Power Designs 2020 Repair

[timb]

Ohhh! What about that stretchy, self-sealing silicon tape? I've used that shit to repair radiator hoses on cars before. A few wraps of that around the outside of the inner can followed by a wrap of cotton insulation.


Sent from my Smartphone

[motocoder]

Ohhh yeah, the stuff they insulate fireplace and oven doors with? Silica cord, I think it's called. It's what most eCigs use for wicking the juice; normally a small bit of nichrome or kanthal wire is wrapped around it to cause vaporization. It won't melt, so I bet you could just wrap the outer tube in a few layers of it.

Another idea I had was hemp. It's pretty resilient.

Finally, for the bit of insulation around the reference board inside the inner tube, why not just use cotton? I've got rolls of 100% pure USP cotton compresses that are nice and fluffy when pulled apart.

In fact, cotton gauze might work on wrapped around the outside too.

Did we ever figure out what temp the OEM thermistor trips at on these? (I can always heat this one up and get a number if still needed.)


Sent from my Smartphone

Hi Tim -

Please don't use cotton. When it burns it releases hydrogen cyanide gas.

My suggestion was actually w.r.t. the inner insulation. For the stuff outside the tube, here is what I used, and it works quite well: One layer of overlapping wraps of 3M glass cloth tape. Outside that is the exhaust wrap I mentioned earlier in the thread. Holding the exhaust wrap in place is some Kapton tape. This works great. In fact, it works too well. I'll probably skip the exhaust wrap in the next version. I did some research into PID loops for thermal systems, and I discovered that you want as little insulation as you can get away with. The reason is it causes a lot of thermal inertia which makes the control loop not so good (pretty obvious if you think about it).

OEM thermistor - there is no thermistor in the original version of either of the PD supplies I have. They both use some sort of mechanical (I assume bimetal) thermostat tucked away in the inaccessible part of that tube. The one that was working made quite a loud mechanical click when it would trip. Maybe they use this in newer models? I am using a thermistor in my replacement temperature control board. It is one I salvaged from a coffee maker. In the next version of this board (the one that will have a PID control loop), I am using this one:

http://www.vetco.net/datasheets/VUPN6551/TKS_TTC05102KS.pdf

I don't think people really know what temp these things were designed to work at. THe one I just blew up had a thermostat that was switching the current on and off, but it was not off frequently enough to keep the temperature under control. Instead it was just heating up to thermal equilibrium - same temp it would have gone to without a thermostat - about 90C (estimated, I actually measure 85C with the case off).

My replacement solid-state thermostat runs it at 49-50C. I don't see an benefit to running it at higher temps, do you?

[timb]

Interesting, information. Thermostat would have been a better word than thermistor; so I just ran some tests on the 2005A thermostat I pulled out last night:



So the bottom plate there is what attaches to the inner can under the socket; I attached my Fluke 87V’s thermocouple directly to one of the mounting holes, hooked another meter up to read the resistance and then slowly heated the top of the plate (with care not to blow hot air directly towards the thermocouple and at a rate to allow the plate and thermostat unit to reach equilibrium) with my hot air reflow gun set to 100c.

The thermostat stays around 0.200 ohms until about 65c, when it starts ramping up to 0.450 ohms, where it ultimately goes open circuit around 70c or just a tad over. It closes again around a tad under 70c.

70c seems about right to me for the inner can temperature. Another thing of note is there’s no audible clicking in my unit.

As for cotton, the fire point is 250c and autoignition around 400c; it starts to breakdown after longterm exposure above 150c. A lot of people use pure organic cotton as a wicking material in eCigs. (Though if the wick goes dry and burns it’s pretty much ruined. Plus the hydrogen-cyanide gas, albeit a tiny, tiny amount in this case.) Anyway, I’m thinking you’re right and cotton isn’t the best choice; wouldn’t want the thermostat to stick and it cause a fire!

I’m thinking if I wrapped the inner tube with that 3M Glass Tape, or perhaps the self-adhesive silicone tape I mentioned earlier, it would be an adequate insulator compared to the 50 year old glass wool, don’t you? I’ll take a look at that exhaust insulation you mentioned; maybe I can use that for inside the inner tube.

As for running at higher temps than 50c, if they *were* in fact running at 70c, than I’m guessing that’s the temp the reference parts were hand picked to optimally have the lowest drift at. That said, run it at 50c and measure the drift, if it’s within OEM specs, seems good to me.

Speaking of references, here’s pictures of the 2005A reference I pulled; the blue components there did have a black shell on them, but they literally crumbled off when I was cleaning the board last night…

[motocoder]

Agree with everything you say. The good thing is we can easily change the set point and try out different temperatures. Even with my current circuit, I can do that by swapping out a resistor (or adding a pot to the circuit).

There was one person in the large power designs thread that said he got better stability with a lower temp, so that is why I am using 50C now.

[timb]

Agree with everything you say. The good thing is we can easily change the set point and try out different temperatures. Even with my current circuit, I can do that by swapping out a resistor (or adding a pot to the circuit).

There was one person in the large power designs thread that said he got better stability with a lower temp, so that is why I am using 50C now.

[To clarify what’s below, I’ve got three PD power supplies: 2020B, 2005, 2005A and a 1025P (100V/250mA Remote Programmable) on the way from eBay. The 2005A is the one with a physically damaged oven mount I posted about above; the 2005 non-A is what I’m discussing below.]

So I had some time tonight to work on my 2005 tonight. It works fine overall, but I noticed the oven lamp never goes off. I just tried heating the outer can with my hot air gun (started at 100c and went up to 180c) and I couldn’t get it to trip, but maybe the heat just wasn’t getting to the thermostat and it’s the coil that’s not working? Next I’ll desolder the heating coil power wire and run it through my ammeter to see if it’s working.  If the coil is fine, I can rig up a replacement thermostat switch. This will mean another can to repack with new insulation!

Speaking of which, I was looking at this stuff.

Here’s my thought, I can cut a round piece for the top of the outer can, then cut a portion to wrap the inner wall. This, combined with wrapping the inner can with silicone or glass thermal tape should keep the vast majority of heat in, without needing to add insulation to the *inside* of the inner can. Though, to keep the reference board at a stable temperature, I was thinking of wrapping the board itself with silicone tape. It’ll take longer to warm up, but once it does it should keep the heat in for a long time.

Thoughts?

(By the way, I didn’t mean to hijack your thread! This seemed like a better place to discuss all things oven. If you want me to move to the main thread, let me know!)

Edit: Yeah, it’s got to be the coil. I’m getting 30MOhms (which slowly goes up) across the coil and thermostat contacts and open circuit between the coil and lamp contacts. (This is with all wires unsoldered from the three terminals.)

[timb]

Hmmm, have you looked into Ceramic Fiber Blankets? That looks very promising. I might order a foot and see how it is.

[motocoder]

timb, I think you're putting too much emphasis on the insulation. More is not better here.

Regarding the temperature, I can't say that this will be true on every model, but on my 2005, I get better stability with 50C than I do with 70C. Since it also uses less power, and heats my lab up less, that is what I'm going with.

Also, here is an update on my experiments with a PID controller. I tend to avoid anything "Arduino", because I find I can get better results using AtmelStudio and configuring the timers and what-not myself. For example, I was able to get several orders of magnitude better accuracy on an ultrasonic distance sensor by creating a very fast timer by keeping the interupt routine simple. On the Arduino, with the sloppy "millis()" functions and what-not, it simply wasn't working right.

However, for this project, I just wanted to try some experiments quickly without having to spend too much time coding and soldering, and for that the Arduino was perfect. I soldered up a small board with a power MOSFET, an NPN BJT to drive the gate (so that the MOSFET would be fully in saturation) and a few passive components to bias everything correctly and to provide a voltage divider with the thermistor. This was connected up to the Arduino such that it can PWM the MOSFET (heater) via one of its outputs and can read the thermistor value via one of the analog inputs. I also used a simple PID library that I found on the Arduino site. Other than the fact that the PID library uses floating point, it's not bad. It's really a pretty simple thing to do - the hard part is tuning the parameters.

The test set-up is a scrap tube from my broken heater with layers as follows: thermal epoxy (very thin), resistance wire, kapton tape, glass cloth tape, exhaust wrap, kapton tape. Inside the tube I have used some more of the thermal epoxy to bond a thermistor to the inside wall of the tube. This greatly lowers the latency between heat changes and works much better than a thermistor separated by an air and/or insulation gap. Also inside the tube, but separated by an air gap, is a thermocouple. The thermistor is used as the input of the control loop, but the thermocouple is used to measure what the actual temperature stability would be at the location where the voltage reference would sit.

So, I spent half of the day today playing around with PID parameter tuning. Lots of time can be spent on this, and I am sure with some more tuning, I could get something that settles out quicker, but what I have is pretty good. The thermistor voltage is steady within about 800uV at 421mV (0.2%) and the thermocouple voltage is steady within 1uV at 0.974mV (0.1%), which means temperature at the voltage reference would be stable to within 0.1%. At the set point I am using (50.2C), steady-state current draw with a 24VDC supply is around 100mA. It should be a little better once everything is sealed up.

So this is an order of magnitude better temperature regulation than my thermocouple + voltage comparator thermostat. At this point, the only open question I have is whether or not the PWM-ing of the heater current is going to add ripple to the output or impact the voltage reference stability. I think next weekend I'll go ahead and solder up a board and try it out in my 2005.

[motocoder]

I've attached a graph of the PID response (just PI actually, no D). This is using some constants generated by the PID Autotune program, and is after I modified the PID program to support setting a limit on the accumulated I value. These are not optimal, but probably usable for this application.

[timb]

Thanks for posting the results, very interesting!

Why do you spread thermal epoxy on the outside of the tube? I think a wrap of glass tape would provide the needed electrical insulation without the mess of thermal epoxy.

What kind of resistance wire are you using? I’ve got several sizes of A1 Kanthal Wire I was thinking of using.

timb, I think you're putting too much emphasis on the insulation. More is not better here.

Can you expand on that a bit? I imagine the more outer insulation you have, the more heat is kept in, the less power you use and the more stable your temperature will be.

Do you have any hysteresis on your PID? You could have the cutoff be, say, 51.2 and the turn-on be 49.2. This way you’re not constantly cycling.

As for PWM, personally, I wouldn’t. Just use a constant current to heat the wire up to a specific point and let it run. How are you getting the 24VDC? Are you tapping off the transformer?

Finally, here’s full specs on the oven of the 2005A that I’ve put together; hopefully it’ll be helpful to others looking to rebuild their ovens.

• 26 x 2 Wraps of Heating Wire
• ~95mm Inner Can Circumference
• ~16’ Total Wire Length
• 670 Ohms
• ~70c Thermostat Cutoff
• 180mA @ 120VAC

[motocoder]

Why do you spread thermal epoxy on the outside of the tube? I think a wrap of glass tape would provide the needed electrical insulation without the mess of thermal epoxy.

A very thin coat of thermally conductive epoxy conducts the heat from the wires to the surface of the tube much better than glass tape (which is really a pretty good thermal insulator), and even better than a single layer of kapton tape. You want the heat to get inside the tube as fast as possible, as that decreases the lag in your control loop.

What kind of resistance wire are you using? I’ve got several sizes of A1 Kanthal Wire I was thinking of using.

30 AWG kanthal. It is about 9 ohms per foot.

timb, I think you're putting too much emphasis on the insulation. More is not better here.

Can you expand on that a bit? I imagine the more outer insulation you have, the more heat is kept in, the less power you use and the more stable your temperature will be.

Because it introduces a non-linearity into the system. The rate at which the system heats up is controlled mostly by the power into the heater. The rate at which it cools down is controlled by the outer insulation. If you have too much insulation, and your control loop has any overshoot, it will take a very long time to cool back down. It will be impossible to tune the PID parameters optimally because of the large difference in heat rate versus cool rate.

Do you have any hysteresis on your PID? You could have the cutoff be, say, 51.2 and the turn-on be 49.2. This way you’re not constantly cycling.

I am not sure how hysteresis applies in a PID controller. A PID controller adjusts the manipulated variable (MV, heater power in this case) periodically or continuously to try and achieve a particular value for the process variable (PV, or temperature in this case). To do this, it multiplies the error (difference between the desired PV and the actual PV) by some constant , the accumulated value of this error by another constant, and the rate of change of this error or of the PV (the "D" term) by yet another constant. It then sums it up as P + I - D (hence the name).

I have hysteresis in my simple solid-state thermostat. However, the large lag in heat transfer between the coils and the thermistor make it impossible to achieve better than about 1C of drift with this approach.

As for PWM, personally, I wouldn’t. Just use a constant current to heat the wire up to a specific point and let it run. How are you getting the 24VDC? Are you tapping off the transformer?

Yes, that is my concern, but let's try it and see. Using constant current means I will have to dissipate significant heat in the MOSFET or BJT controlling the heater coil, which probably means putting a heat sink on it. Would rather not do that. It also means I have to find some way to get an analog voltage out of the 8-bit micro (AVR) I am using.

I have some more results to share and some interesting findings regarding the PID loop, but it's getting late here so I'll post those tomorrow.

[timb]

Why do you spread thermal epoxy on the outside of the tube? I think a wrap of glass tape would provide the needed electrical insulation without the mess of thermal epoxy.

A very thin coat of thermally conductive epoxy conducts the heat from the wires to the surface of the tube much better than glass tape (which is really a pretty good thermal insulator), and even better than a single layer of kapton tape. You want the heat to get inside the tube as fast as possible, as that decreases the lag in your control loop.

I gotcha, makes sense. I actually found a huge roll of NOS double glass insulated nichrome wire on eBay for cheap, so I’ll be using that. (I can send you a bunch if you or anyone else wants it.)

Do you have any hysteresis on your PID? You could have the cutoff be, say, 51.2 and the turn-on be 49.2. This way you’re not constantly cycling.

I am not sure how hysteresis applies in a PID controller. A PID controller adjusts the manipulated variable (MV, heater power in this case) periodically or continuously to try and achieve a particular value for the process variable (PV, or temperature in this case). To do this, it multiplies the error (difference between the desired PV and the actual PV) by some constant , the accumulated value of this error by another constant, and the rate of change of this error or of the PV (the "D" term) by yet another constant. It then sums it up as P + I - D (hence the name).

I have hysteresis in my simple solid-state thermostat. However, the large lag in heat transfer between the coils and the thermistor make it impossible to achieve better than about 1C of drift with this approach.

Right, I suppose I meant just using a simple on/off control with hysteresis, like a solid state controller. You won’t get better than 1C of drift, but it’s simple.

Since you’re going for absolute temperature accuracy and stability, you might think of adding a thermistor attached to the middle of the outside of the inner tube. Hold it down with a single wrap of glass tape.

You can do some measurements from this sensor and the one inside the inner can, plot the data out over time. If you overlay the two graphs, you should get a very clear view of the temperature differential. You can add this as either a constant to the PID’s error, or keep the second sensor installed and use it as a dynamic variable in the PID’s error.

I’ve done something similar to this with multiple thermistors when I was designing the controller for a chicken egg incubator unit (basically a Kanthal wire based heater and 120MM fan that the user installs into refrigerator or other insulated enclosure.

Eagerly awaiting your latest results! =D

[motocoder]

I actually found a huge roll of NOS double glass insulated nichrome wire on eBay for cheap, so I’ll be using that. (I can send you a bunch if you or anyone else wants it.)

What is the resistance per length? I would definitely like some, and would be happy to Paypal you some money for that. This would save me a step and waiting for the epoxy to dry.

Right, I suppose I meant just using a simple on/off control with hysteresis, like a solid state controller. You won’t get better than 1C of drift, but it’s simple.

Ok, in that case, the answer is: yes. My simple voltage comparator controller does have some hysterisis (a few tenths of a volt). Although in practice it isn't really critical, because after it cuts off the heat the temperature continues to rise for some time. For that design I used the thermistor loosely attached to the board, so there is a large air gap between the tube and that. When it cuts off the MOSFET controlling the heater current, the heat in the tube continues to "soak" inwards for some time - this is what causes the 1C range.

Since you’re going for absolute temperature accuracy and stability, you might think of adding a thermistor attached to the middle of the outside of the inner tube. Hold it down with a single wrap of glass tape.

You can do some measurements from this sensor and the one inside the inner can, plot the data out over time. If you overlay the two graphs, you should get a very clear view of the temperature differential. You can add this as either a constant to the PID’s error, or keep the second sensor installed and use it as a dynamic variable in the PID’s error.

I've captured similar data tonight, except it is sensor on inner wall of tube (thermally bonded) and sensor free-floating inside tube (suspended by a piece of cardboard).

The design I am planning is a relatively fast inner PID loop whose PV is the outer thermistor reading and whose CV is the heater current, and a relatively slow outer PID loop whose PV is the inner thermistor reading and whose CV is the inner PID loops setpoint.

Also, I'm thinking a lot about your comment about the PWM. I think you're right that I need a way to control the heater current without PWM-ing on and off the large heater current. One option is to run the microcontroller PWM through a low pass filter to get a "DC" value that can then be used to control a voltage or current regulator. Would be nice to get a regulator that can cut off the voltage/current completely. All I have right now are some LM317 that can only go down to around 1.25V.

[timb]

27.2 Ohms/ft

How much current is your current setup pulling?

Instead of using a voltage regulator, just use an Op-Amp and a Transistor or MOSFET; low pass filter your PWM into the non-inverting input of the Op-Amp and have the output control a transistor or Linear MOSFET.

If you do want to use a regulator, you can get a lot of them down to 0V by either biasing the control pin with a negative voltage *or* sourcing current from it. 1.25V is the internal error amplifier reference voltage of the LM317, hence why you can’t get it lower without one of the above tricks. You could also always hang a relay off the output to completely cut it off.

[motocoder]

27.2 Ohms/ft

Wow, that's even better. I would only need to wind a couple of feet. Can gou send me some of that?

How much current is your current setup pulling?

The input is rectified A/C. I can get by with around 500mA.

Instead of using a voltage regulator, just use an Op-Amp and a Transistor or MOSFET; low pass filter your PWM into the non-inverting input of the Op-Amp and have the output control a transistor or Linear MOSFET.

Yes, I know these techniques, but to use the n-channel MOSFET I have for that, I then need a high-side driver, need supply for the op-amp, have to worry about stability, etc. It is all do-able, but just becomes a bigger experiment. I will take a look into it after I get my PID approach nailed down.

The technique you mention with the 317 might be a good option. I could use a zener and a negative voltage from opposite polarity rectified A/C from the transformer to generate the negative reference.

Or I could,use op-amp, MOSFET feedback loop with small sense resistor to make a current regulator and to keep the MOSFET drive low-side. DEcisions, decisions...

[timb]

27.2 Ohms/ft

Wow, that's even better. I would only need to wind a couple of feet. Can gou send me some of that?

Sure thing! I actually got two rolls, so I’ll send you a bunch. PM your address.

How much current is your current setup pulling?

The input is rectified A/C. I can get by with around 500mA.

500mA seems like an awful lot. The stock oven only runs at 200mA. But I guess you’re heating faster?

Instead of using a voltage regulator, just use an Op-Amp and a Transistor or MOSFET; low pass filter your PWM into the non-inverting input of the Op-Amp and have the output control a transistor or Linear MOSFET.

Yes, I know these techniques, but to use the n-channel MOSFET I have for that, I then need a high-side driver, need supply for the op-amp, have to worry about stability, etc. It is all do-able, but just becomes a bigger experiment. I will take a look into it after I get my PID approach nailed down.

The technique you mention with the 317 might be a good option. I could use a zener and a negative voltage from opposite polarity rectified A/C from the transformer to generate the negative reference.

Or I could,use op-amp, MOSFET feedback loop with small sense resistor to make a current regulator and to keep the MOSFET drive low-side. DEcisions, decisions...

Yeah, tough call…

[motocoder]

Sure thing! I actually got two rolls, so I’ll send you a bunch. PM your address.

PM sent. Thanks!

500mA seems like an awful lot. The stock oven only runs at 200mA. But I guess you’re heating faster?

It's actually a bit more than that (peak). I've attached a schematic and some simulation results from LTSpice. You can see the current in the bottom pane of the simulation image. I think actually this simulation has the values from the aborted PD2010 heater mod, so the heater resistance and voltage across the heater are different. But it's in the ballpark. Keep in mind that the original heater was running off mains voltage. Power is V*I, and V is less than half, so expect current to go up proportionately to keep the same heater power.

[timb]

So I just found official documentation on the oven temp of the 2005A:


(Source: 2005A Manual)

See the orange highlight? It says 75 Degrees.

I also found the same 75 stamped on the bottom of my 1025P's oven board.

Mystery solved I guess.


Sent from my Smartphone

[motocoder]

Good catch.

[timb]

BTW, your wire is in the mail! =)


Sent from my Smartphone

[motocoder]

BTW, your wire is in the mail! =)


Sent from my Smartphone

Thanks, Tim. I'm hoping to put some more work into it this weekend (assuming there isn't another power outage).

[timb]

Cool, keep me posted.

I'm toying with the idea of re-designing the entire 2005A oven board with modern SMD parts. I can actually make the board round and have a 3D case printed so it will fit exactly in the hole left behind from the old unit. The big advantage here is we could build the heater into the PCB itself as an inner or bottom layer via serpentine traces. I've already specced out modern replacements for the two transistors. Working on the diodes now.

The whole thing would be smaller, more efficient and most likely with a good deal less drift.


Sent from my Smartphone

[motocoder]

Cool, keep me posted.

I'm toying with the idea of re-designing the entire 2005A oven board with modern SMD parts. I can actually make the board round and have a 3D case printed so it will fit exactly in the hole left behind from the old unit. The big advantage here is we could build the heater into the PCB itself as an inner or bottom layer via serpentine traces. I've already specced out modern replacements for the two transistors. Working on the diodes now.

The whole thing would be smaller, more efficient and most likely with a good deal less drift.


Sent from my Smartphone

that would be nice. Maybe even use a circuit like Widlar's "ring of two" to get more stability than the single zener.

[timb]

Ohhh, that's a good idea!

I'm working on getting the entire 2005A schematic into LT Spice, but it might take a few days.


Sent from my Smartphone

[timb]

BTW, that wasn't Bob Widlar was it? I thought Jim Willams came up with it.

Either way, we could use a matched NPN/PNP package plus a thermally bonded dual Zener. Hmmm.


Sent from my Smartphone

[motocoder]

BTW, that wasn't Bob Widlar was it? I thought Jim Willams came up with it.

Either way, we could use a matched NPN/PNP package plus a thermally bonded dual Zener. Hmmm.


Sent from my Smartphone

Could be, but the book I have mentions Widlar. Need to make sure we use a zener or other part designed for this purpose. Apparently otherwise there will be long term drift issues due to surface changes.