Adafruit #570 in Inc 5000 & #11 in manufacturing - fastest-growing private companies in America - INC 5000 - Read more!

Re: Component C29 is backwards on all TB-303's... (more inside)

by Brassteacher on Wed Dec 08, 2010 10:36 pm

antto wrote:i'm sorry, i get very angry when someone talks about the 303 like he really knows what he's talking about
about 2 years ago when i started gathering information about the 303.. there was so much info on all kinds of websites
and by following the details i figured something is not right.. they were wrong? very wrong!? extremely wrong!? complete bullsh*t!
starting from the wikipedia article (which afterwards got a little fixed now)
there were things like "The oscillator has two waveforms: Sawtooth and Triangle wave" sometimes Sine wave
lots of weird things said about the slide (which actually is very simple)
"the accented notes modify/increase the filter resonance and vca"
"the square wave is produced by inverting every other cycle of the sawtooth" :shock:
i wasn't so familiar with analog gear back then (not that i am very familiar now but..)
i had no real clue how this thing works and i wanted to mimick the sound of it, so i searched and searched and watched videos, listened to recordings..

now i know more about the 303 (and i have a x0xb0x)
but i had to waste my time with all the bullsh*t written everywhere (and i believed it at first)
in fact, there is very little bits of true told about this synth

so now when someone writes stuff about the 303 that sounds like he is the 303 guru and knows what he's talking about but he's just interpreting another crap-article he's read before.. i just get very angry.. that's me :roll:
EDIT: i mean.. one should be responsible for the words he writes on forums or web articles, other people will be reading it eventually, some might follow it


antto, do yourself a favor, stay away from high-end audio and "audiophile" forums! :D

For example: You will run into some of those guys who will SWEAR that a plain piece of wire conducts current better in one direction than another. As in, that cable doesn't sound very good, are you sure you don't have it plugged in backwards? :lol: (They totally forget that audio signals are AC!)

The sad thing is, that's not the craziest stuff I've seen those guys come up with! :shock: Then again, there is the sometimes really crazy sounding stuff that actually DOES make a difference. Of course, I'm a natural skeptic, if I read or hear about some kind of crazy claim, I'll go grab one of whatever they are talking about (assuming I can afford it!) and try it out.

This reminds me, got to make a couple recordings to post in the other thread to go along with the scope screen shots...
Don Taylor, a.k.a. Brassteacher
"Jack of all trades, master of none, proficient at even fewer..."
Brassteacher
 
Posts: 248
Joined: Fri Oct 08, 2010 3:49 am
Location: Kennesaw, Georgia, USA

Re: Component C29 is backwards on all TB-303's... (more inside)

by bcbox on Thu Dec 09, 2010 6:17 pm

It's nice to see people take the ball and run...
I still have the same 303 on the bench.

To clarify, my original A/B'ing was done on follows.
1. I used a toggle switch to switch the direction of the cap realtime.
2. TB-303 powered with Boss PSA-120.
3. I leave the settings in one position and record a couple bars with the switch in each position.
4. Edit the files in my wave editor so it's looping 4 bars, 2 bars at each setting.
5. Set the controls to a new position and repeat.

In my original post I purposely did not say anything about the cause of the difference, as I don't know and didn't want to add my subjective opinion.
--------------------------------------------------------------------------------------------------------------------------
Here's some follow-up testing.

Setup
Calibrated Fluke 45 used to measure voltage across cap.
TB-303 powered with Boss PSA-120.
Recorded direct into Wavelab through Steinberg MR-816x.

Image

Capacitors
I have thousands of original caps from TB-303's and other vintage Roland gear. For these tests I selected three that came from old TB-303's and a new Nichicon UK-series.

Image
Image

Sound Files
Same as last time, in each sample the first 2 bars is C29 as per actual TB-303, and the last 2 bars is C29 as per schematic.
Nichicon UKW1H010
Sanyo 1
Sanyo 2
Sanyo 3

Measurements
Voltage across cap when placed in each direction.

Nichichon UK:
Backwards (per 303): 3.520
Correct (per schematic): 3.541

Sanyo 1:
Backwards (per 303): 3.313
Correct (per schematic): 3.541

Sanyo 2:
Backwards (per 303): 3.150
Correct (per schematic): 3.541

Sanyo 3:
Backwards (per 303): 2.870
Correct (per schematic): 3.541
bcbox
 
Posts: 587
Joined: Thu May 25, 2006 12:30 am

Re: Component C29 is backwards on all TB-303's... (more inside)

by nilsomat on Fri Dec 10, 2010 9:48 am

Very interesting! So the difference is only there with old caps? Is there a way to reproduce that behaviour with new caps also (like lower capacitance or so)?
nilsomat
 
Posts: 10
Joined: Wed Dec 01, 2010 11:38 am

Re: Component C29 is backwards on all TB-303's... (more inside)

by guest on Fri Dec 10, 2010 2:25 pm

something i noted in my tests
that might be of relevance to your a/b testing
is that the capacitors had really long soak in times
and this was longer in the reverse direction

the voltages would continue to drift for at least a minute
after i placed the capacitor in the circuit
which is much longer than the rc time constant
guest
 
Posts: 3155
Joined: Fri Feb 17, 2006 4:35 am

Re: Component C29 is backwards on all TB-303's... (more inside)

by Altitude on Sat Dec 11, 2010 3:45 pm

Wow. Like the 303 needs any more voodoo associated with it.

So I guess the question is how do we duplicate it?

I wonder if this a question of age, part type, the fact it's been in backward for so long, or a combination?

I'm curious if a cap from a different source but of the same age would yield the same results
User avatar
Altitude
 
Posts: 975
Joined: Wed May 11, 2005 4:17 pm

Re: Component C29 is backwards on all TB-303's... (more inside)

by Tim Stinchcombe on Sat Dec 11, 2010 5:38 pm

bcbox wrote:It's nice to see people take the ball and run...
Well, it is a real interesting discovery and hence a really interesting problem, so someone has got to do it!
To clarify, my original A/B'ing was done on follows.
1. I used a toggle switch to switch the direction of the cap realtime.
...
3. I leave the settings in one position and record a couple bars with the switch in each position.
...
Commendable methodology to take any guess work or accidental slip-ups out of the loop!
Measurements
Voltage across cap when placed in each direction.

Nichichon UK:
Backwards (per 303): 3.520
Correct (per schematic): 3.541

Sanyo 1:
Backwards (per 303): 3.313
Correct (per schematic): 3.541

Sanyo 2:
Backwards (per 303): 3.150
Correct (per schematic): 3.541

Sanyo 3:
Backwards (per 303): 2.870
Correct (per schematic): 3.541
Most excellent stuff! The reduced voltages with the cap reversed leaves me with no doubt that the cause of the sonic difference is thus as we suspected - the reversing of the cap is loading down the biasing resistors sufficiently to cause the gain of the differential amplifier to be reduced by significant amounts. The varied readings (probably) also illustrate how hit-and-miss it will be getting any sonic difference from reversing a cap in an x0xb0x, as there appears to be no guarantee that doing so will give the kind of voltage drop required to hear the difference - probably what x0xb0x owners would now like to see is a predictable means of 'accelerated aging' of reversed 1uF caps, or whatever it is that causes the cap to become sufficiently 'leaky' without going pop!

I have attached some more simulation results, showing frequency response plots of the filter output (i.e. no accounting for the VCA or anything else downstream) with the capacitor as normal (blue), and with the simple 'fudge' used in my post above, of adding a resistor in parallel to emulate the capacitor being reversed (red) (the resistor was chosen as 30kohm, which roughly equates to about 3V at C29; all three plots are for the same cut-off setting). Naturally enough the difference is emphasized at the peak of the higher resonance setting, which is why it is easy to hear in the real examples.

Tim
Attachments
c29_norm_v_rev_resp.gif
Simulation output of TB-303 filter at different resonance settings and with C29 normal vs. 'reversed'.
c29_norm_v_rev_resp.gif (19.54 KiB) Viewed 2996 times
User avatar
Tim Stinchcombe
 
Posts: 11
Joined: Wed Dec 01, 2010 5:25 pm
Location: Cheltenham, Glos UK

Re: Component C29 is backwards on all TB-303's... (more inside)

by guest on Sat Dec 11, 2010 6:19 pm

good analysis tim

im getting the feeling that the reverse capacitor
is merely reducing the gain of the feedback loop
and as a result causes the frequency effects you show

if this is true
the same effect could be achieved by turning the resonance down a bit
guest
 
Posts: 3155
Joined: Fri Feb 17, 2006 4:35 am

Re: Component C29 is backwards on all TB-303's... (more inside)

by Tim Stinchcombe on Sat Dec 11, 2010 7:03 pm

guest wrote:im getting the feeling that the reverse capacitor
is merely reducing the gain of the feedback loop
and as a result causes the frequency effects you show

if this is true
the same effect could be achieved by turning the resonance down a bit
The gain affected by the reversing of the cap is in the forward path, whereas the resonance obviously affects the feedback path: simulation is quickly able to show the difference, attached (one of the many reasons I use simulation so much!). Note that reducing the resonance increases the gain in the passband - how noticeable this would be aurally is anybody's guess, but the difference in the curves is quite marked!

Tim
Attachments
c29_rev_vs_less_res.gif
Effect of lowered gain due to reversed cap vs. simply lowering the resonance
c29_rev_vs_less_res.gif (11.37 KiB) Viewed 2986 times
User avatar
Tim Stinchcombe
 
Posts: 11
Joined: Wed Dec 01, 2010 5:25 pm
Location: Cheltenham, Glos UK

Re: Component C29 is backwards on all TB-303's... (more inside)

by textile on Sun Dec 12, 2010 1:18 am

How would a bi-polar EL or a polyester type go?
Can you do a simulation of these types?
The graphs are very good and are much needed!

The question is of course is , why did roland do it.
Could it not just be a batch (large batch) that slipped through unnoticed?

Also is it recommended to reverse the cap. If the TB had it .. should the x0x too? It seems logical for a true clone.
User avatar
textile
 
Posts: 221
Joined: Mon Nov 23, 2009 8:34 pm
Location: Australia

Re: Component C29 is backwards on all TB-303's... (more inside)

by Altitude on Sun Dec 12, 2010 7:24 am

textile wrote:How would a bi-polar EL or a polyester type go?
..

Also is it recommended to reverse the cap. If the TB had it .. should the x0x too? It seems logical for a true clone.



If you listen to Brian's samples a few posts up, a new modern cap makes no difference if it is reversed or not, only an old used one does
User avatar
Altitude
 
Posts: 975
Joined: Wed May 11, 2005 4:17 pm

Re: Component C29 is backwards on all TB-303's... (more inside)

by antto on Sun Dec 12, 2010 7:55 am

just looked at bcbox's samples right now..
bcbox: :shock:
in your sanyo caps recordings.. the square wave has ugly edges i've never seen before on any 303 or x0xb0x
what the hell was that?!
did you use any sort of mixer or effect before the recording device?
We are here too: irc.freenode.net >>> #x0xb0x
..:: c0nb0x v1.00 ::.. http://forums.adafruit.com/viewtopic.php?f=7&t=29282 (new control app - win/osx)
VCO Tuning: http://forums.adafruit.com/viewtopic.php?f=12&t=24759 (do it the EASY way)
User avatar
antto
 
Posts: 1348
Joined: Thu Apr 15, 2010 2:21 pm
Location: 127.0.0.1

Re: Component C29 is backwards on all TB-303's... (more inside)

by guest on Sun Dec 12, 2010 2:46 pm

it makes sense that increasing the feedback
would decrease the amplitude in the passband
but there really shouldnt be much difference
besides for amplitude differences
with a different forward path gain

imagine you had two gain blocks
and putting them in series gave the same gain as you had before
now you merely take the output from the first gain block
you can then reduce the resonance to null out the second gain block
its the same circuit
just tapped at a lower gain point
basically you should be able to get the same frequency characteristics
with everything shifted down by a fixed amount
guest
 
Posts: 3155
Joined: Fri Feb 17, 2006 4:35 am

Re: Component C29 is backwards on all TB-303's... (more inside)

by antto on Sun Dec 12, 2010 4:20 pm

+1
We are here too: irc.freenode.net >>> #x0xb0x
..:: c0nb0x v1.00 ::.. http://forums.adafruit.com/viewtopic.php?f=7&t=29282 (new control app - win/osx)
VCO Tuning: http://forums.adafruit.com/viewtopic.php?f=12&t=24759 (do it the EASY way)
User avatar
antto
 
Posts: 1348
Joined: Thu Apr 15, 2010 2:21 pm
Location: 127.0.0.1

Re: Component C29 is backwards on all TB-303's... (more inside)

by Tim Stinchcombe on Sun Dec 12, 2010 6:16 pm

textile wrote:How would a bi-polar EL or a polyester type go?
Can you do a simulation of these types?
Reversing the electrolytic cap in location C29 is sometimes causing a difference in sound because the cap is polarized, and it doesn't like being used backwards - thus using a non-polarized type like a polyester or non-polarized electrolytic should not show any difference, as electrically it looks the same either way it is inserted.

Generally capacitors used in simulations are ideal, i.e. they consist purely of capacitance, whereas in reality a cap will have an effective resistance both in series and in parallel to it, and if the circuit is likely to be sensitive to such things, then you must explicitly include them in any simulation (and sometimes you need to go to the extent of including lead inductance etc. as well). There are many 'standard' models for caps that include such things, and when we start talking about modelling a reversed electrolytic, they become more complicated. On seeing guest's data, I simply decided that a resistor in parallel would suffice to show the kind of effect we were looking for - i.e. it would 'leak' lots of DC current and hence impact on the biasing. Incidentally, this means that probably the easiest way to 'replicate' this kind of effect in a x0xb0x will be to simply do as I did in the simulation, i.e. rather than reversing the cap physically, simply solder a 30kohm (or so) resistor across C29! (But my suspicions are that many owner's who might want to do this might not see that as mirroring the real thing in a TB-303, and will want to reverse the cap - the resistor method does of course have the advantage that you might not have to wait for several years - or however long it takes, I don't know! - for the capacitor to start becoming as leaky as bcbox's examples!)

The question is of course is , why did roland do it.
Probably a mistake in laying out the PCB, pure and simple, and it has never been detected. If hundreds of units had started showing up with that cap blown, then I dare say they would have spotted the error, but since it seems the units do still function with C29 reversed, no one would have realised that anything was wrong.

Also is it recommended to reverse the cap. If the TB had it .. should the x0x too? It seems logical for a true clone.
That will come down to personal taste, but so far we have little idea (well I have little idea) what it takes to get the cap into a state whereby you could say the difference in reversing it is noticeable (see above)!

Tim
Last edited by Tim Stinchcombe on Sun Dec 12, 2010 7:28 pm, edited 1 time in total.
User avatar
Tim Stinchcombe
 
Posts: 11
Joined: Wed Dec 01, 2010 5:25 pm
Location: Cheltenham, Glos UK

Re: Component C29 is backwards on all TB-303's... (more inside)

by Tim Stinchcombe on Sun Dec 12, 2010 7:16 pm

guest wrote:it makes sense that increasing the feedback
would decrease the amplitude in the passband
but there really shouldnt be much difference
besides for amplitude differences
with a different forward path gain
The simulation traces say otherwise, unless of course I have made some ghastly error (always a possibility!).

imagine you had two gain blocks
and putting them in series gave the same gain as you had before
now you merely take the output from the first gain block
you can then reduce the resonance to null out the second gain block
its the same circuit
just tapped at a lower gain point
basically you should be able to get the same frequency characteristics
with everything shifted down by a fixed amount
The key bit I'm wary of here is "just tapped at a lower gain point" - we are not moving any 'take off' point in the filter by reversing the cap, all we are doing is changing gains, and if you do some simple algebra you'll see the forward gain turns up in both the numerator and denominator, whereas the feedback gain (the resonance) only appears in the denominator, and it is likely this is what causes the difference. I did try and work out a simple example, but it gets too confusing too quickly so I abandoned it. I often find it difficult to rely on 'intuition', as it is easy to overlook some important aspect of the circuit and go astray, whereas the maths gets it right every time - when the going gets tough, resort to algebra! (search for 'obstacle'!). But if you care to 'algebracize' your example, or draw a block diagram, I'm open to being convinced that I've gotten something wrong somewhere!

Tim
User avatar
Tim Stinchcombe
 
Posts: 11
Joined: Wed Dec 01, 2010 5:25 pm
Location: Cheltenham, Glos UK