[time-nuts] HP3420B calibration

[Dick Moore]

On Mar 18, 2010, at 3:14 PM, time-nuts-request at febo.com wrote:

> Message: 5
> Date: Thu, 18 Mar 2010 22:53:46 +0100
> From: "Dr. Frank Stellmach" <drfrank.stellmach at freenet.de>
> Subject: [time-nuts] HP 3420B calibration assistance
> To: time-nuts at febo.com
> Message-ID: <4BA2A0EA.2080800 at freenet.de>
> Content-Type: text/plain; charset=UTF-8; format=flowed
> 
> Hi Dick,
> 
> I'm sure, I met the 3420B long time ago..that's a nice unit,.. but I do 
> not have the schematics of the 3420B at hand, also did not find them on 
> the agilent site, only the HP catalogue from 1970.
> 
> The instrument - as far as I remember - compares the input voltage 
> directly against 0-10V or 0-1V given by an internal reference and a 
> Kelvin-Varley divider. 100V and 1kV input are divided first down to 10V 
> and/or 1V.

Internal ref is 11V and 1,1V, 100V and 1kV are divided down to 1 V in both cases, whereas I would think that the most stable would be to divide down to 10V.
> 
> So you have to calibrate the linearity of the KV first, best done in the 
> 10V range.
> Then you have to calibrate the 10V, 1V range, then the 100V and 1kV range.
> 
> Using the 5440B only is the best way to handle this process, even better 
> and faster than using an internal Wheatstone Bridge for linearity 
> calibration. As far as I remeber, the missing card did not contain the 
> necessary precision resistors for the Wheatstone Bridge, the user had to 
> solder the Bridge on his own.

Exactly right.
> 
> Linearity of the 5440 is ~0.5ppm, sufficient for the 3420Bs KV.
> Range uncertainty is also superior.
> 
> First two decades should be binary coded, so they should be calibrated 
> each by a 1-2-4-8 sequence from the 5440B.
> The 3458A may be used for adjustments in the 100mV range.
> 
> I do not remember the construction of the KV, especially how the 
> adjustment from one decade to the other is designed. I.e. if the next 
> lower decade is applied directly to the predecessor, or if an adjustment 
> resistor is used (probable, if you mention 5 resistors per decade). 
> Perhaps you may send me a scan of the KV, with the adjustment pots.
> 
> I think, this should work if no additional decade resistor is used:
> 
> Use the normal 10V range of the 3420B for linearity calibration.
> 
> Start with the third decade, which cannot be adjusted, all its digits 
> and lower ones to 0.0999X, which gives 100mV. Set 5440B for this assumed 
> voltage of 100mV in the divider mode, freshly calibrated by 3458A, and 
> set to operate.
> Adjust 10V range cal of 3420B , so that differential VM of 3420B shows 
> Zero to 1?V or less.
> 
> Set DVM un-sensitive and change 3420B setting to 0.10000, ?1`being the 
> first decade resistor which can be adjusted.
> Apply 100mV again, set DVM to sensitive, and adjust this resistors 
> adjustment to Zero deflection. This calibrates the transfer from the 
> third decade to the 2nd.
> 
> Now, apply 200mV, 400mV and 800mV, and adjust the appropriate resistors 
> of the 2nd decade at 0.20000V, 0.40000V, 0.80000V settings acoordingly 
> to Zero, 1?V or less.
> 
> Set 2nd decade to 0.X0000, or to 0.9999X if the first setting is not 
> possible.
> 
> Apply 1V from the 5440B in the normal mode (not divided, control by 
> 3458A) and adjust 3420B 10V range to zero deflection again, to have a 
> more precise adjustment.
> Set 3420B to 1.00000V and adjust the first resistor of the 1st decade to 
> a Zero deflection, <1?V.
> 
> This calibrates the transfer from 2nd to 1st decade.
> May also be different, if an additional decade resistor is designed in.
> 
> Now calibrate 2V, 4V, 8V from 5440B against 2.00000V, 4.00000V, 8.00000V 
> of 3420B to Zero deflection to 2?V, 4?V,8?V, or better, if not too noisy.
> 
> Now apply 10V, setting 3420B to 9.9999XV or X.00000V, whichever is 
> possible, finally adjust 10V range.
> Apply 1V, 100V and 1kV and adjust those ranges at the 3420B.
> As far as I remember, 1000V adjustment may not be not necessary, as this 
> is accomplished already in the 100V range.
> 
> Hope,this description helps.
> 
> Frank

Thanks Frank -- your memory is pretty good! The decade resistors are 1k (fixed), 1k (adj), 2k (adj), 2k (adj), and 4k (adj) in the second decade for 10k total, with an extra 1k (adj) added for the first decade, which has an X position, to bring it to 11k total. The thing is, the switched resistors of the two decades with adjustments are always in series with each other, and are used in every position, being switched from the upper leg of the decade divider to the lower leg (and vv) depending on the dial position, so that adjusting one intrinsically changes the whole string at every setting -- I don't see an easy way to isolate a particular resistor's adjustment. 

You are right, the precision resistors to use the internal Wheatstone bridge system had to be supplied by the user and kludged up by hand.

I think this calibration will involve a lot of back-and-forth tweaks up and down the line. As mentioned above, an X switch position is only on the first (most significant) decade, the others only go from 0 to 9, which is a little counter-intuitive, but then I've been using Fluke 885s, 887s, and 895s for years, so I'm used to them.

Unfortunately, I don't have the manual in electronic form, so I'm not sure how I would get you that section of it. I have a very good quality paper reproduction of the original. If I can get a scan done, I'll email the dividers and descriptions to you, but please don't hold your breath...

Dick Moore