FSR sensor compatible with Megadrum?

[dmitri]

While i understand that I can adjust the sensitivity in MegaDrum, by not taking advantage of the full 0-5V range, I am exaggerating the dynamic range because smaller voltage changes ie hit pressures are extrapolated i.e your mapping 0-3V into velocity 0-127, so small changes in pressure translate to larger increments of velocity whereas if I use gain to max out to 5V then the mapping is 0-5V into 0-127 so the pad is less sensitive to very small differences in hit pressure. this translates into a feeling of being able to hit from very soft to quite hard and hearing the same thing.

With the Gain set to 8 MegaDrum's dynamic resolution for 0-2.5V signals is the same as when Gain set to 0 for 0-5V signals.

[jmcdougall]

My perspective is that this is not a problem with the ability to handle dynamic resolution in the software, it is my dynamic resolution capability. Resolving 0 - 2.5V into 128 steps equals .0195 V per step while 0 - 5V equals .039V per step. Realistically, we probably only hear differences of more than 2-3 steps. What this does translate into is a little more forgiving in the differences between the hits from either hand and the repeatability of one's hits for a certain dynamic. This relates to a problem friends often see in the studio. People who are used to playing drimsets and then move to e-drums often have not adapted to the dynamic limitations of e-drums -- they still hit the pads like a regular drum and then run out of dynamics at the top end. The problem with using gain to compensate means that it is harder to tune gain so that a moderately hard hit is only generating a velocity value of 80 - 90 in order to provide audible headroom to handle hard hitting accents. Having larger quantization steps is helpful.

[dmitri]

Having larger quantization steps is helpful.

The number of steps for 0-2.5V/Gain8 and 0-5V/Gain0 are the same thus making dynamic resolution the same.

[jmcdougall]

yes, I am assuming that they are both quantitized to 128 steps. My point was that the voltage change each step represents is different if I am quantizing 0-2.5 versus 0-5v. Because the voltage steps are bigger, small variations in my hit pressure i.e. voltage variations, is not as likely to cause a step change on a 0-5v scale as it would on a 0-2.5V scale because the quantitized steps are bigger voltage changes.

You are absolutely correct in implying that using FSRs is more complex electronically and more expensive than using piezos. However, there are benefits. First, there is no significant crosstalk so thresholds can be set low. I can absolutely verify this as I have both FSR based units ( the Drumkats and TrapKat) as well as conventional piezo pads (Yamaha, Pintech, Dauz and old Kat pads) on my drum rack. Second, because the voltage range is known and controlled, dynamic range is quantified and consistent. I should have a storage scope by the weekend and will run some tests on hits from both FSRs and some of the piezo pads.

[dmitri]

yes, I am assuming that they are both quantitized to 128 steps. My point was that the voltage change each step represents is different if I am quantizing 0-2.5 versus 0-5v. Because the voltage steps are bigger, small variations in my hit pressure i.e. voltage variations, is not as likely to cause a step change on a 0-5v scale as it would on a 0-2.5V scale because the quantitized steps are bigger voltage changes.

With Gain 0 Atmega uses ~+5V as voltage reference. With Gain 8 Atmega uses ~+2.5V as voltage reference. In each case Atmega quantize voltage in 1024 level steps between 0 and maximum so from Atmega point of view the size of the level step doesn't matter.
Although I must admit that with 0-2.5V range voltage noise has twice as big effect on precision compared to 0-5V range. To sum up, yes, 0-5V signal level range is better but 0-2.5V is not much worse.

You are absolutely correct in implying that using FSRs is more complex electronically and more expensive than using piezos. However, there are benefits. First, there is no significant crosstalk so thresholds can be set low. I can absolutely verify this as I have both FSR based units ( the Drumkats and TrapKat) as well as conventional piezo pads (Yamaha, Pintech, Dauz and old Kat pads) on my drum rack. Second, because the voltage range is known and controlled, dynamic range is quantified and consistent. I should have a storage scope by the weekend and will run some tests on hits from both FSRs and some of the piezo pads.

What is the measured resistance of FSRs when at rest,hit light, medium and hard? If the resistances are in the suitable range than it might be possible to modify MegaDrum firmware to use FSR directly without opamps.

[jmcdougall]

I don't have a good way of directly measuring instantaneous resistance, but I did pick up a storage scope on E-bay that I should get Fri. With it I can measure the voltage peaks and then go back and apply static pressure to get the same voltage, remove the voltage and measure the resistance. In theory, the opamp may not be necessary, but if you look at the intersil application notes, they recommend this as a buffer. I think part of this is the fact that the resistance of an FSR is not infinite at rest so there is a small voltage being passed ( i measured a few hundred millivolts) and with the opamp you can bias this out by adding a trimpot between V+ and Grd and feeding the wiper to the - input on the opamp. With multiple similar FSRs, at least you would only need 1 trimpot as it could feed multiple opamps.

[dmitri]

I think part of this is the fact that the resistance of an FSR is not infinite at rest so there is a small voltage being passed ( i measured a few hundred millivolts) and with the opamp you can bias this out by adding a trimpot between V+ and Grd and feeding the wiper to the - input on the opamp. With multiple similar FSRs, at least you would only need 1 trimpot as it could feed multiple opamps.

FSRs for the drumming application when hit will produce short impulses so we can filter out any bias with a capacitor. In theory.

[jmcdougall]

I believe that it could also be filter just through the threshold setting. I think the bias issue is for more sensitive applications. The DrumKats actually use a 15p ceramic to ground based on my examination of the actual circuitry. There is a 4.7k resistor array on the board at the same point so I am in the process of examaning the traces to see how it is used. I will try to mock up their actual cicuit in my testing this weekend

[jmcdougall]

I received the storage scope yesterday and finally had a chance to warm it up tonight and do some testing. It is a tube unit -:)
I made sure that the calibration was correct and then set about testing serveral pads:
New 6" dauz
Old 6" daux
Yamaha Tp80s
Pintech SE-101
Old 8" Dauz
Old 10" TomKat
Pintech 10" crash cymbal
S&S X-Wing1

The results were very interesting in light of the fact that the 4051 on the input circuit saturates at +5V so anything above 5V is lost

I used both the tip and the butt in testing and as expected I could generate a high voltage hitting a heavy accent with the butt.

The results were as follows:

New 6" dauz - peaks up to 18v
Old 6" daux - peaks up to 6V
Yamaha Tp80s - peaks up to 20V
Pintech SE-101 - peaks up to 16V
Old 8" Dauz - peaks up to 4V
Old 10" TomKat - peaks up to 3V
Pintech 10" crash cymbal - peaks up to 20V
S&S X-Wing1 - peaks up to 3V

The low voltage output on the old pads was not unexpected as over time the piezos degrade but they are all within range for easy tweaking using sensitivity.

Before I comment, I would like to make it perfectly clear that MegaDrum is far from alone in feeding the piezo input into a 4051 for switching inputs to the A/D so the following does not just apply to MegaDrum by any means.

The above numbers mean that most new piezo pads have a dynamic range of about 0 to 16v. However most drum brains only recognize 0 - 5V of that range effectively throwing about 2/3s of it away. The impact of this is that unlike a real drum where the harder your hit the louder it plays, with this constraint, you can run out of dynamic very quickly. There are a couple of ways to compensate for this in the drum brain but none of them are pretty. Solution 1. Feed the input directly to an A/D - adds a lot more chips and increases the complexity of the circuitry trying to read the 8 bit digital output. (very messy !!) Solution 2. Put a trimpot on each pad input and then implement a tuning routine in the software to calibrate the input (adds a bunch of parts and you can easily see why no manufacturer provides for this -- all those tuning holes for a uneducted user to play with and get crud into. Solution 3. Assume a peak input of about 15v and add a resisitor that would knock this down to 5V. Modify the existing sensitivity routines to allow a calibration function i.e. train the brain for each pad -- user does a light hit to set the low point and a hard hit to set the upper point and this is stored by pad. This method does throw away about 20% of the dynamic range but is not a major efort to inplement. Solution 4 --- Retrain the drummer! This only requires a software change in the user's brain.

Unfortunately -- it is hard to break bad habits ! First one has to recognize that you are dealing with a limited dynamic range. That means the style for e-drums is going to have to be different than for a real drumset. This is especially true for practicing. With real drums, you actually vary your hit strength and soften down when your are in a room at home based on the complaints. You count on the fact that when you get into the large space or are playing on-stage you can simply increase the force and play louder. However most e-drummers I have come across, view e-drums as a way to avoid this. This can play just like they do on-stage and just turn down the volume. I view this as a BIG mistake. This mode fosters a habit of playing hard and when you get on-stage, you quickly run out of dynamics on the top end.
practicing on e-drums should emulate what one does on acoutic drums i.e. you should have you amp set to stage power. In order to survive at home, then you are forced to lighten your playing just like you would on an acoustic set. Since most people spend mor etime practicing than playing, you effectively train yourself to play with a softer hand. Now when you get out on the stage, you do have the headroom to push the volume envelope.

In another post, I asked that those who have sequencers do a simple experiment and post the results. I am hoping that there is at least some response. I am fairly certain that the results will support the above comment.

[Ken Forgettable]

...most new piezo pads have a dynamic range of about 0 to 16v. However most drum brains only recognize 0 - 5V of that range effectively throwing about 2/3s of it away.

Positional sensing requires that the initial wave (half - wave) is sampled, this signal is likely to be well below your bottom end levels - any ideas about isolating this could be fruitful.

There are a couple of ways to compensate for this...

Voltage dividers and such are an afterthought to the proper mounting of the sensor. My piezos are mounted in the plastic caps of asprin containers which have a central 'pimple' allowing a good degree of flexing.

Tried to reduce 'ringing' by making a hole in the bottom of the Roland cone which was quite good, but ended up mounting the whole assembly on an inch of pipe lagging which dampened excessive resonance and added isolation from the shell too.