EEVblog Electronics Community Forum
Electronics => Repair => Topic started by: Samogon on March 19, 2018, 04:22:36 pm
-
Got my first 3458A.
According to the dates on the chips it is 1999 year build: most of the chips are 98' and on one of the aluminum shields we it has 1999 print which support this conclusion
It passes self test no errors, but at startup it shows 'err' annunciation. In error list it shows 110 "Calibration required." According to Agilent's service note 3458A-15C
In best case scenario i can get away with replacement of non-volatile CAL RAM ICs.
First step is to change AC Inlet filter to avoid well known IED - Huffner
While i am waiting for RAM chips here are some pictures of unit's guts.
-
Well calibration data is lost i did copy it to new NVRAM U123 was dead and i was not able to read at all.
Rest i read and copied but it appeared to me garbage.
So it will be calibration project.
But before calibration i will test if ADC stability.
Will use 10V 6ppm source and K2002 as control unit. Crossing fingers.
-
Besides the stability with an external reference source, one should also check the stability / drift of the CAL72 values. The ADC board can show a kind of defect (main ASIC drifting) that can shows up first as excessive drift in the CAL72 values.
-
You mean service note 3458A-18A ?
Yah will do that first.
So week of just ACAL every 24 hours.
Thanks
-
Yes, SN18 good to check U180 on A3 too :)
I have fingers crossed. As usual, only me get "lucky" with need of A3 board replacements and all of you guys getting 3458A with just NVRAM swap :-DD.
-
You are more experienced and have guts to take on heavily busted units.
Most of your repairs look scary.
-
Oh c'mon, it's just a boring board swap for $1300 :).
I got burned with one of repairs (scope), which I'm sorta stuck now.. so it's not all gold that shines.
-
Will update message with daily data.
So My K2002 measured source as 10.000525 with last digit fluctuating +/- 3uV
Here is chart i got from 3458A
1. C72 1.00412099 after second ACAL in last 24 hours
2. C72 1.00412262 03/21/2018 2:23pm
3. C72 1.00412304 03/22/2018 2:33pm
4. C72 1.00412199 03/23/2018 2:15pm
----
5. C72 1.00411197 03/28/2018
6. C72 1.00410887 03/29/2018 1:50am
7. C72 1.00410845 03/30/2018 12:25am
8. C72 1.00410734 03/31/2018 1:12am
-
Still reading docs, but looks like my data form last night are screwed, i was doing ACAL and to set unit to local control i used button "local" then screen showed me all dashes, so i hit "reset" and it reset unit and i got control did ACAL read CAL?72 value and then started collecting data. But before this operation i had -0.2 ppm to DC source value measured by K2002, after rest i getting -2.6ppm. It looks like it started again.
Went RTFM :palm:
-
What is the time scale of the graph. Never omit timescale on the graphs ;).
Doesn't look too bad or alarming to me yet, you have just initial settle stuff.
ACAL constants are preserved on resets in 3458A.
-
It is excel chart in this new version i totally lost how to display X axis
start is 19/03/2018 17:37:20
end is 20/03/2018 12:37:22
roughly 19 hours
-
it will be a metrology thread soon.
-
This is chart for three consecutive days but i was doing ACAL every day, so it introduces some inaccuracy and spikes. Temperature this days changes significantly but it is correlates with measurements only last day shows some strange drift in opposite direction.
-
Looks like i am facing A3
Board replacement, drift is 1.5ppm a day according to CAL72 values i measured. :(
-
Uh-oh... feeling your pain.
If you have other DMM with scanner (even 6.5digit will do fine), it can be easier to confirm U180 A3 drift by measuring ADC reference voltages (+12V, -12V) together with A9 7V output. On faulty U180 you will see voltages not tracking each other at same rate.
Here's how A3 drift looked like on my second 3458A
(https://xdevs.com/doc/HP_Agilent_Keysight/3458A/u2/a3u2_fail.png)
You see +12V is stable, but other refs are out of whack.
-
Looks like i am facing A3
Board replacement, drift is 1.5ppm a day according to CAL72 values i measured. :(
Wait a moment,please.
Which reference do you use (10V, 6ppm???).
Is the environmental temperature absolutely constant?
The AN 18 leaves our completely, that CAL? 72 constant depends also on both parameters.
If you monitor additionally the internal temperature TEMP?, you may get a rough idea of the T.C. of your 3458A, for 10VDC, w/o ACAL.
That dependency you have to subtract, before you judge the timely manner drift of your
AD chip.
Rough idea, because the T sensor sits on opposite PCB.
Frank
-
Will upload internal temperature and voltage measured in the morning
Ambinet Temparature was changing drammatically like 7 degree during measurement period
-
Periodically i measured 10v source with k2002 and it drifted itself.
Peobably will have to redo test when AC will start working whole
24 hours. It will ensure that temperature will be in 24C +/- 1 degree
-
modified TiN's script to read data from 3458 which is on his web site https://xdevs.com/guide/ni_gpib_rpi/#setup (https://xdevs.com/guide/ni_gpib_rpi/#setup)
Now my reference will be k2002 measurements instead of fixed ref value. Of course if i had fluke 732A i could rely on single measurement, but my 10V source looks not that stable.
# xDevs.com Python 10V test for 3458A
# https://xdevs.com/guide/ni_gpib_rpi/
# https://xdevs.com/article/hp3458a_gpib/
# https://xdevs.com/fix/hp3458a/
import sys
import Gpib
import time
inst = Gpib.Gpib(0,22, timeout=60) # 3458A GPIB Address = 3
k2002 = Gpib.Gpib(0,7, timeout=60)
inst.clear()
k2002.clear()
#Setup HP 3458A
inst.write("PRESET NORM")
inst.write("OFORMAT ASCII")
inst.write("DCV 10")
inst.write("TARM HOLD")
inst.write("TRIG AUTO")
inst.write("NPLC 200")
inst.write("AZERO ON")
inst.write("LFILTER ON")
inst.write("NRDGS 1,AUTO")
inst.write("MEM OFF")
inst.write("END ALWAYS")
inst.write("NDIG 9")
#setup k2002
k2002.write("PRESET NORM")
k2002.write("OFORMAT ASCII")
k2002.write("DCV 20")
k2002.write("TARM HOLD")
k2002.write("TRIG AUTO")
k2002.write("NPLC 200")
k2002.write("AZERO ON")
k2002.write("LFILTER ON")
k2002.write("NRDGS 1,AUTO")
k2002.write("MEM OFF")
k2002.write("END ALWAYS")
k2002.write("NDIG 9")
cnt = 0
tread = 2
temp = 38.5
inst.write("TEMP?")
k2002.write("TEMP?")
temp = float(inst.read())
temp2002=float(k2002.read())
ppm = 0
with open('10v_3458_nplc200_mm.csv', 'a') as o:
o.write("date;hp3458a;temp;k2002,temp2002;ppm_level;\r\n")
o.close()
while cnt <= 10000000:
cnt+=1
with open('10v_3458_nplc200_mm.csv', 'a') as o:
tread = tread - 1
if (tread == 0):
tread = 20
inst.write("TARM SGL,1")
k2002.write("TARM SGL,1")
inst.write("TEMP?")
temp = inst.read()
k2002.write("TEMP?")
temp2002 = k2002.read()
inst.write("TARM SGL,1")
data = inst.read()
k2002.write("TARM SGL,1")
data2002 = k2002.read()
ppm = ((float(data) / float(data2002))-1)*1E6
inst.write("DISP OFF, \"%3.3f ppm\"" % float(ppm))
time.sleep(1)
print time.strftime("%m/%d/%Y-%H:%M:%S;") + ("[%8d]:3458A %2.8f ,k2002 %2.8f, dev %4.4f ppm, T:%3.1f, T2002:%3.1f" % (cnt, float(data), float(data2002), float(ppm), float(temp), float(temp2002) ) )
o.write (time.strftime("%m/%d/%Y-%H:%M:%S;") + ("%16.8f;%3.1f;%16.8f;%3.1f;%4.3f;\r\n" % (float(data),float(temp),float(data2002),float(temp2002),float(ppm) ) ))
o.close()
-
Ok, strange happens.
As you probably remember i lost cam ram content, basically it was missing on arival
And i was running stability tests for last few weeks with different results even with drifts like 5ppm a day
I have provided calibration with Zero, 10V, 10kOhm and ACAL AC. And instrument changed its behavior it now stays within 0.1ppm referenced to my k2002.
Still has error on power on "110 calibration required". Manual says i need to do SCAL and to do that i have to have thermal transfers :horse:
[size=8pt]Equipment required:
• Keysight 3325A Synthesizer/Function Generator or equivalent.
• 3V Thermal Converter, Ballantine 1395A-3 or equivalent.
• 1V Thermal Converter, Ballantine 1395A-1 or equivalent.
• 0.5V Thermal Converter, Ballantine 1395A-0.4 or equivalent.
• 50? BNC cable (keep this cable as short as possible)
• 50? resistive load (typically a 50? carbon composition or metal film
resistor).
• BNC to Banana Plug Adapter--Keysight 1251-2277 or equivalent.[/size]
It there other way to do SCAL on the unit?
-
Hmm, if i read correctly, then having k2002 for reference and measuring Gen provided RMS values i can eliminate need of thermal transfers.
So procedure will be set gen output to to k2002 tune it to be in 0.2% accurate and then apply it to 3458A. Hehe.
Will do that tomorrow
-
I have done it that way as a rough calibration. I have used a 33120A and used a Fluke 8920A with the rear analog out connected to a meter. Once you get to 2MHz or 8MHz, the dmm will be useless.
I now have a 3325B and also like it better than the 33120A.
If you need to borrow a 8920A, I can arrange it.
-
What is the baseline noise of the Fluke 8920A thermal meter? What is the noise spread/bit 'salad' when hooked into a high res meter?
-
yeah you are right k2002 it up to 2MHz :(
I see tons of Flukes 8920A on ebay. i think we will spend more sending your unit to me and back.
I will save your generosity for the future. :-+
-
What is the baseline noise of the Fluke 8920A thermal meter? What is the noise spread/bit 'salad' when hooked into a high res meter?
3% at 10MHz accuracy
-
I have better idea.
I will use power meter 4418B with sensor 8482A and attenuator 30dB
At least it will give me rougly .5% uncertainty way better than 3% still not that good as expensive thermal converters.
-
Well, overnight it drifted again for almost 2ppm.
-
What drifted? ADC relative to internal reference (CAL72)? Internal reference relative to Keithley 2002 or other external reference (how do you know which of the two drifted)? Is this with our without ACAL every 24h/1°C change in temperature?
-
Conclusion based on collection of data and daily CAL72 value change baded on formula probided in sevice note 18A where seven day drift should be below 0.43ppm per day mine showed 1.5ppm
-
For the initial few weeks after the meter was not used for a long time, there may be some extra drift, that may be not so critical. There can also be some temperature dependence on the cal72 value. To really get reliable drift of cal72 it would need values at the same temperature.
-
Ambient temperature was stable +/- 1 degree. while CAL72 was steadily drifting down and no temperature correlation.
Per SN18 it is enough to run 7 days test to see if CAL72 out of tolerance.
The only thing i did not test is measuring directly internal reference stability.