Great news! I now have a fully working nw2s::io balanced prototype running, fully powered. The signal PCBs arrived last week and the power PCBs this week. They tested out great. There is a slight tweak I’m making to the gain staging, but other than that – perfect! I am now ordering production boards and expect the turnaround time to be about three weeks.
Since so many of you did order the balanced version before they were available, I thought I’d spend a little time to explain the process I went through (i.e. why I’m about a month behind where I expected to be) and I also wanted to chat a little bit about power supply rails and what effect that has on the gain staging and signal levels (i.e. why are we going through this trouble).
So my first attempt at a PCB layout for this thing was a bit optimistic. I decided to try to fit the entire circuit onto two PCBs – one for signal and power and the other for the connectivity. As you know, this thing fits 16 channels of IO into a pretty small 10HP. I was working with one of the other Euro manufacturers during this process and they were a big help, but it turns out that it was just not going to fit on a single PCB 4.35″ x 1.8″ – Technically, yes, it fit, but the PCB was a 4 layer board, components were mounted on both sides, and a few of them were a lot smaller than they really should have been… which all translates into a difficult (read: expensive) manufacturing process.
We did build these, but in addition to just not being comfortable, there were a few other modifications I needed to make to the circuit which I began making right before the holidays.
I decided to separate the power components from the signal components. This would give everything a little more room, allow me to use two two-layer boards with single-sided components. It gave me a little more freedom for using through-hole components when I felt like there was a good reason. It would also reduce the risk of noise with everything so close.
By the time I finished this layout, towards the end of the year, I was ready to start production, but unfortunately, NAMM was nearing and the folks I had been working with were totally slammed. It was going to be February before we’d be able to get anything started. I didn’t want to make anyone wait any longer than necessary, so I found an agreeable assembly house here in the US and had them start on the second prototypes.
Happy to say those are now here and the final architecture is settled. The signal boards will have all SMT assembly done by an assembly house (for now, 100% US made), the power boards will be completely built here along with the panel board, cabling, final assembly and testing.
Why bother with the power?
Some questions I get asked a lot are “What kind of power supply does this thing need?” “Will my Modularator 5000 supply work?” “Why do we even need an onboard power supply?” “Why can’t you just do like all of the other IO boards and use +/- 12V rails?”
To answer that question, you have to understand a couple of things. First, you have to understand that we are using application-specific high-performance input and output devices manufactured by THAT which are engineered to provide peak performance at +/- 18V. While a balancing circuit could be built using op-amps and resistors, the benefit of the THAT devices is that they contain a high-quality audio path as well as laser-trimmed resistors that will provide extremely accurate balancing.
Then, you need to understand why a hotter power rail will result in higher headroom at this most-important stage of the recording or interfacing process.
A typical configuraton, running at these power rails, will provide a maximum balanced line output level of about 24dBu which, coincedentally, is about the same level signal most studio gear begins to clip.
To be clear, a 24dBu is 34V p-p – how can an 18V rail produce a 34V signal? When we’re talking about a 24dBu balanced signal, we are not describing a voltage level referenced to ground like unbalanced signals. A balanced signal is measured against an inverted copy of itself. Sounds a bit like pulling yourself up by your bootstraps, but by doing this, you remove ground from the equation completely. So the 34V p-p signal is really +17V and -17V (relative to the ground signal of the source). The ground signal of the destination doesn’t really matter – it just takes the difference between the hot and cold as the signal level.
Now we know the peak levels that the output can operate at when using 18V rails. If we were using 12V rails, the overhead would be significantly reduced such that the peak levels at which our device is operating would be very near the nominal levels that is our target output.
The distiction between peak levels and nominal levels is one that needs to be made clear.
Modular signal levels, and eurorack specifically, are generally a lot hotter than most other signals. This is by design as most of the analog circuitry is maxed out such that the op amps are operating in a mode where their nominal level is basically the same as the peak level. This range of +/- 8-10V is very near the power rails of +/- 12V. In other words, there is very little headroom – or none at all.
This is perfectly acceptable when you are generating a single sine or triangle tone, but less ideal when you are working with program material that is, for the most part, much more dynamic. With dynamic material, audio gear will have a nominal level (typically +4dBu) and a peak level (typically +20 to +24dBu) – this gives you about 20dB of headroom to play with.
There are a few good explanations on the MW forums establishing a 9dB difference between nominal eurorack levels and nominal studio levels. To meet various design goals, I settled on a 6dB difference. That is to say that when transitioning from modular level to studio level, I would decrease the signal by 6dB (cutting the voltage in half) and when transitioning between studio and modular level, I would add 6dB gain (double the voltage).
Here’s some of my justification of that…
Let’s assume that 8 – 10V p-p (11 – 13dBu) is upper end of what is a typical modular audio signal will be. If we want 8V p-p to translate to about +4dBu, then we would need 7dB gain reduction. Don’t forget that the 1606 adds 6dB gain, so we have to decrease by about 13dB total.
Let’s round that down to 12dB to make the math work out a little better (12dB attenuation can be created using voltage divider with a pair of R and 3*R resistances). Since the nominal input impedence of the 1606 is 10kΩ, we can create a 12dB pad easily by simply putting a 30kΩ resistor in series with the signal.
The our nominal output level ends up being about 5dBu which is just about right.
Why not have adjustable gain?
Simply put, we don’t need it. There’s only two reasons you’d want adjustable gain – the first is that your signals are too hot and need to be attenuated and the second is that your signals are too low and need some gain added.
Let’s take the case where your signal is too hot. This would happen if you are generating a signal that will clip the ‘io gainstage, downstream converters, or would otherwise be too hot for some downstream gear. Assume that the hottest signal that your modular can generate is 24V p-p (+/- 12V). That translates to about 21dBu. Accounting for the 6dB gain reduction of the ‘io, that’s about 15dBu (12V p-p). 15dBu is well below the io’s peak levels and is well below the peak levels of studio gear. In effect, it will be impossible to clip any downstream devices that operate at standard studio levels.
What about quiet signals that need to be amplified? Generally modular levels are pretty hot. If you wanted to amplify something, it would probably be way down around 1V p-p. This can happen in studio gear all the time… reverb tails, soft singers, ukuleles recoded with ribbon mikes. But it’s not a common occurence with modulars. The few times it might happen (as is the point with modulars), I’m sure you have a VCA that will be amplify the signal somewhere in that big box!
That covers the output stage… now let’s find out more about the studio to modular stage.
Because we are decreasing gain, going from modular to studio, we are increasing the headroom and have a little bit of room to work. Going the other direction, we need to increase the gain and are therefore losing headroom, so have to be a little more careful. This also becomes an area where having the higher voltage power rails helps handle the increase in signal level.
First, we have to figure out what the gain difference is. Earlier, we talked about how the established gain difference is about 9dB, but that for various reasons, I settled on a 6dB difference. So we know we need to increase the gain by 6dB coming from the studio into the modular.
Then, we have to figure out how we can increase the gain. One option would be a set of audio op amps, however, THAT comes to the rescue again. Their balanced line receivers come in three models. 0dB, -3dB, and -6dB. Luckily, they configured the inputs in such a way, that by simply rewiring the input, you can change the gain structure to 0dB, +3dB and +6dB. So we’ll be able to increase the gain by 6dB without adding any further elements to the circuit. That’s handy. And sounds better too.
For most pro-audio situations, THAT recommends the -3dB model so that gear can operate on +/- 18V rails and accept +24dBu signals (If you want to see that math, their datasheet is helpful). Unfortunately, the studio signals are already too low and we absolutely don’t want to lower them even further.
Unfortunately as we increase the gain, we lose headroom, so our modular input will clip about 9dB earlier than the 24dBu that we can probably crank out at peak levels. I should be able to show, however that even a +15dBu signal from your DAW is hotter than anything your modular would normally deal with. Then I’ll show you a few tricks about how to maximize the headroom you do have when working with your modular as a hardware insert to your DAW.
In the case where we have a synth line, or guitar or bass track in our DAW that we want to process with the modular, we should have a signal that is fairly steady and not extremely dynamic. This is the simpler case. If this track is being routed to an AUX send and it’s level is about -18dBFS (the digital level), then the analog signal should be about +4dBu. Routed into the ‘io, it will be increased to +10dBu which will produce a 7V p-p signal. To produce a 10V p-p signal, you can increase the output by another 3dB (-15dBFS) and you’ll have a nice round 10V p-p signal. Of course, if it is a guitar track, it will be a tad bit more dynamic than your typical modular signal, so sometimes it will be a bit higher and sometimes it will be a bit lower. This is normal and perfectly acceptable.
Let’s take the case, however, where you may want to process a vocal track or drum track. These signals are extremely dynamic. You may feel that you’re not getting a very hot signal. This also is normal. In reality, the signal is what it is and if you leave it alone, you’ll be able to run the signal through filters, VCAs, wave folders, etc just like any other signal.
How hot is too hot?
If, however, like me, you’re looking for a little more character out of your analog gear, you are going to want to push it a little harder. There’s two ways to do this. First, you can just turn up the output signal. If you let the output peaks hit, for example 0dBFS, your analog signal peaks will be so hot that you’ll likely clip not only the ‘io circuits, but also some of the poor ol TL072’s in the audio paths of your modules.
This hard clipping may be what you want, but if you’re looking for a bit more subtlety and control, you should increase the RMS value of the signal rather than the peak levels. How do you do this?
The first tool I always reach for is a limiter. It will tame the peaks and allow you to easily increase the average levels by 3 6 or even 9 db without much degratation. (Yes, this is how the loudness war started, but we’re not talking about the same thing exactly). Another useful tool is a compressor. Depending on the source material, experimenting with these tools will give you a wide range of options for your own tone palatte.
I hope this has helped understand what the ‘io is, how it got to be what it is, and what some of the design considerations were while I was working through some of the issues. Mainly I appreciate everyone who has supported me through the process and had the confidence to place an advance order.