Alpha Juno DC offset

[xhy3]

I wonder where the final DC offset on the Alpha Juno output comes from and how it is introduced. The image in the attachment shows the output with a simple saw.

In the schematics, it looks like there is already an HP filter after the DCO output (page 11) per voice that should already remove the offset before the filter and AMP envelope:

https://www.synthxl.com/wp-content/uplo ... -notes.pdf

When silent, the first notes envelope attack starts at zero. The ending is also at zero when you hold the note for a longer time or if the note has a long release time. But it's probably not at zero when the note is released after a short time.

The slow DC offset "jump" is there even when the HP with a frequency > 200 Hz is enabled which normally would remove the offset pretty fast within a few samples.

It also looks like the DC offset is not there at the start and the end of the note. It starts and ends at zero. This makes me think that it is probably introduced somewhere in the voice chip before the AMP envelope. But why doesn't the HP filter affect the behavior?

Any ideas are welcome. Thanks

[mystran]

Not sure about the specifics here, but what we see here looks like your typical exponential decay (perhaps one-pole, perhaps two poles) that suggests that for one reason or another the blocker starts at the DC offset before the note. Some amount of DC offset can result from changing frequency, phase resets or even application of a moderately fast envelope (the envelope probably explains the offset we see at the end of the note). It could also be the case that the waveform isn't actually exactly DC centered before the final blocking caps after envelope or something, idk. Similar things often happen even with pure analogs and it could really be a combination of many things... and don't forget that if you're recording this on a typical audio interface, there's probably a capacitor at the input too, so what you see in software isn't necessarily exactly what leaves the synth itself.

Just looking at output alone often tends to be somewhat guesswork though, because "DC" variations through the signal path can look different depending on where you look at, it's not always just one cause, but rather anything non-linear can introduce more and if you truly want to figure out where the offsets come from you'd kinda want to poke a probe into the circuit (or sometimes even rig some opamps to buffer the signals you want to probe if you happen to have a circuit where probing directly changes the circuit behaviour too much)... or if you don't have the circuit (or an oscillator with a probe) around, then perhaps try to see if you can reproduce the results in Spice.

[xhy3]

Thanks for sharing your thoughts. This already helps a lot. I have the device here, but it's probably a good idea to reproduce it in spice. At least some parts of it.
I also thought about a DC blocker that starts with an offset but then I thought this is a bit odd. I will do some more tests.

[S0lo]

If I were to make a quick guess. I'd say that the DC blocker needs time to bring back it's compensation for the DC offset to zero (which is also a DC offset with an opposite polarity). The same way it needed some time to observe the initial input and eliminate it's DC offset. The time required is inversely proportional to the cutoff frequency used in the DC blocker. (Edit: but there is also another factor at play here which is the order of DC blocker).

But this is just a guess relating to how most DC blockers work. I have no idea about how the alpha juno circuitry works. Too many things/variables can be present. So my guess is a good as yours.

[xhy3]

Thanks for the information. I did some tests and got pretty close results when using a DC blocker fed with an offset when a note plays (zero when not playing) before the AMP envelope and one for the mixed output. Both are relatively slow with a low cutoff frequency. The second one on the output is a two-pole to make it "oscillate" a bit without going straight to zero.

[Tone2 Synthesizers]

I did not notice that my Alpha Juno suffers from a DC offset.
Not sure if your measurements are representative for a new Alpha Juno. These machines are old. It might be also be caused by a bad/broken capacitor in the Juno or another component in your signal chain.

What I know for sure is that the is a highpass filter with pitch tracking after the oscillator output. This one should savely remove the DC offset.

[xhy3]

I did not notice that my Alpha Juno suffers from a DC offset.
Not sure if your measurements are representative for a new Alpha Juno. These machines are old. It might be also be caused by a bad/broken capacitor in the Juno or another component in your signal chain.

What I know for sure is that the is a highpass filter with pitch tracking after the oscillator output. This one should savely remove the DC offset.

Thanks for the information. Didn't think about that. That's possible. Can you share a scope image or audio file with a simple saw from your device?

[xhy3]

I did some research and it looks like this isn't a problem with my device. You can also see the DC Offset in the Alpha Juno demos. A good example is the 65 Oboe sound from here:

https://www.synthmania.com/alpha_juno-2.htm

I did some more research and the DC offset is introduced by the DCO Waveforms itself. Every waveform introduces a DC offset, even the Alpha Juno SAW. The DC remover filter probably resets when a note starts or the Amp Envelope changes the DC offset. This probably leads to DC instability at the beginning of a note.

[xhy3]

Here is the screenshot of the Oboe Sound. Even a sound that uses the strongest HP filter setting has this offset at the start of the note.