| Author |
Message |
    
Bill Moore
Intermediate Member
Username: Bill_beerman
Post Number: 317
Registered: 10-2003
| | Posted on Saturday, May 21, 2005 - 05:15 pm: | 
|
I'm more concerned about long term affect on accuracy. |
ScottDeW
Intermediate Member
Username: Scott
Post Number: 309
Registered: 03-2003
| | Posted on Saturday, May 21, 2005 - 06:06 pm: |
|
I've done it with mine for several years now. No problems.
If I'm measuring wort or some other sticky liquid, I'll give it a good cleaning before putting it away.
Scott
http://texanbrew.com
|
Marlon Lang
Intermediate Member
Username: Marlonlang
Post Number: 487
Registered: 03-2003
| | Posted on Sunday, May 22, 2005 - 04:49 am: |
|
Bill,
My son Scott is absolutely correct.  If you wish, you can immerse the naked wire directly into wort, hot liquor, or whatever. Won't affect a thing. Type K is an alloy of chrome and aluminum. |
Richard Nye
Advanced Member
Username: Yeasty_boy
Post Number: 703
Registered: 01-2004
| | Posted on Sunday, May 22, 2005 - 01:39 pm: |
|
I heard it would give you alzheimer's. |
Marlon Lang
Intermediate Member
Username: Marlonlang
Post Number: 490
Registered: 03-2003
| | Posted on Monday, May 23, 2005 - 01:14 am: |
|
Aw, come on "Yeasty_boy".  Everybpdy knows that Alzeimer's is caused by hair on the palm of your hand.  |
Wortgames
Member
Username: Wortgames
Post Number: 136
Registered: 06-2003
| | Posted on Monday, May 23, 2005 - 09:56 am: |
|
Wouldn't liquid, which is conductive, affect the resistance of the thermocouple and therefore the reading?
Melbourne Australia - the other deep south
|
ScottDeW
Intermediate Member
Username: Scott
Post Number: 310
Registered: 03-2003
| | Posted on Monday, May 23, 2005 - 12:19 pm: |
|
No, the voltage produced is at the junction of the two metals. It is influenced by the types of metal wire used and the temperature at the junction.
Many thermocouples are sheathed in metal jackets. These keep the wires from direct exposure to the liquid. While probably better for long-term and constant use, I've found type K to stand up to wort pretty well.
Occasionally I get some wacky readings and have to snip off an inch or two and form a new junction by twisting the wire together. I can almost always attribute this to poor cleaning of the junction.
Of course, I did burn up one of my wires this weekend by letting it hang too close to the burner. Thankfully, the wire is cheap and I've plenty.
Scott
http://texanbrew.com
|
Wortgames
Member
Username: Wortgames
Post Number: 138
Registered: 06-2003
| | Posted on Monday, May 23, 2005 - 12:32 pm: |
|
So the junction actually generates a voltage? Very cool.
I recently bought a supposedly 'accurate' digital thermometer with a type K thermocouple and it seems to be miles off - are there different 'grades' of type K or are they basically all the same?
Melbourne Australia - the other deep south
|
Fredrik
Senior Member
Username: Fredrik
Post Number: 2253
Registered: 03-2003
| | Posted on Monday, May 23, 2005 - 01:46 pm: |
|
The voltage generated by thermocouples are small around 2.4 mV @60C and the sensitivity is about 41 uV/degC.
In my experience the two most common issues with accuracte thermocouple measurements is cold junction temperature correction errors and various noise. The quality of the amplifier will much affect the stabiltiy and noise too.
A thermocouple measurement has to account for the temperature also at the terminal to your amplifier, because in addition to the voltage generated by the termocouple junction you get two addition "termocouples" at your screw terminal. Say alumel-steel + chromel-steel. This skew the reading depending on the terminal temperature. Thermocouple amplifiers should have a temperature sensor to measure the temperature of the terminal junction plate. This is then used to correct for the cold-junction-temperature. Any error here, will carry over into your total accumulated measurement error.
/Fredrik |
Wortgames
Member
Username: Wortgames
Post Number: 141
Registered: 06-2003
| | Posted on Tuesday, May 24, 2005 - 12:12 am: |
|
So my 'brand new' (ebay) $85 Extech digital thermo that reads more than 2C lower than my most reliable thermometers at mash temps is buggered?
(Message edited by wortgames on May 24, 2005)
Melbourne Australia - the other deep south
|
Marlon Lang
Intermediate Member
Username: Marlonlang
Post Number: 492
Registered: 03-2003
| | Posted on Tuesday, May 24, 2005 - 01:52 am: |
|
Wortgames,
Not necessarily. I have several "yellow-back" lab thermometers that are 2-4 degrees off. Best thing to do is to imitate Lord Kelvin. Put both in an ice bath - should be close(within 1F)to 0C/32F. Put both in boiling water - should be close to 100C/212F (adjust for elevation)
But as my son Scott sez, wire is cheap. |
Wortgames
Member
Username: Wortgames
Post Number: 142
Registered: 06-2003
| | Posted on Tuesday, May 24, 2005 - 04:54 am: |
|
Thanks Marlon, I'll give that a try and see what I get.
Melbourne Australia - the other deep south
|
Fredrik
Senior Member
Username: Fredrik
Post Number: 2256
Registered: 03-2003
| | Posted on Tuesday, May 24, 2005 - 06:05 am: |
|
Depending on what is causing the offset - if the CJC is out of proper function, you might see that the offset varies with temperature.
If you want to put it to a harder test, you could while having the probe in an ice-bath - slowly warm your *instrument* say 5-10C and see if you get drift. A good instrument should account for most of this.
As a quick test you can also shortcircuit the thermocouple amplifier at the input using a metalwire, you should get the temprature reading of your instrument, as measured by the internal sensor. This value should change accordingly as you heat the instrument. If you have 1 degC error in this compensation, this will translate right off into 1 degC error in your actual reading.
But chances are good that you can calibrate it away like Marlon says, but while you'r at it, U would check that this calibration is valid if you heat your instrument. You can calibrate anything, but the question is if the calibration holds. For example my cheap pH meter, I have to calibrate it *every single time* just before usage. It is not stable. Clearly this meter would be inappropriate for long term measurements, like for example monitor fermentation temperature during 5 days.
From what I recall from work, some instruments at least a few years ago, did a single CJC reading at the start of the test, or at intervals, but if the instrument temperature changes during the test, or within the intercals, you'd get errors.
A good amplifier would continously compensate for the CJC temperature.
/Fredrik |
Wortgames
Member
Username: Wortgames
Post Number: 143
Registered: 06-2003
| | Posted on Tuesday, May 24, 2005 - 08:21 am: |
|
Thanks Fredrik. I haven't done the ice bath thing yet, but like any good hobbyist I've taken it apart.
The probe connections are inside a small (ceramic?) block which feels noticeably cold to the touch, and there does appear to be a sensor in the centre of this block.
I haven't noticed any way to actually calibrate it, so I can only assume that it would be done through some cryptic combination of keypad buttons. This will require reading the instruction book. Damn.
Before I do anything drastic like that though I'll spend some time with it and a few other thermometers to try and pinpoint the inaccuracies. I just tried it against a cheapo digital 'medical' thermometer (which I would guess is the most accurate one I have given its narrow range) and it was extremely close - so I'm startng to suspect my other thermometers.
Are all K-type probes created equal, or should I try and find some NASA spec wire? Would a J-type probe be better? (the meter will take either).
Ultimately I bought this thing to be my temperature 'point of truth' for calibrating other thermometers, so anything I can do to maximise its reliability would be helpful.
Apologies for the blatant threadjacking BTW. 
Melbourne Australia - the other deep south
|
Fredrik
Senior Member
Username: Fredrik
Post Number: 2257
Registered: 03-2003
| | Posted on Tuesday, May 24, 2005 - 10:28 am: |
|
> The probe connections are inside a small (ceramic?) block which feels noticeably cold to the touch, and there does appear to be a sensor in the centre of this block.
This sounds like it should. Often the CJC sensor is an NTC resistor or something similar immersed in a heat sink along with the screw terminal to make sure the sensor temp is the same as the actual terminal temp. Without the sink you'd get errors. These are cheap and have excellent signal/noise ratio, but the problem is that these NTC resistors (unlike pt100 and thermocoupes wich are fairly linear for practical purposes) are *highly* nonlinear thus the compensation is only valid in a relatively narrow range just to cover normal ambient temperature conditions only.
The sensitivity of the type J is around 30% higher than type K, so it should give slightly better signal/noise ration but other than that, type K is more resistant to corrosion than type J. The linearity in 50-70C is better for type K than for type J. But most dedicated thermocouple amplifiers will linearize the signal and this will be less of an issue. In the 50-70C range the linearity deviation for typeK is around 0.7% and 1.5% for typeJ.
I would personally prefer type K though for these reasons.
/Fredrik |
Wortgames
Member
Username: Wortgames
Post Number: 144
Registered: 06-2003
| | Posted on Tuesday, May 24, 2005 - 12:39 pm: |
|
Awesome! Thanks Fredrik!
My temp controllers use PTC and NTC probes - how do these compare with a type K? Is it reasonable to use a type K meter to calibrate the controllers or would that be like using a candle to calibrate a stopwatch?!
Melbourne Australia - the other deep south
|
Fredrik
Senior Member
Username: Fredrik
Post Number: 2258
Registered: 03-2003
| | Posted on Tuesday, May 24, 2005 - 01:09 pm: |
|
Wort, PTC and NTC have excellent signal to noise ratio and are truly cheap but are typically very non-linear, at least compared to thermocouples and RTD's like pt100. I can't speak for your sensors, or controllers(if the linearized or not).
*nonlinear* generally means that you can be correct at say 0C and 100C but still be way off in the midrange.
OTOH, If you have a decently linear sensor like, thermocouples or pt100 sensor, and you calibrate at 0C and 100C or whatever, then you can be decently assured (unless something is really wrong) that it will also measure correct along the way.
However, some are decent and some are really bad NTC sensors. So I would look into the documentation of your controller to be able to tell. I would assume that if anyone makes use of the NTC technique they would make sure it's decent.
Calibrating a non-linear sensor in a narrow range is easy and you can attain excellent accuracy this way, but to have it accurate over a larger range you typically need more complex polynomial and multipoint calibrations.
I once tried to calibrate a NTC sensor on the 0-100 range against another thermometer, of the type that in some cpu-boards sit in your cpu-card to probe cpu-temperature. This was a damn hassle. I think I tried a 10 point calibration and it still wasn't good. There was just no linearity at all, but awesome sensitivity.
What is worse, is that no two of these damn cheap NTC's where the same. I could do a 10 point calibration on one sensor, and then replace it with another one (same brand/model) and it was way off.
/Fredrik |
Fredrik
Senior Member
Username: Fredrik
Post Number: 2259
Registered: 03-2003
| | Posted on Tuesday, May 24, 2005 - 01:24 pm: |
|
Good signal/noise ratio also means you can get away with less amplification (usually cheaper) which is the advantage, and the noise problem is moderate.
Thermocouples OTOH, require a more sensitive amplifier since the signal is just some mV, and to not get craploads of noise this gets more expensive, and sometimes you need filtering too.
I have a simple amplifier not suitable for thermocouples so I use the single chip temperature sensors, which has the price/sensitivty or NTCs but are linear. It's not a simple resistor depending on temperature like NTC/PTC, it's a semiconductor device. Mine has three legs, power, ground and output. In many applications, like cheap temperature monitoring ciruites, the single-chip sensors are probably replacing the NTC's in many cases.
This works very well for me.
If you make sure that thermocouple meter is good and calibrated chances are it would be better than your NTC stuff, at least potentially superior in linearity which is the most important factor. Offsets are easy to adjust for anyway.
/Fredrik |
ScottDeW
Intermediate Member
Username: Scott
Post Number: 312
Registered: 03-2003
| | Posted on Tuesday, May 24, 2005 - 01:54 pm: |
|
Put the wire in ice water. Check for 0C (32 F).
Put the wire in boiling water. Check for 100C (212F).
Check mid-range calibration with other thermometers (consensus approach).
Have a beer.
Scott
http://texanbrew.com
|
Wortgames
Member
Username: Wortgames
Post Number: 145
Registered: 06-2003
| | Posted on Wednesday, May 25, 2005 - 04:09 am: |
|
Thanks guys, I think I now have a pretty good grip on the subject!
In boiling water I get 99.5C and in ice/water I get 0.9C, which I guess will have to be good enough (bearing in mind the agreement I get with the fever thermometer). So it looks like a couple of thermometers that I had here-to trusted are actually wrong - they read high by a degree or two. Dang.
However, I hereby name this meter God, and all other meters shall be judged by Him.
One more question - presumably with such tiny voltages the effect of volt drop over the length of the wire would be significant? Is there a maximum / optimum wire length?
Way ahead of you on the beer Scott!
Melbourne Australia - the other deep south
|
Fredrik
Senior Member
Username: Fredrik
Post Number: 2260
Registered: 03-2003
| | Posted on Wednesday, May 25, 2005 - 09:22 am: |
|
Usually Voltage drop problems has more to do with currents, rather than with low voltages and usually occurs on the excitation lines (thermocouples have no excitation line), and not the signal lines since the amp input impedance usually exceeds the wire resistance by orders of magnitues anyway. If the current is minimal, like in signal lines, the voltage drops is usually neglectable. On the signals lines noise and EM interference are usually more of an issue especially when the voltage is small. This is why thermocouples are often twisted to reduce some of these effects and sometimes shielded too.
The amplifier is high impedance and take only an absolute minimum amount of current. A low impedance amplifier would might give incorrect drained readings by draining power from the termocouple (which it isn't designed for).
Only you wire impedance starts to go up significantly as compared to the input impedance on your amplifier, then you may get some effects, but these are usally quite small unless your cable impedance and cable lenght is outrageous.
Some thermocouple wires are very thin (low thermal mass) to get quick response times, and are thus a littler higher in impedance. In these cases where you want really long wires, thicker extension cables are sometimes used.
So shorter cable is always better but if you have a couple of meters, this is not an issue. My guess is main issue would be noise/pickup if any.
When using 1-2 m cable to my temp-sensor (single-chip sensor), with noise reduction I get around +/- 0.05 degC in noise after noiser reduction.
With the same cable, but almost 10 meters, no shielding, all the way from the Pc out into the kitchen where I brew, from what I recall from the top of my head I get more than 0.5-1 degC noise after the similar noise reduction. I will later on get a better and shielded and twisted cable for this application and use this to log mash temperature and I think it will improve alot.
/Fredrik |
Wortgames
Member
Username: Wortgames
Post Number: 147
Registered: 06-2003
| | Posted on Wednesday, May 25, 2005 - 12:06 pm: |
|
Good on you Fredrik, thanks again for the crash course - I hereby rename my most accurate meter Fredrik in your honour 
Melbourne Australia - the other deep south
|
Fredrik
Senior Member
Username: Fredrik
Post Number: 2261
Registered: 03-2003
| | Posted on Wednesday, May 25, 2005 - 12:32 pm: |
|
Hehe  My pleasure to be of any help. Considering the amount of questions I asked here in the past and all of you responded to them I am glad to pay back.
/Fredrik |
Brews and Views Archive 2005 
Closed: New threads not accepted on this page