The HP 3400A True RMS AC Voltmeters
There might still be a place in the Lab for this Forgotten Meter
In a previous tech note, we discussed how many modern DMMs, such as the widely used 34401A 6 1/2 digit DMM, are not suitable for measuring AC signals less than about 1 to 5 mV. Since then, we have been investigating low cost solutions for low level AC measurements. Lately I have been thinking about high Crest factor ACV measurements.
As well documented at other sites, there has been an evolution of the HP 3400A series meters from their introduction in 1964, through the late 1990s ending with the 3400B.  The HP 3400A was described in the January, 1964 edition of the Hewlett Packard Journal.  The 3400A uses the well known balancing method in instrumentation. An input AC signal is signal conditioned by an AC amplifier of appropriate bandwidth and applied to heat a heating element. The temperature of the first heating element is registered by a first thermocouple. A "balancing" amplifier applies whatever DC voltage is needed to a second heating element attached to a second thermocouple to match the temperature of the first heating element. Also in a classic approach, the voltage signals from the two thermocouples are added with one of the two having a "reverse" polarity connection, so that when they are at the same temperature, their voltages cancel to near zero.  The input AC amplifier(s) have a bandwidth suitable for the bandwidth of AC voltage measurements (10 MHz for the original 3400A), however the balancing "DC amplifier" only needed a bandwidth wide enough to make the balancing measurement in a time frame on the order of a second! So, these ACV meters with wideband input amplifiers, are non-the-less relatively slow in response time, presumably one of the reasons (beyond cost) that this thermocouple technology is no longer used. The DC output voltage (and corresponding DC meter deflection) provides the AC voltage measurement.
The original 10 MegaOhm input R was provided by a nuvistor (vacuum tube in a small metal can) RF amplifier. Here is a technical data sheet for the 1970 version of the hp 3400A PDF. Note the older style hp front panel logo, the red lensed power light, the silver power toggle switch and the older oval style line cord.
In 1983, the hp 3400A still had the front panel silver toggle switch ($1325): 1983 catalog page PDF. The 1984 catalog page shows the change over to the black handled toggle power switch.
By serial number prefix 2415A (if not earlier), the input board was replaced with a JFET circuit.
The original DC amplifier was a "chopper stabilized" amplifier in order to provide sufficient DC stability at the microvolt measurements needed to measure small differences in thermocouple voltages. The first chopper stabilized circuits used a galvanic break of neon bulbs to cadmium sulphide (cds) photo cells. Later the circuit was updated using the then new Intersil ICL 7650 style of integrated chopper amplifiers. The still available Linear Technology LTC1052 appears to be a good alternative (without modification) where repairs of the later model 3400A A6 boards are needed.
In years past, I avoided the hp 3400A series as too complex and too expensive (~2005) for our noise measuring experiments. We selected the hp 400 average responding meters (hp 400 GL, F, FL, E, EL) as more available low cost alternative AC voltmeters at the time. As well documented, for example, by Jim Williams and Todd Owen of LTC , average responding meters do not directly measure noise correctly. However, as noted by Horowitz and Hill , for Johnson noise measurements, a small correction factor does the job. Recently, the availability of late model (e.g. serial prefix 2415A) hp 3400A has increased, and in many cases the prices have now fallen well below $100, on occassion below $25.
Of the three that I have restored so far (including 2415A S/N prefix, from 1986 33,000s and 1990 37,000s), one just needed calibration and one was DOA. Some hours later (I care not admit how many!), it turned out one of the two film capacitors used by the Intersil chopper amplifier failed with a high D. A working 0.1 uF (100 nF) film capacitor should have a D (ESR/Xc) at 1 kHz on the order of .002 to .003, yet our failed capacitor had a D of over .01. Fortunately, we had two suitable replacements on hand and all is well now. Symptoms included significant movement of the meter needle and low sensitivity. In the process of the repair, we changed out the chopper stabilized amplifier and added a high quality 14 pin socket  In fact, the repaired unit is somewhat more stable (less noisy) than the working unit, which makes us wonder if due to the improved LTC part? (More measurements to follow some day).
Replacing the A2 board foam: With a late model S/N prefix 2415 unit, you can expect to need to replace the foam holding the input JFET A2 board in two plastic end slots. Probably all of the foam inserts are decomposing by now. Note the wire positions and colors and remove the wire plugs from the gold posts. A ground wire is soldered to the bottom of the board and does not need to be removed. Gently push on one side of the A2 board and lift it out. Remove all of the old insulation foam, including the adhesive paper tape at the bottom of each plastic slot. Cut new foam pieces from a high quality open soft packing foam (typically a dark gray color). Work both new long rectangles of foam into the slots and then work the board back into the new foam. If you are going to clean the rotary switch with Deoxit D5 (Caig Chemicals), followed by P5 (highly recommended), then use a small wire tie to tie the A3 board off to the side of the chassis to keep it out of the way and stable during rotary switch cleaning. Place the unit on its side during spraying, so the A2 board is on top and does not get sprayed.
Buying Advice: Although not a fixed rule, the later models can be generally identified by a white neon bulb front panel power light indicator, the more modern hp logo (black and blue) silk screen painted on the panel, the black toggle power switch (not the metal toggle), and a fuse holder with the light grey insert (the removable part) on the rear panel. Unfortunately, I have since seen examples of S/N prefix 1218A (circa 1972) and 2225A (circa 1982) both of which had all of these more "modern" attributes. If you are looking for a late model 2415A series from 1984 and some later years, the only way to be sure might be to see the prefix in a picture of the rear panel or to ask the seller.
With the power off, the meter needle should be on "0". With the power on, the needle should fall between "0" and the first mark. If the description says needle moves and the picture shows a needle indication above the first mark, plan on a high likelihood of repair! The first mark is the 1/10th scale value. The 3400A is designed be calibrated at both the 1/10th scale mark at the meter low end, and at the meter full scale mark. That is why as long as it reads correctly at both points (and in-between), where it falls between the zero mark and the first mark with no signal input, is of no consequence. I have only gone through three late model units (all 2415A S/N prefix), however I would guess that the A6 chopper board, with the 1/10th scale and full scale trimmers, is where most repairs will probably be made.
Of course, the same recommendation we made with the simpler average responding hp 400 series applies, you might be better off with a low time, very well cared for older unit, than with a late model unit that was abused and/or stored in a hostile environment for 10 or more years. I have no direct experience with the earlier models. I do not know what serial numbers changed over from the nuvistor to the JFET input, or from the neon bulb chopper to the intersil integrated chopper (might be in the manual changes somewhere; I think the 7650 came out around 1980, not sure).
These pretty amazing AC voltmeters correctly measure AC waveforms to a Crest factor of 10. While the specifications indicate that the hp 3400A have a useable bandwidth of 10 MHz upgraded to 20 MHz with the hp 3400B, I have found that our S/N 2415A units, at least for a sine wave input, appear to measure fine to 20 MHz (more measurements to follow, not sure yet if the later S/N hp 3400A was update to a wider BW with the new RF input board, or if our observation is in error, perhaps the lower bandwidth limit is for complex waveforms).
Our interest in the hp 3400A is related to an interest in high Crest Factor AC voltage measurements. 
The manuals for the hp 3400A are widely available from the Agilent website and other manual down load sites, so we have not re-scanned or re-posted it here. Also, note that many of the scanned pdfs offered at a modest fee by the e-manual companies, have scanned them at very high quality as compared to most of the free downloads which are generally scanned at relatively low resolution.
Serial Numbers: Note that while a general rule of thumb with hp instrumentation the S/N prefix plus 60 gives the year of manufacture, for many models, it appears that the prefix is more of a "series time frame ". For example, the 3400A prefix series 2415A (some of ours from 1986 and 1990) presumably began in 1984 (24+60). HP 3400A S/N examples include a 2415A S/N prefix from 1986 ~S/N 33,000s and another 2415A S/N prefix from 1990 ~ S/N 37,000s. A particularly good example of the exception to the rule is the hp 3458A, where for some late models, instead of the traditional four digit prefix, "US280" was the prefix (Not, for example "2803"), with the numbers that followed, the specific units S/N. For example, numbers were over 32,000 by the year 2005. Close inspection of the S/N tags shows the US280 was separately colored or highlighted on the label. Early and mid production models used the traditional numbering scheme 2803A nnnnn.
1. HP 3400A RMS Voltmeter at Test Equipment Gallery. (See also their main gallery.) They seem to be missing the S/N 2415A version of the hp 3400A with the newer style open power transformer (black cloth around the windings) mounted off to one side which, in the later units, replaced the shielded rear mounted power transformer.
2 A RMS-Responding Voltmeter With High Crest Factor Rating, Hewlett Packard Journal, Vol. 15, No. 5, January, 1964; follow up note, Hewlett-Packard Journal, Vol. 15, No. 8, April 1964, page 8.
3. Fig. 4, ref .
5. Understanding and selecting rms voltmeters, Failing to verify ac-voltmeter accuracy before conducting rms-noise measurements may cause highly misleading results, Jim Williams and Todd Owen, Linear Technology Corp, EDN, May 11, 2000. (Jim Williams passed away the same Sunday that I was writing my first notes on high crest factor measurements. Jim's absence will be a huge loss to the Analog design community.)
6. Horowitz and Hill, “The Art of Electronics”, Cambridge University Press, 1986, Section 7.10, “Origins and kinds of noise”, pages 288-290; 3. Section 7.20, “Bandwidth limiting and rms voltage measurement, page 306.
7. High Crest Factor AC Voltage measurements, a work in progress, more later.
8. HP AN 124 True RMS Measurements, June 5, 1970.
9. The fear of sockets in uV circuits is in our opinion over-rated. I concede that by the 100 nV region (as noted in the LTC data sheet) a socket can be problematic, however by the uVs with no significant air flow in a closed box, socket to pin thermocouples are less of an issue (yes, it can depend on the particular metals and pin plating) since all of the pins are at the same or a relatively constant temperature. True, temperature gradients across a chip can be an issue. Here, I wanted the option to try a few different versions of this same part, so the socket. The 3400A with the socket easily calibrated to spec. I suppose it would be interesting someday to see if any measurement error can be detected by reading the meter face (and/or at the DC output) with changes in ambient temperature (and if so, can such changes be shown to be caused by having the chopper IC in a socket). This could be done, perhaps by monitoring the DC output level of the chopper amplifier (e.g. at TP4). But, then one would also need to confirm the stability of the input ACV amplitude as well as the thermocouple-heater pair. Another approach might be to do the test with two or more 3400A and move the A6 board with the socket between the units in separate test runs over temperature. It might not be such an easy project!
hp manuals and related catalog pages are reproduced with Permission, Courtesy of Agilent Technologies, Inc.
COPYRIGHT © 2011 JOSEPH M. GELLER.