rhythm wolf teardown: any questions?

From circuitbending to homebrew stompboxes & synths, keep the DIY spirit alive!

Moderators: Kent, luketeaford, Joe.

Post Reply
User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Fri Mar 27, 2020 12:39 pm

i took a bunch of years off from analog design to focus on microcontroller optimization, and when i got back to doing analog, it was such a refreshing feeling.
openmusiclabs.com

User avatar
astrosound
Wiggling with Experience
Posts: 253
Joined: Wed Aug 06, 2014 11:14 am
Location: New England

Re: rhythm wolf teardown: any questions?

Post by astrosound » Fri Mar 27, 2020 2:41 pm

This is wonderful, thanks for sharing those mods guest. The percussion section is the only voice I haven't touched so I'm eager to crack the RW open again.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Fri Mar 27, 2020 3:23 pm

i put some more detail on the noise source in the wiki. if you think you want to have the noise source in there i can help figure out decent values. i did a bunch of mods to it, but never got it into a state that i thought was finalized.

also, ive done a ton of mods to the synth section, but i still havent figured out what i like best, so im waiting on publishing those. the envelope section one is a game changer for me, though.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Sun Mar 29, 2020 10:24 am

i just finished the kick drum. i liked how it sounded, so i didnt make many mods. i adjusted the tone sweep, which pretty much wasnt doing anything before. after doing all the mods, i decided i liked it without the tone sweep for a lot of the range, but that the low end sounded better with the tone sweep, so i did a compromise. ideally, you could put a pot on this parameter. i was thinking of repurposing the attack knob for this, but i wanted that as well, so maybe later ill put one in, or at least a switch.

1. change R181 to 100k. this adjusts the pitch bend amount. the min/max tones are 40Hz/62Hz. with the pitch bend these can go up to 64Hz/83Hz at the beginning of a note. 10k is the lowest i would go here, and that gives the full bend, and gives a sharp transition between the hi/lo end of the note. 47k also gives the full bend, but has a smoother transition. 100k only pitches up to 58Hz on the low end, and gives a longer transition. and 220k goes up to 55Hz and gives the longest transition. going above that makes the effect pretty subtle. i decided to go with 100k as the really high sweeps seemed a little comical.

2. change R188 to 1.5M. anything between 500k and 3.3M would probably be fine. the larger the value, the longer the sweep becomes for a given capacitor value. i went with a middle range value for reasons explained at the bottom of this post.

3. change R205 to 100k. this can be anything from 10k to 1M, with larger values giving longer decay times, but when R205 becomes greater than half the value of R188, the max sweep goes down, so there are diminishing returns here. 100k to 500k are probably ideal. this can be replaced with a pot to make the sweep time adjustable. the 50k pot could be fitted here, with a 2.2k resistor in series to limit the current at max value. ideally, a 500k pot would be used, as the current is still pretty high with the 50k.

4. change C98 to 100nF. this changes the sweep time with R205. the values given here gives a 10ms sweep time, which is very subtle. if you actually want the sweep effect, i would reccomend upping this to 330nF or 470nF. the way ive been measuring sweep time, is the time it takes for the collector of Q29 to go from 10V to ground. i set mine to 10ms, which is really short. 50ms is a pretty good setting, and times up to 200ms are reasonable. after that, its hard to hear the effect anymore, as the tone has decayed out. this voltage drops linearly. by increasing R188 it can be made more exponential, but this is risky for the reason noted below.

a word of caution with all of this: this part of the circuit is entirely beta dependent, which means it will change with temperature and between synths. for this reason, ive picked some conservative, midrange values, so that when beta changes, the circuit is still in a reasonable operating range. this is particularly important if you pick large value resistors, as beta could shift and the transistors dont turn on anymore.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Mon Mar 30, 2020 9:43 am

something wasnt sitting right with me about my findings on the kick drum, so i sat down with them again. i kept asking why did the original 808 settings not change anything? and why couldn't i get the snappy attack of the 808 and the downward pitch shift? i kept trying different components and nothing seemed quite right. i also noticed pretty early on that R39 was way too large to do anything. i tried smaller values, but it didnt seem to do anything either, so it got me wondering what the hell it was supposed to do on the original. which led me to this:

http://www.dafx14.fau.de/papers/dafx14_ ... ed,_ci.pdf

this is an academic paper describing the 808 functionality, and makes it pretty clear why the RW bass drum cant do the same things as the 808. the upshot, is that the impedance of the TUNE pot is too low. if you want to get an exact replica, you will have to replace the TUNE pot and C14,15, and R114. basically, youll have to replicate the 808.

the trigger pulse on the 808 is 1ms, whereas its 5ms on the RW. this is something that has been frustrating me on each voice ive worked on: the trigger is way too long, and changes the envelopes. you cant get nice, sharp spikes. it also causes retriggers of the pinged filters that are audible. its really annoying, and i havent figured out why they did it that way. im wondering if one of the envelope capacitors wasnt charging up enough (because they picked too large of a value), so they needed to extend the trigger time.

at any rate, the decay time for the pitch shift is around 5ms in the 808, which is the full trigger time in the RW. so it doesnt matter what values go into the RW at R205 and C98, as long as they are small enough to not extend the time appreciably. so the pitch up time is actually pretty close to the 808, and gives the nice attack sound, which is what i liked about my pre-mod RW. the beginning of the notes sounded so much better before my mods, but there was no pitch decay. with the mods, i could get whatever pitch decay i wanted, but the attack sounded wrong. so i did a compromise.

the paper proposes that the pitch decay in the 808 is handled by R39 and not R114. im not sure i fully buy this, but if it were teh case it would explain a lot, as the voltage at R39 never gets large enough to cause any effect because the low impedance at that node. which would also explain why lowering R39 didnt change anything, as it was the voltage that mattered. it needed to get above 0.6V, and it only go to 0.2V. im thinking of swapping the attack and pitch knobs, and just rebuilding that section correctly.
openmusiclabs.com

ihav2p
Common Wiggler
Posts: 110
Joined: Wed Apr 28, 2010 2:03 pm

Re: rhythm wolf teardown: any questions?

Post by ihav2p » Mon Mar 30, 2020 10:35 am

I was also reading that paper when you made your first post about the kick section.

In my opinion, there is no reason to get the kick to sound identical to an 808 because new music trends have made the definition of an 808 kick very fuzzy. I used to believe I could ID an 808 kick from a mile away, but after hearing them pitched every which way and bent sideways my mental picture has become so vague.

IMO its enough to just get it sounding decent and call it a day, although I appreciate the lengths you've gone to to figure out what does what. It's definitely good for science!

[edit] The thing about the triggers is a real problem though, I feel like it applies to all the voices.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Mon Mar 30, 2020 11:52 am

yeah, the triggers suck. it actually wouldnt be difficult to mod that, if youre interested. since they are all driven from the same trigger line, you would just need to cut the line, and insert a capacitor. a better mod would be to put a capacitor, resistor, and HC14 or opamp, but the cap should work just fine.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Mon Mar 30, 2020 11:54 am

also, i agree about the kick. i was thinking of adding a transistor so i can control the pitch decay and initial pitch as two seperate parameters. that way i get the nice attack sound, but also a deep, decaying bass note.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Tue Mar 31, 2020 8:54 pm

ok, so now im working on the hihats. ive got the pitch mod working, and i quite like it, so im going to finalize it. i dont really like the analog noise, so im probably going to repurpose the "TUNE" knob, and make it an actual tune knob. the other option is to add a bit of the envelope to the pitch, so they pitch down a bit during the decay.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Thu Apr 02, 2020 4:27 pm

ok, almost done here. i just need to modify the min decay times. so far i have implemented:

1. true pitch shift
2. variable pitch for O/C. either can be higher than the other based on a switch, or they can both be the same.
3. closing of the open hat, when a closed hat is triggered.
openmusiclabs.com

User avatar
Mungo
Super Deluxe Wiggler
Posts: 1256
Joined: Fri Jan 04, 2013 8:12 pm

Re: rhythm wolf teardown: any questions?

Post by Mungo » Fri Apr 03, 2020 1:02 am

guest wrote:
Mon Mar 30, 2020 9:43 am
something wasnt sitting right with me about my findings on the kick drum, so i sat down with them again. i kept asking why did the original 808 settings not change anything? and why couldn't i get the snappy attack of the 808 and the downward pitch shift? i kept trying different components and nothing seemed quite right. i also noticed pretty early on that R39 was way too large to do anything. i tried smaller values, but it didnt seem to do anything either, so it got me wondering what the hell it was supposed to do on the original. which led me to this:

http://www.dafx14.fau.de/papers/dafx14_ ... ed,_ci.pdf

this is an academic paper describing the 808 functionality, and makes it pretty clear why the RW bass drum cant do the same things as the 808. the upshot, is that the impedance of the TUNE pot is too low. if you want to get an exact replica, you will have to replace the TUNE pot and C14,15, and R114. basically, youll have to replicate the 808.
Nice to be cited even if it isn't indexed. Their focus was on recreation/copying the circuit rather than tuning or extracting a parameterization, but great they added some insights along the way.

You could scale the other parts around the P7+R60, C14 and C15 values, so changing R19, R39, C17, R180, R181 and R114, possibly also R188, R205 and C98. Since you are proposing to change most of those anyway for your mod its possibly time to go back and work around the constraint of the P7 value.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Fri Apr 03, 2020 9:29 am

im not sure the analysis in the paper around that particular section is correct, as Q43 could never turn on after the initial envelope closes. but, for the rhytyhm wolf, the low P7 value means that the voltage at that juncture never exceeds 200mV, so the diode cant turn on, and no current can flow (at least after C17 has been exhausted). it would also incur a large thump at the beginning, as the relative amplitude of the diode turning on during the attack phase will be much larger.

there are a bunch of other papers by that author. ill try to get them posted later today.
openmusiclabs.com

User avatar
Mungo
Super Deluxe Wiggler
Posts: 1256
Joined: Fri Jan 04, 2013 8:12 pm

Re: rhythm wolf teardown: any questions?

Post by Mungo » Fri Apr 03, 2020 6:12 pm

guest wrote:
Fri Apr 03, 2020 9:29 am
Mungo wrote:
Fri Apr 03, 2020 1:02 am
guest wrote:
Mon Mar 30, 2020 9:43 am
something wasnt sitting right with me about my findings on the kick drum, so i sat down with them again. i kept asking why did the original 808 settings not change anything? and why couldn't i get the snappy attack of the 808 and the downward pitch shift? i kept trying different components and nothing seemed quite right. i also noticed pretty early on that R39 was way too large to do anything. i tried smaller values, but it didnt seem to do anything either, so it got me wondering what the hell it was supposed to do on the original. which led me to this:

http://www.dafx14.fau.de/papers/dafx14_ ... ed,_ci.pdf

this is an academic paper describing the 808 functionality, and makes it pretty clear why the RW bass drum cant do the same things as the 808. the upshot, is that the impedance of the TUNE pot is too low. if you want to get an exact replica, you will have to replace the TUNE pot and C14,15, and R114. basically, youll have to replicate the 808.
You could scale the other parts around the P7+R60, C14 and C15 values, so changing R19, R39, C17, R180, R181 and R114, possibly also R188, R205 and C98. Since you are proposing to change most of those anyway for your mod its possibly time to go back and work around the constraint of the P7 value.
im not sure the analysis in the paper around that particular section is correct, as Q43 could never turn on after the initial envelope closes. but, for the rhytyhm wolf, the low P7 value means that the voltage at that juncture never exceeds 200mV, so the diode cant turn on, and no current can flow (at least after C17 has been exhausted). it would also incur a large thump at the beginning, as the relative amplitude of the diode turning on during the attack phase will be much larger.
Current continues to pass through Q43 (mostly in one direction!) at such low voltages, the 808 has similarly small voltage swings at that current summing node despite the much higher impedances. The D52 + C39 diode pathway can be removed and the frequency modulation over period+amplitude effect remains, but eliminates the "sigh" they talk about. This all works when you flip to thinking about the currents rather than voltages as the diodes and transistors have exponential conformance across many many decades.

I've quickly run through the tuning here and as predicted it was possible to get that envelope section to replicate the 808 behaviour very closely while keeping the 500+something ohm tune control.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Fri Apr 03, 2020 7:09 pm

there is a pitch up at the beginning of the note, when Q43 is on, and that can be done on the RW, but i dont see how current can flow thorugh Q43 (as the author states) after this period is over, and create the "sigh" as he describes is. as best i can tell, the sigh is a result of current through R161, both charging C39 and conducting through D52. once C39 is charged, its just D52 conduction. this can be seen on his plot of frequency versus time, where the pitch goes up on every half cycle, when D52 turns on. but, there is a discrepency in this graph between his model and the original circuit. you can see that the bottom of the frequency oscillations shifts up on the original for a period of 150ms (exactly 4 time constants of R161 and C39), whereas his model is flat for this period, only going up for the diode conductions. this is the latter portion i dont believe can be accomplished on the rhythm wolf without changing the pot, as the voltage level at the juncture never goes above 200mV, and therefore can not turn on the diode. i removed that resistor from the circuit, and heard no change in the sound.

i worked around this by playing with the time constants of the circuit to get the transistors to slowly turn off, rather than go directly from one state to another. this gives a similar effect, but isnt quite the same.
openmusiclabs.com

User avatar
Mungo
Super Deluxe Wiggler
Posts: 1256
Joined: Fri Jan 04, 2013 8:12 pm

Re: rhythm wolf teardown: any questions?

Post by Mungo » Fri Apr 03, 2020 8:11 pm

Predict the current flowing through D52, it has to be less than 1uA, likely much less. To get audible/useful pitch modulation through the diode you need many orders of magnitude more current flowing (experiment by making the pitch modulation just with a diode and variable resistor alone as in the 606 mods I analysed). The audible pitch modulation in the 808 is current being passed by Q43.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Fri Apr 03, 2020 8:25 pm

are you the author of the paper? im a bit confused as you mentioned being referenced in it previously, but maybe that i misinterpreted that statement and you meant that the paper itself was being referenced in this thread. if so, its great work, but i just dont see how Q43 could possibly be in conduction with its base pulled to ground. in contrast, when the collector of Q42 goes from +12V to ground, so does the charge on C39. This places 12V across R161, which would cause 12uA to flow, which is a sizeable fraction of the current through R166,165. This slowly decays to 0uA over ~150ms. but, the diode then turns on, with ~1uA flowing for every volt at that juncture. this can be evaluated as a percentage of the current flow through R166,165 by doing a ratio of the resistors (50k/1M = 5%). 6% is the next note up in a chromatic scale, and therefore very perceptible, and not so coinciedentally, exactly the same level of pitch shift shown in figure 11.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Fri Apr 03, 2020 8:29 pm

also, i finished the hihats rework. apologies for the wall of text, i just finished writing this up for the wiki and figured i would just copy and paste:

i tried a number of different mods, and ended up using a small set of them. The ones i kept were as follows:

1. Minimum decay time - change R208 to 22k and R202 to 4.7k. This is very much to your taste, so i would suggest trying a few different values to see how they sound.

2. Tune knob rework - i eliminated the analog noise, as it didnt sound as good as the digital noise, and repurposed the tune knob to tune the digital noise. This is accomplished by varying the voltage to the 74AHC14 IC that creates the digital noise. To do this, remove Q72 and replace with an NPN transistor. The trace the 3.3V rail does not need to be cut if that pad is no longer used. Instead, run a trace to the +12V supply. i connected up at the comparator on the backside of the board, but any +12V point would do. Remove R363 and place a 1uF capacitor between the base of the transistor and digital ground. Digital ground can be found closest at C177. If you are not planning on doing pitch bending, i would suggest a 10uF capacitor instead, as it will be more stable. Run a wire to the wiper of P10.

To rework the pot, cut the wiper trace and wire it to the transistor on the bottom of the board. Remove R118 and replace it with a 10k to +12V. Remove R113 and replace it with a 4.7k to ground. You can use a smaller value for the 4.7k, but it might not work. i installed a 3.9k, and it sounded great for a while, but then the digital noise started to cut out at low pot settings. The 74AHC14 is only rated down to 2V, and a 3.9k supplied 1.5V or so. i'm not sure if it was the low voltage or something else i had done that messed it up. A 3.6k is the lowest that will work at all. This will allow a wide frequency sweep. You can reduce the 10k value to get higher frequencies, but it seemed high enough to me at 10k.

Reconnect the noise source to the VCAs with a 10k resistor from the output of U14A (pin1) to the input of the VCAs (C84,127). This value can be made smaller or larger to give varying amounts of clipping in the VCA output. Larger values will give a softer sound, but at some point will make the volume very quiet. Smaller values will give a harsher tone.

3. Pitch shift - you can make the closed hats a different pitch than the open hats by adjusting R364 and R353. When one of these resistors is engaged, it pulls down the voltage on the 74AHC14, lowering the pitch. So, if you want the closed hat to be a lower pitch, put a resistor in R353 and not in R364. To make the open hat lower pitched, do the reverse. The value of the resistor determines the pitch depth, with values from 10k to 100k being reasonable. A 47k would be a good starting point. If you use a larger value than 1uF for the transistor base capacitor in (1), the pitch shift will have a slew to it. its interesting, but not something i wanted as a fixed option. See discussion below for tweaks to this mod.

4. Open hat closure - to close the open hat when the closed hat triggers, run a 22k resistor from the output of U36B (pin7) to C104. This will discharge the envelope of the open hat whenever the closed hat is on, very similar to how the 606 or 808 works. Smaller values of this resistor will close the hat faster, but will also attenuate the hat more on subsequent retriggers when the closed is still sounding. An alternate version of this involves a transistor and a few resistors off the trigger signal. This latter method is far more abrupt, and mutes the open hat whenever the closed is activated. i didn't like it in comparison to the former version, so i didnt explore it completely. R214 would need to increase in order to keep Q32 from over-powering the pulldown resistor on the mute. In theory, it should be >10x the value of the pulldown resistor.

Other mod options

There is a lot of design space here, and i did not explore it fully. i tried adjusting one of the oscillators on the 74AHC14, as there were two set to 1kHz, but after changing it, it didn't sound as good, so i put it back. The main area that could use some work work, is the settings on the comparators. i removed R386 and R410, which eliminated the hysteresis. This raised the threshold for detecting note off, which worked well in keeping the pitch from always hanging low. But, this also had the effect of causing the pitch to toggle on transitions when used with the closed hat. It would probably be good to raise these levels and add some hysteresis back in. The exact level and amount of hysteresis you want is dependent upon how its used. For pitch shifting purposes, it depends upon which note is being shifted down. I found that triggering off the closed hat made the most sense, as it was the shorter of the two envelopes. If you trigger off the closed hat, and an open hat triggers while a closed hat is playing, the pitch shift will not occur. If the pitch shift turns off halfway through a note, that could be unpleasant as well, so in general the thersholds should be low for pitch shifting. But, for muting, you dont want to hold the mute too long, so a higher threshold would be preferred. If the mute comes off too early, it won't matter as much as the envelope will have probably decayed to a very low level by that point anyways.
openmusiclabs.com

User avatar
Mungo
Super Deluxe Wiggler
Posts: 1256
Joined: Fri Jan 04, 2013 8:12 pm

Re: rhythm wolf teardown: any questions?

Post by Mungo » Fri Apr 03, 2020 9:22 pm

guest wrote:
Fri Apr 03, 2020 8:25 pm
are you the author of the paper? im a bit confused as you mentioned being referenced in it previously, but maybe that i misinterpreted that statement and you meant that the paper itself was being referenced in this thread. if so, its great work, but i just dont see how Q43 could possibly be in conduction with its base pulled to ground. in contrast, when the collector of Q42 goes from +12V to ground, so does the charge on C39. This places 12V across R161, which would cause 12uA to flow, which is a sizeable fraction of the current through R166,165. This slowly decays to 0uA over ~150ms. but, the diode then turns on, with ~1uA flowing for every volt at that juncture. this can be evaluated as a percentage of the current flow through R166,165 by doing a ratio of the resistors (50k/1M = 5%). 6% is the next note up in a chromatic scale, and therefore very perceptible, and not so coinciedentally, exactly the same level of pitch shift shown in figure 11.
My prior work was cited in the DAFx-14 paper.

Q43 is only weakly pulled to ground, consider all the transistors as current controlled current sources. The initial envelope controlled frequency modulation is easy to see/understand but then the output impedance of Q42 and values of C39, R159, R161 all interplay for the "sigh" and period+amplitude based frequency modulation.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Fri Apr 03, 2020 11:07 pm

i wish i had an 808, id like to poke around see what the voltages are. i suppose i can easily enough build up this section and test it. i suppose i could see Q43 being "weakly" off. ill give that some thought. something needs to account for the period based modulation, which means either Q42, Q43 or D52, and D52 is probably reverse biased for the first 150ms, so it might not be able to do the job unless the voltage swing is really high at Vcomm.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Sun Apr 05, 2020 6:48 pm

i built up the circuit, and i dont see any pitch shifting after the first 10ms. i dont have the exact transistors, i used 3906 instead (i had to invert the whole circuit to allow for PNP), but these would have lower beta than the 945P used in the 808, and as such would be more likely to allow for "weakly off". but, Q42 is very much saturated at 100mV. its also very clear that D52 does not come into conduction during the decay phase, nor does R161 create a frequency sweep. the current to Vcomm creates a DC offset that decays, but the impedance seen by Vcomm through R161 does not change, and therefore the pitch can not change.

the only possible mechanism i could see, was that Q43 has some small section of reverse transitor action. this makes it a common base amplifier, so no signal amplification, but it does put R159 in parallel with R165 for negative excursion greater than 0.6V. but, the impedance of the "diode" quickly increases, so this only effects the first few periods. this effect is also larger with a larger accent.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Sun Apr 05, 2020 8:16 pm

ok, im very convinced of the reverse conduction theory now. the voltage on the base of Q43 almost identically matches the pitch shift plot in figure 11, which means that it must be the negative going peaks tha cause the shift. since this is out of phase by 90 degrees to the output, this means the output speeds up on the down slope of the sine wave, and goes relatively slower on its up slopes. the thing i like about this mechanism, is that the pitch shift follows the decay time. the longer you decay, the longer you will pitch shift. unfortunately this can not be done with the rhythm wolf, as it never gets above 200mV at that node, and could not turn on a diode.
openmusiclabs.com

User avatar
Mungo
Super Deluxe Wiggler
Posts: 1256
Joined: Fri Jan 04, 2013 8:12 pm

Re: rhythm wolf teardown: any questions?

Post by Mungo » Sun Apr 05, 2020 11:39 pm

I'm not convinced C-E reverse conduction of Q43 is the dominant process, also there is no voltage modulation visible on the collector of Q42 wrt ground in the physical circuit. It is there in SPICE simulation (that paper only confirms its accuracy against SPICE) but not apparent in a real circuit. This is adding to the limitations of the transistor models.

The frequency enhancement does occur on the positive going edge, but this is when the collector of Q43 is positive (in simulation but not physical). There appears to be interplay between contributors to the impedance on the collector of Q42. Shorting D52 doesn't eliminate the amplitude based frequency modulation, while increasing the current into the base of Q42 does. R161 appears to pass enough current to overwhelm any reverse conduction effects.
Last edited by Mungo on Mon Apr 06, 2020 1:14 am, edited 1 time in total.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Mon Apr 06, 2020 12:10 am

i built up a replica and i see voltage modulation on the base of my replica. again, not certain how closely this matches the real thing. this effect is largest at high accent and long decay time. for short decay and and accent, its only visible on the 1st or 2nd period. ive reached out to a friend who has an 808 to take some measurements for me.

after re-reading the paper, the author seems to imply its C-B [EDIT, said C-E before, meant B-E] breakdown, although the wording is very odd:

"When Vcomm swings low enough (below about one diode drop below ground), the base of Q43 gets lifted up and current flows into the base, causing even more current iC to be drawn in through its collector."

so, the first part makes sense (Vcomm going a diode drop below), but Q43 base gets pulled low (in my replica circuit), not lifted up as a result, and the current in the collector does increase, but it neither flows "into", nor is it due to transistor action as the wording implies.

i hadnt realized that their comparison was only to spice, i sort of assumed "physical" meant an actual circuit, not their physical model. that makes me less confident in their comparison. its hard to say exactly when the frequency modulation occurs on their graph, as they do not overlay the output signal with the frequency modulation.

the collector of Q42 stayed in saturation (100mV) on my build for all time outside of the initial 8ms period. i tried playing around with the base current into Q42, and going larger didnt change anything, but going smaller brought Q42 out of saturation, and left Q43 on, pitching all notes up, with no modulation.
Last edited by guest on Mon Apr 06, 2020 10:37 am, edited 1 time in total.
openmusiclabs.com

User avatar
Mungo
Super Deluxe Wiggler
Posts: 1256
Joined: Fri Jan 04, 2013 8:12 pm

Re: rhythm wolf teardown: any questions?

Post by Mungo » Mon Apr 06, 2020 1:12 am

guest wrote:
Mon Apr 06, 2020 12:10 am
so, the first part makes sense (Vcomm going a diode drop below), but Q43 base gets pulled low (in my replica circuit), not lifted up as a result, and the current in the collector does increase, but it neither flows "into", nor is it due to transistor action as the wording implies.
I double checked the physical circuit, care of the large ranges and tiny currents its getting to places where probing and preamp (overload recovery behaviour) all play into the measurements. You may well be correct that Q43 conducting in reverse is the cause of the frequency modulation on the rising edges. There is a lot of voltage modulation at the base of Q43 but its not following through to the same modulation visible in simulation at a smaller on magnitude on the collector of Q42.

SPICE is quite a way off what you can measure in circuit down at these behaviours.

... edited previous post to reflect the measurements taken.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Re: rhythm wolf teardown: any questions?

Post by guest » Mon Apr 06, 2020 10:48 am

just did a quick edit above. meant C-B and not C-E. its like the 606 bass drum mods you have up on your site, with the 2 back to back diodes, except its just one transistor.

the odd thing i found, which originally had me thinking i was messing something up entirely, was that there wasnt any pitch shifting for the low decay/accent settings (on my build). i measured the time between successive periods of the waveform on the scope, and they were all the same (excepting the initial 10ms). as i increaed the decay, all of a sudden it was there, a measurable difference, going from 56Hz or so down to 50Hz, across the length of the note. im wondering if at short decay times, because their isnt enough time for the brain to totally get a sense of pitch, it "smears" the initial pitch and the later pitch to create a psychoacoustic effect of note drop.
openmusiclabs.com

Post Reply

Return to “Music Tech DIY”