0
\$\begingroup\$

Being relatively new to the world of sound and vibration signal manipulation, I am attempting to drive a small vibe table in an open loop configuration. For preliminary trials (AKA learning how to use the darn thing) my planned method is to:

1.) Synthesize time history for my random vibration PSD

2.) Normalize this data in some way (centered about 0 and +/-1 amplitude..?)

3.) Write the data as analog voltage via NI 9263 + cDAQ

4.) Input into my DSI-1000 power amp

5.) ??

6.) Profit! (Successful acceleration response)

I'm still trying to wrap my head around the fundamentals of a power amplifier. Would I merely send my normalized data as voltages, then adjust the sensitivity dials on my PA until my accelerometer reads my desired level (and make note of that sensitivity setting)? Note, I do NOT have a controller, nor do I want to buy one. I believe that open loop control will be suitable for me.

I realize this is a vague thread, so I apologize. Is there someone who has experience with this that could shed light on what my process is missing?

PS: My power amp datasheet does NOT have this information included. It's pretty barebones

\$\endgroup\$
5
  • \$\begingroup\$ I wish: Power amplifier @EugeneSh. \$\endgroup\$
    – Austin Prater
    Jan 19, 2018 at 22:01
  • \$\begingroup\$ You should NOT send any waveform to your PA that is not a sine-wave or LF square wave. The best PA's roll off at 30 KHZ to 50 KHZ. \$\endgroup\$
    – user105652
    Jan 19, 2018 at 22:03
  • 1
    \$\begingroup\$ @Sparky256: Who listens to sinewaves? An LF squarewave's harmonics extend to infinity too. \$\endgroup\$
    – Transistor
    Jan 19, 2018 at 22:11
  • \$\begingroup\$ What set of resonant modes have you calculated for the table-actuator-DUT system? \$\endgroup\$
    – Spehro Pefhany
    Jan 19, 2018 at 22:18
  • \$\begingroup\$ Are you trying to test a product for HALT/HASS? \$\endgroup\$
    – Tony Stewart EE75
    Jan 20, 2018 at 1:13

2 Answers 2

Reset to default
1
\$\begingroup\$

Open Loop is easy. (I worked in vibration testing for over 15 years.) You are right, synthesize your PSD as a time waveform, normalize it, put it through a low pass filter (cut-off freq set to about 2-3 kHz - decided by your vibration table), feed it to the PA, adjust PA level control to get your average "g" level.

Most PA won't pass DC and will have a low cut-off of about 20 Hz, which could be a bummer, high frequency limit required depending on vibration table upper resonant frequency, if its electrodynamic, maybe 2-3 kHz. You don't want to unnecessarily excite the vibration table at its resonance.

Cheers

\$\endgroup\$
4
  • \$\begingroup\$ Thanks for the input @Transistor ! Good to know I'm generally on the right track. What determines how to normalize the data? Would I want to be centered about 0 with amplitudes +/-1, or do I offset by +1? Also, the PSD I'm wanting to generate is 20-2kHz so my ED shaker should be able to handle it. \$\endgroup\$
    – Austin Prater
    Jan 19, 2018 at 23:41
  • \$\begingroup\$ @AustinPrater Offset doesn't matter, since the PA doesn't pass DC. Normalization doesn't matter as long as you respect the PAs max input voltage. \$\endgroup\$
    – τεκ
    Jan 20, 2018 at 0:21
  • \$\begingroup\$ @Sofcrafts: I'm running into an issue where my time history over a span of 5 secs is upwards of 160,000 elements, meaning I need a system that can write at 32kHz. Does this sound correct to you? I can't imagine that my shaker could achieve such a rate (or even the PA's for that matter) \$\endgroup\$
    – Austin Prater
    Jan 20, 2018 at 1:48
  • 1
    \$\begingroup\$ @τεκ: I have only a basic understanding of this application but if I was doing it I would be looking at a DC coupled amplifier (with no input or output capacitors). The output could be single ended or bridge mode. So offset would matter. \$\endgroup\$
    – Transistor
    Jan 20, 2018 at 6:20
0
\$\begingroup\$

Power Spectral Density (PSD) or Acceleration Spectral Density (ASD) relies on the linear relationship between force, acceleration and current but in time and frequency the spectrum is measured in g2/Hz.

Thus band limited flat PSD between f1 and f2 of amplitude W has an overall acceleration \$G_{rms}\$ such that \$ G_{rms}=\sqrt{W(f_2-f_1)}\$

However if your shaker table has a resonance in the middle of your range, then you have to make a compensation filter.

enter image description here

\$\endgroup\$
2
  • \$\begingroup\$ Thanks Tony. I already have pretty good understanding of my desired PSD and the associated response. This thread is more-so to get a grasp on what the process is for turning that PSD into a useable signal for my power amps. \$\endgroup\$
    – Austin Prater
    Jan 20, 2018 at 1:25
  • \$\begingroup\$ OK .. We used to do PSD on MOBO's for HASS of SMD solder joints and before that used swept sine 10 to 5k and then HASS tests on cards in cage with 10Hz rubber bumper tests. \$\endgroup\$
    – Tony Stewart EE75
    Jan 20, 2018 at 2:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.