I'm having a hard time making acceptable quality digital recordings of an 8-bit computer synthesizer - that is with the line-out of the synth plugged into the line-in on my soundcard. The voices of the synth are programmable-frequency square wave generators.
Max level is 8-10 dB below clipping on record, but single discrete notes that sound pure directly from the synth acquire an annoying degree of "graininess" when recorded and played back - like there is a lot of FM modulation.
When playing chords of closely spaced keys in the same octave, additional and distinctly annoying artifacts are produced. It's present on the initial uncompressed WAV recordings but is exacerbated when the recording is compressed, which you can't avoid when uploading to a video sharing site.
The chord playing artifacts (especially after compression) make the digitally recorded audio sound remarkably like something being played back from a scratchy LP record.
The synth analog out has a single-pole roll-off at around 18 kHz. There is no undue noise or anything else.
I am wondering if there is still enough high-energy harmonic content in the squarewaves to play havoc with my audio cards digital sampling. My card is a not too flash Asus "Xonar" thing that set me back in the region of $35 maybe 10 years ago.
Has anyone else had similar experience or has enough insight into the robustness of anti-aliasing filtering on run-of-the-mill soundcards?
Max level is 8-10 dB below clipping on record, but single discrete notes that sound pure directly from the synth acquire an annoying degree of "graininess" when recorded and played back - like there is a lot of FM modulation.
When playing chords of closely spaced keys in the same octave, additional and distinctly annoying artifacts are produced. It's present on the initial uncompressed WAV recordings but is exacerbated when the recording is compressed, which you can't avoid when uploading to a video sharing site.
The chord playing artifacts (especially after compression) make the digitally recorded audio sound remarkably like something being played back from a scratchy LP record.
The synth analog out has a single-pole roll-off at around 18 kHz. There is no undue noise or anything else.
I am wondering if there is still enough high-energy harmonic content in the squarewaves to play havoc with my audio cards digital sampling. My card is a not too flash Asus "Xonar" thing that set me back in the region of $35 maybe 10 years ago.
Has anyone else had similar experience or has enough insight into the robustness of anti-aliasing filtering on run-of-the-mill soundcards?
Does the sample rate setting have an impact? How about a makeshift low-pass filter to filter off signals above 100 kHz? Do you see any suspicious patterns when you inspect the waveform with GoldWave, Audacity or some other audio editing program?
I have experience with a computer sound card driving a dirt cheap audio ADC. It produced a soft beep due to aliasing. It was a sigma-delta ADC with no analogue anti-aliasing filter at all, only the digital decimation chain, so any sound card clock residue at a few kilohertz from a multiple of the sigma-delta clock rate became a tone at a few kilohertz.
Another issue I once had were zero samples inserted every buffer length samples, but randomly, in about 1 out of 20 recordings made with a specific sound card, some fancy semiprofessional multichannel sound card.
I have experience with a computer sound card driving a dirt cheap audio ADC. It produced a soft beep due to aliasing. It was a sigma-delta ADC with no analogue anti-aliasing filter at all, only the digital decimation chain, so any sound card clock residue at a few kilohertz from a multiple of the sigma-delta clock rate became a tone at a few kilohertz.
Another issue I once had were zero samples inserted every buffer length samples, but randomly, in about 1 out of 20 recordings made with a specific sound card, some fancy semiprofessional multichannel sound card.
This sounds like a fairly classic case of the synth outputting loads of ultrasonic energy and the sound card not bothering with an analog anti aliasing filter, seen it before with a rather too cheap ADC and some eurorack nonsense. Easy enough to fix usually with something like a 4th order filter up at 40kHz or so.
It is very tempting when you have a constrained COG to omit this filter in a design as you usually get away with it and you both save money and make the in band numbers look a bit better... Synths outputting square waves are somewhat pathological, as are a set of keys shaken in front of a mic...
It is very tempting when you have a constrained COG to omit this filter in a design as you usually get away with it and you both save money and make the in band numbers look a bit better... Synths outputting square waves are somewhat pathological, as are a set of keys shaken in front of a mic...
Indeed if you have unfiltered square or sawtooth waves (typically these might be from logic chips, with logic chip risetimes) you need a really good anti-aliasing filter as there is likely to be significant signal power upto many 10's of MHz to deal with. The first thing is good RF filtering at the input to stop RF getting inside the box (perhaps feed-through capacitor?), then a competent hgher-order (Butterworth?) opamp-based antialiasing filter. JFET opamps are a wise precaution against residual RF energy as they are much more immune to RF superimposed on the inputs than bipolar chips. You can do 2, 3 or even 4 filter poles per opamp with the Sallen-Key topology which keeps component cost down.
There are SMT 'three terminal capacitor' networks that are good for this when combined with a hundred ohms or so of series resistance to give them something to work against (Don't get carried away, these add Johnson noise) or a suitably lossy ferrite bead, trick is to keep the return path back to the chassis REALLY short (And independent of the internal system reference).
Also remember that the Sallen & Key filter is notorious for coming back up once the thing runs out of GBP, I generally prefer to deal with the MHz+ stuff passively before hitting the active filter.
How the guys designing such synths pass EMC is an ongoing mystery.
Also remember that the Sallen & Key filter is notorious for coming back up once the thing runs out of GBP, I generally prefer to deal with the MHz+ stuff passively before hitting the active filter.
How the guys designing such synths pass EMC is an ongoing mystery.
Yes, you are hearing aliasing.
You may want to try a software synthesizer instead. That should be able to mimic the sound of the old synthesizer.
(I don't have specific recommendations).
Ed
You may want to try a software synthesizer instead. That should be able to mimic the sound of the old synthesizer.
(I don't have specific recommendations).
Ed
Last edited:
- Home
- Source & Line
- Digital Line Level
- Digital recording artifacts