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

[MBY]

Ok, time for a update...

I have worked with the firmware and I'm quite happy with the function now. But PID-optimization is slow and boring as hell and I have to redo it every time something changes, like thermal isolation, setpoint, drive voltage, etc, etc.

Now I'm doing another spin. I have changed the controller to a ATMega328 (Mega88 is enough, but 328 is what I have) to get a 16 bit PWM. I also added current sense feedback to the ADC and improved some other things. I was using a software/USI simplex UART with the Tiny85 for debugging. Now I have HW USART and also test pads for I2C if it ever will be used.

As I talked about before, I do not longer aim at putting the PCB inside the heater, so the PCB area could be greatly extended simplifying routing, measurement and patches.

I also added a second LED "interface" so I can drive a bicolor LED and a optocoupler feedback from the PD2005 so the controller knows whenever the PSU is on or off. This way, I can optimize the PID further still as the electronics inside the heater adds a fair amount of heat.

My design is floating (dedicated transformer), so a optocoupler seemed logical.

I actually already has routed the board, but not etched it. Probably I will make some final adjustment like improving the ground plane. I also promised to add something to commemorate the first or two first forum members that can answer my questions in the other thread. In the other thread, robrenz thankfully gave me some ohms and temp measurements, so I (or he) must figure out something (small text or artwork) on the copper layer before I can etch.

Attached are the new schematics and KiCads 3D-view of the board. The DIL-6 on the board is actually a DIL-4 (PC/EL817), the 3D model is wrong.

Again, its a PITA to do a plot and convert to a picture that this forum takes for upload, so only print screens.

Edit: The HC49 style package is actually not a crystal but a SMD switch that happens to fit on the same footprint. My idea was to set a setpoint with a press of a button by using a common K-type thermometer and press when the desired temperature is reached.

[grenert]

It is interesting to see that this PS has a custom ovenized reference circuit.  Is it the same in both the 2005 and 2020?  The Power Designs C500 (100V 500mA model) just uses an LM199.  I wonder which is more stable.

[robrenz]

It is interesting to see that this PS has a custom ovenized reference circuit.  Is it the same in both the 2005 and 2020?  The Power Designs C500 (100V 500mA model) just uses an LM199.  I wonder which is more stable.

No custom oven on the 2020

[MBY]

It is interesting to see that this PS has a custom ovenized reference circuit.  Is it the same in both the 2005 and 2020?  The Power Designs C500 (100V 500mA model) just uses an LM199.  I wonder which is more stable.

No custom oven on the 2020

That got me thinking. What does the 2020 use? Do you know other units that use a custom oven? I was thinking to document my project well and release it as "open source hardware" to let others improve their PD PSUs. But, of course, that prerequisite that the stability on my PD2005 actually improves. If it worsens or give no benefit then there is not much point besides fixing a broken oven.

Any ideas on text or artwork on my PCB to hail you!? I'm thinking of something like 10*10 mm (approx 0.4*0.4 inches). Remember that I will etch the board at home, so very fine details may be destroyed in the process. I can use a picture (bw of course) or a text (edit: or perhaps a dummy/nonsense component as an Easter egg).

[MBY]

I'm going of the internet for a week or so as I'm switching ISP as well as LAN layout. But I haven't forgot the project, it is still running.

I probably add another LED control line (why just control two LEDs when you as well can drive a full RGB one?) and maybe other features (I'm thinking of double NTCs for redundancy and maybe other HW redundancies. This puppy should be designed for unattended 24/7 running and must err in the right ways).

I'm also thinking of replace the PD 2005s main transformer to some 230 VAC model. I have bad experiences with old transformers. Thankfully I have always happened to be at home when the fire starts, but that's no guarantee.

Edit: Another thing I did was making a new outer shield to be mounted on the actual shield. Made by what? A Red Bull bottle and PU foam! I will cut out the foam to make a chamber to mount on top of the shield.

Will report back in a week or two!

[MBY]

So, it has been much more than a week, but I haven't forgot about the project! I currently don't have a lab, so I can only do computer related stuff now (cad, SW).

Anyhow, attached is the latest schematic. I have added a second NTC as a HW protection using the second half of LM358. I'm currently tries to fit this addition to the PCB in Pcbnew (KiCad).

Sorry for the poor layout of things in the schematics. KiCad is a real PITA when you want to move blocks around to make new room. The result is that you stop caring and just squeeze in a new cap, resistor or whatever and leave it so with overlapping designators etc.

[MBY]

Ok, things always take more time than anticipated... But I had lot to do the last months. Finally my new lab is coming around. I have only a half dozen moving boxes to go through before I will start up the oven project again!

See you all in a few days I hope!

[MBY]

Ok, time for an update. I have etched the PCB but was impatient enough to make two basic and idiotic errors, so I had to make patchwork almost immediately. First, I forget to do a last flooding and DRC before printing the graphics. A last time change made the ground plane and VCC connected. Argh!

The second idiotic mistake was in the design process. I forgot that the Mega48/88/168/328 family does not have a selectable 2.56 V AREF but only the usual bandgap (1.1 V). Argh 2! So I had to cut traces for fixing error 1 and piggyback a TL431 (they sure come in handy often, 2.5 V was perfect!) to AREF.

But now I'm up and running with my new PCB, enjoying 16 bit PWM and real USART, two LED/lamp drive lines and feedback both from NTC and from current. And also my SW independent second NTC interlock.

I'm currently converting the SW from Tiny85 to Mega328. No pictures at the moment, but the design is basically as I showed in earlier posts. PCB is much bigger than the Tiny85 board as I no longer plan to use the PCB inside the oven.

Edit: Language...

[robrenz]

Glad to see you are still at it.  I have not had time to do anything with mine yet but my first project will be to make a run of 10 turn 1 ohm pots for the 0 to 1mV scale.  They dont exist as far as I know but we need them to get the microvolt adjustability that these suppies are capable of.  I also stole a Preamble/LeCroy 1855 100Mhz differential amplifier for $300.00  on ebay. (see my sometimes it pays to be stupid thread) This has 10x gain and will allow meaningfull ripple measurements to be made on these supplies.

[MBY]

Time for a status report!

Now things starting to come together. I have remounted the oven to the PD2005 and started testing. The PID loop regulates very well, much cleaner than the old thermostat. By experiments I have established that the old thermostat were shutting the heater power when the temperature is 70 degrees C and turning it on again at 65 degrees. So there is a hysteresis of 5 degrees. My PID loop keeps the temperature within about 0.3 degrees. The only problem is that the electronics in the oven creates rather much heat in itself, forcing different PID settings depending on if the PD2005 is on or off. Still, there is a tendency to put the temperature up or down by turning the PD2005 on or off. So the "always on" heater is a bit of a dud. But the same goes for the original design. The way to see it, is perhaps to regard the heater (when PD2005 is off) as a "pre-heater", shortening the time for the unit to reach stable temperatures.

My setpoint is 60 degrees, not 65-70. I can adjust the setpoint up to 85 degrees but I think 60 degrees is good enough as its a compromise between power requirements, "effort" for the PID, stable temperature and longevity.

I'm using two NTCs. One of them is used by the MCU controlled PID, the other one is a interlock using only analog circuits to shut the power when the temperature is off (interestingly, the hysteresis here seems to be about 5 degrees C), adjusted by a pot to about 70 degrees.

I had a hard time to fit it all in the unit. My first design where a small board intended to sit in the oven housing, but in my second design I've scrapped that idea and made a board  about 75*90 mm. Then, the new transformer needed to be fitted, as well as the power board (rectifiers and caps).

I ended up using felt as insulation. The felt is inside the oven, and another layer on a paper roll (toilet paper!) that fitted perfectly on the oven.

The stability seems to be improved. Stupid enough, I didn't record the stability enough, so I will ponder this thread to get clues about how bad or good it was before. But during several hours, the max and min recordings seems to be 200 µV apart in CV mode (no load). I will do additional tests under load and CC mode.

Below is some pictures. Note the fitting of the transformer (is a 230->24 VAC @ 1500 mA used as a 115 -> 12 V. The max current used is just under 500 mA). The front panel neon light for the oven is replaced with a RGB diode. Blue is "too cold", red is "too hot", green is "hunky-dory", combinations of the colors indicates various error conditions, pre-heating and such). Also note the dial cover that was originally white but very worn. I'm used my girlfriends nail polish with a "Power Designish" orange color. I think its rather neat!

From the top of the oven (big red cylinder) comes all the signals from the heater. The tick green wires are power, red and green enamel are the two NTCs and the thicker white is an ordinary K-sensor to a external thermometer. Also note the outer felt isolation. Upper in the picture you can see the control board. Between the transformer and the front is the little power board. Its not secured yet and only sits there. I'm using a network cable to connect the oven to the control board.

On the control board (main pcb) you can see my little patch board I mentioned in a earlier post, with the TL431 reference.

Everything is rather ugly, a rats nest, but it works!
 

[MBY]

Okay, another update with some stability figures. My first impression was that the stability goes up when the temperature does the same but I was wrong.

A lower setpoint seems to improve stability. I have tested temperatures of 46.3 degrees C and 70.0 degrees C.

Code: [Select]

~95 mA:
Setpoint K-temp DM3061 avg DM3061 min/max Fluke 45 min/max Meas time Adscans d/d
450 49.4 93.077 mA 92.5296/93.6101 92.485/93.598 10 min 4 1.0805/1.113
450 49.3 92.3951 92.1742/92.6395 92.130/92.586 3 min 4 0.4653/0.456
705 70.0 92.696 91.8280/93.3903 91.795/93.337 10 min 4 1.5623/1.542

~10V:
Setpoint K-temp DM3061 avg DM3061 min/max Fluke 45 min/max Meas time Adscans d/d
705 70.0 9.99533 9.99497/9.99545 9.9953/9.9956 ~5 min 4 480/300µV
400 46.3 9.99633 9.99631/9.99635 9.9956/9.9958 ~ 20 min 4 40/200µV
Some comments: The setpoint is in ADC units, K-temp is temperature reading on the K-type thermocouple I placed in the oven core. The avg and min/max readings are of course DMM readings from my Rigol DM3061 (6.5 digit) and Fluke 45 (5 digit). "Meas time" is the time I gave the values to stabilize and "Adscans" is something that relates to the SW (the ADC-samples per value). "d/d" is just the max-min deltas for the two DMMs.

Now I'm collecting max and min values for much longer times, a couple of hours. Edit: I don't like the discrepancies between the DM3061 and the Fluke 45 in the d/d column. I may have a noise problem and I need to assert if the noise is from the PD2005 or something else.

[robrenz]

Does the Rigol Allow setting the number of power line cycles NPLC to different values?  I used 100 NPLC on the measurements you saw in the other Power Designs thread.  That averages out a lot of the noise.

[MBY]

Does the Rigol Allow setting the number of power line cycles NPLC to different values?  I used 100 NPLC on the measurements you saw in the other Power Designs thread.  That averages out a lot of the noise.

No, not to my knowledge, but the sample time can be adjusted with no reference to NPLC. You can speed up the measurement and the display indicates from 4½ digits to 6½ digits of accuracy (but always show 6½ digits). However, short time noise does not appear to come from the PD2005 as it does not vanish when the unit is powered off (along with the oven controller).

[robrenz]

You might try shorting the inputs the Rigol right at the meter jacks and recording min, max, avg over similar time spans to see what the meter stability/noise is.

[MBY]

You might try shorting the inputs the Rigol right at the meter jacks and recording min, max, avg over similar time spans to see what the meter stability/noise is.

Yes, good idea. I'm just checked both the Rigol and the Fluke for some hours and not a single LSB digit noise on the Rigol (20V range), but +/- 2 counts on the Fluke 45 (10V range).

[robrenz]

IMO you should lock both meters in thier lowest range for the highest resolution to do the meter stability test.  Then set your PD to a voltage that is still within that lowest range to do a stability test.  That will give you the most resolution on your stability tests.  The 20 volt range does not have enough resolution to see the noise and drift of the meter.

[MBY]

IMO you should lock both meters in thier lowest range for the highest resolution to do the meter stability test.  Then set your PD to a voltage that is still within that lowest range to do a stability test.  That will give you the most resolution on your stability tests.  The 20 volt range does not have enough resolution to see the noise and drift of the meter.

That is a valid point! But I'm not really interested in the stability beyond 1-10 µV but will do the test anyway because you never know what will turn up. Actually I did this very test before, but I didn't record the results (and I never measured the PD2005 at this range).

But, what I know this far: The noise is not from the PD2005 as it appears even with it (and the heater) turned off. The noise is rather big on my scopes, about 50 mV PP (AC coupled), so I'm looking for some noise source. Probably some switched wall wart (those f**king things always spur noise on everything!). I also have to check the termination of the measurements and some other stuff. However, the 50 mV PP doesn't show on the DMMs (too fast probably) and since my DMMs show very little noise when the PD2005 is off I suppose that drift is real.

Now I have done some SW changes, added a bootloader so I can fit everything and close the chassis and still update the SW with only the serial port on the outside.

[MBY]

Ok, I checked the DM3061 on the 240.0000 mV range. Measured min and max was +50/-150 nV after several hours (a whole night).

Now I have a very strange problem. My oven has regulated very good both when the PD2005 is on and when its off. No problem. Error part in the PID is zero almost all the time. Now, I have assembled everything, did a calibration and was ready to put the cover on when I discovered that the error jumps around as if it was a loose connection or something, but ONLY when the PD2005 is off. My design is floating and the only connection between the oven controller and the PD2005 is via an optocoupler (so I know when its on and can preset the integral part according to that).

What is going on? No problem when the PD2005 is on, but in off state the NTC seems to jump around. I've tried the most obvious things like measure ohms and voltages between floating things and reversed the power outlet.

(And I "successfully" involuntary tested AC on the ohms range on a DMM as well as the residual-current device [unsure about the terminology, in Swedish we say "jordfelsbrytare" meaning "ground fault breaker"] in my apartment! I'm an idiot! Edit: And, yes my UPS also. Everything worked.)