Teardown of Bliley OCXO module

[dnessett]

Ever wonder what is inside one of those 10 MHz OCXO modules you find all over Ebay? I did a destructive teardown of a BLILEY NV47M1008 and documented the results on my website (Bliley Teardown). Note: figures 7 and 8 can be expanded by clicking on the images. This will provide a higher resolution image that allows the viewer to see the information referenced in the text.

I will try to format this information in a way that allows me to post the complete teardown on the forum. However, I was having some difficulty doing that, so I thought I would get out this post now and work on the problems I have encountered concurrently.

[MK14]

Thanks, I enjoyed looking at that on your website. Nice pictures.
I've sometimes wondered what the heater would be like. I was sort of expecting a small power resistor to be used.
It is more complicated inside, than I was expecting. But I guess they have done it that way, to ensure the accuracy and quality of the oscillation.
I was also curious as to how much (thermal) insulation there was going to be. I could not see any (assuming the insulation you showed was just purely for electrical isolation reasons). But it could serve both roles.

[dnessett]

Thanks, I enjoyed looking at that on your website. Nice pictures.
I've sometimes wondered what the heater would be like. I was sort of expecting a small power resistor to be used.
It is more complicated inside, than I was expecting. But I guess they have done it that way, to ensure the accuracy and quality of the oscillation.
I was also curious as to how much (thermal) insulation there was going to be. I could not see any (assuming the insulation you showed was just purely for electrical isolation reasons). But it could serve both roles.

I think the white stuff around the coil may be thermal insulation.

[MK14]

I think the white stuff around the coil may be thermal insulation.

That makes sense. Somehow, I got it into my head, that the insulation would be between the entire module and the outer case. But really only the quartz crystal, heater and temperature sensor (if present/needed, I bet it is needed, unless it uses PTC material or something), need to be in the "oven/heated" part of the unit.
Hence less thermal mass, hence lower power consumption, smaller heater and faster warm up time. So win win win.

Until you get to design these things and/or open them up (like you just have), it is easy to miss these subtleties in the design.

[texaspyro]

I've sometimes wondered what the heater would be like. I was sort of expecting a small power resistor to be used.

Oven heaters are seldom resistors which heat at P=(I^2)*R.  Transistors as heaters heat at P=I*E which is much easier to control.

[edpalmer42]

I've sometimes wondered what the heater would be like. I was sort of expecting a small power resistor to be used.

Oven heaters are seldom resistors which heat at P=(I^2)*R.  Transistors as heaters heat at P=I*E which is much easier to control.

I think there's a more basic reason that a transistor is used.  If you used a resistor, you'd have a transistor in series with it as the control element.  So you're going to dissipate power in the transistor anyway.  Might as well use that power as the oven power.  The same idea applies to things like the various ovens in Rubidium standards.  The newer ones are small enough that one or two transistors can heat the whole thing quite nicely.

Ed

[texaspyro]

I've sometimes wondered what the heater would be like. I was sort of expecting a small power resistor to be used.

Oven heaters are seldom resistors which heat at P=(I^2)*R.  Transistors as heaters heat at P=I*E which is much easier to control.

I think there's a more basic reason that a transistor is used.

Nope, I came across a paper (or was it a patent?) on (I think) the FRS/M100 rubidiums that talked about the advantage of using transistors as heaters.  I boiled down to I vs I^2.

[Vgkid]

Thanks for posting that , kind of reminds me of the cfp-04 I tore down. I believe in the hp brief on the 10811 they discuss the switch to transistor heaters.

[edpalmer42]

I believe in the hp brief on the 10811 they discuss the switch to transistor heaters.

I don't know about the M100, I didn't see it in a quick look at the manual, but the HP Journal article on the 10811 (March 1981, p. 24) says that the change was to reduce the power dissipation.  i.e.  use the transistors as the heaters so you're not wasting power dissipated in the transistors.  They decided to use two transistors rather than one.

Ed

[MK14]

Oven heaters are seldom resistors which heat at P=(I^2)*R.  Transistors as heaters heat at P=I*E which is much easier to control.

That makes sense.
The relatively slow time constant of heat (compared to electrical changes), means that it is potentially that much harder to control, anyway (i.e. potential for overshoot or oscillations). So improvements like that are important.

I don't know about the M100, I didn't see it in a quick look at the manual, but the HP Journal article on the 10811 (March 1981, p. 24) says that the change was to reduce the power dissipation.  i.e.  use the transistors as the heaters so you're not wasting power dissipated in the transistors.  They decided to use two transistors rather than one.

Ed

I can understand why designers would want to try to reduce the power wasted in heaters. So techniques like that, which put the heat lost in the control circuitry's output driver, into the heater itself. Sounds like a good idea.
It also means that there is less heating of the surrounding circuitry (outside of the insulated hot oven section), and hence less temperature drift for the control and rest of the circuitry.

[Kleinstein]

If the controlling transistor is inside the module, it is not that bad to have a resistive heater at heart. The transistor will still heat the whole module, so the heat used there is not totally wasted. One advantage resistors can have is, that they can be spread out over a large area, so that the heat source can be distributed about the way the heat loss is expected. This can help to keep temperature gradients small. In addition the heat can have a PTC characteristic and this way give a first fast acting feedback - this can help stabilizing the loop.

I am somewhat surprised to see rather little insulation from the circuit board to the outside and such a small crystal case. Also the the rather thick wires to the heater look a little odd. 

[Damianos]

The power that is needed to increase the temperature is always the same, with a resistor or without it...

The differences of using the transistor as the heating element are:
- there is only one component that needs thermal connection to the target
- with a resistor, the maximum current is restricted by the resistor's value; without it it's possible to have higher power initially to speed up the increase of temperature
- if the resistance is low, to overcome the above, then while the current is decreasing the power of the transistor dominates... In other words the main heating element changes from the resistor to the transistor

The above are valid for linear regulation, that gives lower temperature fluctuations and does not add interference.

[dnessett]

I think I have solved the image display problem. Following the suggestions of user Brumby (see https://www.eevblog.com/forum/chat/appropriateness-of-a-post/msg1313493/#msg1313493) I have cropped and downsized the images used on my website. The degradations are not significant, but if readers wish to see the images at full resolution, they can visit the website referenced in the first post of this topic and click on the image embedded in the page. Here is the teardown material formatted for the forum.

BLILEY NV47M1008 Destructive Teardown

Recently, I purchased a Tindie OCXO 10 MHz board that was fully populated and calibrated. My intention was to use it to supply not only the 10 MHz square wave the board generated, but also to feed the square wave into a buffer/filter circuit I designed, which outputs a 10 MHz sine wave. While testing the circuit, I inadvertently touched the Tindie board to part of a power supply generating 5V and 12V from 110 AC. This catastrophically damaged the board. After interacting with Chris from Analysir, which developed the board, I determined that the damaged component on the board was the OCXO module, which in my case was a BLILEY NV47M1008.

I purchased a new Tindie board and also bought off Ebay an additional BLILEY module. This meant I had no use for the damaged BLILEY module. Consequently, I decided to take it apart. I knew from the outset that disassembling the module was going to be a destructive teardown, since it encased its electronics in a sealed metal case. So, I decided to document the teardown photographically and put the pictures up on the forum in case others were curious what the internals of such a module comprised.

Figure 1 shows the bottom of the module with 5 pins (only 4 are functional): 2 serving to supply 5V and Ground; 1 returning the 10MHz square wave; and 1 accepting a reference voltage (Vref) that is used to fine tune the frequency of the output - see NV47M1008 spec. (Note: the 4 blue dots on the bottom are not pins. They are part of the bottom plate itself. The 5 pins are on the left and right of the plate near the edge.)

Figure 1

Figure 2 shows the bottom of the module after I used a Drexel tool with a cutting disk to remove the seal between the bottom plate and containing box. The seal was not solder; rather it was a weld, which required significant pressure to remove (I destroyed one cutting disk during the process).

Figure 2

Figure 3 shows the board located on the other side of the module bottom plate. Note the insulation in the cap that keeps the board from electrically contacting its metal top. Also, just visible, is insulation between the board and the bottom plate.

Figure 3

Figure 4 shows more clearly the insulation between the PC board and the bottom plate.

Figure 4

Figure 5 shows the same substructure with the insulation removed, revealing the components on the PC board.

Figure 5

Figure 6 shows the top of the PC board. Note the oven (the circular device labeled 052r4 on top and with white sides) that surrounds the oscillator (hidden). The chip next to it labeled J117 might be a 2SJ117 (Corrected per later post by CJay from 2NJ117), which is a Silicon P-Channel MOSFET for high speed switching. Chris conjectured that "[t]he mosfet is probably for driving/controlling the current (0.5A+), based on a temperature sensor and probably one or more op amps." He also noted that there is a lot of filtering on the board (see next figure).

Figure 6

Figure 7 shows the bottom of the PC board. There is a 6-pin chip labeled OCV3 that may be a Bliley proprietary part, since I could find no information about it on the web. However, the name suggests something like "Oven Controlled Voltage ..." (Oven Controlled Voltage 3.3V - a voltage regulator?) There are also chips labeled LB40 (5 pin) and C1L (3 pin), .3 C (3 pin) and EB1 (3 pin). I'm not sure what these are. However, Chris conjectured that the board would have some opamps working with the mosfet to control the oven. Perhaps the LB40 5 pin chip contains an opamp.

Figure 7

Added 10-3-2017

I decided to see what was "inside" the oven. It turns out the white material around the can was not thermal insulation. Figure 8 shows the oven can with the white potting material removed. This material held the two copper wires to the body of the can. It is not clear exactly what is the purpose of those wires. Perhaps they are part of the sensor that controls the temperature of the oven.

Figure 8

Figure 9 shows the area between the PC board and the can. Just barely visible between them are pins that are soldered to pads on the bottom of the board.

Figure 9

[Vgkid]

Pictures look good.

[Brumby]

Nice teardown.

That's a great effort on the image sizes! 

[CJay]

Great pics and teardown, a tiny possible correction, I think it's a 2SJ117 not a 2NJ

[Kleinstein]

The pictures are really nice.

The side view shows, that the controlling FET is not mounted direct on the board. There is a kind of heat spreader underneath, that extends all the way under the crystal can. So they seem to actually use the FET as the heater rather directly coupled. 

It looks like all 5 pins are connected to the circuit. It is little off to see the tantalum caps inside the rather high temperature can.

[glarsson]

It looks like all 5 pins are connected to the circuit.

The pin between Vdc and Vc is sometimes NC and sometimes Vref. The Bliley NV47 series seems to have both types.

[gamalot]

The SOT23-5 chip (labeled LB40) is MIC5205-4.0  low noise LDO.

http://www.farnell.com/datasheets/29605.pdf

[dnessett]

Great pics and teardown, a tiny possible correction, I think it's a 2SJ117 not a 2NJ

Yes, you're right. (Added later: data sheet is here: 2SJ117 data sheet)

The SOT23-5 chip (labeled LB40) is MIC5205-4.0  low noise LDO.

http://www.farnell.com/datasheets/29605.pdf

Nice find.

[dnessett]

I have added two pictures to teardown. They show the "oven" with the white material removed. It turns out this material was not thermal insulation. It was potting material holding the copper wires tightly against the can.

[Damianos]

I have added two pictures to teardown. They show the "oven" with the white material removed. It turns out this material was not thermal insulation. It was potting material holding the copper wires tightly against the can.

These wires are used in a way to improve the thermal conduction of the sensor to the can (the copper has low thermal resistance), so this material has also to be thermal conducting.

How did you concluded that the transistor is a MOSFET and not a BJT? For example the MJD117 that its package, agrees better with the photos. Have you done any test on the pins of it?

[dnessett]

These wires are used in a way to improve the thermal conduction of the sensor to the can (the copper has low thermal resistance), so this material has also to be thermal conducting.

If you look closely, you will see that the ends of the two wires not attached to the PC board are thinner than the copper wires. I think as I was scrapping off the potting material I destroyed a thermistor. My guess is these wires are actually part of this thermistor. It was attached to the can by some sort of grey stuff that I removed and therefore is not visible in the photo. I think this "grey stuff" is either the thermal conducting material you suggest or thermal conducting material encapsulating the thermistor device (which I discarded without realizing it held such a device).

How did you concluded that the transistor is a MOSFET and not a BJT? For example the MJD117 that its package, agrees better with the photos. Have you done any test on the pins of it?

I did not test the transistor. It is a surface mount part and I didn't attempt to detach it from the board and try to analyze its characteristics. I only conjecture it to be a 2SJ117 (the text says "perhaps"). So, it could very well be another device. However, if it was a MJD117, I would expect the labeling would be JD117, not J117. But, then again, I don't really know.

[Damianos]

The temperature sensor is (was!) probably a thermistor, as you guess. It is easier manageable for applications like this, with operation at a fixed temperature, than , for example, a thermocouple that needs more circuitry.
For the wires what I mean is that, for any type of sensor, the leads are part of the thermal "connection" and they have to be handled in a way to sense the target temperature instead of the surrounding. Another way is to thermally insulate the leads from the environment. The selected solution, in this application, gives better results, because of the tighter thermal connection to the can.

Relative to the transistor: from the provided datasheets the 2SJ117 seems that is produced in a TO-220 package that is not this case. The marking of the MJD117 is J117and it is from the same manufacturer...
You may try to do some measurements without removing it from the circuit board and, if the rest of the circuit permits it, we may be enlighten, just for the curiosity.

[CJay]

Relative to the transistor: from the provided datasheets the 2SJ117 seems that is produced in a TO-220 package that is not this case. The marking of the MJD117 is J117and it is from the same manufacturer...
You may try to do some measurements without removing it from the circuit board and, if the rest of the circuit permits it, we may be enlighten, just for the curiosity.

Good call, my comment was based on 2NJ not making much sense as a prefix, not any deep research so I concur, it's probably the MJD117