Atmega644 / STM32 sensitivity

[schirru]

Hi all

I ordered a new STM32 based Megadrum which should replace an old Atmega644 based Megadrum. I haven't changed the trigger settings for some years but for a new project I want to use new pads. So I spent some hours tweaking the settings for the old and the new pads and reading the documentation and the forum.

The LCD on the new Megadrum is a great improvement. The detection of the velocity is quite consistent but I had to rise Latency from 15 to 30 to achieve that.

One thing I noticed is that compared to the Atmega644 the new Megadrum seems to be more tolerant to hot pads. In the STM32 based Megadrum I can't reach 1023 anymore so for most of my pads I can discard the voltage divider.

From reading though the documentation I understand that I can either set the gain to 0 (and in my case get a HighLevel of around 600) or rise the gain which leads to higher HighLevel and Threshold setting. Does this influence the playability or any of the other settings? Which one is better or does it matter at all? My understanding is that it doesn't matter.

With the new Megadrum I have difficulties to detect soft hits on all kind of pads (mesh and rubber). On the new Megadrum I have to set the Threshold to 25 - 30 compared to 7 on the Atmega644. Which means that I have to hit the same pad harder to get a signal. If I set the Threshold to a lower value, I get lots of low velocity false triggers without touching the pad. The input signal seems to be more noisy. Is there any setting that can influence low velocity triggering? If not is there a hardware tweak to improve the input signal? I don't mind if I have to solder.

Thank you for your help.

Marco

[airflamesred]

Hi all
With the new Megadrum I have difficulties to detect soft hits on all kind of pads (mesh and rubber). On the new Megadrum I have to set the Threshold to 25 - 30 compared to 7 on the Atmega644. Which means that I have to hit the same pad harder to get a signal. If I set the Threshold to a lower value, I get lots of low velocity false triggers without touching the pad. The input signal seems to be more noisy. Is there any setting that can influence low velocity triggering? If not is there a hardware tweak to improve the input signal? I don't mind if I have to solder.

The figures you are quoting really don't have any value. I use rubber pads with a threshold of 81, high level circa 650. The important thing is that the aural feedback you receive from the pads matches how hard you hit them.

[dmitri]

What signal level MegaDrum registers on a signal from a piezo differs between Atmega and STM32 so the settings for Threshold, Gain and HighLevel also will differ for the same pad. The numbers (and linearity) for these settings are not closely calibrated between Atmega and STM32 so the same signal level will be registered as say 10 on Atmega but as 30 on STM32. Same applies for inputs with a Precision addon and for inputs without.

[schirru]

Thank you for your answers.

I understand now that the values of the Atmega644 and the STM32 are not comparable but I still have to hit the same pad (same cable) harder on the STM32 to get a midi signal. At the moment the new Megadrum ignores my soft hits i d'like to play.
I have both Megadrums configured with the lowest possible Threshold setting (so that I don't get any low velocity false triggers without touching the pad). Is there something that can be done to make the STM32 Megadrum more sensitive in the lower end?

[anttipi]

Do you have AltSamplingAlg DISabled? I made the mistake of enabling it when I upgraded from Atmega (automatically assuming it's better because it's newer...). It messed up the way Gain settings work etc., and according to Dmitri AltSamplingAlg is still work in progress and not extensively tested.

Like you, I also found that thresholds need to be set higher, which initially gave me the feeling of diminished sensitivity. But after tweaking everything from scratch, the sensitivity seems OK now.

In my experience, ARM is much better in suppressing false/double triggering, maybe partly because I can now use higher MinScan values to get proper level detection and latency remains low enough. DynLevel/DynTime can be set very close to - and in same cases at - zero, depenging on pad.

[dmitri]

Can you test something with STM32?
1. Set Gain to 4 and Threshold to 30.
2. Disconnect the pad from an input on the modules end.
3. Start lowering Threshold one tick at a time. At which Threshold do you start getting false triggering?

1. Set Gain to 4 and Threshold to 30.
2. Connect the pad to the input.
3. Start lowering Threshold one tick at a time. At which Threshold do you start getting false triggering?

[schirru]

Hi anttipi, I have AltSamplingAlg disabled so that can be ruled out. But thank you for the tip with MinScan I'll have to look into that again.
Dimitri, I will test that tonight.

[schirru]

OK, I tested with 3 different pads three different cables (two different types) on three different input ports with gain 4.

With no pad connected I can get the Threshold setting down to 4 (on all tested inputs).

With any pad connected I can get down to 24-26 depending on the cable and the pad before the false triggering starts (on all tested inputs).

I even unplugged the power cord from the notebook (where the Megadrum gets it's power from) and let it run on battery with no noticeable change.

[schirru]

Sorry, the Threshold levels mentioned are the last ones I could use reliably without getting false triggers. So one tick lower and the false triggering starts.

[dmitri]

Well, this tells me that the noise is coming from a connected pad. Are you sure you're not getting anti-placebo effect just because you have to set Threshold higher in STM32?
Can you do these two tests with the same pad on both Atmega and STM32?
1. Set Threshold to the lowest possible value without false triggering.
2. Set Gain to 4.
3. Set HighLevel Auto to Yes.
4. Set HighLevel to 64.
5. Hit the pad very hard a few dozen times.
6. Set HighLevel Auto to No.
7. Change Bars Number in the MIDI log a couple of times to clear the bars and set it to 24 after that.
8. Let a stick free bounce on the pad until it rests still on the pad. Do 2-3 times.
9. Post a screenshot of the pad settings and the MIDI Log together in one screenshot.