Some of the 808's instruments/parts have trimmers, as mentioned below.

But generally for analog, component tolerances are either 1, 5, or 10%. 1% is pretty audible on oscillators, and 10% is pretty audible on filters.

So each instance of an untrimmed analog synth and drum machine will sound slightly different, even if they're coming off the same production line and using the same circuitry.

Component values also drift over time, so a contemporary 808 is not going to sound the same as it did when it was shipped in the 80s.

Basically analog's inaccuracies add character, detail, and interesting distortion. You can emulate them digitally, but emulating all of them accurately and convincingly is really quite hard.

I understand temperature affects the frequency of some oscillators as well. Is that a possible explanation as well? (I assume for crystal-based oscillators that would not be the case, but for musical/analog oscillators I assume these are not crystal derived).

Most analog gear has SOME type of temperature control. Tempco resistors (low resistance change with temperature), thermisters (couple a positive response to a negative response to cancel out) or other techniques (e.g. heaters - get the circuit up to max temp asap and keep it there).

But none of that prevents temperature differences, they just control for them. And you wouldn’t use temp control everywhere - just where it’d effect the sound dramatically i.e. keeping your main voice oscillators in tune, not keeping your cowbell at a precise frequency

This is correct. I have a couple of analog oscillators in my modular rack and the case needs to heat up for a couple of minutes before they generate a stable pitch. Not familiar with the exact reasons to be honest.

The temperature sensitive part in a VCO is actually the exponential converter inside, which has a strong temperature dependence. This is because the current through a transistor depends on the ratio V_BE / V_T, and V_T is proportional to temperature.

It's a neat trick that lets you build good temperature sensors out of transistors, which is very convenient, but in these VCOs you have to add temperature compensation to the exponential converter, and the temperature compensation is far from perfect.

If you take linear VCOs (instead of 1V/octave exponential), they'll be much more temperature stable.

Higher quality equipment that needs a stable timebase will use ovenized oscillators with an integrated heater and a control loop to maintain a stable temperature. These are all inaccurate before they are up to operating temperature.

TVs and CRT monitors were common to change the shape and color of the picture as it got to temp. You never made critical color decisions on a monitor that hadn't been on for at least 15 minutes, preferably 30.

Basically all components have temperature dependent values

The 808 in particular has an additional twist. The transistors used in it were out-of-spec rejected parts as a cost savings measure. So recreations and repairs can have difficulty recapturing the signature sound if they go only by part number.

https://secretlifeofsynthesizers.com/the-strange-heart-of-th...

I’ve owned a few 808s and sampled all of them. They were indeed all different sounding. The cowbells are one exception and sounded more like the 2 Live Crew example. Most of the variation was in the hi hat and cymbals. Which IIRC is 6 square waves and a noise source.

Very interesting! Thank you! I did a little looking and found this FWIW...

I noticed in OP's post that the two frequencies for the cowbell are 540hz and 800hz (although I don't see where this is denoted on the original schematics.) But we'll assume they're accurate numbers.

While a perfect fifth is a difference of 7 semitones, this produces a difference of only 6.804 semitones. It would seem that its designers didn't have a perfect fifth in mind.

Also, 540Hz is C#5 -45 cents and 800Hz is G5 +35 cents. So neither note seems intended to correspond to any natural key.

> While a perfect fifth is a difference of 7 semitones, this produces a difference of only 6.804 semitones. It would seem that its designers didn't have a perfect fifth in mind.

Remember that 7 semitones is not a perfect fifth. (Note this depends on your definition of perfect fifth, some people allow for being close, since equal temperament is so common.)

While a perfect fifth is a 3:2 ratio of frequency, seven semitones isn’t exactly 1.5, it’s more like 1.498.

800/540 is ~ 1.48, so I definitely agree the designers weren’t going for a fifth, very likely they wanted some beat interference to give it a nice metallic warble.

You are correct sir.

I was speaking (a bit erroneously) in the context of an equal tempered scale. (I think.)

Indeed, the natural overtone series is the proper context for a true perfect fifth.

One of the diagrams shows the frequencies as 1.85ms and 1.25ms, 540.5 and 800 Hz respectively.

In the original circuit,the same RC network is used for both the attack and decay so that circuit will indeed produce a 1:1 mark:space ratio.

Curious about the term "1:1 mark:space ratio" since it was also in the article. I would have used the term "50% duty cycle".

I purposely repeated the OP's wording but since we're talking about an astable multivibrator,frequency and duty cycle (in percent) makes sense. I did a lot of circuit design for communications systems and I associate the words mark and space with ones and zeroes in a frame/word.

I wondered if it was an English (not American) phrasing.

More on the origins of the terminology here: https://en.wikipedia.org/wiki/Mark_and_space

The schematic in the article shows trimpots controlling the 2 frequencies, so it was probably adjustable with a screwdriver.

You are right!

4:52

https://youtu.be/rPOqYHMpvqY?t=292

I think yours may be the first HN post linking The 2 Live Crew. And, indeed, Mr. Mixx is way way under appreciated.