Calibrating a HP 3400A RMS Voltmeter

[enut11]

I have just finished repairing a HP3400A RMS meter, mainly faulty electrolytic capacitors.
I must say that I am impressed with the build quality of these old HP meters: sturdy metal chassis, quality PCBs with gold plated traces, very accurate taut band meter movement and easy to service as all panels can be removed.

The 3400A is a true RMS responding AC meter using a pair of internal thermocouples.
Voltage range is 1mV to 300V and frequency response is 5Hz to 10MHz.
Accuracy is specified as 1% over most of the frequency range.
Meter movement is calibrated to better than 0.5% at the factory during manufacture. It is worth reading how they did it in the HP service note .
All in all a very nice piece of test equipment.

Now, my problem is how to calibrate it.

I have access to the following:
HP3456A 6.5 digit meter - AC response to about 800KHz
Tekway DST1102B 100MHz DSO - vertical accuracy only 3%?

For a signal source I am using a Wavetek Model 29 10MHz DDS function generator.
The 29 appears identical to the TTi TG1010A which has a stated accuracy of 0.2dB to 500KHz and 1dB to 10MHz. My model 29 is un-calibrated.

Calibrating the 3400A to frequencies below 1MHz will not be a problem, but I want to calibrate it all the way to 10MHz.
The Tekway DSO has sufficient frequency response but poor amplitude accuracy.
The HP3456A has very good accuracy, but poor high frequency response.

Looking for suggestions on how to proceed.
enut11

[enut11]

Some sinewave measurements sourced from Wavetek Model 29:

1MHz     Tek DSO=2.78Vpp     HP3400=1.00Vrms

5MHz     Tek DSO=2.65Vpp     HP3400=0.92Vrms

10MHz   Tek DSO=2.06Vpp     HP3400=0.68Vrms

The Tek DSO dropped 2.6dB (calculated) from 1MHz to 10MHz

The HP3400A dropped 3.3dB on the meter scale from 1MHz to 10MHz

Part of the amplitude drop may be from the Wavetek itself (up to -1dB if it were in spec)

At this point in time I do not know which instrument to believe.

If there is an affordable way to measure HF amplitude to better than 1% accuracy, please chip in with your suggestion.

enut11

[Electro Detective]

I think mine is way overdue for a recap,
or whatever is causing the needle to not lock on to the far left zero position, and drift forward instead = useless

This is/was/might be again? my best TRUE RMS meter   

My personal whining aside    assuming your unit was fairly calibrated to begin with,
and replaced the capacitors with 'exact enough' caps,
does it need calibrating?

i.e. have you also checked it against a second standard or decent TRMS multimeter that kicks to at least 100khz AC ?   

[enut11]

Hi ED
Replacing the faulty electros had a significant impact on meter linearity and accuracy.
Low frequency performance is easy to check with my current gear.
HF >1MHz indication depends on the setting of a variable capacitor on the 3400 so you need a reliable HF source.
enut11

[Electro Detective]

fwiw, is the Range switch in good condition too or a bit crusty?
Better get that sorted properly too before going the full cal route and possibly.. 

There's a post here about that: 
www.eevblog.com/forum/testgear/hp-3400a-indicator-needle-behavior-on-range-change/msg796865/#msg796865

Good luck with it btw 

[Kleinstein]

For the 1 -10 MHz range the scope should have a relatively flat frequency response. I don't expect compensation points in this range.

In the 3400A I would expect the input divider and amplifier to be critical parts that can have some frequency dependence. One should be able to check the compensation by looking at the response to a square wave signal (e.g. probe cal signal of the scope).

[enut11]

Thanks Kleinstein
Should I use the direct input on the Tek DSO or use my 10:1 probe?
enut11

[enut11]

For the 1 -10 MHz range the scope should have a relatively flat frequency response. I don't expect compensation points in this range.

In the 3400A I would expect the input divider and amplifier to be critical parts that can have some frequency dependence. One should be able to check the compensation by looking at the response to a square wave signal (e.g. probe cal signal of the scope).

You believe  the Tek DSO is the best reference I have in 1MHz-10MHz range.
At the moment, the 3400 frequency response is flat to 1MHz.
Therefore I should increase the output of the Wavetek DDS so the 10MHz amplitude matches the 1MHz amplitude before re-setting the 3400 reading at 10MHz?
enut11 

[enut11]

fwiw, is the Range switch in good condition too or a bit crusty?
Better get that sorted properly too before going the full cal route and possibly.. 

There's a post here about that: 
www.eevblog.com/forum/testgear/hp-3400a-indicator-needle-behavior-on-range-change/msg796865/#msg796865

Good luck with it btw 

Yes, I cleaned the range switch with IPA although it was in good condition anyway.

[Kleinstein]

There should be no significant difference between the direct scope input and the 1:10 probe. Maybe use bath cases to check consistency.

Capacitive loading (e.g. cables) could be the reason why the amplitude drop from 1 to 10 MHz is more like -2.6 dB. So it may be important to have the scope / 3400 still connected when doing the measurements. Things should get a little better with a 50 Ohms load / 50 ohms termination at the scope.

[blackdog]

Hi enut11,

I did some tests for you with my HP3400 and also the following measuring equipment.
The generator is a Siglent SDG 2042 of which I know it has a very flat frequency response.

But to be sure I also measured the frequency response of this generator with a Fluke 8920A RMS meter and my Hameg HMO3004 400MHz scope.

On the generator is a 1Meter BNC cable and at the end a 50 Ohm terminating resistor.

First I measured the frequency characteristic with the Hameg scoop with the built-in RMS function.
That did not give the expected results this function no longer works well at the higher frequencies. :-)
So I switched the scope to the PP function as far as the meter is concerned and then the measurements were correct.

After the scope indicated that the Siglent generator was very flat to 40MHz I started measuring with one of my favorite measuring instruments, the Fluke 8920A RMS meter.
Up to about 30MHz this meter also indicated that the Siglent is very flat.

After warming up the HP 3400 and setting the "0" dB level (0.775V @ 1MHz) to a precise level, time to start measuring the frwquence characteristic of the HP 3400.

The following values came from this measurement:
1Mhz= "0" dB
10MHz = -0.15dB
13MHz = -0.5dB
15Mhz = -1dB
18Mhz = -2dB
20.6Mhz = -3dB
25MHz = -5dB

Please note that different input capacities of my instruments above 10MHz can give a few tenths of a dB deviation.
The Hameg scoop is 13pF, the Flue RMS meter is <30pF and the HP 3400 I don't know by heart.

I hope this will help you adjust your beautiful HP3400 RMS meter.

Regards,
Bram

[dom0]

3400A has 50 pF or so of input capacitance. The typical -3 dB bandwidth of a 3400A should be around 20-25 MHz.

[0culus]

Successfully calibrated mine using a 3455A voltmeter, Tek 7904A o-scope with a 7A26 amplifier, and my HP 3325B with option 002 as a signal source. I *also* needed a 50 ohm feed thru termination for some of the ranges due to the characteristics of the 3325B. With that I was able to mostly replicate the manual's cal procedure and the unit seems to be spot on now. I don't have access to a signal source that can provide anything like 1000 Vrms so I just left that part alone as I doubt I'll use that range often.

[enut11]

Thanks @blackdog, those are very impressive figures for your 3400!

So, I have to place my faith in one of my instruments and, according to Kleinstein, that should be the 100MHz Tek DSO.

Calibrating the 3400 at 10MHz will now involve boosting the DDS output by 3dB wrt the 1MHz signal. I will do that and see how it works out.

I am using eBay sourced 1 meter BNC test cables and a T-piece from the DDS to both the 3400 and the DSO.
I am wondering if better quality cables would result in a better calibration?

[enut11]

I just replaced the 1m BNC (90pF) with 0.5m cable (45pF). The signal at the 3400 and DSO went up about 1dB at 10MHz 

Using a 50 ohm load into the DSO also produced a small improvement at 10MHz.
Can I put another 50ohm load simultaneously on the 3400 input. Seems wrong to me.

Getting the correct measurable HF signal into the 3400 is proving to be a challenge, but non-the-less interesting.

Next step is to try even lower capacitance leads.
enut11

[trobbins]

It took me a little while to realise that a Keithley 197 was effectively increasing the meter level shown by my HP3400A when I was checking HP3400 performance down at the sub-1mV level once.  And yeh, I only had 200-300kHz performance meters for cross-checking the HP3400A's frequency response.

[blackdog]

Hi,

I looked up the input impedance of my model of the HP3400A and this is about 40pF.

My HP 3400A has a Nuvistor as an input stage with a bootstrap to keep the input impedance as low as possible over a large frequency range.
This input stage is a really nice design from HP!

I calculated the input capacity and looked at the influence of the 20MHz value I mentioned earlier on my HP 3400A.
The measured value at 20MHz was about -3dB and the input capacity has an influence of about -1dB at this frequency.

The -1db should be subtracted from the -3dB and the meter is actually -2dB at 20MHz.
Above 5MHz, the cable capacity can cause a lot of inaccuracy, just like the input capacity of the HP3400.

It is always wise to take into account the variables that occur in the measuring system.

You can also make a 1:10 probe that can handle the 40pF input capacity of the HP 3400.
Maybe you can even convert a scoop probe so you have more bandwidth at the probe tip.
You may lose 20dB of signal(10x) but if the signal is large enough, this will not be a problem.

Kind regards,
Bram

[Gyro]

You can also make a 1:10 probe that can handle the 40pF input capacity of the HP 3400.
Maybe you can even convert a scoop probe so you have more bandwidth at the probe tip.
You may lose 20dB of signal(10x) but if the signal is large enough, this will not be a problem.

Remember that the HP3400A has a 10M ohm input impedance. I've considered shunting mine down to 1M ohm several times, so that I could use a scope probe but never got around to it. Unfortunately the input capacitance changes with range too... 1mV-300mV, <50pF and 1V-300V, <20pF. Irritating.

[macboy]

I just replaced the 1m BNC (90pF) with 0.5m cable (45pF). The signal at the 3400 and DSO went up about 1dB at 10MHz 

Using a 50 ohm load into the DSO also produced a small improvement at 10MHz.
Can I put another 50ohm load simultaneously on the 3400 input. Seems wrong to me.

Getting the correct measurable HF signal into the 3400 is proving to be a challenge, but non-the-less interesting.

Next step is to try even lower capacitance leads.
enut11

Use a single termination at each end of the cable run. The signal generator is terminated internally. So, connect it with a cable to a "T" at the 3400 voltmeter, then another cable to another "T" at the scope, and terminate at that "T". You can swap the positions of the 3400 and the scope to see if there is a difference. There should not be any significant change. Make sure your cables are 50 ohm (e.g. RG-58) and not 75 ohm (r.g. RG-59).

After terminating properly, cable capacitance will have negligible effect.

[enut11]

Thanks @macboy. That is very handy information. I had not previously fully understood cable loading and its impact on signal levels.
enut11

[blackdog]

Hi Gyro,

You're right, I really looked over that regarding the input impedance when you get below 1V.

Thanks!

Bram