Hi, thanks for such an in depth explanation and attention to details. It's a good thing you are digressing towards 808 realm. In fact, if it wasn't for the 808 stuff I wouldn't have found your website!
I'like to suggest an addition to your list of measurement: the RC-808, a digital recreation by the original team at Roland. It exists as a free VST and standalone application. Here : https://rc-808.com/ I've taken a look at the tuning they use for the snare in the "classic" 808 kit they provide and it is: Low 177.03Hz High 315.44Hz
There's an LPF at 2474Hz, Q 0.550, on the low osc. I've also looked at the noise: BPF at 6974, Q 0.389 and gain 0.4
Have you got anything to share about the envelope? I'm trying to replicate the RC-808 outside of the VST and the standalone program.
Next I will take a look at the tuning in the Reverb.com samples and the D16 Nepheton VST.
Thanks for getting in touch! It's a pleasure to know my ramblings are of use to, or interest to someone else. I am getting closer to building my own tunable 808 snare too. I just got the Soundfource 808 Snare PCB and I'm going to modify that to my requirements.
I was sort of aware of the RC-808, but I think last time I looked it wasn't as advanced as this. In fact, there is a whole post about the development of the 808 which is completely fascinating! Not much on the snare, other than they rushed it a bit and had to squeeze the PCB layout quite a lot.
I've looked at the ClsH8V kit and it seems the SD8VCls snare is tuned to F3 and F4 (#53 and #63) which ought to be 174.61 and 311.13, but I measured and got the same frequencies as you. Not far off the expected, but a bit odd. It seems close to teh revised design, which makes sense. That LPF on the low osc is presumably to attenuate the click off the attach. At least that's what I hear. Maybe it was hard to simulate the bridged T accurately at low frequencies(?).
The snappy envelope is something I've only taken a quick look at. I made one up on a breadboard with the swing VCA and convinced myself that the shape of that and the character of the swing is important to the sound of the snappy. Other than that I just see it as a pulse shaper. The input pulse is 1ms long and the output according to RC-808 is about 150ms. The input pulse charges up C51 and C52, I think. And then it's discharging (RC), but I couldn't see exactly how that bit worked out to 150ms. I'd need to do more simulation to be sure.
Good luck with your replication. Is that software?
I've read the same blog post and it's indeed very informative!
Yeah as a VST the RC-808 is a little limited, the only parameters exposed are volume and panning, nothing can be modified with automation. As for the frequency, there is a "Pitch env" set if you look next to the Oscillator wave selector. There's only one step and it's 1.01388, 174.61 * 1.0388 = 177.03Hz, it isn't obvious in the UI but that's the only possible reason for that tiny shift. It probably was calculated to match some frequency value they picked. I have used the "808VinKitOpnCls.allprm" kit.
Thanks for your input on the env. Again, the Amp env on the right hand side has a few steps which aren't obvious. Also the last step of the curve being -inf, I think the resulting sound is in fact inaudible before that, probably at around -40 -50dB. I wonder if these steps were carefully selected.
Yes it's all software on my end, I know you're doing hardware (mostly?), but this research into deep 808 territory has proved to me that sometimes the characteristics of hardware have to be precisely recreated for it to sound "right".
Btw, how do you like the snare of RC-808? Do you hear it as authentic/faithful?
Totally agree about the limitations. As stated, it's not really an instrument yet. But it is very interesting and useful for us!:0)
I see what you mean about the pitch envelope now. It does seem like a fine-tuning after the cents parameter, although they don't use that. I set up a bunch of examples based on my measurements, but it was limited in accuracy by the cents (although not by very much at all) and of course the Tone setting is not easily mappable. I guess this pitch env setting is better for very fine tuning though.
The snappy env is now bothering me a bit. I tried a few simulations to adjust the length after I noticed that real-world examples seem longer than the RC-808. However, adjusting any of the values in that circuit in isolation alters the amplitude as well as the time. I reckon the Snappy control will compensate for that, but the maximum setting will be compromised a bit and I need to do some more work on it. As for your point about the first 2-3ms(?) I assume this is 'real' behaviour to match the snappy envelope more closely, yes. In a simulation that level is never zero volts, so starting at 0.1 seems about right. Returning to zero seems a bit wrong though. It's the difference between retriggering and isolated hits though. My simulation sends a trigger every 1 second so maybe fading away to zero is correct over longer periods.
Yes, I'm HW only, but trying to determine which parts are important.
I've started some comparisons with my TR-06 now. The first thing to note is that the TAE version on the TR-06 has a very different envelope for the oscillators. It's asymmetrical and has a much sharper transient. It's a bit like it's amplitude modulated actually. I have wondered if this is due to the different values of cap in the bridged -T in the revsied design but it seems to be here too and may just be inherent. Needs more work. But overall I think the RC-808 falls a bit short of accurate.
I've just checked again and the amplitude envelope of the oscillators is definitely not smooth. I think this is where a lot of SW emulations probably start go wrong. It is a decaying sinusoid, but the decay is not continuous and there's a good amount of attack in the first cycle. If you need more on that I could probably help. The good news is that RC-808 should make this easy to mimic, once you know how what the shape is.
Also, I took a look at how the Snappy's swing VCA is handled in SD8VCls and it hasn't been at all. Waveshaping is needed here to get the effect of the diode right and the distortion of the white noise here is crucial, I think. As I mentioned above I built the snappy envelope and filter and it's really noticeable how the character is altered by this and you can't just use any old VCA to get that kind of rattle that the 808 snare has. Again, it seems like the provision is there in the waveshaper to get that 'right' and maybe they will dial that in later. It's a bit annoying that you can't isolate the snappy from the shell on the Roland TAE products. I'd like to get that white noise on is own. I might try a subtractive approach though
Hi, thanks for such an in depth explanation and attention to details.
ReplyDeleteIt's a good thing you are digressing towards 808 realm. In fact, if it wasn't for the 808 stuff I wouldn't have found your website!
I'like to suggest an addition to your list of measurement: the RC-808, a digital recreation by the original team at Roland. It exists as a free VST and standalone application. Here : https://rc-808.com/
I've taken a look at the tuning they use for the snare in the "classic" 808 kit they provide and it is:
Low 177.03Hz
High 315.44Hz
There's an LPF at 2474Hz, Q 0.550, on the low osc. I've also looked at the noise:
BPF at 6974, Q 0.389 and gain 0.4
Have you got anything to share about the envelope? I'm trying to replicate the RC-808 outside of the VST and the standalone program.
Next I will take a look at the tuning in the Reverb.com samples and the D16 Nepheton VST.
Hi Fergus
ReplyDeleteThanks for getting in touch! It's a pleasure to know my ramblings are of use to, or interest to someone else. I am getting closer to building my own tunable 808 snare too. I just got the Soundfource 808 Snare PCB and I'm going to modify that to my requirements.
I was sort of aware of the RC-808, but I think last time I looked it wasn't as advanced as this. In fact, there is a whole post about the development of the 808 which is completely fascinating! Not much on the snare, other than they rushed it a bit and had to squeeze the PCB layout quite a lot.
I've looked at the ClsH8V kit and it seems the SD8VCls snare is tuned to F3 and F4 (#53 and #63) which ought to be 174.61 and 311.13, but I measured and got the same frequencies as you. Not far off the expected, but a bit odd. It seems close to teh revised design, which makes sense.
That LPF on the low osc is presumably to attenuate the click off the attach. At least that's what I hear. Maybe it was hard to simulate the bridged T accurately at low frequencies(?).
The snappy envelope is something I've only taken a quick look at. I made one up on a breadboard with the swing VCA and convinced myself that the shape of that and the character of the swing is important to the sound of the snappy. Other than that I just see it as a pulse shaper. The input pulse is 1ms long and the output according to RC-808 is about 150ms. The input pulse charges up C51 and C52, I think. And then it's discharging (RC), but I couldn't see exactly how that bit worked out to 150ms. I'd need to do more simulation to be sure.
Good luck with your replication. Is that software?
Hi, thanks for your reply!
ReplyDeleteI've read the same blog post and it's indeed very informative!
Yeah as a VST the RC-808 is a little limited, the only parameters exposed are volume and panning, nothing can be modified with automation.
As for the frequency, there is a "Pitch env" set if you look next to the Oscillator wave selector. There's only one step and it's 1.01388, 174.61 * 1.0388 = 177.03Hz, it isn't obvious in the UI but that's the only possible reason for that tiny shift. It probably was calculated to match some frequency value they picked. I have used the "808VinKitOpnCls.allprm" kit.
Thanks for your input on the env. Again, the Amp env on the right hand side has a few steps which aren't obvious. Also the last step of the curve being -inf, I think the resulting sound is in fact inaudible before that, probably at around -40 -50dB. I wonder if these steps were carefully selected.
Yes it's all software on my end, I know you're doing hardware (mostly?), but this research into deep 808 territory has proved to me that sometimes the characteristics of hardware have to be precisely recreated for it to sound "right".
Btw, how do you like the snare of RC-808? Do you hear it as authentic/faithful?
Hi
ReplyDeleteTotally agree about the limitations. As stated, it's not really an instrument yet. But it is very interesting and useful for us!:0)
I see what you mean about the pitch envelope now. It does seem like a fine-tuning after the cents parameter, although they don't use that.
I set up a bunch of examples based on my measurements, but it was limited in accuracy by the cents (although not by very much at all) and of course the Tone setting is not easily mappable. I guess this pitch env setting is better for very fine tuning though.
The snappy env is now bothering me a bit. I tried a few simulations to adjust the length after I noticed that real-world examples seem longer than the RC-808. However, adjusting any of the values in that circuit in isolation alters the amplitude as well as the time. I reckon the Snappy control will compensate for that, but the maximum setting will be compromised a bit and I need to do some more work on it.
As for your point about the first 2-3ms(?) I assume this is 'real' behaviour to match the snappy envelope more closely, yes. In a simulation that level is never zero volts, so starting at 0.1 seems about right. Returning to zero seems a bit wrong though. It's the difference between retriggering and isolated hits though. My simulation sends a trigger every 1 second so maybe fading away to zero is correct over longer periods.
Yes, I'm HW only, but trying to determine which parts are important.
I've started some comparisons with my TR-06 now. The first thing to note is that the TAE version on the TR-06 has a very different envelope for the oscillators. It's asymmetrical and has a much sharper transient. It's a bit like it's amplitude modulated actually. I have wondered if this is due to the different values of cap in the bridged -T in the revsied design but it seems to be here too and may just be inherent. Needs more work.
But overall I think the RC-808 falls a bit short of accurate.
I've just checked again and the amplitude envelope of the oscillators is definitely not smooth. I think this is where a lot of SW emulations probably start go wrong. It is a decaying sinusoid, but the decay is not continuous and there's a good amount of attack in the first cycle. If you need more on that I could probably help. The good news is that RC-808 should make this easy to mimic, once you know how what the shape is.
ReplyDeleteAlso, I took a look at how the Snappy's swing VCA is handled in SD8VCls and it hasn't been at all. Waveshaping is needed here to get the effect of the diode right and the distortion of the white noise here is crucial, I think. As I mentioned above I built the snappy envelope and filter and it's really noticeable how the character is altered by this and you can't just use any old VCA to get that kind of rattle that the 808 snare has. Again, it seems like the provision is there in the waveshaper to get that 'right' and maybe they will dial that in later.
It's a bit annoying that you can't isolate the snappy from the shell on the Roland TAE products. I'd like to get that white noise on is own. I might try a subtractive approach though
I've now cracked the Q vs decay time problem. See https://ss30m.blogspot.com/2021/11/sd-8081-tuning.html
ReplyDelete