Re: Pulsewidth differences

[AlainHubert]

I had two rev2s and noticed that both had pulswidth differences on certain settings and tones between voices. Interestingly in blocks of 4 voices.

One rev2 had voices 5-8 different to 1-4 and 9-16 and similar to 8-12. The other one had voices 13-16 different to the others. In some cases this is bothering me, most notably when playing certain bass notes with wide, static pulse widths.

This is, except the 4 voice blocks, not unique to the Rev2, other units have various discrepancies including filters (where auto calibration does not help) and loudness.

So, I would like to suggest an option so users can manually tune/offset parameters for each voice. Actually not just regarding PW but also filters including other units as well (Prophet 12, 6).

After doing a firmware update on mine (to the latest 1.1.0.3), I did a full calibration but things didn't turn out very well. I had voices on which the PW was really off (by 10 values in order to get a 50% duty cycle), and some voices were softer in volume than others. I was starting to get a little worried. So I performed another full calibration a second time, and it took quite a long time to finish (often getting stuck for minutes in some places), but when all done it was back in full health again. Almost all 16 voices are producing a 50% duty cycle waveform with a PW value of 51. Only one voice is ever so slightly off, but barely audible. And all audio levels are identical between all 16 voices now.

So maybe redoing a full calibration at some point might alleviate some of those discrepancies between different units ? 
Also make sure to check each oscillator independently so as not to get fooled by some phasing artifacts between the two.
 

[Razmo]

I had two rev2s and noticed that both had pulswidth differences on certain settings and tones between voices. Interestingly in blocks of 4 voices.

One rev2 had voices 5-8 different to 1-4 and 9-16 and similar to 8-12. The other one had voices 13-16 different to the others. In some cases this is bothering me, most notably when playing certain bass notes with wide, static pulse widths.

This is, except the 4 voice blocks, not unique to the Rev2, other units have various discrepancies including filters (where auto calibration does not help) and loudness.

So, I would like to suggest an option so users can manually tune/offset parameters for each voice. Actually not just regarding PW but also filters including other units as well (Prophet 12, 6).

After doing a firmware update on mine (to the latest 1.1.0.3), I did a full calibration but things didn't turn out very well. I had voices on which the PW was really off (by 10 values in order to get a 50% duty cycle), and some voices were softer in volume than others. I was starting to get a little worried. So I performed another full calibration a second time, and it took quite a long time to finish (often getting stuck for minutes in some places), but when all done it was back in full health again. Almost all 16 voices are producing a 50% duty cycle waveform with a PW value of 51. Only one voice is ever so slightly off, but barely audible. And all audio levels are identical between all 16 voices now.

So maybe redoing a full calibration at some point might alleviate some of those discrepancies between different units ? 
Also make sure to check each oscillator independently so as not to get fooled by some phasing artifacts between the two.
 

The calibration routine is a bit flakey... it does take longer times once in a while... my own experiences is that the longer my REV2 has warmed up, the slower the process is... but I will suggest anyone to WAIT at least half an hour after power up, before you do the calibration routine because, even if the oscillators themselves are perfectly in tune from the get-go, filter frequencies are NOT! ... I have quite a few presets I have done that utilize playing the filter with high resonance, and if you calibrate the filter from a cold REV2, you can be dead certain, that when the machine has warmed up, those sounds will be way out of tune.

I have actually asked DSI about the calibration routine, and was told that they know that it takes longer sometimes than others, and sometimes it may even freeze completely, never ending... I bet the difference has to do with the warm up period, but that is my own suspicion.

I was also told, that I could press the two transpose buttons down at the same time, and then press the "global" button to enter a service mode... in there you can start a function called "erase calibration data"... if you do this (I was told), doing a calibration of voices afterwards would always finish... I do this before any calibration now, just to ensure that calibration will finish.

I wonder though, why erasing the calibration data is needed at all... I would have thought that doing a recalibration would automatically erase the previous data... this makes me wonder why previous data would pose a problem at all... but I guess that only DSI can explain that.

Summa Summarum; if you have this PW problem, try letting the REV2 warm up first, then recalibrate, and see if this corrects the problem... and have in mind, that even though DSI is stating that no warm up is needed with the REV2 because of the Curtis chips, that it's not quite true... oscillators may, but I can certainly guarantee you, that filter frequencies will change with warm up time... you will probably not notice this very slight difference when not using the filters in selfoscillating mode, but when you want to play the filter melodically you definitely have to let the device warm up first, and make sure that you calibrate it after that warm up period... maybe this is the same with PWM...

[Chimponaut]

Remember to disconnect midi cables and usb when doing the calibration routine. The one and only time my Rev2 didn't complete the calibration routine, that was the culprit. Disconnected the cables and the next calibration went smooth.

[AlainHubert]

The calibration routine is a bit flakey... it does take longer times once in a while... my own experiences is that the longer my REV2 has warmed up, the slower the process is... but I will suggest anyone to WAIT at least half an hour after power up, before you do the calibration routine because, even if the oscillators themselves are perfectly in tune from the get-go, filter frequencies are NOT! ... I have quite a few presets I have done that utilize playing the filter with high resonance, and if you calibrate the filter from a cold REV2, you can be dead certain, that when the machine has warmed up, those sounds will be way out of tune.

I have actually asked DSI about the calibration routine, and was told that they know that it takes longer sometimes than others, and sometimes it may even freeze completely, never ending... I bet the difference has to do with the warm up period, but that is my own suspicion.

I was also told, that I could press the two transpose buttons down at the same time, and then press the "global" button to enter a service mode... in there you can start a function called "erase calibration data"... if you do this (I was told), doing a calibration of voices afterwards would always finish... I do this before any calibration now, just to ensure that calibration will finish.

I wonder though, why erasing the calibration data is needed at all... I would have thought that doing a recalibration would automatically erase the previous data... this makes me wonder why previous data would pose a problem at all... but I guess that only DSI can explain that.

Summa Summarum; if you have this PW problem, try letting the REV2 warm up first, then recalibrate, and see if this corrects the problem... and have in mind, that even though DSI is stating that no warm up is needed with the REV2 because of the Curtis chips, that it's not quite true... oscillators may, but I can certainly guarantee you, that filter frequencies will change with warm up time... you will probably not notice this very slight difference when not using the filters in selfoscillating mode, but when you want to play the filter melodically you definitely have to let the device warm up first, and make sure that you calibrate it after that warm up period... maybe this is the same with PWM...

Yes indeed. Even if those OnChip/Marion/Curtis/DSI chips have DCOs, the internals are all analog (PW, VCA, VCF) and require to be at operating temperature to be most stable.
BTW, thanks for the info on clearing the calibration table Razmo. I suspect that, for speeding up purposes, the calibration routine might try previously entered data in that table for a given parameter/voice, and if the data is giving good results, it doesn't take the time to calculate and store a new one ? Hence the time differential sometimes seen between voices and oscillators during calibration ? Just guessing...

[Razmo]

The calibration routine is a bit flakey... it does take longer times once in a while... my own experiences is that the longer my REV2 has warmed up, the slower the process is... but I will suggest anyone to WAIT at least half an hour after power up, before you do the calibration routine because, even if the oscillators themselves are perfectly in tune from the get-go, filter frequencies are NOT! ... I have quite a few presets I have done that utilize playing the filter with high resonance, and if you calibrate the filter from a cold REV2, you can be dead certain, that when the machine has warmed up, those sounds will be way out of tune.

I have actually asked DSI about the calibration routine, and was told that they know that it takes longer sometimes than others, and sometimes it may even freeze completely, never ending... I bet the difference has to do with the warm up period, but that is my own suspicion.

I was also told, that I could press the two transpose buttons down at the same time, and then press the "global" button to enter a service mode... in there you can start a function called "erase calibration data"... if you do this (I was told), doing a calibration of voices afterwards would always finish... I do this before any calibration now, just to ensure that calibration will finish.

I wonder though, why erasing the calibration data is needed at all... I would have thought that doing a recalibration would automatically erase the previous data... this makes me wonder why previous data would pose a problem at all... but I guess that only DSI can explain that.

Summa Summarum; if you have this PW problem, try letting the REV2 warm up first, then recalibrate, and see if this corrects the problem... and have in mind, that even though DSI is stating that no warm up is needed with the REV2 because of the Curtis chips, that it's not quite true... oscillators may, but I can certainly guarantee you, that filter frequencies will change with warm up time... you will probably not notice this very slight difference when not using the filters in selfoscillating mode, but when you want to play the filter melodically you definitely have to let the device warm up first, and make sure that you calibrate it after that warm up period... maybe this is the same with PWM...

Yes indeed. Even if those OnChip/Marion/Curtis/DSI chips have DCOs, the internals are all analog (PW, VCA, VCF) and require to be at operating temperature to be most stable.
BTW, thanks for the info on clearing the calibration table Razmo. I suspect that, for speeding up purposes, the calibration routine might try previously entered data in that table for a given parameter/voice, and if the data is giving good results, it doesn't take the time to calculate and store a new one ? Hence the time differential sometimes seen between voices and oscillators during calibration ? Just guessing...

That would probably be one of the best reasons I've heard thus far... seems very likely that it would be faster to just search for the correct tuning based on a plus/minus from the calibration data already there, instead of going thru the whole pitch range...

[Indigo V.]

Thanks Razmo for sharing how to reset the calibration table.

I did several firmware updates and recalibrations on both units (not that it matters but I kept and own only 1 ) and it doesn't change the problem on a fundamental level. The only thing that would help would be an option to add offset values to each voice. Say voice 8 PW offset +3 between 48-52 so that when you dial in 50 you get 53 which sounds like 50 of say voice 1-7.

This PW issue is pretty prominent on this synth but actually most if not all analog synths I know have this to some degree. It's not even a bad thing per se but beyond a certain point and in some scenenarios it can be annoying. However, where it is most annoying to me is when playing basslines. In that scenerio however, an easy workaround is possible: Setting the layer(s) to single voice. Obviously, you don't get any wrong sounding voices then! But still, I hope there will be some improvements possible.

[AlainHubert]

Thanks Razmo for sharing how to reset the calibration table.

I did several firmware updates and recalibrations on both units (not that it matters but I kept and own only 1 ) and it doesn't change the problem on a fundamental level. The only thing that would help would be an option to add offset values to each voice. Say voice 8 PW offset +3 between 48-52 so that when you dial in 50 you get 53 which sounds like 50 of say voice 1-7.

This PW issue is pretty prominent on this synth but actually most if not all analog synths I know have this to some degree. It's not even a bad thing per se but beyond a certain point and in some scenenarios it can be annoying. However, where it is most annoying to me is when playing basslines. In that scenerio however, an easy workaround is possible: Setting the layer(s) to single voice. Obviously, you don't get any wrong sounding voices then! But still, I hope there will be some improvements possible.

My Behringer Deepmind 12 also suffers from this PW inconsistency between voices, by the way, and even at a greater extent than my REV2.

With the very latest OS on the REV2-16, it took 2 calls to the calibration routine to get it almost perfect. Only 2 voices show a very very slight deviation from a perfect square (50% duty cycle) when compared to all other voices for a PW value of 51.

Don't forget to wait at least 15 to 30 minutes after power-up before calling the calibration, since the part controlling the wave shape, hence the pulse width, is analog (yes even with DCOs) and needs to be warmed-up to be the best it can be.