Two 4338B high resistance meters fail on the same range - Keysight can't adjust EEPROM of either.
I"m pulling my hair out over an identical problem on two instruments. I am
hoping someone might think of a possible cause I have not thought of. Sorry
the post is a bit long, but this is not any easy problem to describe.
I have an Agilent 4339B high resistance meter. This instrument essentially
consists of a variable power supply (0.1 V to 1000 V) and a very sensitive
ammeter. It works out resistance using Ohms Law. It can display resistance,
current, surface resistively and volume resistively.
The service manual states there are no adjustable components in this - all
calibration is performed using software that updates an EEPROM. The EEPROM
is on the CPU board.
This was sent to Keysight in the UK for a firmware upgrade and calibration.
They updated the firmware (stored in a ROM), but the 4339B failed
calibration.
-
All the output voltages from the internal PSU were within specification.
-
The 10nA current range was slightly out of specification. It was reading
about 0.8℅ high, but the specification is about +/- 0.6%, so it was only
slightly out of specification. (All other ranges were within specification,
but some were not far from the limits. One range might have been 0.5% off)
I think the full scale of the current ranges are 10 pA to 100 uA, so 10 nA
is not at either extreme.
- All resistance measurements were within specification. (Keysight test up
to 1e11 ohms, but it can read up to 1.6e16 ohms. I guess they simply
can't get accurate resistors above 1e11 ohms).
Calibration at Keysight includes any firmware upgrades if you request
updates. It also includes the cost of any adjustments needed - unlike most
calibration labs.
Keysight said the EEPROM could not be adjusted to bring the 10 nA range
within specification, so it needed a new CPU board. I never received a
formal quote for repair, but I was told about £2000 (GBP), which seemed a
lot considering the CPU board is about $600 (USD) from Keysight.
These meters sell for around $3000, but other instruments available for far
less use the same CPU. I was intending repairing my 4339B by swapping CPU
boards from a cheaper instrument, and using a new EEPROM, just in case it
was the EEPROM faulty, as that goes in a socket on the CPU board.
However, I managed to find another 4339B at a good price, so that was
purchased and plans to repair the first instrument were put on hold.
I asked for a quote for calibration based on it having a blank EEPROM. I
thought this would be advantageous, as Keysight could put each range "spot
on". I expected the cost to be a bit higher but it was not.
Much to my surprise, the instrument worked and seemed reasonably accurate
even with the blank EEPROM.
I sent this second 4338B to Keysight for calibration. A couple of days
later i received an email from Keysight telling me the second instrument
has a fault. The fault is on the 10 nA range (as the first instrument) and
the EEPROM can't be adjusted (like the first instrument). This time it is
reading about 0.7% low, which is not much out considering the specification
is about +/- 0.6%.
So I now have two 4339Bs, both being within specification on all ranges
except 10 nA, and neither being adjustable! So naturally I queried why both
instruments appear to have the same fault.
I then received an email from someone st Keysight who had noticed I said
the EEPROM was blank. He asked where did I get the EEPROM from. LUCKILY I
had bought the EEPROM directly from Keysight, despite I could have got s
very similar one from Mouser for a tenth of the price or a supposedly
identical one for even less from China on eBay.
I am hoping to speak to someone at Keysight tomorrow,, but does anyone
have any ideas what could cause two instruments to be slightly out of
specification on the same range, but neither instrument will allow them to
update the EEPROM?
Note one instrument reads high and the other low. I can understand that
perhaps the resistors used in the current to voltage converter on the 10 nA
range might be a bit less stable than used on other ranges, but I can't
understand why Keysight can't bring the meters in spec just by updating the
EEPROM.
Dave
My bet is that they've lost the "secret sauce" for updating the EEPROM :( So it's measuring as best it can with no calibration adjustments stored.
Dave
-----Original Message-----
From: volt-nuts [mailto:volt-nuts-bounces@febo.com] On Behalf Of Dr. David Kirkby
Sent: 02 May 2018 00:47
To: hp_agilent_equipment@yahoogroups.com; Discussion of precise voltage measurement; RSGBTechnical@groups.io
Subject: [volt-nuts] Two 4338B high resistance meters fail on the same range - Keysight can't adjust EEPROM of either.
I"m pulling my hair out over an identical problem on two instruments. I am
hoping someone might think of a possible cause I have not thought of. Sorry
the post is a bit long, but this is not any easy problem to describe.
I have an Agilent 4339B high resistance meter. This instrument essentially
consists of a variable power supply (0.1 V to 1000 V) and a very sensitive
ammeter. It works out resistance using Ohms Law. It can display resistance,
current, surface resistively and volume resistively.
The service manual states there are no adjustable components in this - all
calibration is performed using software that updates an EEPROM. The EEPROM
is on the CPU board.
This was sent to Keysight in the UK for a firmware upgrade and calibration.
They updated the firmware (stored in a ROM), but the 4339B failed
calibration.
-
All the output voltages from the internal PSU were within specification.
-
The 10nA current range was slightly out of specification. It was reading
about 0.8℅ high, but the specification is about +/- 0.6%, so it was only
slightly out of specification. (All other ranges were within specification,
but some were not far from the limits. One range might have been 0.5% off)
I think the full scale of the current ranges are 10 pA to 100 uA, so 10 nA
is not at either extreme.
- All resistance measurements were within specification. (Keysight test up
to 1e11 ohms, but it can read up to 1.6e16 ohms. I guess they simply
can't get accurate resistors above 1e11 ohms).
Calibration at Keysight includes any firmware upgrades if you request
updates. It also includes the cost of any adjustments needed - unlike most
calibration labs.
Keysight said the EEPROM could not be adjusted to bring the 10 nA range
within specification, so it needed a new CPU board. I never received a
formal quote for repair, but I was told about £2000 (GBP), which seemed a
lot considering the CPU board is about $600 (USD) from Keysight.
These meters sell for around $3000, but other instruments available for far
less use the same CPU. I was intending repairing my 4339B by swapping CPU
boards from a cheaper instrument, and using a new EEPROM, just in case it
was the EEPROM faulty, as that goes in a socket on the CPU board.
However, I managed to find another 4339B at a good price, so that was
purchased and plans to repair the first instrument were put on hold.
I asked for a quote for calibration based on it having a blank EEPROM. I
thought this would be advantageous, as Keysight could put each range "spot
on". I expected the cost to be a bit higher but it was not.
Much to my surprise, the instrument worked and seemed reasonably accurate
even with the blank EEPROM.
I sent this second 4338B to Keysight for calibration. A couple of days
later i received an email from Keysight telling me the second instrument
has a fault. The fault is on the 10 nA range (as the first instrument) and
the EEPROM can't be adjusted (like the first instrument). This time it is
reading about 0.7% low, which is not much out considering the specification
is about +/- 0.6%.
So I now have two 4339Bs, both being within specification on all ranges
except 10 nA, and neither being adjustable! So naturally I queried why both
instruments appear to have the same fault.
I then received an email from someone st Keysight who had noticed I said
the EEPROM was blank. He asked where did I get the EEPROM from. LUCKILY I
had bought the EEPROM directly from Keysight, despite I could have got s
very similar one from Mouser for a tenth of the price or a supposedly
identical one for even less from China on eBay.
I am hoping to speak to someone at Keysight tomorrow,, but does anyone
have any ideas what could cause two instruments to be slightly out of
specification on the same range, but neither instrument will allow them to
update the EEPROM?
Note one instrument reads high and the other low. I can understand that
perhaps the resistors used in the current to voltage converter on the 10 nA
range might be a bit less stable than used on other ranges, but I can't
understand why Keysight can't bring the meters in spec just by updating the
EEPROM.
Dave
volt-nuts mailing list -- volt-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
and follow the instructions there.
Perhaps if you share EEPROM P/N we can understand why KS cannot replace or update it?
On May 2, 2018 5:37:23 PM GMT+08:00, "David C. Partridge" david.partridge@perdrix.co.uk wrote:
My bet is that they've lost the "secret sauce" for updating the EEPROM
:( So it's measuring as best it can with no calibration adjustments
stored.
Dave
-----Original Message-----
From: volt-nuts [mailto:volt-nuts-bounces@febo.com] On Behalf Of Dr.
David Kirkby
Sent: 02 May 2018 00:47
To: hp_agilent_equipment@yahoogroups.com; Discussion of precise voltage
measurement; RSGBTechnical@groups.io
Subject: [volt-nuts] Two 4338B high resistance meters fail on the same
range - Keysight can't adjust EEPROM of either.
I"m pulling my hair out over an identical problem on two instruments. I
am
hoping someone might think of a possible cause I have not thought of.
Sorry
the post is a bit long, but this is not any easy problem to describe.
I have an Agilent 4339B high resistance meter. This instrument
essentially
consists of a variable power supply (0.1 V to 1000 V) and a very
sensitive
ammeter. It works out resistance using Ohms Law. It can display
resistance,
current, surface resistively and volume resistively.
The service manual states there are no adjustable components in this -
all
calibration is performed using software that updates an EEPROM. The
EEPROM
is on the CPU board.
This was sent to Keysight in the UK for a firmware upgrade and
calibration.
They updated the firmware (stored in a ROM), but the 4339B failed
calibration.
-
All the output voltages from the internal PSU were within
specification. -
The 10nA current range was slightly out of specification. It was
reading
about 0.8℅ high, but the specification is about +/- 0.6%, so it was
only
slightly out of specification. (All other ranges were within
specification,
but some were not far from the limits. One range might have been 0.5%
off)
I think the full scale of the current ranges are 10 pA to 100 uA, so 10
nA
is not at either extreme.
- All resistance measurements were within specification. (Keysight test
up
to 1e11 ohms, but it can read up to 1.6e16 ohms. I guess they simply
can't get accurate resistors above 1e11 ohms).
Calibration at Keysight includes any firmware upgrades if you request
updates. It also includes the cost of any adjustments needed - unlike
most
calibration labs.
Keysight said the EEPROM could not be adjusted to bring the 10 nA range
within specification, so it needed a new CPU board. I never received a
formal quote for repair, but I was told about £2000 (GBP), which seemed
a
lot considering the CPU board is about $600 (USD) from Keysight.
These meters sell for around $3000, but other instruments available for
far
less use the same CPU. I was intending repairing my 4339B by swapping
CPU
boards from a cheaper instrument, and using a new EEPROM, just in case
it
was the EEPROM faulty, as that goes in a socket on the CPU board.
However, I managed to find another 4339B at a good price, so that was
purchased and plans to repair the first instrument were put on hold.
I asked for a quote for calibration based on it having a blank EEPROM.
I
thought this would be advantageous, as Keysight could put each range
"spot
on". I expected the cost to be a bit higher but it was not.
Much to my surprise, the instrument worked and seemed reasonably
accurate
even with the blank EEPROM.
I sent this second 4338B to Keysight for calibration. A couple of days
later i received an email from Keysight telling me the second
instrument
has a fault. The fault is on the 10 nA range (as the first instrument)
and
the EEPROM can't be adjusted (like the first instrument). This time it
is
reading about 0.7% low, which is not much out considering the
specification
is about +/- 0.6%.
So I now have two 4339Bs, both being within specification on all ranges
except 10 nA, and neither being adjustable! So naturally I queried why
both
instruments appear to have the same fault.
I then received an email from someone st Keysight who had noticed I
said
the EEPROM was blank. He asked where did I get the EEPROM from. LUCKILY
I
had bought the EEPROM directly from Keysight, despite I could have got
s
very similar one from Mouser for a tenth of the price or a supposedly
identical one for even less from China on eBay.
I am hoping to speak to someone at Keysight tomorrow,, but does
anyone
have any ideas what could cause two instruments to be slightly out of
specification on the same range, but neither instrument will allow them
to
update the EEPROM?
Note one instrument reads high and the other low. I can understand that
perhaps the resistors used in the current to voltage converter on the
10 nA
range might be a bit less stable than used on other ranges, but I can't
understand why Keysight can't bring the meters in spec just by updating
the
EEPROM.
Dave
volt-nuts mailing list -- volt-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
and follow the instructions there.
volt-nuts mailing list -- volt-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
and follow the instructions there.
On 2 May 2018 at 10:37, David C. Partridge david.partridge@perdrix.co.uk
wrote:
My bet is that they've lost the "secret sauce" for updating the EEPROM :(
So it's measuring as best it can with no calibration adjustments stored.
Dave
Dave,
Your comment got me to check something, which reveals something VERY
interesting, and is perhaps the source of the problem! I owe you a beer!
I initially assumed the calibration procedure required some software to
check the performance of the 4339B, and if out of specificaton update the
EEPROM. But upon reading the user and service manuals, I find that's NOT
the case.
The user manual describes how to check the performance, which requires no
software. The service manual describes how to adjust the instrument if the
performance is not right. The adjustment which needs software.
Furthermore, it seems to me the procedure used by Keysight to verify the
performance maybe wrong, although I am not going as far to say it is, as
maybe they are doing something that's not obvious from the calibration
certificate. Looking in the user manual, it would appear one is supposed to
verify the performance on the 10 nA range by setting the output voltage of
the 4339B to 1 V, measure the voltage on a 3458A, set a resistance box to
10^8 ohms and calculate the current, which should not be assumed to be 10
nA unless the 3458A indicated 1.00000000 V, I have not checked the
specification of the 4339B when generating 1 V, but when set to 0 V, it
should output 0 +/- 0.1 V and when set to 10 V it should output 10 +/- 0.12
V. So at low voltages, this is not an accurate voltage source. At higher
voltages it is a lot more accurate. This reflects my experience when using
it to measure close tolerance resistors - measuring them at low voltages
gives poor results, but at higher voltage, the resistances are measured
more accurately.
I am wondering if Keysight are assuming the 10 nA is generated when the
output voltage is set to 1.0 V.
If out of specification, the EEPROM should be updated with software. I'm
assuming that software is written to compute current based on what voltage
the 3458A measures, rather than what voltage the 4339B is set to output.
To answer Illya's questions, the EEPROM is a Xicor X28C64P-20. I can't find
any reputable supplier with those - plenty on eBay, but I am always
concerned about counterfeits on there. Someone suggested a Mouser P/N.
AT28C64B-15PU would probably do, but I decided to buy the EEPROM directly
from Keysight (P/N 1818-4808). I'm glad I did, as it is removed any doubt
about the suitability of the EEPROM. I'm sure if I had bought the EEPROM
from Mouser or eBay, Keysight would have suspected an incompatibility of
the EEPROM.
This has certainty got me thinking about what might be happening, although
the fact Keysight have said they can't update the EEPROM, suggests they
have the software to do it.
Dave
"This reflects my experience when using
it to measure close tolerance resistors - measuring them at low voltages
gives poor results, but at higher voltage, the resistances are measured
more accurately."
I noticed similar results in general, over years ad several instruments.
Always wondered why, but we have a rule-of-thumb: if the part is to be
used at X KiloVolts, test it at X Kilovlolt.
My 34470A struggles with measuring 1 G Ohm resistors, placing it in "low
power" helps, but not to my satisfaction. The HP/Yokogawa 4329A we have
is stable as a rock, but of course analog output/ limited resolution. I
will probably set up an old school work station to measure these 1G Ohm
with a bridge and standard resistor made up of 10 99M precision
resistors which we also use a lot, with a trim pot.
On another instrument at the other end of he spectrum, my HP4328A
milliOhm meter is having battery issues.
Does anyone know if it can be run safely without the battery i it? I
hate to destroy the calibration labels to get inside and look for
myself.
George Dowell
On 2 May 2018 at 14:25, geoelectronics@rallstech.net wrote:
"This reflects my experience when using
it to measure close tolerance resistors - measuring them at low voltages
gives poor results, but at higher voltage, the resistances are measured
more accurately."
I noticed similar results in general, over years ad several instruments.
Always wondered why, but we have a rule-of-thumb: if the part is to be
used at X KiloVolts, test it at X Kilovlolt.
I am sure your advice is good, but I based measurements on the best
resistors I could find at sensible prices from Farnell.
The issue on the 4339B is the accuracy of the internal 0-1 kV voltage
source, if set to low voltages. I don't have the time at the moment to hunt
for the exact specifications, but the calibration certificate give the
output voltages the meter is set to, and the voltage limits. I calculated
them as a percentage.
0 V -> +/- 0.1 V (infinite percentage)
10 V -> +/- 0.12 V (+/- 1.2%)
25 V -> +/- 0.14 V (+/- 0.56%)
50 V -> +/- 0.18 V (0.36%)
100 V -> +/- 0.26 V (+/- 0.26%)
...(I will miss out 200 V, 201 V, 250 V for safe of brevity)
500 V -> +/- 1.3 V (+/- 0.26%)
1000 V -> +/- 2.1 V ( +/- 0.21%)
The basic uncertainty of the instrument is 0.6%, but clearly if the output
voltage is set to a low value, the percentage error in the voltage is high,
so the percentage error in the resistance will be high. The above would
suggest using a voltage under 50 V is going to compromise accuracy and
using 100 V or more is better.
The uncertainty of the ammeter also depends on the range it is on, and not
surprisingly that has a higher percentage error on its lowest range (100
pA) than on its highest range (100 uA).
The Agilent 4339B is said to work from 1e3 to 1.6e16 ohm. Clearly to
measure 1000 ohm, any voltage above 0.1 V would exceed the full scale of
the ammeter on its least sensitive range (100 uA). Setting the source
voltage at 0.1 V is likely to result in significant errors reading a 1000
ohm resistor. But clearly a "high resistance meter" is not designed to
measure 1000 ohm resistors. I expect a £5 handheld multimeter from China
would do a better job at measuring 1000 ohm than what this instrument
does.
Sorry, I don't know about the HP 4328A. Personally I would have no concerns
about breaking the seals on an instrument of that age. I would consider it
prudent to check for any leaking electrolytic capacitors or other nastiness
that may reside inside an old instrument.
Dave
Thank you I will open it. I was not trying to give advice,
rules-of-thumb are simply shop guidelines we use to deal with our
inadequate equipment (which are may when it comes to high precision).
George Dowell
On 2018/05/02 09:36 AM, Dr. David Kirkby wrote:
On 2 May 2018 at 14:25, geoelectronics@rallstech.net wrote:
"This reflects my experience when using
it to measure close tolerance resistors - measuring them at low voltages
gives poor results, but at higher voltage, the resistances are measured
more accurately."
I noticed similar results in general, over years ad several instruments.
Always wondered why, but we have a rule-of-thumb: if the part is to be
used at X KiloVolts, test it at X Kilovlolt.
I am sure your advice is good, but I based measurements on the best
resistors I could find at sensible prices from Farnell.
The issue on the 4339B is the accuracy of the internal 0-1 kV voltage
source, if set to low voltages. I don't have the time at the moment to hunt
for the exact specifications, but the calibration certificate give the
output voltages the meter is set to, and the voltage limits. I calculated
them as a percentage.
0 V -> +/- 0.1 V (infinite percentage)
10 V -> +/- 0.12 V (+/- 1.2%)
25 V -> +/- 0.14 V (+/- 0.56%)
50 V -> +/- 0.18 V (0.36%)
100 V -> +/- 0.26 V (+/- 0.26%)
...(I will miss out 200 V, 201 V, 250 V for safe of brevity)
500 V -> +/- 1.3 V (+/- 0.26%)
1000 V -> +/- 2.1 V ( +/- 0.21%)
The basic uncertainty of the instrument is 0.6%, but clearly if the output
voltage is set to a low value, the percentage error in the voltage is high,
so the percentage error in the resistance will be high. The above would
suggest using a voltage under 50 V is going to compromise accuracy and
using 100 V or more is better.
The uncertainty of the ammeter also depends on the range it is on, and not
surprisingly that has a higher percentage error on its lowest range (100
pA) than on its highest range (100 uA).
The Agilent 4339B is said to work from 1e3 to 1.6e16 ohm. Clearly to
measure 1000 ohm, any voltage above 0.1 V would exceed the full scale of
the ammeter on its least sensitive range (100 uA). Setting the source
voltage at 0.1 V is likely to result in significant errors reading a 1000
ohm resistor. But clearly a "high resistance meter" is not designed to
measure 1000 ohm resistors. I expect a £5 handheld multimeter from China
would do a better job at measuring 1000 ohm than what this instrument
does.
Sorry, I don't know about the HP 4328A. Personally I would have no concerns
about breaking the seals on an instrument of that age. I would consider it
prudent to check for any leaking electrolytic capacitors or other nastiness
that may reside inside an old instrument.
Dave
volt-nuts mailing list -- volt-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
and follow the instructions there.
Well, that "fixed" my instrument. There was no battery inside. Rather
the "battery test" was reading the charge on the main power supply
capacitor. It must be of high quality because it keeps a charge for a
considerable time.
Everything works just fine ad tracks precisely with my bench instrument.
Because I have now 2 instruments, I feel secure. Today I had to test
some new
IET 1 Ohm and GR old stock .01 Ohm decade switch modules
George Dowell
On 2018/05/02 09:45 AM, geoelectronics@rallstech.net wrote: Thank you I
will open it. hop guidelines we use to deal with our
inadequate equipment (which are may when it comes to high precision).
George Dowell
Dave wrote:
This has certainty got me thinking about what might be happening, although
the fact Keysight have said they can't update the EEPROM, suggests they
have the software to do it.
Or does it mean they have lost the software to do it? Does the cal
certificate have "as received" and "as adjusted" error readings for
other ranges? If not, perhaps they adjusted nothing, which would leave
open the possibility that they simply cannot adjust these meters anymore
due to lack of software.
Best regards,
Charles
On 2 May 2018 at 18:27, Charles Steinmetz csteinmetz@yandex.com wrote:
Dave wrote:
This has certainty got me thinking about what might be happening, although
the fact Keysight have said they can't update the EEPROM, suggests they
have the software to do it.
Or does it mean they have lost the software to do it? Does the cal
certificate have "as received" and "as adjusted" error readings for other
ranges? If not, perhaps they adjusted nothing, which would leave open the
possibility that they simply cannot adjust these meters anymore due to lack
of software.
Best regards,
Charles
I don't have a cal certificate - only a report of how it was received
http://www.kirkbymicrowave.co.uk/tmp/Measurement_report_As_received_1-9957310275-1.pdf
and an email saying it can't be updated.
================
Hi David,
We’ve a High Resistance Meter for calibration, but the instrument is faulty.
Model: 4339B
Serial Number: JP1KD01746
Service Order: 1-9957310275
The calibration has failed, I’ve attached a copy of the calibration
results.
Automated adjustments were attempted. Source Voltage Adjustment could be
completed ok, but the Ammeter Adjustment could not be completed.
I’ve passed the instrument to our repair team who will investigate the
fault, and arrange for a repair quote to be sent out.
The fact the source voltage could be adjusted suggests they could update
the EEPROM, but I'm puzzled why the ammeter adjustment could not be
completed. I'm not sure if they have updated the source voltage at this
point. Every voltage is low, but all are in spec. I'd hope they would be
able to increase the output voltage a little there.
For what it is worth, here's the cal certificate of one I sent in a month
or so ago.
http://www.kirkbymicrowave.co.uk/tmp/1-9690444179-1-combined-file.pdf
I was told that one needs a new CPU. I might have some more information
tomorrow, as Keysight (UK) are contacting the USA for support.
Dave