Modifying the HP465A Amplifier (the ongoing saga - updated 8/13/05)
The HP465 amplifier is a handy device, but the one which I purchased used had "issues".  It had gotten noisy over time, the power supply was wonky and there were more than a few tenths of DC offset voltage on the output.   Furthermore, the input was single ended and I wanted to allow the option of balanced input.  In order to make the unit useful for testing regulator circuits and power supplies I decided that the option of bandpass filtering the output from 0.1 Hz to 10 Hz would be helpful for measuring Voltage Reference Noise. Linear Technology on the other hand specified 10Hz to 100kHz for testing low dropout regulators and power supplies.  To remedy the perplexing choice I put both filters in, separated by opamp buffers. The pictures illustrating the device prior to its "circuit-board-ectomy" are shown below:

 

hp465_5.jpg (37660 bytes) hp465_6.jpg (33686 bytes)
hp465_3.jpg (32362 bytes) hp465_1.jpg (17243 bytes)

 

HP465's sell for $10 to $50 on EBay, and are quite nice to have around.  The first unit which I purchased was "wonky".   Rather than try to fix the unit I decided to rebuild the printed circuit board with a very low noise operational amplifier and a well regulated, low noise power supply.  I chose the SSM2019 from Analog Devices, a programmable gain amplifier which boasts 1nV SQRT Hz low frequency noise performance.  The design follows that published by Audio Amateur's series of articles on low noise regulators, using the SSM2019 for the now discontinued SSM2017.  Gain settings of 20, 40 and 60dB are obtained by switching resistors in parallel with the principal feedback resistor. The lower f3 point of the input circuit is 100mHz.  Input protection is provided by a 1/16th amp fuse and clamping diodes.  Radio Frequency Interference is mitigated by the 100uH inductors on each input.  While not shown on the current schematic, each power pin is decoupled with a 10uF 25V tantalum and 100nF ceramic capacitor. 

I chose to maintain as much of the original hardware of the HP465A as possible.  Of course, the cabinet and chassis are absolutely first rate vintage Hewlett Packard construction and would be uneconomical to reproduce for the average DIYr. This is the primary reason that the amplifier was chosen in the first place.  Retaining the transformer is a bit problematic, however.  It provides 65-0-65 VAC on the secondary, well outside the limits of most conventional regulators after rectification and filtering.  To solve this I switch the input setting of the transformer to 230 VAC.  I use a a pseudo-ground and a POOGE 5.51 Regulator, the components of which can be modified for the high voltages from the diode bridge and electrolytics. 

The input stage of the amplifier has an f3 point of 100kHz owing to the RC filter comprised of capacitors C1, C2 and R3 (C3, C4 and R4).  The dual relay U18 is connected to the 5V supply through a front panel push-button.  This allows the user to discharge the input capacitors, rather than draining to ground.  The input is protected with 1/16th amp fuses on each channel.  RFI and EMI are reduced with the 10uH input filter chokes L1 and L2, and capacitor C5.

Protection diodes in the SSM2019 are supposed to prevent the device from failing if an overvoltage should appear on the input.  They don't.  Use the paired 1n4148's as illustrated above.    Perhaps the next generation will use a second relay to completely cut out the input if the overvoltage condition occurs. 

Gain of the amplifier is switchable to 20, 40 or 6dB by placing the appropriate resistor into the gain circuit of the SSM2019.  As the original amplifier has only 2 gain settings, it is necessary to replace the DPDT original switch with a SP3T-unit.  The design allows you to takeoff the unfiltered signal directly from the opamp, or access the signal through one of two filters.  The first is a 4-pole 10 Hz Low Pass Filter, the second  is 4 pole 10Hz High Pass Filter, followed by a 100kHz Low Pass Filter.  The output of both stages is boosted by 20dB.  The 10Hz-100kHz circuit is adapted from Linear Technologies Application Note 83 "Performance Verification of Low Noise, Low Dropout Regulators".  The LT1562 does the job which would otherwise require several high speed amplifiers, but it does require that the input be protected from exceeding +/- 5V.  The output of the LT1562 is connected to the respective BNC jack through a 330uF capacitor and 100 ohm resistor.  This acts as an additional 5Hz high pass filter.

Not shown on the schematic are decoupling capacitors for the SSM2019, operational amplifiers and comparators.  I decouple each power pin with a 100nF ceramic chip capacitor.

In my case, the amplifier is fed to the input of a Hewlett Packard HP3403C "True RMS Meter".  This meter uses a thermopile as its sensing device, providing very accurate RMS noise measurements.

 

The circuit: (for a clearer image, the illustration is hyperlinked to a larger version)
 

The HP465A uses 5-way binding posts for the input and output. I replaced these with chassis-insulated BNC bulkhead connectors. Since the original plugs were 0.75" apart you can use a pair of BNC-female to-banana adapters if you decide to use patch cables.  For unbalanced operation place a male BNC shorting plug on one of the inputs. 

The Power Supply:
The Printed Circuit Board: The printed circuit board fits exactly into the space allotted in the HP465.  The illustration below (from Ultiboard) does not show the groundplane.  The yellow lines indicate where a trace has been run on both the top and bottom.  If you are building a single-sided board these are the jumper locations.  In the prototype boards I made I ran a copper layer on the top and analog groundplane on the bottom.  Note that there is a "keep out" area running horizontally along the top and bottom (with 4 illegally placed mounting holes).  This is to assure that the PCB does not actually come into contact with the chassis of the HP465
 

 

Modified HP465A printed circuit board including low pass filter, clipping indicator and power supply (Sept 21, 2004)  Note, Ground Plane is not shown for the purposes of clarity.