Products > Test Equipment

Repair of a Hewlett-Packard 6632A, output voltage in error and drifts, display v

(1/1)

Daniel Dufresne:
Unit description
The Hewlett-Packard 6632A System DC Power Supply was designed in the early 1990’s. It is part of a family of three similar models, using the same printed circuit boards, the main differences are the power transformers and some part values. These are two quadrant power supplies; they can source current and sink current. It is an analog power supply with digital voltage limit and current limit. The values are entered via keyboard or HPIB. The 6632B and the other B model are much improved power supplies and are significantly different from the A models. For easy identification, here is the front panel from my unit (fig 1).

The document I refer to here is the Hewlett-Packard 6632/33/34A Power Supplies Service Manual, Manual Part number is 5957-6365, dated November 1987.  This version is available from the Keysight page and contains the errata list and backdating information.

Symptoms
This unit was given to me (thanks JLL), and at first test looked OK. Further tests showed that the output voltage did not match the set value but the front panel displayed to proper value. My unit is serial 3326A07249 manufactured in 1993.

When requesting 10,00 V the output, measured with a separate digital multimeter, DMM, was 10,75 V and dropping slowly towards 10,00 V but always out of specifications. Requesting 5,0 V produced 5,32 V and dropping. With the output voltage set to lower than approximately 4 V, the error is negative and the output drifts upward. The front panel displayed the requested voltage correctly with the available resolution.

Troubleshooting
I removed the cover and saw that I first had to brush and vacuum the accumulated dust. I do not blast compressed air to remove dirt and dust because the air will push dirt, dust and other contaminants deeper into the unit and in all connectors, switches, relays, potentiometers and other nooks and crannies making it very hard to remove. After the cleanup, visual inspection revealed nothing out of the ordinary.

Unless otherwise noted, all tests were done without a load, with overvoltage set at 22 V, output current set at 2,5 A and rear panel normal/fast switch set to Normal and the positive output, (+) connected to +S and the return (-), connected to –S, unless otherwise noted the output was set at 10 V.
I measured all the internal power supplies and all were found to be within specifications both for DC and AC values and limits. The test points locations are shown in figure 4-3, on page 4-6 (PDF file page 49) and bias and reference voltage are listed in table 4-9, page 4-21 (PDF page 64).

I decided to check the VOLTAGE CONTROL CIRCUIT built around U107; this would allow me to isolate the problem to the digital parts, the low power analog circuits, the power circuits or the feedback control. Figure 6-4, page 6-6, PDF 105. Fig 2. Fig 3.

Voltage measurements showed that the reference signal for the loop, CV prog out of CV 12-bit DAC/AMP (figure 6-3, page 6-5, PDF page 102) was stable and of the appropriate value, so did the Vmon signal, out of Voltage Monitor Circuit (figure 6-4 page 6-6, PDF page 105) and the offset voltage source, same page.
If the Vmon signal is stable and of the appropriate value but the output voltage is too high or too low and drifts, then the problem is in the Voltage Monitor Circuit since the output voltage is the input to the Vmon circuit an the circuit is a wideband DC linear amplifier. Because the Vmon signal is converted and sent to the display, it explains why the front panel display reads the appropriate voltage even if the output voltage is in error because it is the Vmon signal that is converted to a digital value and sent to the display.
The Voltage Monitor circuit is a differential amplifier with 8 resistors (2 not shown in figure), 3 capacitors and one operational amplifier U107, LF411 a JFET input device. It amplifies the +Sense and the –Sense, the voltage gain is actually 0,5 or an attenuation. Fig 4.

In-circuit resistance measurements and capacitor measurements revealed nothing out of order in the Vmon circuit. Since the feedback loop keeps Vmon at the appropriate value. I concluded that U107 was defective.

I measured the output voltage and observed the Vmon signal with a Tektronix TDS2034B oscilloscope and two P2220 probes set to 10X. It clearly shows that the Vmon circuit is not functioning properly. The bump in Vout at the beginning is not reproduced in Vmon and the long term gain is in error. Fig 5.

I removed the main assembly, unsoldered U107, installed a machined DIP 8 pin socket and a new LF411, and replaced everything back together.
Powering up, I got the same results, the problem is still there! 

Some of the resistors and capacitors in the Vmon circuit are close to the output connector on the back panel and not easily accessible when the main board is inside the unit. I decided to remove the main assembly a second time. Visual inspection around the parts related to Vmon circuit showed a brownish residue around C106 and R126 not seen in the initial inspection but caught in the photographs I made. Fig 6. Fig 7.

Closer inspection revealed that the residue is a little sticky. This explains why it was not removed by the soft brush used for cleaning. Inspection also revealed that some terminals were oxidized, and that the center of this mess was located at capacitor C110. I unsoldered C110 and saw that a little drop of electrolyte had leaked from the capacitor. Capacitor electrolyte vapors can oxidize solder and metallic terminals. Fig 8.

Repairs
Testing C110 out of circuit with a Tesla BM 591 automatic RLCG meter at 1 V and 1 kHz, showed capacitance was 465 µF with dissipation factor of 0,708. Nominal value for C110 is 470 µF rated for 35 V, so it is electrically operational. Original capacitor is made by Nichicon, series PR(M) 105 °C.
I cleaned both sides of the board area with a solution of baking soda, also known as sodium bicarbonate (IUPAC name: sodium hydrogencarbonate) diluted in water. I rinsed with deionized and demineralized water, dried the area with a paper towel, with a final flush of Isopropyl alcohol. Fig 9.

I replaced C110 with a Nichicon PF series 330 µF 35 V 105 °C I had on hand. I measured it before installation as 324 µF D = 0,400 at 1 V and 1 kHz. Fig 10.

Final test
I replaced the original LF411 for U108. I reassembled the unit and powered it up. I measured the output and success; all is well within specifications, 22 mV error at 20 V. Fig 11.

Corrections to the manual

The service manual contains a few errors and omissions. Make sure you read the four pages of the MANUAL CHANGES the beginning of the pdf document and also Appendix A, MANUAL BACKDATING CHANGES page A-1, PDF page 110.

In section 4-3, Removal and replacement, the tools needed are not listed, you will need a medium size Philips screwdriver, a Torx T15 bit with a flexible adaptor and handle, a 7 mm nut driver and, to remove the wires from the output terminals, a flat blade screw driver.
In figure 6-1 page 6-3, PDF page 97, in the ±15 V BIAS SUPPLIES, C104 is labelled 1.35V is should read 1 µF.
In the PDF document, page 98, showing Components Location, R157 is missing, also missing is C162. See image below for location. Fig 12.

Useful information

The voltage gain from the base of Q113, referred to the secondary common, to the output measured from +out to -out is 48,6 times.  A voltage of -5,0311 V at Q113-b produces + 5,0024 V at the output and a voltage of -4,8253 at Q113-b produces +15,0067 V at the output. The base of Q114 is at -5,26 V. Current flowing in Q115B-e is 8,83 mA.

To do list

This unit contains RIFA X class and Y class capacitors that will have to be replaced for safety reasons, see other posts. It also contains a few tantalum capacitors that are prone to failure and have to be replaced. This will be covered in another post.
Finally after the above repairs, I got the following message: «Error test 17». This message appears at turn-on every now and then. So the third item on my «To do» list it to repair that fault.

Regards.

Daniel Dufresne:
I hit the limit for pictures, here are the last two.

NoisyBoy:
I have a 6632A as well, the RIFA cap were definitely in need of replacement in my case after all these years, there were cracks on most of them and they are accidents waiting to happen.

I did recapped the entire unit as well, in my case, all the caps I pulled were within spec with no leakage.  I attributed it to the high flow fan that runs at constant speed, it is loud, but it does keep the unit cool and might have preserved many of the caps. 

Pull your fan, if the unit is dirty, the inside of the heatsink are likely to be full of dust and reduced cooling capacity.  I replaced my fan with a new Papst fan, which helped to reduce the noise somewhat.

Lastly, if supply accuracy is important to you, it may worth the effort to recalibrate the unit, which was easy to do as long as you have the right equipment.  Good luck.

Swake:
Changed the fan for a quieter and speed controlled model too. Makes a world of difference. The speed control 'upgrade' is documented somewhere else on the forum. Personally I used cheap speed controller from ali
https://www.aliexpress.com/item/1005003189975248.html

Have also installed front panel connectors. very easy to do as the plastic of the front is already foreseen of holes behind the sticker. Don't forget to connect the sense inputs to those new connectors or even better bring them as separate connectors to the front too.

Haven't replaced the RIFA caps yet as none of them seemed cracked. Nevertheless will replace them over the winter if time permits.

Navigation

[0] Message Index

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod