Would you buy a MAX1668 based Temp monitor(kit)?

[MMZ_TimeLord]

This is a poll to see how many people I can get in on this deal.

Based on this article...

http://www.madhacker.org/mbmsensors.htm

and this one...

http://www.overclockers.com/tips1059/

I am considering a run of boards similar to what Starbuck3733T did here...

http://episteme.arstechnica.com/eve...3631&f=77909585

I have had an offer from a friend of mine that works at a PCB manufacturing plant. They do up to 42 layer backplane boards for telecom. He offered run me a batch as their "dummy" batch for calibration of the equipment. He's talking 220 boards of my design PER dummy board. They run a MINIMUM of 48 boards for a dummy run! (Thats a total of 10,560 boards for free.)

Please let me know how many people would want one BEFORE I give him an answer next month.
I would like a minimum of 24 orders before I do the run.

Tomorrow I'm going to tour one of the plants to see how they do this.

I have already secured pricing from Digikey on the parts.

UN-soldered kits will cost $20. I will include SOT-23 transistors for probes and two 20 ft lengths of 30 guage insulated wire. (Blue and White)

Soldered kits will cost $40. I will include probes with SOT-23 transistors pre-soldered to five foot lengths of 30 guage insulated wire.

If you purchase a soldered kit and wish longer probe lengths PLEASE MAKE A POST WITH YOUR REQUIREMENTS!

Here is the probe design... SOT-23 Package (That's 115 x 51 Thousandths of an inch! )



Here is the design for the board. It uses ALL surface mount components. Capacitors and Resistor are all 0805 size components. (That's 80 x 50 Thousandths of an inch! )

[pHaestus]

MMZ: If you want to sell these then I highly suggest a socket for the 1668s. QSOP is a BITCH to solder and the adapter sockets run I think 5-8 in quantities of one. Not a big deal if you are doing it in a dutch oven I guess but if you're thinking about doing this with an iron (as I suspect the people buying kits would) then it's well worth it.

I see the power and SMBus wires aren't next to one another...Won't this mean a somewhat more complicated cable to go from SMBus to your PCB? I'd suggest putting pwr and SMBus together. You should also pimp the fact that you can link multiple 1668s together to get a LOT of temperature readings into MBM.

Longer traces = more resistance = lower temperature readings. Diodes are a bitch. Looks like it's properly designed though with resistors and caps where they are needed. You can add a third header for each probe and connect it to ground at the header and leave it unconnected on the probe end. This is supposed to reduce interference and noise (dunno if it works).

[Brians256]

pHaestus stole some of my comments.

Just let me know when/if this becomes final. I can paypal the $20 whenever you do.

[starbuck3733t]

*Scratches head*

10,000+ boards for free. Wow. That would take me forever to put the QSOPs on. The QSOP to DIP16 adapters, in small quantities, were $7 each.

Any ideas on h ow you'd mass produce the probes (wires with transistors on the end). My experience is desigining electronics, not manufacturing them.

Thanks for the credit too!

Edit: don't be a weakling, make these single sided

Ehehehehehe!

[pHaestus]

Also I should mention that I'm fairly sure that 30 gauge wire is not in spec for remote diodes. Twisted pair (scavenged from ethernet cable) is what I use, and I try to keep leads as close as possible to same size. Diodes suck to work with

[starbuck3733t]

Ditto, the wire in mine is also scavenged cat3, or cat 5. it depends on when in the run I made it. I ran out of spare cat3 and had to dive into the cat 5.

Mad editor!!

Also, from the 1668PDF in re: probe wire length

For remote-sensor distances longer than 8in, or in particularly
noisy environments, a twisted pair is recommended.
Its practical length is 6ft to 12ft (typ) before noise
becomes a problem, as tested in a noisy electronics laboratory.
For longer distances, the best solution is a
shielded twisted pair like that used for audio microphones.
For example, Belden #8451 works well for distances
up to 100ft in a noisy environment. Connect the
twisted pair to DXP_ and DXN_ and the shield to GND,
and leave the shield’s remote end unterminated.

Given, cat 5 doesn't have shielding. for a true testing environment one might want to use the suggested belden cable.

And my design doesn't have the EMI stuff, naughty me. TBH, I didn't know at the time how the hell to make EagleCAD do a ground poor

[nicozeg]

I'm very interested in this, but prefer one built arround the max6655 chip. I can give up on some sensors in favor of the greater accuracy and voltage monitoring.

there are higher accuracy chips at maxim like the 6646, but have only one remote sensor.

[bigben2k]

I'm in for an unsoldered kit.

pHaestus comments are spot on. Nothing I can't fix myself though.

[MMZ_TimeLord]

Re-design is not an issue at this time as I have 30 days to revise the circuit diagram. I plan on hand etching a prototype before then to evaluate the product.

I will defintely consider the ground pins and a second header for daisy-chaining the units. The main reason the power and SMBus headers are separate is because I could NOT find a standard as far as +5V being provided at the SMBus header on the mother boards. So, I left them separate to allow folks the option of making a molex tap cable or some such.

Anyhoo... ideas are good... I may throw up some optional designs and manufacture them all just so I can try them. Again the boards are free... but the components cost money.

Also as stated, the ground for each remote tempreture sensor was left out because I was going to twist the 30 guage pair and only make them five feet each. Under the maximum before grounding is required.

[starbuck3733t]

I knew i should have tried to sell them here too

[pHaestus]

You're going to really need to calibrate either way nicozeg; even the 6655s are only +/-1C over a range we aren't interested in with water (60c eh?).

I just thought of one thing that you might want to consider adding MMZ: A power switch! My 6655 EVSYS board has one and it's really handy when you are setting everything up. Particularly with 4 sensors that you may be moving around a lot while putting into water and such. I can take pictures of the size of the package for the remotes tonight; I have a couple of them lying around at home.

[Brians256]

Starbuck3733t, this would have been a prime place to sell stuff, true. Many ubernerds are here.

[MMZ_TimeLord]

Many more components and the board will grow in size pHaestus. Right now that drawing at the top is scaled up about 400%. The board is LESS than 1" square.

As to the power switch. Why wouldn't you be able to just unplug the power? It's already on a separate header remember?

[pHaestus]

you have no sense of elegance However I guess I could put the power on a switch anyway

[nicozeg]

Originally Posted by pHaestus

You're going to really need to calibrate either way nicozeg; even the 6655s are only +/-1C over a range we aren't interested in with water (60c eh?).

Of course it will need calibration, but 6655s +/-1º is a lot better than 1668s +/-2º

Besides the volt monitor has an aditional advantage: Readings can be conpensated for variations in supply voltage. dont know if you have made this on your test setup, but it seems very usefull if using the pc power supply. We know that psu rails are not that steady.

Four external sensors can be made with two 6655, it will require very little aditional board space, and wont add too much to the cost. (6655 is cheaper than 1668)

[Cptn. Foo Foo]

I looked into making one of these kits a while back but I couldn't find the Maxim IC chip anywhere. Also, the process of calibration seemed rather vague after reading over the various web pages. However, if the guys here in this forum will be getting some of these kits too then I am sure the process will be clearly explained

--------

I am really interested in this. I would be willing to buy 2 unsoldered kits - one for me and one for a friend.

I voted for the unsoldered kit...but I can only vote once...not twice. Just something to remember if you are thinking the vote total equals how many kits are wanted.

One suggestion MMZ_TimeLord. What about mounting holes in the PCB? Sure we can always drill our own but if you are getting the PCB's made then you could just add one (or multiple - in the corners maybe) mounting holes in somewhere. Just a thought.

[BillA]

several perhaps
but I would want to know the final configuration
- twisted shielded cable, always, any length, even if no apparent difference

[MMZ_TimeLord]

Bill has spoken... I will change the probe wires and add grounds.

I will also put the power and SMBus connector together and make two for daisy-chaining.

I'll also add a power switch. I'll leave shunts for the address settings though.

That won't leave much room for the components... but I think I can manage.

I'll see if I can post an updated design tonight. Maybe even include one for the MAX6655. That way if you buy one of each you will have two +/- 1°C temp probes, five +/-3°C temp probes and the additional voltage monitoring.

Let me know if you guyz agree... I just want you to get what you need/want.

[BillA]

sounds good
probably 2 soldered of the 6655s, I like pHaestus' suggestion of 2 together

EDIT: mil spec teflon/silver plated shielded cable is cheap on eBay, I buy it all the time (search "teflon silver")

[pHaestus]

speaking of the voltage monitoring nicozeg:

Is there one of these lines on the 6655 I can monitor a 1.5-2.2V signal on? It'd be nice to just dump vcore into my monitoring app directly.

[nicozeg]

Originally Posted by unregistered

sounds good
probably 2 soldered of the 6655s, I like pHaestus' suggestion of 2 together

Cough cough, it was me

In fact I had plans on making a test system based on the 6655 and a cpu as external sensor just like ph's. I was about to order a bunch of this chips to make some for other friends interested, but if there's the possibility of a pro grade pcb, we can wait for them


pHaestus: Looking at several parts of the datashet can bring an answer.

First:

Four Voltage Channels
+12V, +5V, +3.3V, +2.5V
Three External Monitors
One Internal Supply Monitor

Then says about the external monitors:

ABSOLUTE MAXIMUM RATINGS

VIN1 to GND............................................................-0.3V to +16V
VIN2 to GND..............................................................-0.3V to +6V
VIN3 to GND..............................................................-0.3V to +6V

so it can measure a wide range, but at what accuracy?

ADC Total Error
VIN1, VIN2, VIN3 between 30% and 120% of nominal ±1 ±1.5 %

Then if you use the 2.5V sensor, Vcore will be well inside typ. +/- 1% error margin.

[pHaestus]

excellent. So I just solder to any of the VCC and VSS pins on back of socket with twisted pair? What's the catch?

[Groth]

The catch is, it isn't a differential measurement between the Vcc and Vcc pair. It's a single measure relative the the IC's ground. If you're using separate computers for your taget system and for your measuring/logging, you'll have problems/errors.

[Incoherent]

I'd take two. Especially interested in the 6655 version, one of each.

Incoherent

[murray13]

Yes interested in a 6655 version. 2ea

As for cable, Belden 8451 as per Maxim recomended, is $14 per 100' or $56 per 500' from Newark. Nice cable, only 5/32" dia. Shielded cable for less than $0.15 per foot (if you get enough people interested).