New Agilent 34461A 6.5 Digit Bench Multimeter

[plesa]

I checked similar WinCE based generator (33500B) and there is no such a port inside.

[quantumvolt]

That Keithley line looks too erratic,  I would check your cell formulas and spec interpretation

I checked the formulas and had a mix of 1 + 10 PLC accurracy numbers, but there is not much difference (now with 10PLC data only), see new Chart in original post plus specs here

Very interesting and informative plot. A simple and intuitively correct behavior of a plot of  "accuracy" vs. voltage (range).

As one poster said the higher ranges are determined by the quality of resistor arrays for voltage dividers. At the lower end it is noise and the quality of the DC amplifiers that rules. Somewhere in between the AD converter is "naked" and the instrument is at it's best. Also it is clear for everyone to see why it is better to measure near full scale.

An interesting thing is that better millivolt amplifiers can be bought or built, and the same goes for resistor dividers. But the inherent quality of the AD converter and it's minimum front end determines the instrument's final quality and can not be changed. Also the huge drop in one of the curves represents as said by another poster only a precision increase from around 0.005% to 0.001% and shows up in the naked AD converter range. Shows that the converter is very good but the front amplifier for lower ranges and the dividers for higher ranges are so so.

One can also compare this graph to the "history" of voltage references from 1+ Volt cells to today's 10 Volt arrays. This is the span where the AD converter and it's reference voltage operates and where the instrument's are at their best (I include lower voltages even in modern equipment because there is a lot of portable battery powered calibration / measurement stuff that references much less than 10 volt).

Very nice graph.

[quarks]

Very interesting and informative plot. A simple and intuitively correct behavior of a plot of  "accuracy" vs. voltage (range).
...
Very nice graph.

The credits for that go to robrenz. He was the first I saw this kind of chart from!!! I only adopted it and used it since then for almost all gear I own and for most I liked and tried to buy. 

[robrenz]

IMO the only way to objectively compare specs. Banner specs and huge +LSD counts are quickly equalized between instruments using this method. It is a little tedious getting all the formulas correct for each range of each instrument for all measurement types. You have to make sure you have test values that occur on the range changes/maximum reading for each range for all the instruments compared.

[fpga]

With all the talk of compatibility with the 34401 I'm surprised there isn't an RS232 option.

..well it appears that there is,all they need to do is add software support for it...

..but at the moment it's a debug port.....

Wow! I'm amazed that they would expose all these goodies!

On my projects I'm always required to turn off all debug info in the production build. Perhaps they are shipping the instruments with beta firmware.

[Citizen]

hello. After viewing  teardown tuesday ihave some questions:

1.Is it normal that most vias on the backsdie of the PCB are filed wih solder?

This dual slope  converter.
 Why do you use an opamp to charge the capacitor? All you do is just charging a cap and measuring the time. You dont need an  opamp for it.  And is the charging curve      not exponential ( DaveCad showed a linear increasing of voltage).

[adh]

The ability to display both V and I at the same time is nice in theory, but not sure if it is practical. First, you need to be able to attach multiple leads to your circuit, and second, most important, chances are that the V and the I would be measured using different ranges. I'm not sure about the Fluke and Keithley, but according to the manual, the 34450A performs the two measurements by switching back and forth. If this involves different ranges, the relays will switch back and forth, which I presume will impact the measurement accuracy and speed. Any experience with doing these measurements in practical use?

On most of bench meters (and probably even on reasonable handheld ones) there is no need to switch any physical switch when going between current and voltage measurements. On 34401 going between current (except 3A, because 3A range relay also switches in current source for R measurements) and voltage ranges involves switching two MOS switches in the input switching hybrid (U101 in service manual). And actually the thing by default does moral equivalent of switching ranges between each sample - "autozero", which involves measuring voltage on input shorted by 100k resistor to ground.

It is truly surprising how simple the 34401A's input circuitry is  (essentially whole real input range switching is on sheet 2 of schematics). You essentially get two ranges each for I (5R an 0R01 shunts) and U (small voltages going through few series high-value protection resistors and large voltages going through 9M9:0M1 divider, with both connected at once for "low" impedance mode) and any other ranges are done by adjusting gain of input amplifier after that (U106 and discrete JFET pair U104, with another pins of U101 used for changing divider ratio in its feedback network). Notice that you only have 3 relevant (ie. ignoring K101) relays in this whole mechanism. But because K102 serves dual purpose of switching current ranges and measurement current source for R ranges it is not exactly practical to measure U and I at once, but only slight redesign (probably without any additional components) will enable that, and without any relay clicking between samples.

On the other hand, I would like to not only be measuring U and I at once but also RMS power, which seems to not be practical by switching inputs of relatively slow ADC.

[EEVblog]

Why do you use an opamp to charge the capacitor? All you do is just charging a cap and measuring the time. You dont need an  opamp for it.  And is the charging curve      not exponential ( DaveCad showed a linear increasing of voltage).

You need an opamp to created the virtual ground at the input and steer all the (constant) current through the capacitor.
It is linear because it's a constant current.

[AlfBaz]

This dual slope  converter.
 Why do you use an opamp to charge the capacitor? All you do is just charging a cap and measuring the time. You dont need an  opamp for it.  And is the charging curve      not exponential ( DaveCad showed a linear increasing of voltage).

It's why they call this circuit an integrating op amp. It performs the mathematical integral of the applied voltage over the time. If it wasn't linear (for a fixed dc input) then it wouldn't be called an integrator

[ben_r_]

Man, after watching Daves tear down and review of this unit Im kinda glad I didnt get it. That boot time (or anything other than instant on) would annoy the heck out of me! That takes WAY too long. When I pull out or power up my meter I want instant access to take my measurements and Im out. That Windows CE boot sucks.

[AlfBaz]

Presumably you want to pull out the 6.5 digit meter for some accuracy. You should be waiting an for an hour for it to warm up. So what's the big deal?

[ben_r_]

Presumably you want to pull out the 6.5 digit meter for some accuracy. You should be waiting an for an hour for it to warm up. So what's the big deal?

lol I guess we all look at thing differently. I would use it as my daily/normal bench meter but not necessarily leave it on all day long and I think in this day and age with all that technology crammed in there we shouldnt have to wait at all for the thing to boot up.

[c4757p]

Even an accurate, expensive meter like that can (or, should be able to) serve well as a general-purpose meter in addition. Sure, it'll be warming up for a while if I need that much accuracy, but if I don't, I still want to be able to use it without waiting for that damn startup sequence.

[grego]

Even an accurate, expensive meter like that can (or, should be able to) serve well as a general-purpose meter in addition. Sure, it'll be warming up for a while if I need that much accuracy, but if I don't, I still want to be able to use it without waiting for that damn startup sequence.

Yeah but remember Dave specifically said his had pre-release firmware on it with extra debugging and that Agilent told him it would take longer than normal to boot.

[c4757p]

Remember? I missed that entirely!  Thanks.

[ben_r_]

Even an accurate, expensive meter like that can (or, should be able to) serve well as a general-purpose meter in addition. Sure, it'll be warming up for a while if I need that much accuracy, but if I don't, I still want to be able to use it without waiting for that damn startup sequence.

Yeah but remember Dave specifically said his had pre-release firmware on it with extra debugging and that Agilent told him it would take longer than normal to boot.

I did catch that in the video review, however being that it does run CE, Im sure the boot up is still long, just maybe not AS long.

[mikeselectricstuff]

Even an accurate, expensive meter like that can (or, should be able to) serve well as a general-purpose meter in addition. Sure, it'll be warming up for a while if I need that much accuracy, but if I don't, I still want to be able to use it without waiting for that damn startup sequence.

Yeah but remember Dave specifically said his had pre-release firmware on it with extra debugging and that Agilent told him it would take longer than normal to boot.

The current production firmware takes 50 seconds to boot. Way too long.

[robrenz]

8846A running Linux takes 14 seconds

[grego]

Even an accurate, expensive meter like that can (or, should be able to) serve well as a general-purpose meter in addition. Sure, it'll be warming up for a while if I need that much accuracy, but if I don't, I still want to be able to use it without waiting for that damn startup sequence.

Yeah but remember Dave specifically said his had pre-release firmware on it with extra debugging and that Agilent told him it would take longer than normal to boot.

The current production firmware takes 50 seconds to boot. Way too long.

If 50 seconds is what they consider "faster" than the debug software then yeah - that's a long-ass time for a DMM to start up.  Gotta love WindowsCE.

[ben_r_]

Man, after watching Daves tear down and review of this unit Im kinda glad I didnt get it. That boot time (or anything other than instant on) would annoy the heck out of me! That takes WAY too long. When I pull out or power up my meter I want instant access to take my measurements and Im out. That Windows CE boot sucks.

A lot of manufacturers choose WinCE for political reasons-- rather than technical ones.  If you are ever working on a project that requires a GUI, file-system, and Ethernet connectivity, please don't use WinCE (unless you are forced to).  There are many paths that lead to the same destination, but some of those paths are the dumbest way to go.  For the executives reading the Microsoft sales brochures, WinCE sounds like a dream-- [paraphrasing]: "Practically everything is already done for you-- just install it, and wham!  Now you can hire Windows programmers that know absolutely nothing about embedded systems programming, and have your product ready in half the time!"  The reality is that no, you still have to hire embedded systems programmers that know what the hell they are doing, and because you have now added many layers of over-bloated code, the memory size and boot-times are going to be much bigger, and the firmware project is going to take 3-5 times as long (in my experience-- I've only worked on 5 projects involving WinCE before I gave up on it forever).

PLEASE-- FRIENDS DON'T LET FRIENDS DESIGN EMBEDDED SYSTEMS WITH WINDOWS !!!

And Agilent (I know you are reading this thread)-- you should be ashamed.

All we can do is try and vote with our dollars. Dont buy devices running Windows CE if you dont want them on the market anymore.

[mikeselectricstuff]

Even an accurate, expensive meter like that can (or, should be able to) serve well as a general-purpose meter in addition. Sure, it'll be warming up for a while if I need that much accuracy, but if I don't, I still want to be able to use it without waiting for that damn startup sequence.

Yeah but remember Dave specifically said his had pre-release firmware on it with extra debugging and that Agilent told him it would take longer than normal to boot.

The current production firmware takes 50 seconds to boot. Way too long.

If 50 seconds is what they consider "faster" than the debug software then yeah - that's a long-ass time for a DMM to start up.  Gotta love WindowsCE.

I don't think the production SW is much, if any different to what Dave has - most of the boot-up message build dates are the same. Hopefully this means an update is coming to improve it. Most of the time is before the main DMM app starts & relays start clicking etc.
I suspect they may need to  go back to MS to address some of the time-consuming stuff - I suspect a lot of it is stuff like inappropriate timeouts waiting for non-existent interfaces and unnecessary selftests.
Or they could change the way that powerdown works to go into a standby state - as it has a soft power switch this ought to be possible. Not ideal as you'd still have the long startup from plug-in, but should  be a much faster restart from the front panel, which would be better than nothing. 

[mikeselectricstuff]

The problem is that Agilent will now have a big investment in their own developments under CE - for example all the remote LXI stuff, calibration & production test, so it's not just the OS, at least not now.
I'm sure they would have looked long & hard at the options, and that there were good reasons to move away from Vxworks (used on 5/6/7000).
An obvious reason for choosing CE over Linux is support from the manufacturer, althoigh Agilent are big enough that they could have set up an internal Penguin department.
 

[grenert]

Or they could change the way that powerdown works to go into a standby state - as it has a soft power switch this ought to be possible. Not ideal as you'd still have the long startup from plug-in, but should  be a much faster restart from the front panel, which would be better than nothing.

This is what the Fluke 8846A does.  Boot from plug-in is 14 sec as explained by robrenz above, but subsequent turn-on from the front panel is pretty much instant.

[don.r]

Ahh.. the simple life. My Keithley 196 boots in less than 3 seconds.   

[jonese]

WinCE wasn't that great on a phone platform either.