Tesla MAC199/MAE299/MAB399 and MAC01 : teardown and tests

[TiN]

Toy box from Czesh republic got into my hands from one of my readers
Knowing my precision-sickness there are no less than 12 x MAB399 ovenized 7V references, which are on first glance clone of LM399 series, one MAC01 (10V trimmable reference in metal can 8-pin LTZ-style package and pair of Pt100 RTDs.



I was wondering if these MAB399 any good, as they are rather cheapies (if one can find them .



Let's take a look on datasheet first..



Package is much taller (both zener can and outer plastic box), but following same style as LMx99. Nominal output voltage is bit higher, 7.25V +/-5%, same dynamic impedance <1.5Ohm, <5ppm/K TC (LT LM399 rated <2ppm/K for -A version, <1ppm/K f or regular), no max specification for noise on MAB399.

Now let's look under the hood. I milled top of the can on one of MAB399's and took quick photo:



Stay tuned..

P.S. I know there is another MAB399 thread in here, but decided not to necropost.

[TiN]

TBD.

We can see die marking has MAC199, so it mean all three chips are same, just the sorting/binning and/or ageing is different on those SKUs.

For comparison LM399 photo, provided by Gyro in LM399 thread:

[Vgkid]

I wait for updates.

[quarks]

Looks interesting
Thanks for sharing

[branadic]

There has been another thread about MAC199 before:

https://www.eevblog.com/forum/projects/tesla-mab399-mac199-voltage-references/

[saturnin]

@TiN

Thanks for great teardown of MAB399 reference! I always wondered how its layout looked like. I would like to add some information:

These references were produced by Tesla, a national semiconductor and electronics manufacturer in former Czechoslovakia, Roznov pod Radhostem (It was a company with very long history - established in 1946). Unfortunately, MAC199/MAB399 production didn't last for long - probably only a couple of years around 1989, then an economic collapse came to Eastern Europe (triggered by collapse of USSR).

The semiconductor fab in Roznov was shut down and it seemed to be the end of production of semiconductors in Czechoslovakia. Hopefully, shortly after the shut down, Tesla semiconductor division was bought by Motorola. The reason was obvious: manufacturing costs were ridiculously low, yet quality was good. Next milestone came in 1999, when Motorola semiconductor division split into two companies - Freescale and On Semiconductor. Both companies are still present in Roznov, although only On Semiconductor has manufacturing facilities there (both silicon wafers and ICs are manufactured in Roznov).

Back to references: I have several pieces of MAC199/MAB399 and there is a difference: MAC199 got a little bit better care since it has gold plated legs (see attached photo, I added LM199 for comparison too). I also enclose MAC199 and LM199 internal schematics - MAC199 is not a direct copy of LM199, but an "inspiration" is obvious .

[TiN]

Thank you for your infos, great read! I always struggle to find history on most of legacy items. May I use this information in my upcoming article about these TESLA references (surely credit to you and link to your post will be given!)?

I remember TESLA parts since I was a 10-year old school kid, some BJTs and resistors. I remember elder saying that's a real stuff when I was poking around if these TESLA things any good.

[saturnin]

Yes, sure you can use it.

As concerns quality, I wouldn't say it was bad. Yes, TESLA was 5-10 years behind western countries and their products were bulky and often with low level of integration, but they still work nowadays. Some measuring equipment - oscilloscopes and multimeters - are still popular among hobbyists. It is better to buy old TESLA equipment (you can rely on its specs) than new Chinese crap produced nowadays.

To the references:

1/ Some time ago, I made several tests with MAB399 and if I remember correctly it took it longer to stabilize than LM199 (I mean after power-up, not long-term).

2/ MAC01 is a clone of REF01. Its datasheet is here: http://www.teslakatalog.cz/MAC01.html

I also attached several pictures from a book about history of Tesla (Notice that there were working mostly women in positions where precision was needed - men are not patient enough )

[TiN]

Got some of MAB399's going with just 10.7K 25ppm/K resistor in series, hooked to same K2400 +15V output.



All output are normalized to common point, by subtracting sample at 21/04/2016-06:19:26

Code: [Select]

var dsvw = d3.dsv(";", "text/plain");
dsvw("/datashort/scan_2002_ref10_nplc10.csv", function(error, data) {
    data.forEach(function(d) {
        d.date = parseDate(d.date);
        d.temp = +(d.ch1 / 0.0002) - 273.15;
        d.ch2 = +(d.ch2 - 6.9241967 ) * 1E6;
        d.ch3 = +(d.ch3 - 6.8754749 ) * 1E6;
        d.ch4 = +(d.ch4 - 7.1905522 ) * 1E6;
        d.ch5 = +(d.ch5 - 7.2799687 ) * 1E6;
        d.ch6 = +(d.ch6 - 7.3355209 ) * 1E6;
        d.ch7 = +(d.ch7 - 7.2061466 ) * 1E6;
        d.ch8 = +(d.ch8 - 7.2229355 ) * 1E6;
        d.ch9 = +(d.ch9 - 7.3331865 ) * 1E6;
        d.ch10= +(d.ch10 - 7.184060) * 1E6;
    });
6 x MAB399's from CH2 to CH7
1 x LM399 on CH8
1 x LM399 from my HP3245 on CH9
LTZ1000 output of A9 HP3458A PCBA on CH10 connected by long bodge wire (it's in another side of room)

All sampled by Keithley 2002 with TCSCAN-2001. TEMP? is CJ output temperature of onboard scanner AD590.

URL for Live data graph

RAW DATA.

[zlymex]

Very interesting device.
From the scanned data(btw, it is not a CSV-Comma Separate, but a semicolon separated, but I can read anyway), I can see that three thing drifting all together: the meter, LTZ(ch10), and those MAB399, this makes things difficult to seel which is which. However, there are four things I can tell:
- The meter(2002) has a sudden downward jump at 18?15
- LTZ1000(ch10) is drifting upwards all the time
- The drift of the meter is no better than those MAB399
- Those MAB399(and LMs on ch8 and ch9) perform similar except ch5 from 16:56 to 18:15.

[TiN]

I do agree that it's impossible to tell what is what.

Here are some extra points - LTZ module itself is stable, it was tested over 2 weeks non-stop, both with temperatures ramp from +23.4C (calibrated point) to +50C and back. Here's older graph data, taken with 3458. Sorry, graph JS does take a bit to load, it's showing every sample over 7 days period. So I verdit effect of ramping from thermal EMFs on connections and ambient rise. To confirm this, I just turned on TEC SMU to bring LTZ chamber box temperature to stable +23.4°C (ambient is +30.4°C right now).

Dip on 18:15 was aircon in the room, which was turn on for 30 minutes. All other times - no aircon, just ambient temp. You don't see temperature plot (AD590 on TCSCAN board) on the graph in post above, but it's on live page, blue line.

This 2002 should be rather stable at given temperature, as I have it often turned on and running at least half of the each week.

I'll have this running this night as it is, and perhaps tomorrow night I'll change meter to 3458, just for comparisons. Scan card and setup will be still in 2002, the only difference would be output of scan card routed not to back of 2002, but to 3458 instead.

[zlymex]

It's amazing that you have a live chart.
The thick orange curve is going up all the time, indicating either the LTZ1000 is going up, or the meter is going down.
I redraw the chart regarding ch10 as constant. Left vertical is in ppm.
1. The noise a bit high at about 1ppm pp
2. The downward change must be the meter
3. The wave-shaped curves must be the meter too as those 399 are unlikely to perform in the same pattern.
4. ch5 is the worst
5. My chart was drawn in Excel and show all the raw data, no smoothing or averaging involved.