You are here

Testing & Improving PC Soundcard Audio Performance

Clear Signals
Published November 2002
By Martin Walker

Rightmark's Audio Analyser provides a simple way to assess various aspects of the audio performance of your PC soundcard and compare it to others, and best of all, it's freeware.Rightmark's Audio Analyser provides a simple way to assess various aspects of the audio performance of your PC soundcard and compare it to others, and best of all, it's freeware.

An hour spent fine-tuning the performance of your soundcard can yield permanent improvements in background noise, distortion levels and frequency response.

Over the last couple of months I've noticed increased interest by SOS readers and forum users in checking the audio quality of their soundcards. As a result, I wrote a reply in August's Q&A explaining how I normally measure the RMS background noise that I quote in my reviews. However, as I said at the time, there's a lot more to soundcard quality than this, so this month, I thought I'd detail some ways for you to test more aspects of your soundcard, to make sure you're getting the best performance from it. After all, your recordings may be compromised by unsuitable settings in the soundcard's mixer utility, or by using the wrong type or wiring of cables. They may also be improved by looking more closely at the earthing of your PC.

Traditionally, all audio equipment has been tested with a combination of external hardware: signal generators to produce suitable test signals, and oscilloscopes to examine the result. However, as you might expect, the advent of personal computers has resulted in quite a few automated systems, still using dedicated hardware as a front end, but with software to analyse the results, and a computer monitor to display them. Among the most popular manufacturers of such test equipment are Audio Precision (www.audioprecision.com), whose systems seem to be used by the majority of soundcard manufacturers. Their site also provides a wealth of useful information on audio quality and testing.

Hardware such as theirs should ideally be an order of magnitude (10:1) better in audio performance than the soundcard it's testing, which explains the cost of such systems. However, you can still get useful results by using your own soundcard to play some suitable test signals, looping its output sockets back to the inputs using a short high-quality loopback cable, and then using suitable software to analyse the results.

Rightmark Audio Analyser

I've tried quite a few software audio test utilities over the years, most of which mimic individual pieces of test gear. Audio Precision's freeware PC LevelCheck utility is one such, and I discussed others in SOS September 2000 as part of my 'Test Card' feature. Others perform simple tasks such as channel identification and record and playback checks.

Commercial loopback test software like HpW Works (www.hpw-works.com) and SpectraLab (www.soundtechnology.com) is excellent for the professional, but tends to be expensive if you'll only use it to perform occasional tests on your soundcard, especially if you want the versions that support 24-bit/96kHz. It also benefits from some specialised knowledge to select the most suitable settings from a wealth of options, and to interpret the results.

I was therefore very pleased recently to find an impressive piece of freeware specifically designed to test your soundcard's performance, which requires very little technical know-how to operate. Although it can generate detailed graphs if required, at its simplest you can simply read off a set of five verdicts ranging Poor to Excellent for each individual test. Rightmark's Audio Analyser (RMAA) was written by Russian developers Alexey Lukin and Maxim Liadov, and is currently at version 3.4 (shown above).

Having connected up suitable loopback cables between the output and input sockets, you first need to ensure that input and output levels are properly calibrated using the Adjust Levels button — once the test signals have started you may need to tweak your input and output sensitivity, soundcard mixer faders and so on, so that input levels are approximately equal to output levels. You can then perform a range of tests that comprise frequency response, noise level, dynamic range, THD+noise, intermodulation distortion and stereo crosstalk. The test results for my Echo Mia running at 24-bit/96kHz made very interesting reading. It had an extremely flat frequency response of +0.03/-0.05dB from 40Hz to 15kHz, and only dropped off to -0.17dB at the wider extremes of 20Hz to 20kHz. Background noise measurements are presented in both flat and A-weighted versions, and the Mia managed an excellent -109.8dBA, while the dynamic range, measured in the presence of a -60dB 1kHz sine wave, is fairly similar at 107.2dBA.

Some of the sound differences between different soundcards can be directly attributed to their frequency response, and as you can see here, this can vary quite a bit. Notice, however, that the vertical scale is quite magnified to show the various ripples and peaks.Some of the sound differences between different soundcards can be directly attributed to their frequency response, and as you can see here, this can vary quite a bit. Notice, however, that the vertical scale is quite magnified to show the various ripples and peaks.THD+noise is measured with an input signal level of -3dBFS (dB relative to digital full scale), and proved very low at 0.0041 percent, while IMD+noise uses two sine-wave tones of 19kHz and 20kHz at -9dBFS, which are subsequently notched out to see what nasties remain (a low 0.002 percent in my case). Finally, stereo crosstalk sees how much signal leaks from one channel to the other at various frequencies, and overall was a very low -103.1dB on my card, although the result will depend to some extent on the quality of your loopback cable — I used a short (one-foot-long) pair of balanced cables made from Van Damme low-noise microphone cable.

Rightmark's Audio Analyser 3.4 is an extremely useful tool, and being freeware is an ideal way for musicians to test their own soundcards and compare results — I intend to include some results in my future soundcard reviews alongside my other measurements and findings. I did have a few application crashes due to 'invalid page' faults, but since the entire suite of tests typically takes only 20 seconds to perform, it's easy enough to reboot and run them again on the off-chance that you do hit problems.

It also doesn't recognise all the supported recording/playback modes of some soundcards — it wouldn't let me test the review model WaMi Rack 192X at either 24-bit or 96kHz, or the ST Audio Media 7.1 at 24-bit. My screenshots show RMAA version 4.0 beta 2, which seems compatible with a wider range of soundcards, as well as providing improved crosstalk measurements, and supports test sample rates up to a higher 192kHz. You can download both versions from http://audio.rightmark.org.

Measurement Tweaks

RMAA test results are already taken seriously enough to be quoted on web sites such as Tom's Hardware Guide (www.tomshardware.com), but it's easy to misinterpret figures without a modicum of experience. I've spotted postings on the Rightmark forums worrying about the 1kHz peaks in the Dynamic range and THD measurements, for instance, but these are in fact the test signals, which are subsequently notched out before measurements are taken of the remaining signal.

Are you getting the best audio performance from your soundcard, or do some utility settings shown here need adjusting to flatten the frequency response, lower background noise and distortion? RMAA can provide confirmation of all this and more.Are you getting the best audio performance from your soundcard, or do some utility settings shown here need adjusting to flatten the frequency response, lower background noise and distortion?It's also possible to get strange results from many consumer cards if unused inputs are left open with the faders up, rather than muted. In some cases this can even give rise to ripples in the frequency response due to feedback. Creative Labs have now posted a set of instructions explaining how to set up their Audigy Mixer and EAX Control Panel to avoid such anomalies, but it demonstrates the dangers of leaping to conclusions!

Using such software can, in fact, help you check the optimum settings for your soundcard's Control Panel utility, as these will of course give you the best measurements in RMAA. Make sure you mute any unused inputs (particularly from RIAA and mic preamps), disable any internal DSP effects if they're not required, and avoid using the internal DSP mixer faders unless you're sure they operate in the analogue domain — most are of the digital variety, and while useful for setting up monitoring mixes, generally result in throwing away digital resolution at any position other than 0dB.

Many soundcards (for example the entire M‑Audio range) let you bypass the internal DSP mixer altogether, and this can sometimes give you a cleaner signal if you don't need to mix sources together. Just run the tests, tweak a setting in your soundcard utility, and try again — you'll soon find out which controls make no difference to audio performance, and which need adjusting to provide the lowest background noise and distortion, flattest frequency response and so on.

Like all loopback test suites, RMAA's results cannot isolate the recording performance or the playback performance alone, and instead report on an amalgam of the two. While this can be extremely useful in comparing one card with another, it can never adequately measure soundcards whose ADCs provide better dynamic range than their DACs, or vice versa. Neither can it be used to check manufacturers' results, which normally measure the ADC and DAC performance independently using external lab gear of a higher standard than the soundcard, to minimise errors.

However, one way round this is to use a higher-quality soundcard as a reference, looping its outputs back to the test card's inputs to measure the ADC, and the test card's outputs to the reference card's inputs to measure the DAC. I've already done this in my review of ST Audio's Media 7.1 card, which displayed a poor frequency response, and using my Echo Mia confirmed that it was only the Media 7.1 DAC side that caused the problem.

Since the loopback cables are part of the signal chain, you can also try patching in different ones to see if they noticeably affect audio quality. Despite some of the wilder claims by the hi-fi press, you're most unlikely to notice any change in either frequency response or distortion levels, but I discovered a significantly worse stereo crosstalk result using an old pair of audio leads that I made up years ago using 75Ω digital co-axial cable, which were about 4dB worse than a similar length of quality audio cable.

Soundcard Wiring

Consumer soundcards generally use 3.5mm stereo 'mini' jack sockets, which although somewhat less robust than their quarter-inch relations, take up significantly less space on the backplate. If you have such a card, you'll need to buy or solder up stereo-to-twin-mono leads to connect to most music gear. Phono (RCA) sockets are also common, both for analogue and co-axial digital use, and tend to be more reliable, especially when used with good-quality plugs. However, once again, you'll probably have to use phono-to-quarter-inch-jack adaptor leads for most duties, although hi-fi equipment does use phonos for most line-level connections.

In a complex setup, using balanced I/O or a DI box can make a huge difference to background mains hum levels, as seen in this example where the 50Hz, 150Hz and 250Hz components drop significantly.In a complex setup, using balanced I/O or a DI box can make a huge difference to background mains hum levels, as seen in this example where the 50Hz, 150Hz and 250Hz components drop significantly.Both of these are examples of unbalanced connectors. I don't intend to explain the virtues of balancing yet again (you can find a good description in Mallory Nicholls' article on Studio Installation in SOS September 2002), but since more upmarket soundcards often feature balanced I/O, this seems the ideal opportunity to discuss some of the perils of ground loops that relate specifically to soundcards, and how using balanced I/O can often avoid them.

Most entry-level soundcards designed for musicians with smaller setups have a couple of unbalanced inputs and outputs, often along with a single digital input and output. Examples include popular models like M-Audio's Audiophile 24/96 and Terratec's EWX 2496. However, once you have four or more input and outputs, the possibilities of increased hum due to ground loops multiply alarmingly. This is the case with such models as Creative Labs' SB Live! and Audigy, Marian's Marc 4, M‑Audio's Delta 410 and Terratec's DMX 6Fire, all of which have a larger selection of unbalanced analogue I/O. The situation is made even worse when surround sound and video enter the equation, especially where cable TV is involved, since yet more ground connections may appear.

In contrast, most eight-in/eight-out soundcards I've reviewed, including Aardvark's Q10, Echo's Mona and Layla, ESI Pro's WaMi Rack 24 and 192X, and M‑Audio's Delta 1010, have inputs and outputs that can be used either unbalanced or balanced, as do some less complex cards such as Aardvark's Direct Pro 2496, Echo's Mia, M‑Audio's Delta 66 and Duo, and the Lynx One and Two. This can make connecting them up to a variety of other gear simultaneously much easier, at least in theory. In practice, those with a mixing desk featuring balanced I/O can connect suitable balanced cables between its ins and outs and those of the soundcard, with a high level of interference suppression and no possibility of ground loops. Those with digital mixers and soundcards with compatible ADAT I/O have a similarly easy time, since two optical cables take care of eight ins and outs simultaneously, again with no possibility of ground loops. However, those connecting up multiple unbalanced sources such as keyboards and synth modules directly to soundcards with balanced I/O will need to make up special pseudo-balanced cables, as described many times before in the pages of SOS (see, for instance, SOS January 2002 Q&A). I've done this for all my synths, and it makes a significant difference to background hum levels.

Many synths have two-wire mains cables without a safety ground, while any case metalwork is double-insulated. If this is the case, one of the cables between the synth and soundcard will need its screen connected at both ends to provide a ground connection (I usually ground the left channel), while the other stereo channel and any additional outputs can have it broken, or connected via a suitable small resistor (a few hundred Ohms is the right ballpark).

PC Earth Purity

Most musicians understand that if two items of gear each have their own earth connections, and you connect a cable between the two, you create a ground loop that generally results in higher levels of hum, but in the case of a computer other perils lie in wait. You may hear continuous background digital noise like a host of tiny lawnmowers, along with regular low-level ticking, plus buzzing whenever graphics are redrawn or the mouse is moved. Even at a very low level, such noises are often more obvious and annoying in a mix than hum, and far more difficult to remove once embedded in a recording.

I recently suffered from this phenomenon when reviewing Terratec's DMX 6Fire soundcard, but completely cured it on my system by bolting its front mounting module into my drive bay using insulating nylon washers. This broke my loop, and probably won't be necessary on the majority of PC setups, but plenty of musicians do suffer from similar problems. Laptop and notebook PCs often suffer more in this respect, since their switching-mode power supplies are often floating and left completely ungrounded — sometimes the noises disappear when you switch to battery operation, or become audible only when you do this!

In general, desktop PCs are earthed via their three-wire mains cable, which ensures that the metalwork of the case is safely grounded, and your monitor screen may also have its own earthed mains cable. These should always be left in place for safety reasons. Most soundcards also have their main earth point on their metal backplate, so it's important to bolt this down tightly to ensure a good connection.

If, like me, your studio's electrical wiring relies on a copper earth spike outside in the garden to ensure good grounding, you should also water the surrounding ground occasionally during hot weather. Tell-tale signs of a poor earth connection include your PC case feeling 'furry' or 'tingly', normally due to the leakage current of the RFI filter capacitors in the PSU, as well as unreliable modem performance.

If you clean up the connection between the earth wire and spike with wire wool and DeOxit you'll also achieve a lower-impedance earth connection, which should provide a tighter and more extended bass response from your loudspeakers. Having a well-grounded case also ensures good RFI rejection, both from unwanted digital signals interfering with your soundcard audio, and digital emanations from your PC emerging to pollute other digital devices in your studio.

Digital Tests

Most musicians assume that once analogue converters are out of the way, and audio connections are made via a consumer S/PDIF or more professional AES-EBU connection, digital transfers will yield a bit-for-bit clone of the original. However, as I've found over the last few years, this often isn't the case.

Sometimes, as in the case of Creative Labs' soundcards, every input signal, including digital ones, undergoes real-time asynchronous sample-rate conversion (ASRC). This has the advantage that the digital signal can be mixed in with other analogue signals in real time, and you don't have to rely on the external clock signal, since the card carries on using its own internal one. However, this design approach always places a conversion layer between the original and final signals, and since no conversion is perfect, a tiny amount of audio quality is always lost — normally, some small ripples are introduced into the soundcard's frequency response near its upper limit.

However, in my experience, most soundcards that don't provide bit-for-bit copies fall into one of two other categories. The first add a tiny amount of dither noise to cope with digital inputs of varying bit depths — if, for instance, you input a 24-bit digital signal but are recording at 16-bit, it makes sense to dither the lower bits rather than truncate them, to extract the maximum possible detail from the original file. If you're recording a 16-bit digital signal with 16-bit resolution, there's no need for any dither noise to be added, but unfortunately some soundcards still do this. Some (like my Echo Mia) provide a tick box for Dither Input allowing you to disable this function and get bit-accurate digital copies; others don't have this option, but it's not a disaster, since the noise is likely to be at a level of about -90dB, and therefore scarcely audible.

The second category is more worrying, and nearly always relates to driver or firmware bugs. It seems quite common for digital transfers to end up phase-inverted, which is confusing during checks, but rarely audible in practice. Some cards I've reviewed instead had their left and right channels out of sync by one or more samples. At first, this would seem fairly benign — after all, with the speed of sound being 1120 feet per second, a shift of one sample at 44.1kHz is a delay of only 22.7 microseconds, equivalent to moving that channel's loudspeaker backwards by about a third of an inch. However, if you employ any further processing, or subsequently mix down to mono, even a one-sample shift between left and right channels can be audible at high frequencies. I've also measured the occasional soundcard that changes the gain of digital output signals by anything up to 6dB. This implies a DSP process at work, and will certainly make a mess of your digital transfers.

Some digital audio transfers suffer thanks to bugs in soundcard firmware design, as in this example. The source file at the top has been played back through the S/PDIF output of the test soundcard, and then recorded back into its S/PDIF input. As you can see, it has returned inverted and with its right-hand channel delayed by one sample.Some digital audio transfers suffer thanks to bugs in soundcard firmware design, as in this example. The source file at the top has been played back through the S/PDIF output of the test soundcard, and then recorded back into its S/PDIF input. As you can see, it has returned inverted and with its right-hand channel delayed by one sample.

Happily there's a way to test your digital I/O to weed out such problems, and this is what I do with each and every soundcard I review for SOS. What you need to do is send an audio file out of the S/PDIF output, then back to the S/PDIF input, and compare the final result with the original source. I normally do this as a two-stage process using Wavelab, first selecting the S/PDIF output for playback, making sure the soundcard's clock is set to Internal. I then replay a suitable WAV with an easily identified start point, such as a drum beat, and record this on an external DAT recorder. Then, after switching the soundcard's clock to External so that it locks to that of the incoming digital input, I play back the recording from the DAT machine and rerecord it using Wavelab with the S/PDIF input selected.

Finally, I use the Windows 'Tile Horizontally' option to place the original and final files one above the other, and then trim off any excess from the rerecorded version so that when zoomed into to 1:1 magnification they both start at exactly the same sample. Wavelab's File Comparer function can then be used to see and hear any differences between the two. A steady low-level hiss indicates added dither, while a quiet and often 'comb-filtered' residue may indicate a free-running clock not locked with the external device, or one channel being time-shifted.

It may also be possible to perform some quicker initial checks with Rightmark's Audio Analyser, although not all soundcards permit this: some won't let you simultaneously use the internal clock and record correctly from the S/PDIF input, and I've found some that give false readings even if they appear to work, since you get quantisation noise as the test signal isn't synchronous with the clock.

By the way, don't forget that although some upmarket soundcards have transformer-coupled co-axial digital I/O, most cheaper ones don't, so this is yet another source of ground loops. Toslink optical connections don't suffer from this, although the cables do tend to be more expensive.

Using DI Boxes With A PC

If you don't fancy the hassle of soldering up special pseudo-balanced leads for your soundcard (and I don't know of any companies that sell ready-made ones), you can instead use a 1:1 isolating transformer or DI box, which will also be ideal to break a synth/soundcard ground loop where both source and destination are unbalanced. Don't disconnect the earth wire from the synth's mains plug, since this is potentially lethal.

To add balanced I/O and completely avoid ground loop hum, some musicians I know have bought two stereo DI boxes, one to plug direct into their stereo soundcard's input and the other to attach to the output — but at around £80 for the pair it must surely be cheaper to buy a soundcard with balanced I/O in the first place.

DI boxes are most often used between a high-impedance device such as a guitar and a low-impedance balanced mic input (some engineers insist that using this approach can improve the sound of keyboards compared to plugging them directly into a mixer line input). You'll probably need one with a 20dB pad to avoid overloading the mic input. However, you don't need this impedance conversion for synth/soundcard connections, and there are now also purpose-designed DI boxes for connecting line-level gear to line-level inputs, with 1:1 ratio windings. Examples include Whirlwind's Direct2 and pcDI, both of which support stereo signals.

If you do use a more general-purpose isolating transformer, make sure it's designed for high-quality audio work. I've noticed several people on the SOS forums recommending a 'ground loop isolator' cable, particularly for use with PC laptops, which was originally designed to cure in-car hum and buzz problems. While the hum will certainly disappear, you generally get what you pay for, and while a £7.99 transformer will sound wonderful for in-car entertainment, its low-frequency response may well drop off below 40Hz and its distortion characteristics are likely to be poor. Again, this is an ideal task for a software utility like Rightmark's Audio Analyser — just place the isolator in line in one of the loopback cables, and see how it affects frequency response and distortion. The isolator may also be unshielded, and while this will not necessarily be a problem inside a car, it may pick up significant hum from nearby mains cables, which rather defeats the whole object of the exercise in the studio!

Final Thoughts

With a little effort, along with suitable software utilities like Rightmark's Audio Analyser and your normal audio editor, you may be able to noticeably improve every PC recording you make, achieving a cleaner signal with lower background noise. The solution may simply involve a few tweaks in the soundcard's utility to remove the unwanted contributions from unused inputs or DSP, or it may involve getting out your soldering iron and making up a few special cables.

As I write, I'm actually listening to the results of following my own advice. While my synths already had pseudo-balanced leads, and my Echo Mia has balanced inputs and outputs, I've cleaned up various plugs with DeOxit and ProGold, cleaned up and watered my earth spike, and recalibrated the output levels of my mixer against the input sensitivity of my soundcard. Doing this has provided me with subtle improvements in bass response and top-end clarity and 'air', taken one hour, and cost me nothing.

Published November 2002