AOUT redesign

[NorthernLightX]

Hello everybody,

Lately I am busy with how I will connect external filters and stuff to the MB6582. I have a stash of MAX525 chips and want to use them, and these DACs perform slightly better than those of the the AOUT_NG. For my purpose I would need the bipolar option, which by default is not on the Original AOUT board. I began designing an add-on board for the bipolar option, but soon found out that there is no easy solution of interfacing anything to the AOUT board. :wacko:

I also found out that the Original AOUT board could be split. This is useful when interfacing it with MB6582; this way every core gets 4 analog output channels, 2 for each SID, for example for Cutoff and Resonance. You need 1 AOUT module per core, and if that module has the Original amount of 8 channels there is a big chance you are under-using the module, an idea I didn't like. To work around this this a trace-cut was proposed. I only have one AOUT board yet and cutting it up would yield me a board for only 2 Cores, so I would need to have another one made just to cut it up later. :unsure:

All changes summed up I came to the conclusion that it would be probably just as easy to redesign the original AOUT board to include the proposed trace cut and the bipolar option. I did that and even went a little bit further by also including an option to chain the second part to the first with a few jumpers (thereby undoing the cut), so with this module 8 channels for one core is still an option. I also brought all CV and Gate outputs to the edge of the board, the original module had pins all over the place. :frantics:

The only downside is that it is no longer a single sided board. I have not yet tested it, and I even have a few questions before I'm going to make a prototype. First a few pictures:

[edit]11-10-2012: Updated versions of these pictures have been uploaded, the current versions on my server (and pictured here) are the most recent revisions[/edit]

I have also uploaded the eagle board file and the eagle schematic. :flowers:

Any and all comments are welcome, especially if you see something that could be wrong. :whistle:

I also need to know the resistor values of R_x and R_y with 500 Ohm trimpots. The original AOUT PDF suggests values of 10K and 2.2K, but post by Seppoman points out that these values could be wrong when running in bipolar mode. I know a little about Ohm's law, but not enough that I'm confident I'll come up with good values. Help is appreciated a lot! :thumbsup:

Cheers, Alex.

Edited October 11, 2012 by NorthernLightX

[Shuriken]

Couple of comments from me:

• Looks good
  
• Power connectors seem to be a bit bulky for so little current (why not use regular white pcb connectors?)
  
• I am pretty sure Seppoman's comment only applies to AOUT_NG (See this from TK)
  

As for the resistor values you want to end up with a range of -5/5V. To do this you need a value of 1,44140625 (5V /2,048 -1). With R_x at 3,9K, R_y at 2,2k and R_p 500, you'll end up with 5,006. Which has an overshoot...So you use R_p value of 510.

Edited October 9, 2012 by Shuriken

[ilmenator]

Very nice indeed - I am in a similar situation and would like to make use of my 525s, so I like your approach very much :-). I would probably chain the power connector, so that there is an in and an out connector on the PCB. A simple SIL header would probably do fine. And for the CV outs, it might be nice to have an additional DIL connector? Not sure if this would be feasible though...

[NorthernLightX]

Power connectors seem to be a bit bulky for so little current (why not use regular white pcb connectors?)

I wanted to use a molex connector (the kind you find on IDE harddisks) on the PCB, as I'm going to use those for more modules in my synth. I guess adding a 4 pin connector the size of a SIL pin header is not too hard if there is demand.

I am pretty sure Seppoman's comment only applies to AOUT_NG (See this from TK)

Seppoman's post was about the addition of the bipolar option influencing the resistor behaviour of R_x and R-y, also on the Original AOUT, hence my question. If the AOUT is not influenced by this, that's an answer to my question. Are you really sure about this or is it still an assumption?

As for the resistor values you want to end up with a range of -5/5V. To do this you need a value of 1,44140625 (5V /2,048 -1). With R_x at 3,9K, R_y at 2,2k and R_p 500, you'll end up with 5,006. Which has an overshoot...So you use R_p value of 510.

Are these ideal values for the bipolar option? R_p is a 500 Ohm trimmer, what trimming range will your suggested values provide? Please excuse my ignorance... :whistle:

I would probably chain the power connector, so that there is an in and an out connector on the PCB. A simple SIL header would probably do fine.

I will be making a power bus board for my synth, so daisy chaining power is not important for me. Where would you chain the power to? Is a bus board perhaps a better solution for you as well?

And for the CV outs, it might be nice to have an additional DIL connector? Not sure if this would be feasible though...

It was very hard to route the CV traces the way they are now, and the dimensions of the current board are the maximum that free Eagle will allow. I don't think a DIL header for the CV outs is possible. I will use jacks on the frontpanel to patch the CVs to their intended destination, so 2-pin connectors are fine for me.

More general questions:

Now, I still do not understand why some modules need bipolar and others don't, or some modules even use both at the same time for different functions (Seppomans SSM filter for example). Will a bipolar CV work with all CV devices, or are there devices that will only work with unipolar CV? I mean CV devices in general (for example Moog modules), exceptions can of course exist.

I don't mind adapting the CV inputs of additional modules to comply with one standard, I thought the -5 / +5v to be a good standard to use, or am I incorrect? Additional reading about CV voltages and standards is much appreciated. :flowers:

[Shuriken]

Seppoman's post was about the addition of the bipolar option influencing the resistor behaviour of R_x and R-y, also on the Original AOUT, hence my question. If the AOUT is not influenced by this, that's an answer to my question. Are you really sure about this or is it still an assumption?

Well i quote:

BTW to all SSM2044 builders: The module is designed with this (wrong) bipolar range, so you don't need to change anything on your AOUT_NGs. I guess for proper +/- 5V operation I'll have to try out a few other resistor values on the SSM module, but this will only apply if you if you're using a different CV source than an AOUT_NG.

So he's specifically talking about the AOUT_NG, which makes sense as Seppoman (T.Stoeckl) designed it. The older AOUT (MAX525 based) design comes from TK.

Are these ideal values for the bipolar option? R_p is a 500 Ohm trimmer, what trimming range will your suggested values provide? Please excuse my ignorance... :whistle:

:pinch: Overlooked the fact that it is a trimmer. And you are right, 510 Ohm is not a valid value for a trimmer. This should be correct then: R_x at 3,9K, R_y at 2,7k and R_p 500

You would then have a adjustable range from 4,544 - 5,006V or use R_x at 4,1k then range would be 4,67 - 5,15. Last one is probably better.

Edited October 9, 2012 by Shuriken

[NorthernLightX]

Well i quote:

BTW to all SSM2044 builders: The module is designed with this (wrong) bipolar range, so you don't need to change anything on your AOUT_NGs. I guess for proper +/- 5V operation I'll have to try out a few other resistor values on the SSM module, but this will only apply if you if you're using a different CV source than an AOUT_NG.

So he's specifically talking about the AOUT_NG, which makes sense as Seppoman (T.Stoeckl) designed it. The older AOUT (MAX525 based) design comes from TK.

I know Seppoman's real name; I have communicated with him before. He does not reply to PMs anymore though, so I guess he's too busy with more important stuff. I guess he got swamped with questions about his filter by n00bs like me.

I was referring to this part of his post:

And BTW2: I've just had a short look at the old/original AOUT, and the bipolar mod is the same concept like what I'm doing on the NG (guess that's where I got it from ;)), so this module will probably show the same behaviour.

Apparently the resistors of the bipolar mod half the voltage coming out of the MAX525, that needs to be taken into account. Hence my question. The calculations you are doing are explained in the Original AOUT PDF, I can do those myself (but thanks anyway :flowers:). I'm going to look into the bipolar mod now and see if I can come up with a similar formula. Would it perhaps suffice to just half the reference voltage of 2.048V in the calculation?

:pinch: Overlooked the fact that it is a trimmer. And you are right, 510 Ohm is not a valid value for a trimmer. This should be correct then: R_x at 3,9K, R_y at 2,7k and R_p 500

You would then have a adjustable range from 4,544 - 5,006V or use R_x at 4,1k then range would be 4,67 - 5,15. Last one is probably better.

For the standard unipolar mode you are right (but the standard suggested values give a better spread), for bipolar I'm still not convinced you're not overlooking something :blush:

[NorthernLightX]

More info on Control Voltages (this time from the Midibox CV page)

• 1V/Octave: the maximum voltage should be 10.67V. This value is derived from the available number of notes (128): 10.67V / 128 = 0.0834V, an octave consists of 12 semitones -> 12*0.0834V = 1.00V
  MBHP_AOUT_LC users have to calibrate the offset first: select "Min.", and adjust the offset trimpot until the CV out reaches 0.00V
  Thereafter select "Max." and adjust the gain trimpot until the voltage reaches ca. 10.67V
  Now doublecheck the gain: select 1.00V, 2.00V, 4.00V and 8.00V and adjust the gain until the selected values are reached as good as possible.
  
  
• Hz/V: the maximum voltage should be 10.24V. This value has been choosen to simplify the calculation of the required voltage levels for each note. With 10.24V and 12bit resolution, each DAC step can increase the voltage by 0.005V.
  Important: first go into the "Output Curve" page and select "Hz/V", otherwise you will measure the wrong results. This curve has to be selected for each CV separately which should use this characteristic.
  Thereafter go into the calibration menu, adjust the offset (if required), select the maximum voltage, adjust the gain until 10.24V is reached, then try 1.00V, 2.00V, 4.00V and 8.00V
  
  
• Bipolar voltages (e.g. -5V/+5V): some synths require a bipolar voltage to control sound parameters like Finetune, ADSR, CutOff/Resonance, etc.
  Users of a MBHP_AOUT module need to add the Balanced CV extension to each output in order to get the possibility for adjusting the output voltage below 0V
  An offset trimpot is already available on the AOUT_LC and AOUT_NGmodule.
  In order to calibrate the balanced CV, first select "Middle" and adjust the offset trimpot until 0.00V is reached. Thereafter select "Min.-" and "Max." and change the levels there by adjusting the gain trimpot. Check the "Middle" voltage again and iterate, until you've found perfectly matching settings.
  

This does still not answer my question of what control voltages are used most of the time, but I have often read about 1v/oct so I'm assuming this is more or less the standard. How many (3rd party) modules actually use a bipolar CV? (I'll investigate this further myself, but any info is welcome).

[NorthernLightX]

Some info from a post on MuffWiggler (great forum about modular synths, I've seen some familiar names from the Midibox forum over there as well)

CV signals vary depending on the format of the modular, and the function of the specific module. For example, in Euro modules it is generally safe to plug any signal in the range of +5 to -5 into any input (there are a few exceptions, on poorly designed modules). The CV input, while accepting +/-5 might react to the full range of that signal, or in some cases might react only to +/-2.5 volts. Or in some cases might react only to 0 to +5 volts.

So in the end there is no real standard in CV land.

The next thing I need to ask myself is if I want to go through the hassle of adapting Seppoman's SSM2044 filter for proper CV response (per default it responds to +/- 2,8V (the design flaw mentioned earlier) for frequency and +0/+10,67V for resonance, and it lacks the control summer and manual controls I would like to have). An other option would be to go for a different design that used the SSM2044 chips, but I have a feeling that Seppoman's design is very nice in a fashion; for example the CV controllable resonance.

To adapt the inputs of Seppoman's SSM filter to both respond to +/- 5V external analog circuitry could be the answer. I suspect for the Frequency the +/- 5V could be tuned down to +/- 2,8V easily (voltage devider?). I don't know at all if converting the +/- 5V to +0/+10V is even possible, is it maybe possible to "bias" the CV with a +5V source from the PSU? The control summer and manual control are no more than a formality; the schematic is in the datasheet and manual controls could be added with a potentiometer that has +5v on one side and -5V on the other.

Disclaimer: most of the terminology I use I have only read or heard elsewhere, I do not fully understand the principles behind them (I should have paid more attention in school :sweat:). The voltage devider and bias for example are terms of which I only vaguely know the function.

I do think however that more people could profit from this information once it is worked out, and with Midibox CV V2.0 under development proper interfacing with the modular synth world is something to consider.

[Shuriken]

If you look hard, you may even find a post from me on Muff's :whistle:

I came to the same conclusion :wink: It has been an educational day :thumbsup:

Which other stuff did you find which uses bipolar CV? I didn't stumble across much.

Also Google: Full Wave Rectifier

Edited October 9, 2012 by Shuriken

[NorthernLightX]

Another addition today:

1V/Octave Unipolar is the Roland standard

1V/Octave Bipolar is the Moog standard

Hz/V is the Korg standard

This clears up things for me a lot. I don't need/want interoperability with Roland gear, Moog is enough for me. If someone wants to use the Unipolar mode there are already 3 other options available so I will not add an option to switch channels to unipolar mode to the proposed board.

Next up is investigating if Seppoman's module can be enhanced by external circuitry to adapt it to proper +/- 5V operation for both channels, but that is another topic, this one was about the proposed AOUT changes.

Any other comments about the proposed modules is still welcome!

[NorthernLightX]

Iploaded new versions of all files; added a 4 pin header for power. Shuriken can use it to mount a standard 4 pin connector, ilmenator can use it to chain power :flowers:

Any other comments?

I still need proper resistor values for R_x and R_y in the bipolar mode :whistle:

[Shuriken]

Iploaded new versions of all files; added a 4 pin header for power. Shuriken can use it to mount a standard 4 pin connector, ilmenator can use it to chain power :flowers:

Any other comments?

I still need proper resistor values for R_x and R_y in the bipolar mode :whistle:

Maybe TK can comment on this?

[NorthernLightX]

In the mean time this post on MuffWiggler implies that conversion from bipolar to unipolar and vice versa is possible. Back onto the waves of the interwebz to steal me a schematic, harrrr! :pirate:

[jojjelito]

The rabbit hole goes a bit deeper: The Yamaha CS series also use Hz/V as well as the old monophonic preset boxes Yamaha SY-1 (needs added sockets) and SY-2. Then you have the exotic stuff like Buchla that uses 1.2V/Oct. All of that won't have much bearing to you maybe, but there's weird gear out there for sure... :sorcerer:

However, uni- to bi-polar: Can't you just add or subtract an offset depending on which way you convert? The slope at 1V/Oct should be the same no matter what.

[nebula]

It's nice to see some initiative taken with the original AOUT design. Years ago I bought a board manufactured to the original design (but with plated though holes - nice), and I managed to destroy it by getting overzealous with my mods. I also didn't like that there were some omissions on the original layout. So I would probably purchase one if you get some made, since I already have the SMT shunt and a couple 525's.

Thanks for adding the 4 pin SIP header. I am concerned it is too close to the neighbouring IC though, as an IDC connector might collide with the DIP socket. I suggest you move it closer to the Molex connector, since a builder would probably only populate the board with one style of power connector.

This isn't a really big deal to me, but I would like to take a moment to mention a preference of mine. Others may not care or agree: I'm not a big fan of the way lots of boards (including your proposed layout) has eight 2-pin headers when there are really only 8 unique pins. If it was my design I would either use a 9-pin (or 10 if that is unavailable), or more likely an 8x2 header with the 8 unique pins together, and the other side just ground. The reason is that it makes for a neater end project - the panel-mounted jacks could be wired to a single ribbon cable and connector, instead of having 8 dangling connectors. It also is a little less expensive to buy one big connector instead of 8 little ones.

[NorthernLightX]

However, uni- to bi-polar: Can't you just add or subtract an offset depending on which way you convert? The slope at 1V/Oct should be the same no matter what.

Yes, I found that a "Positive voltage clamp circuit using an opamp" is what we want; this will act as a voltage doubler and force the signal at or above 0v. The opamp part is because is introduces percision over the standard diode clamp, which suffers from the voltage drop of the diode.

Again a bit of theoretical knowledge without knowing the how to apply it in practice. Wikipedia has an article about it but I still don't know which value of components to choose :pinch:

Link to the article: http://en.wikipedia.org/wiki/Clamper_(electronics)

Thanks for adding the 4 pin SIP header. I am concerned it is too close to the neighbouring IC though

I will move it a bit just for you :flowers:

I would either use a 9-pin (or 10 if that is unavailable), or more likely an 8x2 header with the 8 unique pins together

OK, you're the second one to ask for it, I will see if it's possible. A quick look at the traces tells me that for 6 of the 8 traces it's not a big problem, but the devil is in the details (or in the last traces, so to say)

Will try this today and upload my results.

Edited October 10, 2012 by NorthernLightX

[jojjelito]

Yes, I found that a "Positive voltage clamp circuit using an opamp" is what we want; this will act as a voltage doubler and force the signal at or above 0v. The opamp part is because is introduces percision over the standard diode clamp, which suffers from the voltage drop of the diode.

Again a bit of theoretical knowledge without knowing the how to apply it in practice. Wikipedia has an article about it but I still don't know which value of components to choose :pinch:

Link to the article: http://en.wikipedia....er_(electronics)

Errm, I think that's over-complicating things. Why not just use an non-inverting mixer where you add or subtract 5V?

You can use a stable precision +-5V reference if you're really picky, but that's overkill IMHO.

Something like:

CV In->100k (R3)

5V offset->100k (R4)

R2=100k

100k (R1) plus 22pf cap across (in parallel) to prevent oscillation in the feedback connection

v_o =v₁ + v₂ (for all resistors equal)

v_o = (R₁+R₂)/R₂ (v₁ R₄ + v₂R₃)/ (R₃+R₄)

This makes -5V become 0V, +5V becomes +10.

By adding a -5V you can go the other way.

Or, you can use an inverting mixer plus a unity gain inverting op-amp to make this less impedance sensitive at the cost of more op-amps used.

This makes the circuit look like this:

followed by a standard inverting op amp.

[NorthernLightX]

Errm, I think that's over-complicating things.

Don't get me wrong; I'm very grateful that you provide input, but according to the stuff I read yesterday the voltage clamp is not complicated at all, and it uses less parts than your suggestion :wink:

Look at this picture: http://en.wikipedia.org/w/index.php?title=File:Precision_op-amp_clamp.svg&page=1

I see one opamp, two resistors, a diode and a capacitor. Or am I missing something?

[jojjelito]

Don't get me wrong; I'm very grateful that you provide input, but according to the stuff I read yesterday the voltage clamp is not complicated at all, and it uses less parts than your suggestion :wink:

Look at this picture: http://en.wikipedia....lamp.svg&page=1

I see one opamp, two resistors, a diode and a capacitor. Or am I missing something?

No, the less parts (marginal, it's only a couple of resistors) is fine. Less parts is always nice.

But, please run a Spice or similar simulation of this (iCircuit or...) and you'll see the effect of the RC constant. Also, I don't think the clamp is OK with slowly changing or more or less constant DC voltages due to the influence of that RC element. It introduces a frequency dependency that might affect any FM or lag or constant voltage CV. The mixer won't exhibit those effects. You might even want to breadboard those circuits for checking.

For the mixer you only need to worry about op-amp offset (a small problem) plus the bandwidth product of the op-amp.

Have fun!

Edited October 10, 2012 by jojjelito

[NorthernLightX]

Ah, now it's starting to make sense; the clamp thing is used to DC offset an AC voltage, we want to offset a voltage that's already DC!

OK, I'm going to quote a few lines and ask more questions if you don't mind :angel:

Errm, I think that's over-complicating things. Why not just use an non-inverting mixer where you add or subtract 5V?

You can use a stable precision +-5V reference if you're really picky, but that's overkill IMHO.

No a precision 5v is not necessary, the standard 5v of the PSU will do fine. I was only concerned about the diode voltage drop mentioned earlier.

Something like:

CV In->100k (R3)

5V offset->100k (R4)

R2=100k

100k (R1) plus 22pf cap across (in parallel) to prevent oscillation in the feedback connection

v_o =v₁ + v₂ (for all resistors equal)

v_o = (R₁+R₂)/R₂ (v₁ R₄ + v₂R₃)/ (R₃+R₄)

This makes -5V become 0V, +5V becomes +10.

By adding a -5V you can go the other way.

Looks nice and simple.

Or, you can use an inverting mixer plus a unity gain inverting op-amp to make this less impedance sensitive at the cost of more op-amps used.

This makes the circuit look like this:

followed by a standard inverting op amp.

Is impedance an issue in a modular synth? I don't mind using another opamp, a dual opamp is still only a DIP8 package.

By the way the inverting mixer looks a lot like the "control summer" for the frequency CV from the SSM2044 datasheet, only here there is no extra inverting opamp to change the signal back to its original phase in that design?

For the mixer you only need to worry about op-amp offset (a small problem) plus the badwidth product of the op-amp.

Could you perhaps elaborate a bit on the opamp offset and the bandwidth product you mentioned? How much influence is expected?

[jojjelito]

Basically the op-amp offset is the error voltage when the input voltage difference is zero you expect the output to be zero. That is not the case as the transistors inside the op-amp aren't perfectly balanced. Here's an explanation. You can either use external trimming, or the op-amp is trimmed at the factory. There are some low-offset op-amps, for instance it might be a good idea to use TL082/084 for CVs instead of 072/074 (which are generally better for audio) since the 08x has lower offset. Anyway, the error is usally very small but one should be aware that it exists and that there are ways to mitigate it if necessary.

The gain-bandwidth product is basically telling you something like x Mhz from the datasheet. For the TL08x it's min 2.5MHz, typ 4.0 Mhz. It's typically 3.0MHz for TL074. The 074 is less noisy though so it's usally preferred for audio. OK, what does this mean?

Let's say you need to make sure that an AC signal of 100kHz can be amplified without too much distortion. In this case you can amplify up to 25-40x before you lose signal integrity. More for a signal of lower bandwidth.

In the example with the CV mixer we have a gain of 1. A signal of 2.4 to 4.0MHz will pass trough. Good enough for audio, less so for analog SD video.

All in all, I would take a look at the offset (which is likely to be small), I don't need to worry about the gain-bandwidth in this application since we don't amplify and don't have a high-frequency signal to worry about.

[Shuriken]

As i said, google full wave rectifier :wink:

Here is an schematic:

Not my own idea, but from a Utility Module

Comes pretty close to

Edited October 10, 2012 by Shuriken

[NorthernLightX]

full wave rectifier

OK, I was skipping it before because normally a full wave rectifier (the one using 4 diodes) just converts AC to DC (*1,4) and that's not what I was after. I'll look into the schematic of the tellun module - tellun designed a lot of good stuff :cool:

[edit]

OK, the full wave rectifier uses diodes as well. The schematic snippet you posted uses already half-wave-rectified signals (the HWR takes place in a previous stage of the module).

The Tellun Dunsel schematic however does contain exactly the thing I was after: the A+5 schematic. I just need to replace +15V through 300K Ohm with +5V through 100K Ohm for a +5V supply (right?), and place the part where A gets inverted to A- before it bacause it expects an inverted signal.

...man I'm really beginning to understand this shiz. :D

Edited October 10, 2012 by NorthernLightX

[jojjelito]

Cool! You're getting it! Congrats

A+5 going one way, A-5 going the other. Simple

[NorthernLightX]

As I was reading the schematic of the Tellun Dunsel something else came to mind: Maybe the -5V level shifter is a better bipolar option than the way it is done now.

Currently the DAC output is biassed before it hits the opamp. This causes resistor values and calibration to be off when someone wants to switch between unipolar and bipolar. What would happen if we would just let the AOUT output only unipolar signals, and bias them using a -5V shifter afterwards (on modules that need a bipolar input)? To me this sounds better than making an unipolar signal bipolar and then shifting it back to be usable.

Again: am I overlooking something?