We'll call the two variable resistor settings x and y because low makes it high and high makes it lower and it gets to confusing otherwise.
Rx = 470K, Ry = 1.47M
Zx = 470K x 2K22 x 6n8 x 6n8
Fx = 1/2PI x SQRT (Zx) = 724 Hz
Zy = 1.47M x 2K22 x 6n8 x 6n8
Fy = 1/2PI x SQRT (Zy) = 409.7 Hz
Well, that's no good at all! The lower pitch is still higher that the low setting of the newer 808s. Also the range is less than an octave. Clearly the 606 is tuned differently and I need to start over for the original 808 design values.
The MS Shell 1 pitch is tune-able over the range 176 Hz and 1141 Hz. I don't want or need to match that (2.5 octave) range for Shell 2, but I do need to get it down to the lower end of the 300s of Hz.
As the Zy value wasn't high enough, let's try again with an 820K fixed resistor with the 1M variable.
Fx = 1/2PI x SQRT (Zx) = 548 Hz
Fy = 1/2PI x SQRT (Zy) = 368 Hz
Dammit! Not there yet.
Again with 1M for both resistors? No. Only 351 Hz.
OK, so what if I change the C2 capacitor to 15n, like Roland did, AND add a variable resistor? Back with original resistor mods of 470K + 1M.
Fx = 1/2PI x SQRT (Zx) = 487 Hz
Fy = 1/2PI x SQRT (Zy) = 275 Hz
That works for the lower end, but now the higher end is not as high as it was. It would be nice to keep that up to the same level as before - around 500 Hz. So, splitting the difference of the C1 value and choosing 10nF.
Fx = 1/2PI x SQRT (Zx) = 597 Hz
Fy = 1/2PI x SQRT (Zy) = 338 Hz
Ah, just a shade too high. 12n?
Zx = 470K x 2K22 x 6n8 x 12n
Fx = 1/2PI x SQRT (Zx) = 545 Hz
Zy = 1.47M x 2K22 x 6n8 x 12n
Fy = 1/2PI x SQRT (Zy) = 308.4 Hz
That's the one! Phew.
So, to recap. Removing R82 and replacing with the 470K and a 1M pot and also replacing C2 with a 12nF give a usable range for the pitch of the 'high' oscillator of 308 Hz - 545 Hz. That spans both the factory fit 808 component options.
The fact that R2 is starting in roughly the same proportion to R1 as before, might help with what comes next.
Decayed Counter
Evidently these changes will also affect the Q factor, or decay, of the snare too.
I'll introduce here an equation for the decay time of the bridged T, based on this Knuts Acoustics article on Q.
T = 2.2Q/f
Where T is the time to decay 60dB, Q and f are as per the equations in the 808 SM (above).
The Roland Service Notes provide a guide for decay time, but no real explanation.
 |
| TR-808 Voice Decay Time |
What I really need now is a conversion of that handy equation above to work out dB drop which is equivalent to one tenth.
Voltage to decibels is the easy part. dB = 20log10(V1/V2). In our case V1/V2 = 1/10 so the dB drop is -20dB (remember that taking the base 10 log of any exponent of 10 is equal to that exponent).
The harder part is finding (or, eek!, deriving) the basic form of the equation I found at Knuts Acoustics so that I can plug in the correct value for -20dB instead of -60dB. Well, I can work out that -60dB is the same a 1/1000 easily enough. What about that 2.2 though? Where is that from? Knuts say that this is an approximation, so any number of assumptions may have been made.
Deriving Me Mad
What is for sure is that deriving these equations is a devil of a job. Even though my university engineering maths allows me to at least recognise some of what is going on, things like inverse Laplace transforms are beyond what I can pick up and do again now. I know it's about impulse response and I know that stuff is hard. I've dug around the internet but most discussion of the bridged-T is related to the frequency, not the time domain so, this leaves only simulation and guesswork. I like a bit of guess work though and there's enough here to be getting on with!
If the basic equation for decay time is T = xQ/f (where x is some variable for the rate of decay in this case), then this can be restated as
T = xQ x 1/f.
1/f is the period of one cycle of the oscillation and we can then see that xQ is a count of the number of oscillations from the start of the impulse. That ties up with definitions for Q which derive it from the change in amplitude between consecutive cycles.
As calculated above, the original high oscillator, with components as per the schematic had and f and Q of
f = 499 Hz
Q = 10.66
If we assume (hmm!) that T is 60 ms, as per the postulated specification (in the 'typical' values table), we can rearrange to find x in this case.
T = x x Q x 1/f
x = T/(Q x 1/f) = 2.81
However, if I do the same sums for the low oscillator, again assuming that the decay time is designed to be 60 ms, then x = 0.86. This means that the 808 cannot be designed such that the oscillators have equal decay time. Is that right? Or is this x factor a variable too?
Note that the Q for the high and low oscillators are not in the same ratio as the frequencies. In this case although the frequencies (499/249.6) are 2:1 the Q factors (10.66/17.36) more like 3:5. When the revisions were made to the tuning these ratios stayed in place too, which is probably important.
There's not much more I can do now except wonder what the real maths is here. None of the other bridged-T oscillators in the 808 are much help. So, it's back to simulation? Or... Simply measuring my Mutant Snare? The calculated f and Q of the high oscillator are slightly different - 497 Hz and 10.61 - but the actual f is 492 Hz. Or is it? I'm going to digress in 3, 2, 1..
Digression
The MS oscillators distort. Well, they do unless you are very, very careful. I had to attenuate the trigger and set the Drive control just high enough otherwise the peaks of the first few cycles are squared off. If I'm really careful I can then get a cleaner shape, although the noise floor is higher on a quieter signal. In this case I can measure the frequency at 495 Hz.
WAIT a minute though. Don't the 808 oscillators behave the same? The trigger is conditioned already, but is the MS really getting this wrong?? To check that I need an 808 or an ACB clone. Fortunately Roland provide videos with clean audio and demos of the tone control so I took a sample of a low oscillator and this is what it looks like.
This is 334 Hz and as you can see the wave form is nice and smooth. You can also see that it's not symmetrical. I did wonder if mismatching the capacitors might do this, but I might have to come back to that. The decay time for Vpp to 1/10 Vpp (as quoted by Roland) is from ~1.4 to 0.14~ which is about 25 ms -'ish'. That is much shorter than the 60ms 'typical' value, so I think we can safely ignore that from now on.
Let's look at the MS tuned to a similar frequency (366 Hz)
It's about the same in terms of decay time. So, that's good, I think. It is far more symmetrical, but it has matched capacitor values, which might explain that. I'm only going to find out more on that with simulation or actually doing the mod myself. It also looks quite rounded. Now I'm going to punch up the trigger.
And just look at the wave-shape. The only thing I touched compared to the previous was the trigger level. The peak in the frequency has actually shifted down to 345 Hz too, which seems... anomalous. I assume this was a design decision and the idea is to give the snare plenty of oomph, but personally I'd prefer to be able to dial this. The Drive seems to kick it up almost immediately. It's a not a major complaint either, as the sound is pretty clean at 9 o'clock. But still...
Decay Sustain
What I can say is that altering Q will do something to the decay time, so I need to be careful with this.
Looking at the decay time changes when altering the Shell Pitch it's quite clear that the decay gets very short at the highest frequencies. The below screenshot is of the lowest, middle and highest Shell Pitch settings.
 |
| Mutant Snare Shell Pitch - Low, Middle & High |
State of Decay
At this point I have found a resistance range for R2 that tunes the desired range with a capacitor slightly lower that used by Roland in their revision. Unfortunately, this alters Q such that the the decay is now not the same for the same frequency. The reason I've arrived here is that I was trying to keep the range wide enough to cover the original frequency and the revised. In conclusion, keeping both might not be possible without a switched capacitor and possibly resistance range.
That kind of decision can now be deferred to the actual implementation, but in short you cannot have your cake and eat it when you start changing the values of these bridged-T componentsFIL IN
So far I've been analysing the circuits to see how I might alter the Mutant Snare. There is another solution though. If I turn the Shell Tone control fully anti-clockwise I only have one of the two MS oscillators, of course. What if I then bring another oscillator, from another source entirely, into the mix? If it's on the same trigger, what difference does it make that it's not coming from the MS?
I just happen to own a Future Sound Systems Convulsion Generator (FSS FIL1). In fact I won it in a raffle and was handed it by mister FSS himself, Finlay Shakespeare! The FIL1 is simply a bridged-T resonator/oscillator, just like the ones in the 808 snare. R1 and R2 are adjustable; C1 and C2 are fixed at some unknown value (although I could work this out if I wanted to). With R2 set to minimum and R1 at the bottom end of it's range the FIL1 will generate convulsions at around the right frequency. The only difference is that they are very, very short.
As noted above, you cannot simply dial in any decay time and frequency whilst only altering the R values. The ADDAC System T-Networks module (ADDAC103) has switches for the range of each of it's four bridged-T oscillators and the middle one is called 'mod' not 'mid' and provides a longer decay.
As I have the FIL1 to hand though, I patched together a mix of it and the MS. With the Shell Pitch set at around the 335 Hz area and the FIL1 and somewhere around half that I finally had a lower pitched 808 snare! Hooray! What's more I can choose any combination of frequencies and replicate any 808 snare you care to name.
Wait! Can I do that? I mean, is that still the same as the 808? The low oscillator is now mixed after the noise. It's entirely outside the MS. Well, the Shell/Snappy mix is just a mixer and then a buffer. There's no extra filtering or envelope control, just mixing. The order in which elements are mixed doesn't much matter here. The main difference is the length of the decay on the FIL1. It's surprising how little that appears to matter. It's not correct, compared to the 808, but it makes a serious difference to the sound.
Apart from the inconvenience of patching in the additional module and the short decay, the only difficulty with this solution is tuning in the FIL1 and MS Shell Pitch. An accurate and fast frequency counter or spectrum analyser is needed to do that easily.
Conclusion
I have shown that it the Mutant Snare is tuned rather high, in compassion to typical TR-808s and although some people like the original tuning it's not as common and not as I would like it. At least not all the time. The MS could, in principle, be modified to match the revision B 808s built after mid-1980, or even taken further to give variable control over both snare shell pitches. Along the way I have pointed out a small number of other modifications that could be made to extend the features of the Mutant Snare, making a Mutoid Snare, if you will.
What I do next remains to be seen. What I have really done here is point out how the very capable and good sounding Mutant Snare differs from the 808. I could almost decide to build my 808 snare with the controls that I want to have. Something more like the Tip-Top SD808, which is more like a clone, but with some of the control from the Mutant snare, perhaps. I could force the two modules together into a mad team up. Even that would not really give me the ultimate solution though. The ultimate would be
Here's a table of values of I've collated from the above research.
Model |
f Low (Hz) |
f High (Hz) |
Ratio |
Q High |
Decay High - by 60dB (ms) |
Comments |
|---|---|---|---|---|---|---|
| TR-808 spec |
238 |
476 |
2 |
? | 60* | As per the Service Notes table of typical values and looks like the specification. Decay value is from this table and it's not certain how that was calculated |
| TR-808 Early |
249.629 |
498.999 |
1.96 |
10.66 | 42.7 | As per the schematic values – early units |
| TR-808 revised |
173.334 |
335.976 |
1.93 |
9.88 | 64.6 | As per the notes on schematic – most(?) units |
| TR-808 High |
247 |
526 |
2.13 |
Example of older high batch which is very high. Possibly rather unlucky values |
||
| TR-808 Avg. |
220 |
443 |
1.989 |
Example of later batch number with 'average' pitch. Still rather high and closer to the original than the revised schematics |
||
| TR-08 |
167 |
327 |
1.94 |
Measured from YT video |
||
| TR-8 |
167 |
326 |
1.94 |
Measured from YT video |
||
| Cloud 808 |
168 |
326 |
1.94 |
Measured from YT video |
||
| Behringer RD-8 |
171 |
331 |
1.72 |
Measured from YT video |
||
| Yocto |
? |
? |
? |
The Yocto schematic is exactly as the original 808 schematic, with no amendments |
||
| Tip-Top SD808 |
177 |
345 |
1.949 |
Measured from YT video |
||
| Mutant Snare |
176+ |
491 |
|
10.66 | 42.7 | High is nearest to early 808s. Low only just as low as later 808s. |
| CS1x |
202 |
374 |
1.85 |
|
||
| A Sample |
173 |
338 |
1.95 |
|
||
| Ad Rock's | 175 |
326 |
1.86 |
As heard on Brass Monkey | ||
| New Order Confusion | 189 | 334 | Arthur Baker 1983 | |||
| Planet Rock | 244 | 486 | "Man with drum machine, 20 dollars a session" Joe | |||
| Dr 808 The Earthshaker | 254 | 477 | SN: 010731 Dec 1980 | |||
| Cybotron - Alley's of Your Mind/Clear | 159 | 332 | Juan Atkins 1981. | |||
| Synthmania | 177 | 340 | SN - 209265 - 1982 |
Comments
Post a Comment