A new, improved, oven for the PD 2005 Precision power supply

[MBY]

As you may have read in the thread https://www.eevblog.com/forum/product-reviews-photos-and-discussion/looking-for-goodinexpensive-bench-power-supply/ my PD2005 PSU has let some of its magic smoke away. Thankfully, all of the "unobtainium" parts of the electronics seems to work, only the stability oven heater has failed. The oven contains all the reference stuff and to eliminate temperature drift, this electronics is heated to a specific (to me unknown) temperature. The heater is simple: a heating filament in series with a bi-metal thermostat. The heater runs directly on 115 VAC mains (I use a 230->115 VAC transformer as Sweden has 230 VAC in the mains).

First, I wanted to keep things simple and build my new heater to run directly from 115 VAC. The schematics attached shows that concept. A couple of resistors and zeners and a LDO creates 5 VDC for the electronics and a mcu is to be controlling the power on the heater (heater not shown, but I probably gonna use power resistors rather than a filament).

But this idea was not so simple after all. You see, I wanted to be able to control the oven lamp, so I needed another bridge rectifier and a optocoupler. The problem is that a optocoupler need several mA, and several mA means I need to burn away non-trivial amounts of heat in the "resistor step-down".

The heater is controlled in a CC-style circuit with a PWM to set the level. Also the heater is turned on via a AC-copuled high side transistor, to lessen the probability of a software bug that turn the heater full on forever.

Another problem with this design is that most of high voltage FETs runs in depletion mode, meaning they need "beyond rail" drive on the gate (what I mean is that to turn off a depletion mode N-FET you may need a voltage lower than source).

All in all, that makes for a complicated circuit. So I'm gonna redesign this for low voltage use, so I will use another wall wart inside the 2005 (its plenty of room) to drive the heater from something like 12 volts.

All suggestions is appreciated.

The schema for my first (scrapped) design attached.

Edit: If you own a Power Designs model 2005 precision power source you can help! There is a few measurements on a working unit that I could use. The temperature and power of the oven is two unknowns.

[SeanB]

I would guess the setpoint is between 70-80C. This would be doable with a 10W wirewound resistor, as it will be able to get this hot on it's surface without insulation.

Try to wind your new element to be around 8-10R when cold, and then use the insulation of the original to act as a housing. For a controller just use a 70C normaslly closed thermal switch, it will operate at around 75C to disconnect the heater, simple and probably will fit in the position of the original. If the original is still working then reuse it anyway, most are capable of switching 1A or so.

[muvideo]

If this can be of any help, or you want just to play:



this is a very simple oven heater/controller I built some monthes ago.
It is without hysteresis, but it is a proportional one.
The values are for 45°C with a 15-18V working
voltage, at that voltage the maximum heating power
is something around 5-7W. The heating element is Q1.
RV2 is a trimmer that regulates the range of interest
and RV1 regulates the setpoint more finely.
If you put a trimmer in series with R4 you can tune
the integral term, the one i built is slow enough
that doesnt oscillate.

Fabio.

P.S. The reason I used TL431 was simply because I had many
PC psu broken...

[robrenz]

I suggest not changing from the wound filament style heater. It gives a much more uniform temperature input.  What good is a super precision PID control if you have hotspots at the discrete resistors. The resistance wire is cheap on ebay.  I have used this supplier often with good results.  http://www.ebay.com/itm/165-ft-Nichrome-Heating-Resistance-Wires-5-sizes-kit-/310212043161?pt=LH_DefaultDomain_0&hash=item483a142199  They ship from The EU to me in the US.

[codeboy2k]

First, I wanted to keep things simple ...

ok.. simple. right. got it..

...and build my new heater to run directly from 115 VAC. The schematics attached shows that concept. A couple of resistors and zeners and a LDO creates 5 VDC for the electronics ...

ok... so far so good...

... and a mcu is to be controlling the power on the heater

I choked and spewed food at this point.  Sorry to get you all messy.  You don't need an MCU to control a PSU reference heater.  The simple PI controller that muvideo posted is a great starting point.  Don't overcomplicate it with MCU PID control.  And you almost never need the derivitive (D) term.  PI is usually sufficient and easy to do in hardware.

You can easily get enough nichrome wire from an old toaster, $1 at a garage sale, or a hair dryer, $5.00. Or the local junk dealer.. or Salvation Army, etc...
No need to buy it from eBay. 

[MBY]

I suggest not changing from the wound filament style heater. It gives a much more uniform temperature input.  What good is a super precision PID control if you have hotspots at the discrete resistors. The resistance wire is cheap on ebay.  I have used this supplier often with good results.  http://www.ebay.com/itm/165-ft-Nichrome-Heating-Resistance-Wires-5-sizes-kit-/310212043161?pt=LH_DefaultDomain_0&hash=item483a142199  They ship from The EU to me in the US.

The problem is the mechanics. The filament needs to be isolated and at the same time conduct heat to the inner tube. A few word on the ovens mechanical design: The outermost part is a shield tube, and inside there was the filament attached to a inner, thicker tube. The inner tube holds the electronics. The isolation between the two tubes seems to be some sort of glass wool.

I didn't say I would use a few resistors. I plan on use basically the same "tracks" but with small thru-hole resistors instead and a lot of them. But if I use nichrome/kanthal filament or resistors is a little bit besides the point, as the electronics will be the same. The inner tube "will be a good conductor. Because I don't want to adhere something to this inner tube, I will anyway make my own tube in copper to thread on the inner one.

[MBY]

First, I wanted to keep things simple ...

ok.. simple. right. got it..

...and build my new heater to run directly from 115 VAC. The schematics attached shows that concept. A couple of resistors and zeners and a LDO creates 5 VDC for the electronics ...

ok... so far so good...

... and a mcu is to be controlling the power on the heater

I choked and spewed food at this point.  Sorry to get you all messy.  You don't need an MCU to control a PSU reference heater.  The simple PI controller that muvideo posted is a great starting point.  Don't overcomplicate it with MCU PID control.  And you almost never need the derivitive (D) term.  PI is usually sufficient and easy to do in hardware.

You can easily get enough nichrome wire from an old toaster, $1 at a garage sale, or a hair dryer, $5.00. Or the local junk dealer.. or Salvation Army, etc...
No need to buy it from eBay. 

Well, a single MCU is hardly to "complicate" things. Why not? The problem is not that I cannot find resistor wire. It has almost nothing to do with it. What I use in the end is dependent on mechanics. I have lot of good old TL431 (used in muvideo's design above) but they are all in a TO-92 package. They don't cost much, but if I have to find SOT-23 ones, I still need to buy something. Of course PI may be perfectly adequate, but a mcu is sime enough. One "dumb" thing about the 2005 is that the oven lamp indicates "oven on". I want it to indicate "temperature regulated" and/or "heater error". A mcu in a SO8 package is perfectly fine here I think, as I can do more intuitive indications and I already has the board real estate. The PCB must be thin (only smds) and cannot extend more than a dozen or so mm in width. It can, however be several cm in length. The current design uses two SO8 packages and a couple of SOT-23.

And I may very well use nichrome wire, but that depends on the mechanics. I don't want another episode of filament short circuit to the inner tube. Resistor heating is a feature, not a "cannot find nichrome"-situation.

[MBY]

If this can be of any help, or you want just to play:



this is a very simple oven heater/controller I built some monthes ago.
It is without hysteresis, but it is a proportional one.
The values are for 45°C with a 15-18V working
voltage, at that voltage the maximum heating power
is something around 5-7W. The heating element is Q1.
RV2 is a trimmer that regulates the range of interest
and RV1 regulates the setpoint more finely.
If you put a trimmer in series with R4 you can tune
the integral term, the one i built is slow enough
that doesnt oscillate.

Fabio.

P.S. The reason I used TL431 was simply because I had many
PC psu broken...

TL431 is a nice one and I use them as often as I can! But the ones I got is in TO-92 package and I need smaller things. But your design is neat but not entirely usable for me as the heat source is the transistor. I need many sources of heat to avoid "heat spots" even if the inner tubing and my planned copper tube will spread the heat. Of course, heat will be generated in my transistors as well, especially at lower setting (I don't want to use PWM as I don't want to create noise, so the analog control will burn some heat in the transistor, but that a side effect).

But maybe a small adjustments to the circuit can make resistors (R7 in your schematics) the primary source of heat.

[muvideo]

But maybe a small adjustments to the circuit can make resistors (R7 in your schematics) the primary source of heat.

R7 and D1 limit the current in Q1, perhaps you can try to put the wire heater between Q1 collector and ground.
As for hot spots, Q1 and R7 are problem candidates.

Fabio.

[robrenz]

Just a suggestion.  Wrap the inner tube with a layer of Kapton tape to electrically insulate from the resistance wire.  Attach insulated leads to the correct length of resistance wire. Wind the element over the Kapton tape at the correct spacing and put another layer of Kapton tape over the windings to hold the wire spacing and also over the insulated leads junctions.  From the looks of the stuff that was on the original heater wires the Kapton won't be any worse thermal conductivity.

Another thought,  If the slow oscillations we are seeing are from the heater on off cycles.  a PWM heater drive with a fixed Hz would help diminish that.

[MBY]

Good tip, robrenz. I have kapton tape and thought of using it. But it does not conduct heat well, does it? Maybe its not critical, as you said, the old stuff may not have been ideal either. But I don't know what the material was. Maybe some sort of paint that has degraded to a powder.

PWM is the most obvious solution but I wonder if it may cause noise. So I designed a analog control. But PWM is not out of the question entirely. Also, as some have noted, a analog HW PI-loop is sufficient. I have not ruled that out, but I'm still thinking in terms of MCU because I want to control the oven lamp to indicate heater status (error, temperature stable, etc).

No I have scrapped the design running of 115 VAC directly. It got to messy. So I will include a own transformer an run off everything at low voltage. A side effect is that the oven lamp has to be replaced by a LED. 

[robrenz]

PWM is the most obvious solution but I wonder if it may cause noise. So I designed a analog control. But PWM is not out of the question entirely. Also, as some have noted, a analog HW PI-loop is sufficient. I have not ruled that out, but I'm still thinking in terms of MCU because I want to control the oven lamp to indicate heater status (error, temperature stable, etc).

I was thinking an analog input PWM IC with a analog HW control loop not a MCU.  There are some used for motor drives IIRC.

[MBY]

PWM is the most obvious solution but I wonder if it may cause noise. So I designed a analog control. But PWM is not out of the question entirely. Also, as some have noted, a analog HW PI-loop is sufficient. I have not ruled that out, but I'm still thinking in terms of MCU because I want to control the oven lamp to indicate heater status (error, temperature stable, etc).

I was thinking an analog input PWM IC with a analog HW control loop not a MCU.  There are some used for motor drives IIRC.

Ah, I figured. However, I found that the worst case power dissipation of the power controlling transistor will be "only" about 1.8 watt (and it will be trivial to avoid that operation area) if I use a analog control instead of PWM (actually, a PWM->DC-converter) based on a 7.5 watt heater. It may be sufficient to create a "hot spot", but I don't think it will be an issue. Bu hey, its easy to disconnect the PWM->DC step and use PWM directly, MCU or no MCU. See the pwm control part of the scematics (even if this design is obsolete now, the heater drive part is essentially unchanged).

[MBY]

Ok, now I'm about to home-etch my first "low voltage" design. I still going with an MCU as I have explained before (I want some "intelligent" oven lamp control), an ATtiny45/85.

As you can see in the schematics I have somewhat redundant control over the heater. First, an AC-coupled high side switch to make sure only a PWM signal can activate the heater. Second, the power control using a simple pwm->DC level converter and a common CC-style circuit to drive the heater. The downside is heat in Q1, but if there will be a problem (I don't think so, I did the math) I just patch the PCB to make use of pwm directly.

I haven't decided exactly what to use as a heater, but I need about 20 ohms to get around 7 watts. I think 7 watts is reasonable. Either nichrome wire or maybe 20+ resistors. I even had a crazy idea of using loooots of small 0603 in a sandwich between two DIY copper tubes, but I don't think I can afford the space needed. I tested a portion of the old heater and it seems to handle the bigger current of lower voltage needed but if I want anything of this design it is to _lower_ the intermittent temperatures if possible, not rise it.

Never mind the really idiotic drawn schematics. It was a quick and dirty job and I used components already in the library. I personally don't like the way the mcu is drawn at all and I usually use named nets and "off page" gaps, but I didn't bother this time (I'm relative new to KiCad as well and haven't learned it yet. All my old libs and such is for Cadsoft Eagle).

The layout is captured from the "3D view" in pcbnew (KiCad). The same story here: I could not bother spending hours to get a decent ground plane or unify the use of different packages (this ugly beast uses 0603, 0805 and 1206 depending on what I have in my possession). It was important to make it single sided, as I have 0.6 mm single side PCBs, and size really matter here as I plan to stick the whole thing between the inner and outer tube of the oven. That explains the elongated PCB, less than 18 mm side but near 70 mm in length.





Edit: The ISP connector is a micro-match. A good thing of this land pattern is that its easy to "touch-program" without needing to actually mount it.

Edit2: Some of the values of the components are from the old design and will be adjusted.

[muvideo]

Given the fact that you want to activate the  heater with an AC signal from the mcu, you can also
build a simple charge pump and charge the gate of an Nmos, the mos will be connected like the
one you have drawn (that is Pmos but connnected like an Nmos, check it now I have to
go but I dont understand the way you connected Q2, seem that the source voltage will be lower than
input voltage... I'll check better later).
Also I would invert R25 with TH1 so the adc values will go up with higher temperatures
only a "cosmetic" effect   .
There ia a nice murata tool to chose quickly the series resistor so you can esily put the
themistor voltage in an easy range for the adc (I usually start with a series resistor
value in the range of the thermistor working resistance).
http://www.murata.com/products/design_support/mcnvs/index.html

Fabio.

[MBY]

Given the fact that you want to activate the  heater with an AC signal from the mcu, you can also
build a simple charge pump and charge the gate of an Nmos, the mos will be connected like the
one you have drawn (that is Pmos but connnected like an Nmos, check it now I have to
go but I dont understand the way you connected Q2, seem that the source voltage will be lower than
input voltage... I'll check better later).
Also I would invert R25 with TH1 so the adc values will go up with higher temperatures
only a "cosmetic" effect   .
There ia a nice murata tool to chose quickly the series resistor so you can esily put the
themistor voltage in an easy range for the adc (I usually start with a series resistor
value in the range of the thermistor working resistance).
http://www.murata.com/products/design_support/mcnvs/index.html

Fabio.

Great tool! It seems to work ok in wine (running linux here). I have a junk box full of NTCs but I have no data on them, but they seems to be fairly standard with a room temp resistance of about 10k.

You are absolute right, the high side PFET is reversed. I hate the KiCad symbols, I always usually use the implicit DS-diode to orient the device, sloppy me.

Anyway, a charge pump is a good idea, but since I need to pump 5 VDC to 12+, I need several steps, so I think the common NFET-PFET circuit is easy enough. Since I already routed the board and succeeded with a single side PCB without extra wiring I will only change "bugs" and really bad circuits, such as the wrong-way PFET.

But, when I think about it, a "charge pump" will be pretty easy. You see, I don't need to generate any voltage, only sink. A "sink pump" should be easy. Will check that out.

Apart from being in the wrong direction, it is nothing strange with Q2. This is a very common way of making a high side switch.

 

[muvideo]

The murata tool:
it's very nice, if you go in setting you can activate other simulations like for example self
heating effect. For a micro design it's very useful "Temp. Approximation", that suggests the
approximation polynomial and shows the error. I suggest you to characterize the thermistor
mesuring it at 3 temp points far apart and use an online calculator for extracting the beta from these.
So you will have the temp conversion  almost done as you first write the fw (probably for
you it's enough to figure the working temp and use a linear interpolation).

The strangeness i found was because of me:
at first sight I thought that the mosfet was a P part with D and S swapped in the schematic,
later I made a quick google search that returned a seller (Future) reporting it as an N channel,
so I didnt understand. Now reading the datasheet it's clear that the first thought was correct

The charge pump:
I was thinking to something like this:

The advantage is less parts and an Nmosfet that it's easier to find with good parameters
(I mean in general N vs P, dont mind the parnumber in my schematic, it was choose only for the simulation).
You can add a small capacitor on the gate, depending on the output frequency of the
mcu pin and refine the component values.
It should work, but I didnt tested it

Fabio.

[MBY]

The murata tool:
it's very nice, if you go in setting you can activate other simulations like for example self
heating effect. For a micro design it's very useful "Temp. Approximation", that suggests the
approximation polynomial and shows the error. I suggest you to characterize the thermistor
mesuring it at 3 temp points far apart and use an online calculator for extracting the beta from these.
So you will have the temp conversion  almost done as you first write the fw (probably for
you it's enough to figure the working temp and use a linear interpolation).

The strangeness i found was because of me:
at first sight I thought that the mosfet was a P part with D and S swapped in the schematic,
later I made a quick google search that returned a seller (Future) reporting it as an N channel,
so I didnt understand. Now reading the datasheet it's clear that the first thought was correct

The charge pump:
I was thinking to something like this:

The advantage is less parts and an Nmosfet that it's easier to find with good parameters
(I mean in general N vs P, dont mind the parnumber in my schematic, it was choose only for the simulation).
You can add a small capacitor on the gate, depending on the output frequency of the
mcu pin and refine the component values.
It should work, but I didnt tested it

Fabio.

Yes, it seems to work. Thanks for your help, I like the idea. But I selected components mainly on what I already possess and as the PCB is routed I think I stick with the current design.

[MBY]

Ok, I have etched and mounted the PCB and checked the electrical. It fits in the old heater mount, but I still don't have any new heater. Time to start on the firmware.

The wire on the pic is not a fix, its only to strengthen the ground plane. I'm sure it will work just fine without it. I did not found the DPAK transistor I plan to use, so I took a MPSA13 NPN darlington in a TO-92 just for testing. It handles a lot for the package, but not enough. Will be enough for testing though.

The "resistor bank" is NOT the heater, its only the current sense. Slightly over-engineered. The non-pupulated 1206 (under the wire) is where the NTC goes, but it will use leads (I will place the NTC into the "electronics package" inside the inner tube). This PCB is to be mounted along with the heater between the outer shied (the red tube in the picture) and the inner one. Also the zener is not mounted as I could not find it, but its not needed really, just some overvoltage protection for the LDO that takes no more than 16 volts.

It may seem to be strange to mount the PCB together with the heater(!), but I think I figured all out. If space or temperature does not permit it in the end, there is of course possible to mount the PCB on the outside.

[saturation]

Thanks for all the fine work, MBY, been following your progress keenly.  As you mentioned earlier, when its up and running I'd be curious to see how much noise it injects into the outputs.

[MBY]

Yes, the noise will be interesting to see. I worked little with the firmware and started with a clean "P" controller or "droop controller". It is only slightly better than thermostat on-off with hysteresis. But I soon went for a full PID (maybe I drop the D term later on). One problem is that Tiny85 only got a 8 bit PWM. I found that a really fast updating control loop acts as a overlayed low frequency PWM. But I haven't got the full PID to work yet and I know why. The input range is 1024 steps, the output range is 256 and the integral part must exist in a much bigger interval not to be completely "damped" by windup-protection.

In later years I have become aware of that broking the "float taboo" is often helping a lot in these circumstances. So, I started to use floating point math (please don't flame me on this, I'm perfectly aware of using left shifts and fix precision integers but I also know that it may actually _save_ code space to use floats. I have seen it many times. Floats is not a n00b decision).   

Anyhow, when I was leaved the project for work I still hadn't got the PID to work but I have only spent like one hour trying to make it work. I will resume the project in a day or two.

A short word on the power supply. I don't want to load the transformer of the PD 2005 (and its off when the PSU is off) so I will use a 230->24 VAC @ 1.5 A ordinary transformer witch gives 12 v at 115 volts and sufficient current. Hopefully that also reduces noise. But I also have thoughts of using a SMPS with PFC (the power factor can be bad with a common transformer when a rectifier is used).  Any ideas? Should I pursue a (PD2005 internally mounted) SMPS or go with the trafo?

Saturation: Do you have an idea about the temperature (or the heater power) of the original heating system? Of course, I do not need to replicate that in detail, as a recalibration theoretically could use any temperature at all, as long as it stable, but it would be nice to know (I primarily want to settle for a little lower temperature to spare the electronics).

I'm amazed that the electronics survived. Damn it was hot! I almost thought the transistors were germanium, but it cannot be, it was so hot! It took ages to get the temperature down and I had time to smell the smoke from meters away, turn it off, unplug it, take it down from the shelf, go find a screwdriver, open the upper cover, the lower one, spray-freeze it from above, from below (the base), notice it to warm up again (the thermostat clicked a few times back and forth), and it was still to hot to touch and smelled "funny". It must have been at least 5 minutes but may have been 15! And when I first noticed something was wrong it must have been several minutes of not-so-regulated heater action. I still don't know what happened. The only data I have is that the oven lamp started to light brighter a couple of minutes before disaster, as it happens that I had mounted a LDR on it to log the oven duty cycle. Well, if someone wants to know the duty cycle they have to do it themselves as my PD2005 gonna be unique.   

 

[T4P]

I'll say SMPS, remember to use some`extra filtering as i don't think the china smps's have much noise filtering

[MBY]

Yep, I will need one with PFC and of some quality as it would go (nearly) 24/7 unattended. I will test with my trafo first and if I experience noise I shall try a SMPS. I will have to find a good one, as apart from line noise I cannot tolerate much EMI either (as it will be mounted inside the chassis of the PD2005).

[SeanB]

Use the magnetic, more reliable, and less EMI. Also easier to filter out the bridge rectifier noise and it will last decades like the original. Most SMPS designs will die after a few years of operation.

[MBY]

Hi again! I haven't have the time to work on the project lately but when I leaved it the last time the PID worked perfectly. I did a mini-oven with thru-hole resistors and some kapton tape and stuck the NTC inside it. Of course, it has not the thermal inertia or power as the real deal, but that will be adjusted by adjusting the PID constants. And yes, the D therm is rather superfluous but I think its a bit cool to have it anyway.

The new oven will be about 6.5 watts I think and I settled for heating wire (not the original one) and kapton as isolator as suggested in this thread.

But, as I shortly mentioned a while ago (maybe in the other thread) the heating filament goes all the way down to the base of the heater assembly, so I cannot reach it or replace it with ordinary wire. And, as I'm using lower voltage and higher current, I cannot stick to my original plan to use only "I/O"s already present as pins in the socket. Therefore, I need to make a small hole in the outer shield to stick the cables inside. That means there is no longer any need to have the control electronics _in_ the oven. So maybe I do another spin at the board to fix the two "bugs" I detected. It would be nice with current feedback to ADC, using the U2B opamp. I also realised I never did post the second schematics, so here it is:

I'm sorry it is a crappy print screen. KiCad doesn't have a function to export/print to pictures (besides SVG). Some comments: The six 43R resistors in the current shunt is actually 10, not 6. I'ts only a matter of mounting "towers". I also did have to swap some pins to enable a debugging SW (USI-driven) uart. If I do another spin I might as well put in a bigger MCU as the space can grow outside the heater assembly. A cap on the NTC was removed - stupid me, of course you have trouble programming with a 100 nF load on SCK!   

Edit: The rectifier on the left is on a own board. I draw and routed them on the same PCB, and cutted it to two pieces. That way I saved some PCB material.