Fisher 1266 PART 2

[Eugene52]

Fisher 1266 PART 2

Great Work Ivconic AND KingJL !!!!!!!!...............Regards..........Eugene52

[porkluvr]

1265 & 1266 power supplies

Here are LTspice simulations for two versions of 1266 and one version of 1265 power supplies. I tried to include necessary third party symbols, models, etc. There is a jpg to explain what's what.

There is a problem with one of the 1266 power simulations, and another problem that shows in all three.

Problem #1:
Where the 1265 used LM385-2.5 voltage reference, the 1266 uses a B-C junction of TIP126. (The part in the simulation is named TIP127, but ignore that becauses TIP125,6,7 use the same model.)
The 1266 simulation (with TIP127) gives +/- 6V with a 12V battery, +/-7V with 14V battery, 12.5V with 25V battery, and so on. That can't be right. Replacing the TIP127 symbol with LM385-2.5 puts the output voltages at +/- 4.4V, (or, +8.8V, +4.4V if you prefer) and increasing the battery voltage no longer affects the output voltage like before. That's more like what's expected.

Problem #2:
In all the simulations, so long as the (+) and (-) outputs see equal loads, the output voltages remain relatively stable. However, if you vary either supply rail's load and not the other, even by just a few mA., you can see a substantial deviation from the expected symmetry. One voltage goes up, the other goes down. Arggh.

I am pretty sure that problem 1 is a "test fixture problem", a common weakness among transistor spice models. (I foggily recollect reading about where B-C junctions are usually not correctly modelled. What can be done about it?) I tried all my PNP darlington models. I even tried some NPN darlington models, swapping the B-C junctions for correct polarity, but the output was still +/-6V with a 12V input, just as before.

I think it is interesting to lay the 1265 and 1266 power circuits side-by-side for comparison, but I am not schooled in electronics enough to be able to explain how these convoluted circuits work. And, I don't have either detector so I can't take measurements to say whether the poor load regulation (problem #2) is just a simulation bogey, or really exists. But it is SERIOUS. Take a look.

Various voltages have been renamed but their relationships have not changed. I followed my schematics as closely as possible (omitting extra capacitors) but some of the features aren't clear in my factory schematic photo, and too, the reverse engineered diagrams may have errors. You may notice differing resistor values around the comparator, either by design, or from being mis-read. I haven taken great care to create a package where anybody who has a reasonable familiarity with LTspice can run the simulation (from the unzipped folder) with very little hassle, but look at the screenshot.jpg and read the text file before you start. If there is a problem please let me know.

If anybody has any ideas about these power supplies I would appreciate a shout. And, if any of you spice savy engineers can give some pointers as to how to make these simulations more accurate, well, I'd really like to hear about it. (What to do about that darlington??)
Cheers. Happy hunting.

[KingJL]

Problem #1:
Where the 1265 used LM385-2.5 voltage reference, the 1266 uses a B-C junction of TIP126. (The part in the simulation is named TIP127, but ignore that becauses TIP125,6,7 use the same model.)

In the official schematics that I have for the 1265 and the 1266, both use a LM336 2.5V voltage reference.

[KingJL]

I think it is interesting to lay the 1265 and 1266 power circuits side-by-side for comparison, but I am not schooled in electronics enough to be able to explain how these convoluted circuits work. And, I don't have either detector so I can't take measurements to say whether the poor load regulation (problem #2) is just a simulation bogey, or really exists.

I just looked at both schematics in detail. They are identical! No differences between the 1265 and 1266 power supplys.

Regards,
J. L. King

[porkluvr]

I just took another look in my Fisher stash and found a 1266 "low resolution" scat that I had forgotten about. I can't read some part values but I'll figure them out. The pwrsply agrees with the 1265 factory scat that Qiaozhi posted and uses LM336. I don't have a LM336 model and I had to hack LM385-1.2 to give me 2.5V, but it puts out 2.5V and so it be good enough (for me). I have just now renamed the scat to xxx"factory".bmp so I won't over-look that I have it in all my clutter.

As I take another look (at my previous post) it looks like the 1265 power that I posted has two resistors transposed at the comparator (-) input (compared to the 1266) and there is an extra pull-up resistor. The newer 1266 scat that I just found has the pull-up but it is a different value from what my 1265 scat shows. No big deal.

I believe the 15k resistor I showed on the 1266 comparator should be 18k but I can't be sure. I'll make the change in my files but I expect no visible performance difference.

I recently took some time to re-arrange the components in my 1265 power circuit (but I just now took another look at it) and now I see that except for mostly trivial differences (and that dad blasted TIP126) it is the same as 1266. I didn't see that before because I had some things turned around.

KingJL, I looked at your TX simulation. I am interested in the audio transformer. Did you get the coil parameters from actual measurements?? I appreciate that because it's kind of difficult to guess a reasonable inductance value just from looking at an impedance specification or a part number.

[porkluvr]

I feel kind of stupid expecting a power darlington C-B junction to zener at 2.5V so that it would work in the 1266 power supply. It may be a reality, but now that I think about it, wouldn't it need to be a minor miracle?

I don't like breadboarding stuff, but I think I may need to test this concept. I'll add TIP126 to my next parts order.

Let's digress... I never saw TIP126 in the schematic. I saw TI126 and a PNP transistor. Then, a light went off in my head (or maybe it was a fuse blowing).

Stick with what you know works.

[Aziz]

Hi friends,

I have catched the spice model for LM336-2.5 from NI (National Instruments). It doesn't support the adjust pin in the model but could help further.

Aziz

[KingJL]

KingJL, I looked at your TX simulation. I am interested in the audio transformer. Did you get the coil parameters from actual measurements?? I appreciate that because it's kind of difficult to guess a reasonable inductance value just from looking at an impedance specification or a part number.

I got the audio coil parameters from an audio coil specification that matched the actual physical size and all of the electrical specifications that I found on the schematics.

Regards,
J. L. King

[KingJL]

In all the simulations, so long as the (+) and (-) outputs see equal loads, the output voltages remain relatively stable. However, if you vary either supply rail's load and not the other, even by just a few mA., you can see a substantial deviation from the expected symmetry. One voltage goes up, the other goes down. Arggh.

With the circuit set up using the LM336-2.5 voltage reference. The symmetry should remain for a load difference of a few milliamperes. The load variance can't be very great as the symmetry and the analog ground reference is maintained by the LM339 "switching" between the rails as the load/voltage of each rail changes. The current sink capacity of the LM339 is ~mA, so that is the max current for each rail. This is not a problem in the 1265/1266 because the 0 volt analog ground is only used for the Rx signal processing Inc's. The Rx IC's are all ultra-low current devices and draw a max ~15 ma total. All the high current/load circuits (TX, audio) do not use the 0v analog ground reference, but use the chassis ground (-4v, battery -, etc ...all the same) and +4v rail (8v referenced from chassis gnd).

When trouble-shooting/testing a 1265/1266, you must remember which ground reference you want to use to view the signals correctly. And if you have multiple ground connections (e.g. from multiple probes) you must make sure that all are connected to the same reference so that you do not short circuit the -4v rail. Now, how do you think that I came to that conclusion?!!

I actually think the 1265/1266 PS is a clever implementation for what was needed. I believe that if you need a beefier current capability, you could use the output of the LM339 to control a totem of PMOS/NMOS to regulate the 0 volt reference.

Regards,
J. L. King

[porkluvr]

Hi friends,

I have catched the spice model for LM336-2.5 from NI (National Instruments). It doesn't support the adjust pin in the model but could help further.

Aziz

Thanks, Aziz. I think it's amazing that for a simple two or three pin device like these voltage references, how difficult it can be to find working models. I had to flog the ever so dear PSpice to get LM385. I waded through gobs of temperature dependence models before I found something simple enough for my use. And then I had to hack the code to get the right voltage (I think I originally found 2.5V but needed 1.2V, not the other way around). Perhaps it is beneath the dignity of the manufacturers such as TI and NS to construct models of those devices. And even LTC - having its own version of spice, does not offer a model for the LTC1004-1.2 which I believe is very popular and widely used. Waaaahh?

[Aziz]

Hi porkluvr,

that's really difficult to get the right model. I am often using the National Instruments spice software, which has lots of spice models. The models can be exported into a text file (copy & paste). Symbols will not be exported. I often had to edit and check the symbols and pins.
NI software is expiring every 30 days. So every month, a new fresh windows system without the NI software will be re-backupped. It has another benefits: no spyware, no virus, ... it takes only 20 minutes by just almost one mouse click.


Just let me know, if you still missing some other important parts model. I probably could find it. I have downloaded as much as I found in the internet.

Aziz

[porkluvr]

I have realized that I forgot to include the LM385-2.5 model with my simulation file ("vref.lib.txt" was left out). I am sorry about that.

Anybody can create a symbol, but it won't work for anything without some code, uuh-huuh. I have re-edited the symbol's attributes so that it looks for the file LM385-2.5.sub. This is different from the original symbol's annotation, but again - without the code that symbol is NO GOOD.

I am not including the larger file because, even though it had some more good models for other references, it also had some that may not have been tested. And I'd also feel obligated to send appropriate symbols and that takes more time.

Of course, if Aziz's LM336 works, you could just use that instead. I haven't tried the LM336 yet. If it's good, then save the LM385 for a rainy day.

This is the pain about using LTspice, but I still think I like it better than other simulators I've tried. It's not perfect but it ain't bad for being free of charge.

I haven't tested the symbol-model combination I just made and probably won't because I'd need to reconfigure some stuff that i don't want to. But it should work. Let me know if there's trouble.

[Aziz]

Found the LM385-2.5 spice model with symbol file and test schematics (showing reference voltage dependend on temperature). It is based on the LT1004-1.2 model from a german guy (not me). It is obviously supporting the feed back pin (reference). See the .sub or .txt file for convergence parameters.
Happy simulating...

Aziz

[porkluvr]

... based on the LT1004-1.2 model from a german guy (not me).
Aziz

Helmut Senewald is the moderator/(administrator?) of the LTspice users group.
http://tech.groups.yahoo.com/group/LTspice/

WHERE did he get the LT1004-1.2 model, and why isn't it included with LTspice? Well, actually since Helmut's the the LTspice guru next-in-line to Mike Englehardt, he could pull a model out of thin air.

I just looked at the Yahoo group and found the LT1004 series that Helmut posted. Yep, he wrote them.

[ivconic]

No, it is not voltage issue here, at my handmade. I desoldered all the supply parts out from pcb. Than made another on small pcb, providing regulated +4.5 and -4.5 . Handmade is acting the same!? In nonmotion mode it is working excellent (same as original) but in Disc mode detection is weak and phase looks out of range to much. On iron acting like detecting hotrock and no detection or very poor detection on some coloured items. Turning GEB trimmer i can achieve some so so detection on masive colured item with famous Fisher 1265 "bang" in audio, but than it is not detecting other types of coloured metals. For example if it detects masive bronze washer than it is not detecting 2cm coin or Al plate...etc...
Back to nonmotion mode it is detecting nicely all the metals!?
More and more i am sure that there is some mistake on pcb. Need good concentration and to recheck pcb's again and again. I am kinda "used" to these pcb's so it is hard for me to spot the mistake. Need to take few days rest from it...
All other thing seems to be ok, power drain is approx. 40mA, voltages are as noted....in nonmotion mode audio is pretty ok while in Disc mode it is a bit silent..?
Huuuh!

P.S.
Like i assumed; fully simmetrical supply is not needed here. I alredy measured (long time ago) +6.35 and -3.89V on one accurate and good 1265 which worked excellent... So.... kind a returned back to the place where i started!?? Right? Sheeeesh!