Using A Second PC As A Standalone MIDI Soft Synth

Some soft synths, like NI's Pro 52 shown here, can be run in stand–alone or VST Instrument modes. However, the stand‑alone version utilises more CPU power and requires more memory.

Modern software allows you to turn a spare PC into a synth or sampler that's perfectly tailored to the needs of your studio. Martin Walker tells you how.

Many musicians are now in the position of owning two PCs, usually because it's not worthwhile attempting to sell an older computer when you upgrade to a new one, and more and more musicians are using their older PC to run software synths and samplers, leaving their main one to run a MIDI + Audio sequencer. Some are even buying or building a second PC specifically with soft synths in mind, or investing in a laptop as well as their main desktop PC.

If you're intending to use two PCs simultaneously you'll need to decide what software to run on each one. Most musicians tend to devote the more powerful machine to their main MIDI + Audio sequencing software and software effect plug‑ins, with the less powerful model effectively used as a MIDI synth for playback purposes only, running a selection of stand‑alone software synths/samplers or a set of VST Instruments and a suitable VST‑compatible host application.

However, if your music uses lots of soft synths you may be better off working the other way round, since these require far more processing power than audio recording and playback, especially if you don't use loads of plug‑in effects on your audio. If you want to process the output of the soft synths, you may also want to install a second set of software plug‑ins in the soft‑synth PC, unless you use a multi‑output soundcard to add rackmount effects, or link the two PCs using digital audio connections so that you can pipe soft synth outputs digitally into the audio PC and add plug‑in effects there.

If you have a laptop‑and‑desktop combination, it may be preferable to run your sequencer on the laptop. Your soft synths can then be run on the desktop PC, which lets you take advantage of a much wider range of soundcards. The chances are that you'll achieve much greater polyphony on the desktop unit too. Some musicians even use a Mac to run their favourite Mac MIDI + Audio sequencer alongside a PC for soft synths, since, at least until recently, there were far more soft synths available in PC format than on the Mac. This is the only way Mac users could, for instance, take advantage of Nemesys' Gigasampler.

Most software synths specify a minimum of a Pentium 266MHz processor or similar and 32Mb of RAM. As I've pointed many times before, this is woefully inadequate when running a MIDI + Audio sequencer as well, but when running one of the many stand‑alone soft synths, or a VST Instrument via a cut‑down host application (see later), a Pentium 266MHz PC may be perfectly sufficient to give you reasonable polyphony.

PC Processor Choices

If you want to run Nemesys' GigaStudio on your second PC you'll need plenty of RAM (notice that this selection of instruments already consumes 78 percent of my 128Mb of RAM for buffering purposes), as well as a suitable GSIF‑compatible soundcard.

If you are lucky enough to be buying a new second PC specifically for soft synths, what feature set should you go for? Most importantly, it should have the most powerful processor you can afford, since this is what determines the maximum number of simultaneous notes and effect plug‑ins that you will be able to run. Both Intel and AMD processors provide good performance at today's high clock speeds, although they can perform differently depending on the type of audio processing being carried out — the Intel range seems to be better at running reverbs, for instance, particularly when compared to AMD's new Duron and Thunderbird models, whereas AMD's Athlon provides excellent performance with frequency‑related plug‑ins such as Emagic's Spectral Gate. Some developers have already recompiled their software to take advantage of the ISSE instruction set of the Pentium III, and claim huge drops in CPU overhead, but very few music applications seem to have been optimised for AMD's 3D Now! Instructions.

The important thing is that your choice of processor will also determine the choice of motherboard chipset, and this is where many musicians run into conflicts with certain soundcards including models from Aardvark, Echo, Guillemot, Lexicon, and Soundscape, mostly due to incompatibilities between the Motorola DSP chips on the soundcard and the motherboard chipset. The situation seemingly changes month by month as new chipsets and soundcards are launched, so it's still safer to stick to Intel processors and Intel chipsets if you want to avoid possible problems. If you really want to buy an AMD processor and support chipset (they're certainly better value for money and in some cases faster), just make sure you double‑check that your chosen soundcard will work before you get your credit card out.

RAM Requirements

Native Instruments' Reaktor is a complex and demanding soft synth, but has the flexibility to run either as a stand‑alone program or as a VST Instrument.

It would be foolhardy to install less than 128Mb of RAM in a soft‑synth PC. Windows really needs 64Mb to itself, but soft synths, soft samplers, and plug‑in effects vary in their requirements, so I made various measurements to gain an overall picture.

Soft synths will always consume significantly more RAM in stand‑alone mode than when running as a VST Instrument inside a host application. For instance, simpler hard‑wired designs like Native Instruments' Pro 52 seem to need about 10Mb when the stand‑alone version is launched, but only about 5Mb when used as a VST Instrument. So, the chances are that if you're running more two or more soft synths in stand‑alone mode you would be better off running them all from a single VST Instrument host if this is an option.

As the flexibility of the design increases, so does both the CPU overhead and RAM consumption. I found that Software Technology's stand‑alone VAZ Modular, for instance, took between 13Mb and 20Mb, depending on the polyphony selected. This variation makes sense, since each note will require buffering in RAM: if there is a user polyphony setting in a soft synth, you will reduce both RAM and CPU requirements by keeping it down. One of the most versatile, and therefore RAM‑hungry designs is NI's Reaktor, which takes between 26Mb and 32Mb of RAM in stand‑alone mode, or 12Mb to 23Mb when launched inside Cubase.

If you're running VST Instruments instead of stand‑alone programs, you will also need RAM for the sequencer — a fully fledged application like Cubase VST 5.0 seems to need around 32Mb, plus another few Megabytes for the song itself — say 35Mb in total. Bear in mind that once you select any VST Instrument it will occupy RAM whether or not its Power button is on in the panel, so make sure you replace any VST Instrument currently unused in your song by 'No VST Instrument' to make the most of your memory. Plug‑ins also use RAM: simple effects like Autopan and Choirus tend to consume only a few hundred kilobytes each, but this rises to several Megabytes for more advanced effects such as reverbs and vocoders.

In general, then, 128Mb of RAM will let you run Windows, a host application like Cubase VST, plus three or four VST Instruments and some simple plug‑in effects. However, if you're also running Reaktor or more than a couple of advanced effects then 192Mb would be preferable. As soon as you run any soft sampler that uses system RAM alongside a sequencer you will require 256Mb of RAM or even more, depending on how many simultaneous sounds you intend to play back. Even Gigasampler and GigaStudio benefit from adding more RAM, despite the fact that they stream their sample data from your hard drive. This is because each of the samples being currently played uses its own buffer to keep the latency down. This buffer size isn't affected by the length of the samples, but stereo sounds will require twice as many buffers as mono ones, and velocity‑layered ones will need a buffer for each sample layer too. Your maximum polyphony may therefore be affected by the amount of RAM installed, depending on which Giga instruments you are playing.

You can now buy 128Mb of RAM for about £60, so there's little reason to cut corners on memory.

Other Spec Issues

Low‑cost multi‑channel digital soundcards such as RME's Hammerfall allow you to transfer separate soft‑synth outputs from your slave PC back into your master PC for processing and recording.

Any reasonably modern hard drive will be more than adequate for most soft‑synth PCs: as long as you've installed sufficient RAM, the hard drive is unlikely to be accessed at all when running the synths, so its performance becomes largely irrelevant. The exception comes when you use a program such as Gigasampler, Steinberg's forthcoming HALion, or the latest update to Emagic's EXS24, since these are almost totally reliant on a large, fast hard drive to stream all their sample data in real time.

The remainder of the spec is largely up to you. There's no point in buying an expensive 3D graphics card — any low‑cost AGP model should do fine. If you're on a tight budget then it's possible to share one monitor screen between the two PCs using a switcher box, but now that you can buy a 15‑inch monitor for around £100 this should be unnecessary. You will definitely need a CD‑ROM drive, if only for the installation of applications, but again with the average drive only costing about £35 this won't break the bank. Some people like to have a CD‑R drive in their second PC, so that they can burn CDs while working on other projects on their main machine, but this approach is reliant on being able to transfer data easily between the two computers (see the Transferring Data box).

Soundcard Driver Options

Fxpansion's RT.FX is a lean and sleek VST Instrument host application that specialises in real‑time playback of VST Instruments and plug‑in effects — perfect for your second PC.

If you want the notes on your soft synths to play back in time with your key‑presses on your master keyboard, it's vital to install a soundcard that provides low latency. However, unlike audio recording and playback, which really require ASIO or EASI drivers for low latency, it's perfectly possible to achieve satisfactorily low latency with soft synths using DirectSound and even MME drivers. This is largely because you're only dealing with the latency between a MIDI input and audio output, rather than audio input to audio output.

As a guide, I often list the latency values I achieve with my Pentium II 450MHz PC and different driver types in my soundcard reviews. Using Native Instruments' Reaktor, for instance, I've managed excellent 'real‑time' settings of 10 to 15mS with most recent soundcards using both DirectSound and MME drivers. Many soundcards now let you tweak the latency at driver level too. By the way, don't forget that drivers tend to be developed as an ongoing process, and an update may enable you to significantly lower your latency value — it pays to visit the soundcard manufacturer's web site occasionally.

If you want to run Gigasampler or GigaStudio then there's another important consideration: GSIF (Gigasampler Interface) compatibility. Both products will run with DirectSound drivers, but doing this will restrict you to a single stereo output and compromise latency to some extent. If your soundcard drivers support GSIF you will automatically get 6mS to 9mS latency and the option of multi‑channel outputs. Such is the popularity of these products that many soundcard manufacturers have added GSIF support to their drivers. Nemesys maintain an up‑to‑date list of tested soundcards on their web site, so it's a good idea to check this before choosing your soundcard.

Audio Output Options

Whether or not you need multiple hardware outputs will depend largely on which soft synths you intend to run. If they are all VST Instruments, or if they are ReWire‑compatible, then you will only need to run a single host application (more on choosing this later), and can still add real‑time effects to individual synths and MIDI sounds using its mixing facilities, without ever needing more than a single stereo output. Even GigaStudio users could also manage with a single stereo output if they are happy using that program's internal NFX effect plug–ins.

Another approach is to buy a card with a single stereo output, but a larger number of virtual outputs. Yamaha's SW1000XG is one example, with six stereo audio outputs in its DSP mixer that are mixed down internally to emerge from its single hardware output. Unfortunately its drivers still don't fully support DirectSound, so MME is the only choice for multiple soft synths, and the lowest possible latency is fixed by the hardware at 23mS. A new contender I've yet to test is the Echo Mia, which is GSIF‑compatible and has eight virtual outputs feeding a single hardware output.

Where you have to run two or more stand‑alone applications simultaneously, you will either need one stereo soundcard for each application, or a soundcard with multiple outputs. The first approach can be easier to implement, since modern Plug and Play soundcards give few conflicts (I've regularly run up to four in my PC). However, a multi‑output soundcard is generally a far cheaper and more elegant solution, as long as you don't run into configuration problems. Some soundcards refuse to let a second application get a look in, while others have driver options to let each output pair be allocated to a different application without conflicts. Never assume that a particular multi‑output soundcard can be used with several applications simultaneously unless this is specifically stated in its spec, or you know of someone else who has already managed it!

MIDI Requirements

If your MIDI interface has multi‑client drivers you could perhaps use the same MIDI input port to play several stand‑alone soft synths in real time, each using a different MIDI channel, but if not you will need to use a utility like Hubi's Loopback. The beauty of running several VST Instruments inside a single VST Instrument host application is that this is all taken care of automatically — you could run up to 16 VST Instruments from a single MIDI port, as long as you don't overdo the automation.

However, sending everything down a single MIDI cable will compromise MIDI timing. Remember that although VST Instruments play back with sample‑accurate timing from their internal time‑stamped track data, when responding to external MIDI data in real time they will not only be subject to MIDI jitter (see PC Musician November 2000 for more details), but also the serial nature of MIDI itself.

The best approach is to have a multi‑port MIDI interface for each computer, and simply connect a set of leads from one to the other. If you also have external MIDI synths then you'll need more ports on your main sequencer PC to connect these. For instance, you might need an eight‑in/eight‑out interface for your sequencer PC, and a four–in/four‑out interface for the soft‑synth one. If you're using GigaStudio, remember that the 96 version supports two MIDI inputs, and the 160 version up to four. If you're stuck for an extra MIDI port, I'm told that you can use the serial‑port drivers available from synth manufacturers like Yamaha on each PC, and connect the two serial ports using a null modem cable, although I haven't tried this myself.

Linking Your System Together

There are a variety of ways to link your two computers together. To simplify the following explanations let's refer to the PC running the MIDI + Audio sequencer as the master, and the one running the soft synths or soft sampler as the slave. If you are solely using the slave as a stand–alone synth or sampler, the only MIDI connection you need make between the two is a cable between the master MIDI Out and the slave MIDI In. You then use an external keyboard to record MIDI performances on the master PC, and the data is forwarded in real time to the slave PC. If you have a MIDI controller box with knobs or sliders you can also record automation data on the master PC, and use the same cable to send it.

What this approach lacks is a way to record real‑time movements of the soft‑synth controls on the master PC. For this you will need a second MIDI cable from the slave MIDI Out to the master MIDI In. Most software synths can generate standard MIDI controller information when their controls are moved, and once this cable is in place, this data can be simply recorded onto your MIDI + Audio sequencer.

On the audio side, the simplest way to combine the audio outputs from two computers is to connect them both to an external hardware mixer — you simply patch both sets of soundcard analogue outputs into different channels of your mixer, where the sounds from other hardware MIDI synths can be added, along with hardware effects such as reverb. For those with no hardware mixer, there is another approach, but it relies on having more advanced soundcards in both machines. The simplest version of this is to use a slave soundcard with an S/PDIF digital output. You mix all your soft synths along with any plug‑in effects you need down to a single stereo output on the slave PC. Then you connect its digital out to the digital in of your master PC, and make sure that the master soundcard is using the external clock signal embedded in the S/PDIF signal. The audio signals on both machines will be locked together to sample accuracy, and you will be able to add your stereo soft‑synth mix to the main mix on the master PC as a live input. Once you're happy with your soft–synth submix you can record it as a stereo audio track on the master PC.

This technique can be taken further if you have two soundcards with ADAT or TDIF digital I/O. Each TDIF or ADAT connector supports up to eight simultaneous tracks, and piping the ADAT output of the slave PC into the ADAT input of the master will let you transfer your soft synth sounds en masse, so that you can add plug‑in effects on the master PC to individual soft–synth sounds coming from the slave. This approach avoids having to install any effect plug–ins on the slave machine, so you can maximise soft–synth polyphony. It also has the added advantage that you can create your final mix entirely on the main screen display of your MIDI + Audio sequencer. Alternatively, you could use an ADAT‑equipped soundcard to connect your soft synths to a hardware digital mixer.

Setting Up

In most cases a soft–synth PC should be set up in exactly the same way as your main music PC, except that you won't need much extra software such as Internet, office, or graphics applications.

However, there is one area where a little tweaking may be in order: buffer size. In general, there's a trade‑off between a high number of simultaneous playback/record tracks from disk, which may require a high buffer size, and a low latency, which demands small buffers. On your audio PC, then, you may want to keep the buffer sizes fairly high to help you achieve the maximum number of disk tracks. On the soft‑synth slave, however, you can keep them as low as possible, to minimise real‑time latency. The hard disk cache settings can also be tweaked in the same way: use a larger setting for Vcache such as 10240 (10Mb) on the audio recording/playback PC, and a smaller one such as 4096 (4Mb) on the soft synth PC where it isn't needed, to make the best use of your RAM.