EEVblog #406 - Keithley 480 Picoammeter Teardown & Calibration

[EEVblog]

Dave.

[CarlG]

Yeah, you get a real big 

I love that you show all of these "good old" instruments. You can get a lot of old instruments having features that may be hard to find in modern equipment. One of the first things I got added to the lab when I started my current work, was a Fluke 8921A true RMS voltmeter, first design late in the 70's I think (manual dated Oct -78), but I see there are newer versions (still 8921A).

The 8921A is 4 1/2 digits, has up to 20 MHz bandwidth, 700V max input, and resolution down to 0.1 1 uV on lowest range, and also selectable dB scale. Great stuff! Very useful for measuring residual noise on (switching) power supplies for instance.

//C

[SeanB]

I would say the power switch had dirty contacts. Looking at the insides you can see corrosion products on the back of the gold plated contacts inside the shield, so there was likely an exposure at some time to an atmosphere that was humid and contained thiols ( probably downwind of a coal power plant or industrial plant that used coal or a oil refinery that was processing heavy crude oil or near a harbour) that contaminated the boards. They would benefit from cleaning with a contact cleaner, along with the range switches, as you can see on the video they are noisy when operated ( decimal point flicker when range changing and flicker on power on) but have wiped themselves clean after a few operation cycles.

[McMonster]

The strange cable mounting is there so you can nicely wind the cable around the case stands. I don't like equipment with permanent mains or any other cables sticking out of it, and I carried a lot of it at work. Really annoying unless you have a third hand to hold the cables, otherwise it is easy to trip over the cable.

[Bloch]

I am looking at the schematic and there must be some thing i do overlook

How is it possible to have so low burden voltage then the resistor array go from 1K to 99M ?

[robrenz]

Great video Dave! As always I learned something new

[CarlG]

I am looking at the schematic and there must be some thing i do overlook

How is it possible to have so low burden voltage then the resistor array go from 1K to 99M ?

Well....here's a clue: Q104 + U104 acts as a difference amplfier. So what voltage do you get at the gate of Q104-A?

//C

[SeanB]

I am looking at the schematic and there must be some thing i do overlook

How is it possible to have so low burden voltage then the resistor array go from 1K to 99M ?

The resistor arrays are the feedback loop of an opamp arranged to invert. The input is connected to this virtual earth point, and the opamp tries hard ( via open loop gain) to keep this voltage equal to ground ( non inverting terminal) by sinking/sourcing current via the resistors in the feedback loop. This means any input current is translated to a voltage on the opamp output  while the input is kept very close to ground. The meter then simply shows this output voltage suitably scaled and you interpret it as a current.

[bitwelder]

The mysterious empty DIP connector near the big IC seems to be another diagnostic connector (check J1002 on the right size of diagram) with basically all the digital communication with the display board available on its pins.

Looking at the manual, the apparently unused upper ground clip maybe has a real function when the battery pack gets installed.

[quarks]

Hello Dave,
I just watched this very nice video  and I hope you will do the video you are talking about accurate low value resistance measurement (45:40 min.), because that is what I am just evaluating myself.
 
BTW Happy new year to all.

[Bloch]

Ok I see.


But only if S107 is always closed. So my question is now that is S107 funktion?

[Bloch]

But only if S107 is always closed. So my question is now that is S107 funktion?

Zero check.

Well....here's a clue: Q104 + U104 acts as a difference amplfier. So what voltage do you get at the gate of Q104-A?

I am not sure my FET knowledge is not that good.

But the Q104 was a part off the difference amplifier ... right ?

That are Q102+Q103 doing ?

[LLR1967]

Bloch: "That are Q102+Q103 doing ?"

If you're asking what the purpose of Q102/Q103 is, they are used as low-leakage diodes for input protection.

[CarlG]

But the Q104 was a part off the difference amplifier ... right ?

Right. Q104-U104 act as an OP-amp. So if the gates of Q104 were the inputs to an ideal OP-amp, what is the voltage between the gates?

//C

[CarlG]

Bloch: "That are Q102+Q103 doing ?"

If you're asking what the purpose of Q102/Q103 is, they are used as low-leakage diodes for input protection.

And that's also a clue to how the circuit operates. As stated above, Q102/Q103 are for protection, not needed for the basic operation. Maybe easier to understand if you remove them and short the 1M?

//C

[Bloch]

Thanks guys

I have the flu today not a excuse    just saying.

In 1mA range selected

Gain = R110/R107 = 1M / 1K = 1000
Vout =  Gain * (Vin+ - Vin-) =  1000 * (0 - Vin-)

Not sure that the Vin will be. `

But Vout will be inverting voltage. That will cancel out the burden impedance..... Am i close 

[Rick]

Time to buy a "for parts or repair" type Keithley 480 and try a repair...
At this trend I shall financially ruin myself.
I simply love those repair sessions. It's crazy how many devices I threw away whereas they could have been fixed very simply...
I have acquired a dead Fluke 8800A, that I shall try to resuscitate with my "big" repair experience when it arrives. That will be my first 5 1/2 digit and bench multimeter, if it works. Fluke not Keithley of course.

[CarlG]

Thanks guys

I have the flu today not a excuse    just saying.

In 1mA range selected

Gain = R110/R107 = 1M / 1K = 1000
Vout =  Gain * (Vin+ - Vin-) =  1000 * (0 - Vin-)

Not sure that the Vin will be. `

But Vout will be inverting voltage. That will cancel out the burden impedance..... Am i close 

Yep, you've got it  Good job! Have a look at Fig 2-3.

Q104 and U104 will ideally sink/source current in the FB resistor in order to keep Q104-A base at the same potential as Q104-B, which happens to be at ground. I.e the instrument input will be at ground potential. The burden voltage will mainly depend on the matching of the Q104 pair and Q104 source resistors (U104 trimmed to zero readout by R110). The offset voltage due to load/temperature dependent leakage current in Q102, Q103, and Q104-A via R110 is more or less negligible (a wild guess ~10pA * 1M << the specified 200uV).

//C

[sorin]

what is the part number of Q102-Q103 ?

[LLR1967]

I have acquired a dead Fluke 8800A, that I shall try to resuscitate with my "big" repair experience when it arrives. That will be my first 5 1/2 digit and bench multimeter, if it works. Fluke not Keithley of course.

I have one similar to that one at work (it's been there for >30 years), and I fixed it by replacing, as one might expect, all of the electrolytic capacitors. I was told that A.C. was an option on that meter; it has the switch, but it is locked and can't be moved. Also, in my case, the meter wasn't dead. It was very unstable.

Larry

[Rick]

I have acquired a dead Fluke 8800A, that I shall try to resuscitate with my "big" repair experience when it arrives. That will be my first 5 1/2 digit and bench multimeter, if it works. Fluke not Keithley of course.

I have one similar to that one at work (it's been there for >30 years), and I fixed it by replacing, as one might expect, all of the electrolytic capacitors. I was told that A.C. was an option on that meter; it has the switch, but it is locked and can't be moved. Also, in my case, the meter wasn't dead. It was very unstable.

Larry

Ok.
Mine is completely dead I think:
http://www.ebay.com/itm/380444720310?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1439.l2649
Note the shipping cost is more than twice the cost of the item. So I bought it for fun really.
A dead transformer, a blown fuse or something in the power supply, or may be an empty lunch box (if it has cannibalized)... We shall see.

[Bloch]

what is the part number of Q102-Q103 ?

ITE4392

Well i am not 100% sure ! I did search for TG-77 and found out that is a Keithley Part Number in an other Keithley see here http://physics.ucsd.edu/neurophysics/Manuals/Keithley/Model%20227%20Current%20Source

[CarlG]

what is the part number of Q102-Q103 ?

ITE4392

Well i am not 100% sure ! I did search for TG-77 and found out that is a Keithley Part Number in an other Keithley see here http://physics.ucsd.edu/neurophysics/Manuals/Keithley/Model%20227%20Current%20Source

At least it makes sense. I've seen 2N4392 = ITE4392 being used this way before.

//C

[nixxon]

There is a Keithley 440 Picoammeter for sale on ebay that obviously needs som tweaking:
http://www.ebay.com/itm/KEITHLEY-440-DIGITAL-PICOAMMETER-/360550605394

How does the 440 compare to the 480?

[NickS]

Why can't you compensate for the burden voltage?

You can measure the burden voltage, pump it in to a formula and you can tell what the output is supposed to be.
So why don't ammeters do that and adjust the output number?

[McMonster]

Why can't you compensate for the burden voltage?

You can measure the burden voltage, pump it in to a formula and you can tell what the output is supposed to be.
So why don't ammeters do that and adjust the output number?

I think it is not actually related to the measured number, but to the output voltage supplied to the device you're testing. It may seem like nothing, but try this with something powered by a LDO 3.3 V or even 2 V regulator at marginal input voltage.

[robrenz]

It's name comes from the fact that it is a burden on the circuit that is supplying the voltage.  Whatever the burden voltage is it is an additional voltage drop that was not in the original circuit before measuring the current. Therefore the circuit is disturbed by the act of measurement and that may change the way the circuit functions. This becomes more critical with very low system voltages.

[NickS]

Therefore the circuit is disturbed by the act of measurement and that may change the way the circuit functions. This becomes more critical with very low system voltages.

Yes I understand that.

It means the measurement is inaccurate, but if the burden voltage is known then it is inaccurate by a known amount.
Tiny bit of math and you've got an accurate reading again.

[vl400]

The shield spring shown at approx. 10:30 in the video makes contact with the optional battery module.

[alm]

Assume the DUT is a diode, how would you correct for burden voltage? Forward current through a diode can be approximated with an exponential function of the forward voltage (Shockley's diode equation), which depends on several parameters that vary with temperature and device. If the DUT is a switching power supply, then the voltage drop of the burden voltage might cause the current to go up. How is the ammeter supposed to know whether the DUT is a resistor, a diode, or a more complex device?

There is also the problem of showing a hypothetical situation as opposed to showing the current reality. If the ammeter increases the current to compensate for burden voltage, then other concurrent measurements like voltage and heating, will be off.

This is similar to the two competing strategies that Agilent and Tektronix used to have (not sure if this is still current) for probe design. One of them would calibrate a probe so the amplitude from a well defined source impedance as measured by the probe was equal to the unloaded amplitude. The other would calibrate it to the amplitude of the signal with the probe applied. The former strategy has the advantage that it will correct for the probe loading as long as the source is a well behaved 50 ohm source, but falls apart once you apply multiple probes to the same test point. The probe response will correspond to neither the unloaded response (since it's being attenuated by the loading from the second probe), nor the response of the source loaded by two probes. See this appnote (under oscilloscope probing philosophy) for more information.

[Rick]

I was told that A.C. was an option on that meter; it has the switch, but it is locked and can't be moved.
Larry

I was not aware of that.
If it is the same problem with my device, I shall use it only for DC then.
No way to unlock it really?
You mean the hardware is there but there is a physical lock preventing the use of it?

[PuterGeek]

Dave, this was one of your best videos ever! I loved you complaining about it not needing to be repaired or calibrated!

Right the opposite of what anyone else would want!!
 

[SeanB]

Burden voltage is because current is measured by voltage drop across a known low value resistor. The Keithly does this instead by an active measurement of using an opamp virtual earth to generate a voltage proportional to the current internally to keep the input voltage close to zero. TO do this for higher currents than the 1mA offered means you need an opamp capable of sinking or sourcing the current on it's output thus you will need one capable of doing 1A linearly for a 1A measurement, 10A for a 10A unit. This current has to come from a power supply, so will make it both a very big and very expensive unit, especially if you are looking for high accuracy, as high current power amplifiers normally do not have the best offset and drift characteristics at low currents, and definitely not at full power. not much good having a 10mV burden voltage if the opamp has 100mv of drift over a period of 1 minute, or 200mV of offset as it goes from 0 to 10A.

[Salas]

Happy new year! What is that resistor doing in a fuse? Is it behind it and just looks weird?

[SeanB]

It is a delay fuse, where the resistor is there to go open circuit at massive overload as it is a film unit, but where with moderate overload it gradually heats up till the low melting point solder that is used to hold the one end to a spring ( a tin Gadolinium alloy IIRC) softens and the spring breaks the circuit. It was common as a time delay fuse. A 100% load would be fine. 200% would trip in about 5 hours and 400% would trip in about a minute, while a short ( more than 20A) would blow the resistor element in under a second. It does have significant drop at low current ratings, all values of this construction drop around 5V across them, but they are perfect for mains transformers where they limit inrush current and can be rated at the full load draw of the transformer with little risk of them failing randomly. Normally they were available in ratings up to about 630mA ( biggest I have seen) but came in as low as 15mA ( seen that value) and possibly lower.

[Salas]

Aha, are such fuses still widely available?

[SteveyG]

The strange cable mounting is there so you can nicely wind the cable around the case stands. I don't like equipment with permanent mains or any other cables sticking out of it, and I carried a lot of it at work. Really annoying unless you have a third hand to hold the cables, otherwise it is easy to trip over the cable.

It's on the bottom because the front and the rear cards on these instrument cases are removable.

[LLR1967]

I was told that A.C. was an option on that meter; it has the switch, but it is locked and can't be moved.
Larry

I was not aware of that.
If it is the same problem with my device, I shall use it only for DC then.
No way to unlock it really?
You mean the hardware is there but there is a physical lock preventing the use of it?

I do have the manual w/ schematics at work someplace, so I can confirm whether or not A.C. is indeed an option (and possibly a module) and not just a stuck or "locked" switch. In that case, the engineer before me didn't know what he was talking about.
When I had it open to replace the caps, I didn't take notice what prevented the switch from operating, as I only use the meter for D.C. and resistance. I really don't use that meter that much, but it works quite well...

[Rick]

I was told that A.C. was an option on that meter; it has the switch, but it is locked and can't be moved.
Larry

I was not aware of that.
If it is the same problem with my device, I shall use it only for DC then.
No way to unlock it really?
You mean the hardware is there but there is a physical lock preventing the use of it?

I do have the manual w/ schematics at work someplace, so I can confirm whether or not A.C. is indeed an option (and possibly a module) and not just a stuck or "locked" switch. In that case, the engineer before me didn't know what he was talking about.
When I had it open to replace the caps, I didn't take notice what prevented the switch from operating, as I only use the meter for D.C. and resistance. I really don't use that meter that much, but it works quite well...

I am just reading the manual right now. It looks like it does not measure currents AC or DC!
haha
That's sounds great;)
Only AC and DC voltages and resistances can be measured. There is no current button on the front panel anyway.
Next time I shall read the manual before buying.
The only Option they mention is the data output option.
May be that engineer was talking about the current measurement, right?

[robrenz]

The 8921A is 4 1/2 digits, has up to 20 MHz bandwidth, 700V max input, and resolution down to 0.1 1 uV on lowest range, and also selectable dB scale. Great stuff! Very useful for measuring residual noise on (switching) power supplies for instance.

CarlG,  the manual I looked at showed 180µV as the lower functional limit independent of resolution. Is that what you find in practice. If you use it below 180µV how low does it go and what did you use to verify its accuracy below the 180µV. I am not disputing your info, I just may be interested in one if it works well in the single digit µV.

[LLR1967]

I was told that A.C. was an option on that meter; it has the switch, but it is locked and can't be moved.
Larry

May be that engineer was talking about the current measurement, right?

Nope, A.C. volts...  no mention of current...

[CarlG]

The 8921A is 4 1/2 digits, has up to 20 MHz bandwidth, 700V max input, and resolution down to 0.1 1 uV on lowest range, and also selectable dB scale. Great stuff! Very useful for measuring residual noise on (switching) power supplies for instance.

CarlG,  the manual I looked at showed 180µV as the lower functional limit independent of resolution. Is that what you find in practice. If you use it below 180µV how low does it go and what did you use to verify its accuracy below the 180µV. I am not disputing your info, I just may be interested in one if it works well in the single digit µV.

I'll have to check when I get back to work on Monday, but as I remember it, the noise floor is better than that. On the other hand, my memory has a tendency to be a bit optimistic in some situations

Note that I'm only speaking of resolution, not accuracy at that level.

In what paragraph did you find that figure? Is it Table-3?

//C

[robrenz]

I'll have to check when I get back to work on Monday, but as I remember it, the noise floor is better than that. On the other hand, my memory has a tendency to be a bit optimistic in some situations

Note that I'm only speaking of resolution, not accuracy at that level.

In what paragraph did you find that figure? Is it Table-3?

Yes and table 3-4 shows <180µV as a Underload condition.

[CarlG]

Yes and table 3-4 shows <180µV as a Underload condition.

Ok. I haven't read all of the 8921a manual, but I see that the underload condition is used by the Auto ranging function to step to the next range. I don't see it (underload cond) directly related to the accuracy?

[free_electron]

The mysterious empty DIP connector near the big IC seems to be another diagnostic connector (check J1002 on the right size of diagram) with basically all the digital communication with the display board available on its pins.

Nope. That's the tap into the BCD data coming from the A/D convertor. This goes to the optional GPIB board. what they do is simply catch the digit BCD data and digit select signals. this gets stored in a block of circuitry so it can be read over gpib.

Looks like an interesting machine. just found one on ebay for 69$ including the gpib board... on its way now ... 

[Bored@Work]

Looks like an interesting machine. just found one on ebay for 69$ including the gpib board... on its way now ... 

This guy just wants $500 http://www.ebay.com/itm/Keithley-480-Picoammeter-/221161535990
How great are the chances the seller watched the eevblog video and thought it is time to make a killing?

[Rick]

Why does this device use a BNC connector rather than the usual banana jacks used in multimeters? Smaller losses?

[robrenz]

Yes and table 3-4 shows <180µV as a Underload condition.

Ok. I haven't read all of the 8921a manual, but I see that the underload condition is used by the Auto ranging function to step to the next range. I don't see it (underload cond) directly related to the accuracy?

Table 1 specifications shows the measurement ranges are all multiples of 0.18 thru 1.999 except the 700V top end.

[SeanB]

BNC is used because it can have lower leakage current when clean, as well as having inherent shielding capability.

[free_electron]

Looks like an interesting machine. just found one on ebay for 69$ including the gpib board... on its way now ... 

This guy just wants $500 http://www.ebay.com/itm/Keithley-480-Picoammeter-/221161535990
How great are the chances the seller watched the eevblog video and thought it is time to make a killing?

Yikes. You can get a 485 for that price.. 10 years younger and autoranging...

[ivan747]

Looks like an interesting machine. just found one on ebay for 69$ including the gpib board... on its way now ... 

This guy just wants $500 http://www.ebay.com/itm/Keithley-480-Picoammeter-/221161535990
How great are the chances the seller watched the eevblog video and thought it is time to make a killing?

After making the video on frequency standards, Dave says rubidium frequency standards are now 50% more expensive than before 

[CarlG]

The 8921A is 4 1/2 digits, has up to 20 MHz bandwidth, 700V max input, and resolution down to 0.1 1 uV on lowest range, and also selectable dB scale. Great stuff! Very useful for measuring residual noise on (switching) power supplies for instance.

CarlG,  the manual I looked at showed 180µV as the lower functional limit independent of resolution. Is that what you find in practice. If you use it below 180µV how low does it go and what did you use to verify its accuracy below the 180µV. I am not disputing your info, I just may be interested in one if it works well in the single digit µV.

I have started a new topic and moved my previously posted answer to that one. Sorry for the confusion.

//C

[robrenz]

Thanks for all your help CarlG.  I am looking for something that goes below my Tek 7A22 differential amplifier plugin at 10µV/div. When you mentioned the 8921A I thought it might be the ticket for accurate very low power supply noise measurement. That is why I was pressing for the performance at single digit µV levels.  Thanks for all your help and I hope I didn't come across as argumentative.

[CarlG]

Thanks for all your help CarlG.  I am looking for something that goes below my Tek 7A22 differential amplifier plugin at 10µV/div. When you mentioned the 8921A I thought it might be the ticket for accurate very low power supply noise measurement. That is why I was pressing for the performance at single digit µV levels. 

No problem, I guessed that you were looking for something sharper.

Thanks for all your help and I hope I didn't come across as argumentative.

Oh no, not at all

[free_electron]

Mine came in yesterday.

Here's the guts WITH the GPIb installed :

so essentially the trap the BCD datastream wtween the 7103 and the 4511 , serialize it using some standard 4000 series cmos stuff ( counter + shifter ) , send it over two optocouplers (so the gpib is earth referenced but the machine input remains floating) where a motorola 6821 traps it. they use the same 6821 to do GPIb . i think this design predates the availability of real GPIb interface chips like the TMS9914 or uPD7210 or even motorola's 68488.

It uses an AMI made 6802 processor and a 2716 EPROM made by INTERSIL (i never knew those guys made eprom's ... )
Probably handcrafted assembly...

[EEVblog]

Mine came in yesterday.

Here's the guts WITH the GPIb installed :

so essentially the trap the BCD datastream wtween the 7103 and the 4511 , serialize it using some standard 4000 series cmos stuff ( counter + shifter ) , send it over two optocouplers (so the gpib is earth referenced but the machine input remains floating) where a motorola 6821 traps it. they use the same 6821 to do GPIb . i think this design predates the availability of real GPIb interface chips like the TMS9914 or uPD7210 or even motorola's 68488.

It uses an AMI made 6802 processor and a 2716 EPROM made by INTERSIL (i never knew those guys made eprom's ... )
Probably handcrafted assembly...

Wow, totally old-skool!

Dave.

[free_electron]

jep. totally old-skool. When real programmers wrote 1's and 0's straight into rom.

Not that fuddy duddy ladida erasable flash stuff programmed by 50$ wingpangpong programmers bought from ebay 

It's kinda funny. they use an intersil 71c03 duals slope controller with an 8052 analog front-end. not a microprocessor in sight in the whole meter section.
Want GPIB ? vooom. 6802, eprom , 6821 and still a pile of 4000 series to trap that data stream ... no wonder that gpib option was so expensive !

[SeanB]

Intersil started out with memories, I think I still have some around in a tube somewhere.

[JoannaK]

6502 chip + GPIB sounds a lot like those old Commodore Pet Floppy drives, except that those had two 02 chips.. One did bus protocol and another did disk-IO routines.

[ttp]

After making the video on frequency standards, Dave says rubidium frequency standards are now 50% more expensive than before 

I can certainly confirm that. Bought one days after Dave's video for about $AU36, after about 4 weeks emailed the seller that the item hasn't arrived. Next day I had my money back in Paypal account without any dispute (every other time I've complained on Ebay the sellers ask you to wait a bit in case od postal delays). Found another one soon but this time was just under $AU50, couldn't find it any cheaper.

[free_electron]

6502 chip + GPIB sounds a lot like those old Commodore Pet Floppy drives, except that those had two 02 chips.. One did bus protocol and another did disk-IO routines.

Not 6502 (mos) but 6802 ( Motorola ) completely different beast. 6821 is an io co troller

[SeanB]

Like the isolation transformer on that GPIB board as well, driven from the existing mains transformer secondary.

[free_electron]

nope. incoming power cable has a molex that plugs into main board. to install gpib : unplug molex from mainboard. install board plug molex from gpib on main board , plug power cable molex on gpib board.

so that little transofrmer on the gpib board is actually 110 volt to low voltage.