| Author |
Message |
|
|
 Posted: Jun 09, 2012 - 07:13 AM |
|
Joined: Nov 24, 2011
Posts: 32
|
|
I am attempting to measure the amount of time between the initial impact of a bullet and when an accelerometer "feels" the impact. I am trying to figure out which MCU can handle this type of precision with timing.
I am worried about timing because the propagation of sound through generic steel is roughly 6100 m/s. The size of the object I will be shooting at will be at most 18 inches.
So my questions are:
Is there an AVR MCU than can handle these timing restraints? If not what should I use?
Also what measurement do I look for when browsing different MCUs? Bandwidth, resolution, etc.
Thanks |
|
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 07:35 AM |
|
Joined: Jan 03, 2006
Posts: 4417
Location: Hemel Hemsptead, UK
|
|
I'd suggest that this is not in the initial stages a job for a microcontroller but rather a high speed storage oscilloscope.
You don't say how you're detecting the initial impact though your mention of the speed of sound suggests that you're considering a microphonic approach; equally, you don't indicate how the accelerometer will be attached.
The acoustic pickup will 'hear' the impact at a time determined by its distance from the strike point, so between zero and 1/(6000/0.5) seconds - about 85 microseconds - after the impact. However, it will also hear internal reflections of the sound, and probably the sound transmitted through air. This could be a very complex waveform...
The accelerometer though is also constrained by speed of sound; as far as I understand that's as fast as you can deform the target material. So that's also likely to have a complex waveform and one which will depend on the mass as well as the material from which the target is formed.
Without more information about the mechanical setup and in particular the waveforms from the sensors I don't think you're yet in a position to start considering the measurement technology. I'd start with a high speed storage oscilloscope to look at the waveforms and their relative timings. I might also consider detecting the bullet in flight - at 200m/s to 400m/s this should be reasonably easily done - and calculating the impact time from that. |
_________________
Neil Barnes
www.nailed-barnacle.co.uk
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 07:38 AM |
|
Joined: Nov 02, 2009
Posts: 3239
Location: Zelenograd, Russia
|
|
~0.5 m / 6700 m/s = ~ 75 uS, so using any 20 MHz AVR with timer input capture module you get 1500 clocks per 75 uS which means at least 0.1% resolution which is ~ 0.3 mm per single clock tick. Is this enough for your application?
BTW in any case "incredibly fast" are picoseconds, but not tens of microseconds at all  |
_________________
Warning: Grumpy Old Chuff. Reading this post may severely damage your mental health.
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 07:41 AM |
|
Joined: Nov 24, 2011
Posts: 32
|
|
|
barnacle wrote:
I'd suggest that this is not in the initial stages a job for a microcontroller but rather a high speed storage oscilloscope.
You don't say how you're detecting the initial impact though your mention of the speed of sound suggests that you're considering a microphonic approach; equally, you don't indicate how the accelerometer will be attached.
The plan is for the accelerometers to be attached with a screw style mount. I haven't considered microphonic approach as the accelerometer seemed easier.
The acoustic pickup will 'hear' the impact at a time determined by its distance from the strike point, so between zero and 1/(6000/0.5) seconds - about 85 microseconds - after the impact. However, it will also hear internal reflections of the sound, and probably the sound transmitted through air. This could be a very complex waveform...
The accelerometer though is also constrained by speed of sound; as far as I understand that's as fast as you can deform the target material. So that's also likely to have a complex waveform and one which will depend on the mass as well as the material from which the target is formed.
Well the waveform should have one peak followed by slowly decaying oscillations which would be the steel reverberating. I planned on simply capturing the peak. and using that as the identification of an impact event.
Without more information about the mechanical setup and in particular the waveforms from the sensors I don't think you're yet in a position to start considering the measurement technology. I'd start with a high speed storage oscilloscope to look at the waveforms and their relative timings. I might also consider detecting the bullet in flight - at 200m/s to 400m/s this should be reasonably easily done - and calculating the impact time from that.
How would we detect it in flight? |
|
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 07:43 AM |
|
Joined: Nov 24, 2011
Posts: 32
|
|
|
MBedder wrote:
~0.5 m / 6700 m/s = ~ 75 uS, so using any 20 MHz AVR with timer input capture module you get 1500 clocks per 75 uS which means at least 0.1% resolution which is ~ 0.3 mm per single clock tick. Is this enough for your application?
BTW in any case "incredibly fast" are picoseconds, but not tens of microseconds at all
So from your calculation an AVR should be able to handle this? What type of module would you recommend? |
|
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 07:45 AM |
|
Joined: Nov 02, 2009
Posts: 3239
Location: Zelenograd, Russia
|
|
| I recommend the AVR datasheet type of module |
_________________
Warning: Grumpy Old Chuff. Reading this post may severely damage your mental health.
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 12:27 PM |
|
Joined: Jul 18, 2005
Posts: 62299
Location: (using avr-gcc in) Finchingfield, Essex, England
|
|
|
Quote:
So from your calculation an AVR should be able to handle this? What type of module would you recommend?
It might just but surely resolution (and overhead) are key here? You can get some Cortex that hop along at 120MHz or even 150MHz. That'd give you far more headroom for about the same silicon cost. |
_________________
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 12:58 PM |
|
Joined: Sep 04, 2002
Posts: 21258
Location: Orlando Florida
|
|
| Can you clear up the block diagram for me? I think it sounds like: an 18" steel plate hanging in the ballistic range lane with an accelerometer attached. If the bullet is 1" long and travelling 1000 ft per sec (12K inches per sec) it would take 83usec for the bullet to pass through the target (if the target was a piece of paper for example). Since the bullet probably splats all its momentum into the steel plate instead, I think the impulse detected by the accelerometer is m of bullet x dv/dt. We know the dv is 12K inches per sec and the dt is 83usec. When the accelerometer detects the impact, we know the bullet hit 83 usecs earlier. Useful? |
_________________
Imagecraft compiler user
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 08:20 PM |
|
Joined: Jan 03, 2006
Posts: 4417
Location: Hemel Hemsptead, UK
|
|
But surely the delay to the accelerometer is dependent on where the bullet strikes in relation to the sensor?
Detecting the bullet might be done in a number of ways: optical or acoustic are the obvious ones (http://arxiv.org/pdf/physics/0601102 is concerned with flight time of the bullet but is worth a read) but you might consider a sheet of conducting material close to the target which makes an electrical contact as the bullet strikes.
A microphone a known distance from a target that goes 'bong' would give an acoustic signature which can be largely independent of target strike and whose time delay can be accurately calculated and therefore adjusted for an accelerometer signal. But I think you're going to have to look at the signals themselves before you can decide what processing is required and therefore what hardware. |
_________________
Neil Barnes
www.nailed-barnacle.co.uk
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 08:25 PM |
|
Joined: Sep 04, 2002
Posts: 21258
Location: Orlando Florida
|
|
| I dont think the speed of sound in the steel plate has anything to do with it. The mv of the bullet is transferred to the plate, and it starts accelerating back the microsecond that the bullet hits. I assume the accelerometer output starts rising at its max rate, which might not be able to slew as fast as the plate is accelerating, but by golly it starts rising at the same microsecond the plate starts moving. Start the timer when the accel output starts moving! |
_________________
Imagecraft compiler user
|
| |
|
|
|
|
|
|
Posted: Jun 09, 2012 - 10:48 PM |
|
Joined: Dec 11, 2007
Posts: 6849
Location: Cleveland, OH
|
|
I, too, don't understand the physical setup.
A diagram of the target would help, how "deep" in the target?
Does the bullet bounce off the target, lodge within the target, or pass through the target?
With a "thin" metal disk, with the acceleromter mounted on the back side of the disk, the impulse time through the disk is calculatable. It would seem the response time of the accelerometer wuld be defined in the data sheet under Impulse Response.
JC |
|
|
| |
|
|
|
|
|
|
Posted: Jun 10, 2012 - 01:31 AM |
|
Joined: Jun 17, 2005
Posts: 472
Location: New Zealand
|
|
What is the real purpose of this excercise ?
Are you trying to measure the speed of a bullet or are you solely interested in the time lag of your average accelerometer ? |
|
|
| |
|
|
|
|
|
|
Posted: Jun 10, 2012 - 02:37 AM |
|
Joined: Nov 22, 2002
Posts: 12049
Location: Tangent, OR, USA
|
|
As pointed out, the time intervals you are talking about are NOT that fast.
The real question, as JimK asked, is effectively "What are you trying to determine?":
Is it delay between impact and arrival of the shock wave at the accelerometer?
Is it delay between shock wave at the accelerometer and accelerometer output?
Do you have reason to expect a big delay between shock and accelerator output?
As others have pointed out, there IS a delay between impact and accelerometer arrival, depending on where the impact occurs. It depends linearly on distance. Halve the distance, halve the delay.
I would expect the internal accelerometer delay to be fully characterized by its bandwidth with little or no internal transport delay. This can be measured by non-impact means.
Jim |
_________________
Jim Wagner
Oregon Research Electronics, Consulting Div.
Tangent, OR, USA
"The only thing standing between us and victory is defeat" P.G.Wodhouse in Wooster & Jeeves series
|
| |
|
|
|
|
|
|
Posted: Jun 11, 2012 - 05:24 AM |
|
Joined: Nov 24, 2011
Posts: 32
|
|
Hey everyone, the point of my experiment is to try to find where on a circular target a bullet hits based on the differential arrival time of the vibration wave between accelerometers arranged in a square or a inverse triangle. I am basing it off this project that used an expensive USB DAQ by National instruments.
Here is the math
https://decibel.ni.com/content/docs/DOC-6126
Here is the video.
http://bulletin.accurateshooter.com/201 ... n-targets/
Any more advice will be helpful - especially on picking an accelerometer and an mcu |
|
|
| |
|
|
|
|
|
|
Posted: Jun 11, 2012 - 12:48 PM |
|
Joined: Sep 04, 2002
Posts: 21258
Location: Orlando Florida
|
|
| Seems like this is a job for a simultaneous sampling hi speed a/d. Analog devices, Linear Tech make them. |
_________________
Imagecraft compiler user
|
| |
|
|
|
|
|
|
Posted: Jun 11, 2012 - 01:39 PM |
|
Joined: Sep 05, 2001
Posts: 2497
|
|
Choosing an mcu was the absolutely last part of this task.
At first you must decide, which sensors you use and how they represent the masuring value as an electrical signal.
Then you must decide the wanted measurement range (min, max value), the resolution and the accuracy.
Then you can select ad converters or other converters to get data for the mcu.
And only after this you can pick an mcu, which was able to perform the measuring task.
Peter |
|
|
| |
|
|
|
|
|
|
Posted: Jun 11, 2012 - 03:06 PM |
|
Joined: Nov 02, 2009
Posts: 3239
Location: Zelenograd, Russia
|
|
| A simple quad fast comparator chip is enough for conditioning the accelerometers' (or any other appropriate sensors') signals and converting them to discrete pulses. Any MCU with 3x input capture modules will easily do the rest of a job. |
_________________
Warning: Grumpy Old Chuff. Reading this post may severely damage your mental health.
|
| |
|
|
|
|
|
|
Posted: Jun 13, 2012 - 02:54 AM |
|
Joined: Nov 24, 2011
Posts: 32
|
|
|
MBedder wrote:
A simple quad fast comparator chip is enough for conditioning the accelerometers' (or any other appropriate sensors') signals and converting them to discrete pulses. Any MCU with 3x input capture modules will easily do the rest of a job.
Do you have any recommendations for comparators - I have never used them before. Also, by input capture module do you mean using inturupts? |
|
|
| |
|
|
|
|
|
|
Posted: Jun 13, 2012 - 02:55 AM |
|
Joined: Nov 24, 2011
Posts: 32
|
|
|
danni wrote:
Choosing an mcu was the absolutely last part of this task.
At first you must decide, which sensors you use and how they represent the masuring value as an electrical signal.
Then you must decide the wanted measurement range (min, max value), the resolution and the accuracy.
Then you can select ad converters or other converters to get data for the mcu.
And only after this you can pick an mcu, which was able to perform the measuring task.
Peter
Could you assist me and figuring out how to find a suitable resolution I have never done a project with such a small time base. A google link or anything would be of much assistance. |
|
|
| |
|
|
|
|
|
|
Posted: Jun 13, 2012 - 06:41 AM |
|
Joined: Nov 02, 2009
Posts: 3239
Location: Zelenograd, Russia
|
|
|
johnjkb wrote:
by input capture module do you mean using inturupts?
By input capture module I do mean using datasheet. |
_________________
Warning: Grumpy Old Chuff. Reading this post may severely damage your mental health.
|
| |
|
|
|
|
|
|
Posted: Jun 13, 2012 - 12:58 PM |
|
Joined: Sep 04, 2002
Posts: 21258
Location: Orlando Florida
|
|
| Have a look at simultaneous sampling a/d converters? |
_________________
Imagecraft compiler user
|
| |
|
|
|
|
|
FAQ
Search
Register
Log in to check your private messages
Log in
Maximize
RSS
