A developer says its plugin and the hardware "sounded the same side by side." Should that sentence make you spend money? Not until you have heard the plugin under conditions the marketing never controlled: matched loudness, identical patches and no knowledge of which is which.
The hook is discoDSP's TED-8, a plugin that models the Oberheim TEO-5. According to the gearnews Synth Journal roundup, the developer says TED-8 is "built from thousands of measurements of a real unit," with every knob, curve and effect tuned until the plugin and the hardware sounded the same side by side. That is discoDSP's statement, not an independent finding.
The same post lists an introductory price of $99, rising to $199, with a $49 loyalty tier. It also notes that the TEO-5 is still on sale, so the hardware is at least obtainable for a comparison.
The protocol below works for any analog-modeled plugin and any hardware synth.
Why casual A/B tests mislead
Three things skew most comparisons. The louder sample sounds better, patches differ in ways you cannot see, and you know which one is the expensive one. The steps below address each problem.
Step-by-step test protocol
- Build identical starting patches on both instruments.
- Record the hardware into your DAW.
- Capture the plugin the same way.
- Align and level-match the clips.
- Switch blind.
- Score what you heard.
Step 1: Build identical starting patches
Start from an init state on both instruments. Use one sawtooth oscillator, with no sub, no noise and no second oscillator. Switch off every effect, including chorus, delay and reverb, and set modulation depths to zero.
Set filter cutoff to the 12 o'clock position and resonance to a fixed mid position, then write both values down. Choose a simple amp envelope: fast attack, medium decay, full sustain, short release. Knob positions rarely map identically between a panel and a plugin GUI, so do not trust the picture alone.
Verify the match with a spectrum analyzer on a held note. Compare where the filter rolls off and where the resonance peak sits, then nudge the plugin until they agree.
The gearnews post says TED-8 can link to the hardware over USB and act as an editor. If your plugin offers something similar, it may help sync settings. Check the result with the analyzer rather than assuming the patches are identical.
Step 2: Record the hardware
Send the same MIDI clip from your DAW to the hardware every time. Use C2 and C3 at fixed velocity, with one long held note, one staccato run and one repeated-note pattern.
Record the hardware's output through the same interface input for every take. Set gain staging once and never touch it again.
Capture at least four passes of each clip. Analog oscillators and filters wander, so a single take can mislead you about what is normal.
Step 3: Capture the plugin the same way
Play the identical MIDI clips through the plugin. Bouncing in the box is the simplest route. See related: internet speed for remote music collaboration explained for additional background.
A fairer route is a re-amp loop. Send the plugin's output out of one interface output and back into an input, so it passes through the same converters as the hardware. That removes a difference you were never trying to test.
Keep the plugin's quality or oversampling setting fixed and write it down. Many plugins offer a higher-quality mode, so decide in advance whether you are judging the default or the best setting.
Step 4: Align and level-match
Trim each clip so its first transient lands on the same sample position, or as close as your editor allows. Latency offsets can make one source feel snappier than the other.
Pick a short, representative section of about five seconds. Measure it with a loudness meter in RMS or short-term LUFS. Add or subtract gain with a utility plugin until both read within about 0.1 dB of each other.
Apply the gain to the clip, not to the synth's own volume control, which would change the saturation behavior you want to hear.
Check that neither clip clips or hits a limiter. If matching leaves one source with higher peaks, leave it that way. That difference is part of the sound.
Step 5: Switch blind
Have your DAW or an ABX tool hide the identity of each clip. If you are working alone, rename the files with random labels, shuffle them and look away from the track names. A friend can also assign letters and keep the key. Label everything before you listen, never during.
Listen at the same monitor volume every time, and keep each switch to a few seconds so memory does not blur differences. Run about ten trials per trait, and write down your guess before checking the key.
Pure guessing gets 8 or more of 10 right about 5% of the time, and 9 or more about 1%. A result of 9 out of 10 or better is therefore a stronger sign that you can hear a real difference.
What to listen for
Use a separate clip for each trait so you know what you are judging.
Filter resonance. Sweep cutoff slowly with resonance high. Listen for where the peak sits, whether the low end thins as resonance rises, and what happens near self-oscillation.
Envelope snap. Use the staccato clip. Focus on the attack click and how quickly the decay curves into silence. Check the release tail at the end of each note. This pairs well with our guide to how to test budget dj headphones and speakers at home.
Drift and detune. Play the long held note and the repeated-note pattern. Does pitch move between takes? Does the oscillator phase change each time you retrigger? Listen for drift that is too regular, which can sound modeled, and for pitch that is so stable it sounds sterile.
Saturation as level rises. Record the same note at three velocities or input levels, then level-match each pair. Listen for harmonic thickening and compression that a clean plugin may skip.
Stereo effects. Test these last, and only after the dry patch passes. Turn on one onboard effect at a time in both instruments, and compare width, modulation rate and tail character.
Step 6: Scoring sheet
Score each trait from 0 (clearly different, identified every time) to 5 (could not tell). Copy this table onto paper.
| Trait | Blind trials correct (out of 10) | Score 0-5 | Notes | |---|---|---|---| | Filter resonance | | | | | Envelope snap | | | | | Drift and detune | | | | | Saturation at higher levels | | | | | Stereo effects | | | | | Total (out of 25) | | | |
Reading your result
These thresholds are suggestions, not industry standards. Adjust them to the traits you care about.
Also read: serato sample alternative: chop breaks in your daw
- Buy if you score about 20 or higher and your blind results sit near chance on the traits you use most.
- Demo longer if you land between roughly 14 and 19, or if a single trait pulls the total down. Test that trait again in a real mix before deciding.
- Skip if you identify the hardware reliably in the filter or envelope tests, since those shape most of what a synth does.
Weigh price too. A $99 plugin that scores 17 may be a better value than a pricier one that scores 20. Check current pricing on the developer's page, because the figures above come from a gearnews post dated October 29, 2026.
If you do not own the hardware
You have three workable routes.
Borrow a friend's unit. Ask them to record your MIDI clips using your written patch settings and send back the WAV files. You then render the plugin side. Better still, have your friend label the files randomly and hold the key.
Use a demo version. Check its limits first, since some demos mute audio, add noise or block saving. If you cannot record clean takes, the test is invalid.
Compare two plugins. This will not tell you how close either one is to hardware, but it uses the same method and shows which plugin you prefer blind. It also lets you practice the protocol before a real purchase.
In all three cases the rules stay the same: identical patches, matched levels, blind switching and a written score.
If your first run looks off
If the first pass shows big differences, check the patch before blaming the plugin. Redo step 1 with the analyzer, then repeat the capture. Once the patches match, your scoring sheet gives you a result you can defend instead of a hunch.
