| Electronics > Repair |
| Agilent E4421B RF generator restoration project |
| (1/9) > >> |
| nctnico:
About a year ago I bought a 'for parts' E4421B from Ebay. The most common fault I found when researching these machines is the output amplifier going bad. I more or less assumed this machine was most likely to have the same problem. -Spoiler alert- I was wrong :scared: The most obvious problem was the display. Did someone stick a post-it behind it? After some reading I discovered these DSTN screens can also create a yellow color when overdriven. After opening it up I got a bad feeling. Everyone who has worked at a repair shop knows the look of a test unit. The unit with the boards which can't be fixed but are good enough for experiments or to try something quickly. Every module has a blue cross on the outside... To sum up the problems with my E4421B: * Frequency slightly off * Output level too low and distorted * Display not working * No AM modulation * Dead backup batteryThe first thing I attacked was the frequency problem. According to the service manual (which only has block diagrams and no schematics) the reference board should be the place to look for that problem. |
| nctnico:
The reference board The reference board generates the reference frequency and a 1GHz signal which is used on the output board to generate frequencies in the range from 100kHz to 250MHz. The AM and FM modulator are also on this board. After opening the reference module it quickly becomes clear a chip labelled U13 is missing :'( That is about the worst that can happen when trying to repair something without a diagram. Oh crap... :palm: Some probing around learns me that the missing chip is connected to the reference input of a DAC (circled in red) which seems to have been replaced given the later production date than the other chips and lots of flux residu. So U13 must be some kind of reference. Fortunately Mike (Mikeselectricsstuff) did a teardown and he took a closeup of the chip: Some Googling for a Linear Technology device labelled 2110 revealed it is an LT1021 10V precision reference. Let's order one but first see why it was removed. I hooked up some wiring to supply 10V from a bench PSU and measured the current: It shows more than 11mA is drawn from the reference chip. Since the maximum current is 10mA it is no wonder why the reference chip was removed. With the bench power supply set to 10V the output frequency is spot on though. I tried to find other loads connected to the LT1021 but the circuit gets fuzzy quickly. Now how to proceed?? Since the reference can source and sink current I decided to live with the 11mA and added a 470 Ohm bypass resistor from the input to the output (this is a suggestion from the LT1021 datasheet). This should limit the current through the LT1021 to around 5mA. I suspect the AM modulation not working could have something to do with the high current draw on the reference but I think I just leave that. I declare the reference board 'fixed' for now. |
| G0HZU:
Interesting thread :) It's a shame that HP/Agilent/KS won't release schematics of these old signal generators. They do seem to be suffering from various reliability issues and I'm not sure I envy your task in terms of repairing your list of faults... --- Quote ---I suspect the AM modulation not working could have something to do with the high current draw on the reference but I think I just leave that. --- End quote --- In case you don't already know this, the AM often comes for free on a sig gen that uses ALC to keep tight control on the RF output level. This is because the AM modulation is often injected into the ALC system as a modulated reference for the ALC to aim for and track. So as long as the ALC response time is faster than the AM modulation then the ALC will track and AM modulate the RF output faithfully. So if the levelling is faulty then I'd expect the AM to be faulty too. But I don't have any experience of working inside these E44xx series generators even though I own a couple of them. So maybe this doesn't apply. What I will say is that these aren't really lab grade generators in terms of AM modulation quality. So don't expect too much in terms of low distortion if you dial up AM modulation from the menus after you repair it. Both of my E44xx generators are quite poor in this respect. |
| KJDS:
I've recently found a couple of dead units in the back of my storage lot. One has a label on it saying "display broke". No information on the other so I've just tested it and the display works but it looks like it has a lot of other faults on it. |
| nctnico:
The output board The picture below shows the output board and the signal paths (0.1 to 250MHz in blue, 250MHz to 3GHz in purple). There is a lot going on here. Besides the end amplifier this board also contains the filters to filter unwanted harmonics from the output signal and the auto-level circuitry. The big giveaway was the blue cross on the end amplifier (1GMI-420). Ofcourse this chip is unobtanium so I had to somehow replace it. The first thing I tried was to replace is with a wire between the input and output. This gave some signal but at very low (and distorted) levels. I needed something stronger so I made a small board with an ADL5602 MMIC as an amplifier replacement. This gave me an idea on the amount of amplification needed. This little board had a major problem though: it sang like a bird due to the poor layout. From this board it became clear there was also something wrong with the auto leveling. After taking parts of the log-amp circuitry apart it became clear the detector diode (2 diodes in one package; blue circle) was broken as well. Fortunately there was a diode with the same markings on the board and even more surprisingly only one diode was used (red circle) and that diode wasn't the one that was broken in the detector diode. After swapping these over the automatic levelling started to show some signs of life. I did some measurements on the input and output levels of the amplifier and it appeared that a lot of harmonics at the output where produced because the input signal was already overdriven. This was especially true for the <250MHz range because the input signal is around 3dB lower. I guestimated the original amplifier had an amplification of at least 30dB. I was unable to find a low noise MMIC with that much amplification. I also needed a low noise (low noise figure) and low distortion (high IP3) part. I decided to cascade 2 MMICs with a gain of around 15dB each. I choose the Triquint TQP3M9028 devices because of their low noise and low distortion figures. A problem with MMICs is that they don't perform well at low frequencies which isn't surprise because their outputs are shorted to the power supply. I did a test with a resistor in series with the inductor and this greatly improved the low frequency performance. I had some PCBs made by Seeedstudio which have the same size as the original amplifier. The replacement PCB has two layers; the bottom layer is a solid ground plane. The attached PDF shows the schematic of the amplifier. I also found the root cause of the amplifier failing: The aluminium block it is mounted on has a very rough surface. This is extremely bad for heat transfer. I don't understand how HP/Agilent/Keysight managed to overlook this! If the aluminium block and the module housing where milled so the block and the housing had smooth surfaces the amplifier most probably wouldn't have died. Anyway, with this fix the output signal works over the entire range albeit with some distortion at the lower frequencies (100kHz to 5MHz). The maximum output level is about +3dBm (more between 250MHz and 2GHz) but that is good enough for me. |
| Navigation |
| Message Index |
| Next page |