It does, but it is integrated in the voice coil and manifests itself as back emf. This provides an automatic feedback mechanism which linearizes the driver at its resonance frequency. Only works with voltage drive though.
many ways to do motional feedback
I have one group wanting me to disclose an approach, but without disclosing to me what they want it for, they at least said it was NOT for reduction of (nonlinear) distortion. But the technique I have in mind either requires a significant design detail be incorporated into a given driver, or relies on specific types of transducers, specifically with overhung VCs. It was an idea that came out of some small drivers designed primarily for laptops (initially) and desktop speakers, and although there may be some devilish details, should work quite admirably. However, without applicability in general I don't think it is likely to be executed. And someone said (sorry it's so far back, when this came up the last time in here, I have forgotten who) that when I mentioned the "plant model" approach of Klippel's, supposedly applicable to any transducer, that he had heard a demonstration and it did not sound good at all. I suppose it may have suffered from trade show demo-itis.
Although what I have in mind could be done in the analog domain, it cries out for decent realization at least in part in the numeric domain. And once you're there you're apt to forget where the door out is.
I have one group wanting me to disclose an approach, but without disclosing to me what they want it for, they at least said it was NOT for reduction of (nonlinear) distortion. But the technique I have in mind either requires a significant design detail be incorporated into a given driver, or relies on specific types of transducers, specifically with overhung VCs. It was an idea that came out of some small drivers designed primarily for laptops (initially) and desktop speakers, and although there may be some devilish details, should work quite admirably. However, without applicability in general I don't think it is likely to be executed. And someone said (sorry it's so far back, when this came up the last time in here, I have forgotten who) that when I mentioned the "plant model" approach of Klippel's, supposedly applicable to any transducer, that he had heard a demonstration and it did not sound good at all. I suppose it may have suffered from trade show demo-itis.
Although what I have in mind could be done in the analog domain, it cries out for decent realization at least in part in the numeric domain. And once you're there you're apt to forget where the door out is.
You would find out that what the voice coil is doing, is often very different from what the cone is doing. Or parts of the cone are doing.
Below a few 100 Hz it could work, but so would motional feedback with the driver in a bridge to extract the back EMF. Cheaper probably.
Jan
Philips had a small piezo-element in the dust cap.
Bu you can do it purely electronic too. Put the driver in a balanced bridge and you can extract a signal that represents the movement/EMF; you can use that for motional feedback.
VERY diy-friendly!
Jan
Last edited:
Richard,
So glad to hear you have water for the dispenser and ice maker!(A great uncle was one of the two fellows who laid out the route for the dam at boulder to L.A.)
Canyon,
Attached is a picture of the heatsinks being used. One per 12706 device. The fan attaches directly to the fins. It was designed to cool a pentium so the 20 watts each one sees doesn't appear to be an issue. At 200 watts it is designed to keep the contact surface be under 47C.
Now the use of a linear power supply is actually a bit much. All I really need is a center tapped transformer and pair of low Vf rectifiers. I use the devices well below ratings at 6 volts. They seem quite happy there.
But by all means do let me know where I am foolish. (Not counting spending time here.)
Also attached is my basement lighting under construction.
We use what is called an ACR test to determine if we have damage a TEC during a build (all eutectic - even do this test after soldering leads). The test consists of a full voltage swing + and - tem oscillating tem times and then we measure the impedance. Whish should be on the data sheer 2 to 5 ohms. this is repeated ten times. If the impedance varies by 5% it is a fail.
If you want it to last more then a couple years, you need to use a linear PSU. or a switch with mega filtering. I would use those nice reflex condensers on the cold side and a water cooled cold plate on the hot. Tap water is pretty constant. You just need a trickle of flow, it's amazing. A dT of 47 degs coupled with the Q on the cold side, you will be running these pretty hard. I would want to limit the dT on the cold side, the inefficient side. You will need a fan on the inside, no?
Another thing you need to worry about is a control. Always put your sensor on the surface of the TEC, they will and do run away with slow feed back loop. Then bias the temp so the fin will be where you want to cooled ambient.
Once you hit thermal run away there is nothing you can do but shut the system off. reversing the polarity will make it hotter until the TEC melts. They make almost perfect Kelvin heaters.
If you are going to [put the time in, do the math, TEC system design on first order usually fail.
Canyon
I suspect the data given for these very low cost peltier devices is just copied from quality units. The current draw is quite non linear. The 12707 really draws 4 amps at 12 volts and 2.8 amps at 6 volts.
Making measurements in use with an ir thermometer shows a temperature at the hottest spot to be 39 C. The coolest spot on the fan cooled heat sink is 27 C. The cold side runs 4 C. This is four devices running at 6 volts or 91 watts total.
I use 4 devices each with the fan cooled 6 heat pipe copper heat sink shown. One heat sink per device. One fan per heatsink.
I do have enough devices so I can test them for delta r and use just the least bad.
Comments welcome.
I suspect the data given for these very low cost peltier devices is just copied from quality units. The current draw is quite non linear. The 12707 really draws 4 amps at 12 volts and 2.8 amps at 6 volts.
Making measurements in use with an ir thermometer shows a temperature at the hottest spot to be 39 C. The coolest spot on the fan cooled heat sink is 27 C. The cold side runs 4 C. This is four devices running at 6 volts or 91 watts total.
I use 4 devices each with the fan cooled 6 heat pipe copper heat sink shown. One heat sink per device. One fan per heatsink.
I do have enough devices so I can test them for delta r and use just the least bad.
Comments welcome.
Last edited:
When I used Peltiers for the final cooling of a detector and preamp front end for a spectrometer, the thermistor for feedback was as close as I could get it to the cold side---and this was done to maximize the speed of the stabilization loop. The heat exchanger used a pump and ethanol as the working fluid, with the other end of the system being copper cooling coils immersed in an isopropyl/dry ice slurry in a big stainless dewar. It required fairly frequent additions of dry ice. A closed cycle refrigerator would have been nice.
Once during lab tests of the system, I forgot to turn off the controller when the pump was turned off. It was a meltdown of course, but the Peltiers were not damaged---just one wire desoldered and stopped the current. The thermal insulation did melt back a lot and had to be replaced. Messy but could have been a lot worse. I added an overtemp shutdown after that.
It's hard to imagine how a well-filtered class D controller would screw things up, but it was not considered for this application as we were trying to see very tiny charge signals, and the less switchmode crap around the better. Efficiency was not high on the list of concerns, although turbulence from heat sources in the telescope dome was also undesirable.
Once during lab tests of the system, I forgot to turn off the controller when the pump was turned off. It was a meltdown of course, but the Peltiers were not damaged---just one wire desoldered and stopped the current. The thermal insulation did melt back a lot and had to be replaced. Messy but could have been a lot worse. I added an overtemp shutdown after that.
It's hard to imagine how a well-filtered class D controller would screw things up, but it was not considered for this application as we were trying to see very tiny charge signals, and the less switchmode crap around the better. Efficiency was not high on the list of concerns, although turbulence from heat sources in the telescope dome was also undesirable.
And I just use a large thermal mass and limit the power per device. The cold plate is 1/2" aluminum and the thermostat is inside the drawer top center.
It was interesting to run a test with the unit powered from 4 volts to 12 volts and using a small 45"**2 aluminum plate as the thermal load (room temperature). Increasing the voltage lowered the temperature up to 8 volts then it leveled off rather dramatically. You could watch the increasing internal resistance and loss of efficiency.
It was interesting to run a test with the unit powered from 4 volts to 12 volts and using a small 45"**2 aluminum plate as the thermal load (room temperature). Increasing the voltage lowered the temperature up to 8 volts then it leveled off rather dramatically. You could watch the increasing internal resistance and loss of efficiency.
Last edited:
BarleyH20, Ok so I got the roomEQ wizard....
How do I calibrate it from a laptop. Only
one output that goes to the pre amp.
So the lap top mic is picking up sound....not ideal but I only have
one input/output, that is that littel mini type jack on the DELL.
I've gone through the manual a few times....
You would be surprised at how far down the food chain folks are considering motional feedback.
3rd time isn't a charm either. If I do it again, I'll go deaf.
Trying to calibrate it match the number, the software won't
take it and wants notes etc.
There is a noise type tones.
Then there is a freq type tone...my ears are still bleeding from.
you can't enter data stopped or it reverts to something else,
at least that is what software says.
Maybe it will let me post a waterfall.
Not sure how to read these:
Trying to calibrate it match the number, the software won't
take it and wants notes etc.
There is a noise type tones.
Then there is a freq type tone...my ears are still bleeding from.
you can't enter data stopped or it reverts to something else,
at least that is what software says.
Maybe it will let me post a waterfall.
Not sure how to read these:
Last edited:
What sorts of things Scott? I'm almost afraid to ask.
This reminds me of a Chinese person with whom I went to a local restaurant many years ago, and who had almost become a restaurant owner himself in NYC before pursuing a successful career as a choral conductor in Europe. He thought some of the dishes weren't too bad, but wouldn't eat some after tasting because they were too sweet (he had a rapid exchange in Mandarin with the waiter about it, who agreed that it was catering to "American taste" and was a shame).
When the mu shu pork came out with the side of hoisin sauce, he said "Come out of can. Good to know, and not do".
Can't say but some of the most ordinary computer speaker manu's are considering embedding things into the magnet structure. My last trip to China completely reoriented my idea of the food there. You can't explain it you can only share it.
Last edited:
Unfortunately Australia has a thing about sweetness, and blandness - when a new Asian restaurant opens, get in fast before the feedback from the local crowd gets them to steadily tone it all down, turning it towards pap ...
A young Indian couple started a restaurant in a nondescript cafe setting, locally - the best Indian I have ever had - of course, that place didn't last long ... far too much flavour, and freshness - down under can't handle that,.
A young Indian couple started a restaurant in a nondescript cafe setting, locally - the best Indian I have ever had - of course, that place didn't last long ... far too much flavour, and freshness - down under can't handle that,
.
down under can't handle that,
They seem to handle the obnoxiously over the top wines well enough.
D'ARENBERG THE DEAD ARM poop shoot you have to be kidding.
Last edited:
Just like your audio musings you need to zero in on the flaws in the face of all the praise.
- Status
- Not open for further replies.