Putting back to life a broken Tektronix 2430 scope

Notes on a little repair project that some people may find useful

Storage oscilloscopes are useful when analyzing non-repetitive or non-periodic signals and signals with complex patterns. Having a need for such a scope for my shack I picked a used Tektronix 2430 digital storage scope from a local seller for a nifty price of $150. This is a digital unit with 50MHz bandwidth when used in real time sampling mode, and 150MHz when used in equivalent time sampling mode for repetitive signals. Overall impression is it is a nice and solidly build piece of equipment. Tek made two of these 2-channel digital scopes: 2430 and 2430a which differed in some specs (GPIB capabilities in particular) and, as I found out later during troubleshooting of my scope, in some critical parts of the schematic and calibration procedures - more about that later in this article.

The beauty of this scope from the era of 80-s (Tek discontinued support in 1996) was its self-calibrating capabilities. At power-up the scope performed a self-check routine of all the critical components and displayed the status if any of the parts of the check did not pass. This information was useful for troubleshooting the problems. I guess it was a state-of-the-art technology back in 80s. The self-calibration procedure was using the built-in DAC that generated the analog test voltages, switched the signal paths, analyzed the response and updated the non-volatile RAM with the calibration coefficients. The scope has a internal lithium battery that keeps the settings and calibration coefficients stored in the NVRAM. By the way, when I troubleshooted my scope and checked the battery I was amazed it was still right within the spec after 20-25 years (the startup screen on my scope says the firmware is from 1987) and after several years of storage on the shelf with the previous owner.

The scope was functional but had a intermittent problem when it would not pass the self-check at power-up or all of a sudden the trace would go wild in the middle of a measurement or when just running idle on the bench. Beside that, the traces appeared to be excessively noisy and the Channel 2 trace was twice as noisier as the Channel 1 one. At some point I decided I need to look into it and the original printed Service Manual that came with the scope proved extremely handy. I could not even imagine how much I will learn about digital scopes and how much fun I will have during this repair exercise.

Apparently, the Tektronix scopes from that era used Hybrid technology with their proprietary ICs. This was probably a strength at the time but may have become an issue as time passes. I say later why I think so. The heart of the 2430 is the charge-coupling device (CCD) IC. It clocks the input signal in at a high speed until the IC buffer is filled and then clocks the samples out at a lower speed. Therefore, a slow analog-to-digital converter (ADC) can be used to digitize the samples coming out of the CCD. The ADC in the 2430 scope is a 8-bit one. The scope has two CCD arrays, one per channel, and each of the CCDs has two pipes and that allows doubling the sampling rate. The output CCD signal samples are interleaved so at any given time a frame of data coming into the ADC is composed of 4 interleaved samples in 11-21-12-22 manner. The digitized samples are then logically separated by the waveform processor, recombined and displayed as two separate traces.

This picture shows the Tek 2430 with the case removed and the scope turned upside down. The main board is therefore now on top. The two scope channels can be seen going in parallel through the center mid line of the board going from the front panel to the rear of the device. First goes the shielded attenuators, then two power amplifiers, then two peak detectors followed by two CCDs. The two smaller ICs with the black finned heatsinks after the CCDs are the Phase Clock Array and the Trigger Logic Array ICs. The two big heatsinked ICs along the left side of the board are the Calibration Amplifier (by the front panel) and the A/B Trigger IC.

Interleaved architecture has its strong and weak points. The obvious strong point is only one ADC is needed, and the CCD sampling rate is doubled. The weak point is the sampling clock on the two pipes on the same CCD must be precisely aligned to meet the Nyquist criteria on equally-spaced sampling, or signal distortion will occur which also appears as noise on the trace. Read this Agilent Application Note that explains the issue in depth. This is critical to understand because one of the 2430 adjustment procedures deals exactly with this and the Tek Service Manual does not tell you why it is needed. I will give more details about it later in this article.

WARNING: When troubleshooting the 2430 scope with the case removed, care must be taken to provide adequate air flow to cool the main board that has the CCDs and other ICs with the massive heatsinks on it. This can be done using a external fan positioned in such a way that the airflow blows the heat away from the board. If this is not done, the ICs may overheat and die. Securing a fan to the scope chassis may be a non-trivial task because for troubleshooting the scope may need to be turned on the bench to have access to the top, bottom and side boards. I used a household floor fan that I placed beside the bench and that provided sufficient cooling all times. The scope should not be turned so the main board is at the bottom facing the bench, that will prevent the air from cooling the main board.

Troubleshooting of the Power-On Self-Diag not passing

  The problem

As I mentioned before, the main problem with my scope was failing the Self-Diag routine at power-on (Tests 7000, 8000, 9000), and sudden loss of sampling. When the latter happened, the screen displayed somewhat strange looking traces with spikes and ringing, even with no signal on the scope inputs. Both channels were affected at the same time.

  The analysis

My troubleshooting revealed the Phase Clock Array IC U470 (schematic 11 in the Service Manual) developed a internal failure on one of the four driving clock outputs that drive both CCDs. The clock outputs are supposed to have voltage swing from close to ground for a logical "0" to close to 5V for a logical "1". The problem was with one of the outputs suddenly losing more than half of the voltage swing, with the logic "0" being at +3V level. The pin therefore was switching between +3V and +5V as shown in the following screenshot on the left.

Bad clock line		Good clock line

The clock lines coming out of the Phase Clock Array U470 drive the bases of the switching transistors Q450, Q460, Q550, Q560 (numeration per the Tek 2430 Service Manual) to amplify and level shift the clock to the levels required by the CCDs. When the affected clock line was failing, it could not turn its transistor on and because of that the CCD was losing one of the four sampling clocks. Because the both CCDs share the clock lines, both channels were affected. That caused the display to go wild as shown in the screenshot in the next section.

  The proposed solution

To restore the CCD clock I reduced the value of the current limiting resistor in the affected transistor's base circuit by soldering a 2.2K resistor parallel to the existing 3K one. That pulled the base of the transistor a bit more to the ground potential when the Phase Clock Array presented a logical "0" (which was +3V on the failed clock line coming out of U470) and was sufficient to make the transistor switch again. As soon as I have done that, the scope's display was back to normal as shown in the following screenshots.

The scope display before the repair with a sinewave signal applied to the scope input. Only spikes and ringing can be seen.		The display after the repair with the same signal applied as in the left screenshot. Clean sinewave as it should be.

I have found though the raise time on the repaired clock line increased 3-fold from 30nS to about 100nS compare to what the other 3 phased lines had, and that may have affected interleave timing and could have contributed to some trace jitter that I am still having on both channels, which becomes more noticeable as the frequency of the signal into the scope input increases. The Adjustment procedure described in the next section could not remove it. I eliminate the jitter by turning samples averaging on. The averaging factor as little as 8 eliminates most of the trace jitter and noise, with the heavy averaging factor of 256 the traces being rock stable and smooth.

Because the problem was caused by a internal failure in the Phase Clock Array IC U470, the radical solution to the problem would be to replace the IC, but after doing a exhaustive search on the Net and contacting Tektronix I was not able to find a replacement part. So until I possibly find one in future I will stick with the 1-resistor repair mod that I have done. And by the way, the scope now passes the Self-Diag power-on test and performs Self-Cal just fine.

If you have a 2430 or 2430a scope that fails all three Extended Diagnostics tests 7000,8000 and 9000, and if the Cold Start procedure (see the Service Manual) does not cure it, I suggest checking all the digital clock output lines that come out of the Phase Clock Array IC U470 for the full voltage swing. If you find one that behaved like mine shown in bad clock line switching screenshot before in this section, chances are it is what is causing the problem. In my case it was a CCD clock driver line but there are others that may cause issues other than the one I had, up to the screen being completely blank. If a suspicious line is identified and replacing the U470 is not an option, it may be possible to cut some traces on the board or lift one end of the series resistor to insert a appropriately biased fast comparator IC or a shaping circuit made of 1 or 2 TTL buffers and convert the reduced swing signal back to the normal rail-to-rail amplitude. There is enough room inside the scope and sufficient clearance between the main board and the chassis top plane to install a small veroboard or mount the circuit dead-bug style.

Troubleshooting of the trace noise problem

  The problem

Excessive trace noise on one or both channels, even if the input is grounded.

  The analysis

I analyzed and searched excessively what could cause the trace noise. I looked at the inputs and outputs of the analog stages from the input attenuator to the CCD input and could not find anything suspicious. Also, the analog part employs differential type architecture which suppresses the common mode noise. To check the differential lines I used my other 2-channel Tek scope with the Add function turned on. That displayed a sum of the signals on that scope's channel 1 and channel 2 inputs, therefore acting as a differential probe. Everything looked fine. I then thought that this may be the ADC which is noisy probably because of the noisy reference voltage or the ADC itself. So I moved right to the ADC and managed to shift the input signal outside of its working range. Having the input into the ADC eliminated I observed a perfect straight noise-free line on the display and that told me the ADC is not the culprit. It was not noisy at all. I then traced backwards from the ADC through the demultiplexer and track-and-hold circuits back to the CCDs and finally realized the noise garbage comes out of the CCDs. The signal at the CCD input was clean but at the CCD output it was contaminated with noise, and that was not common mode noise, that was noise on top of the differential signal. So the issue apparently was with the CCD.

Was it possible that my both CCDs failed? That did not look likely. Also, the scope was passing the Self-Diag test on power-on. Could the noise be originated inside the CCD because of the internal crosstalk? Because of aging? I did not know and I could not do much about it anyways. At least I was able to isolate the trace noise problem to the CCDs. So I decided to leave it alone and move on with the manual part of the calibration. Suddenly, one of the cal operations that I performed has largely cured the trace noise problem. This was as follows.

  The proposed solution

I performed the Sample Skew Adjustment procedure per the page 5-7 of the Tektronix 2430 Service Manual. This equalized the clock timing delays between taking the positive and negative samples, resulting in more accurate interleaving of the CCD channels. I recommend reading the following Agilent Application Note: "Evaluating Oscilloscope Sample Rates vs. Sampling Fidelity" that explains the theory behind this in great details.

It may not be possible to achieve a perfect result for the 2430 scope because it uses one trimming resistor control for both channels and as such the adjustment is about finding a compromise that is acceptable for both channels. Chances are the CCDs may have different internal delays that would prevent the two channels from finding a good match. The 2430a scope has separate Sample Skew Adjustment controls for Channel 1 and Channel 2, and it has additional controls separate for each channel to align the CCD pipes gain within each channel, so my assumption is the 2430a scope can be tweaked better and would have performed better after the adjustment then the 2430 model.

Verification

The following screenshots show how the 20nS pulse on my repaired digital 150MHz Tek 2430 scope (in REPET mode) compares to one on my analog 400MHz Tek 2467B scope. The repaired 2430 digital scope did well in this test even if it had almost 3 times less of the bandwidth than the 2467B.

While the Tek 2430 has 50MHz limit in real time sampling mode, the bandwidth in REPET equivalent time sampling mode is extended to 150MHz if the signal being analyzed is periodic or repetitive. In this case turning the REPET mode on and performing samples averaging does wonders. The scope will randomly sample the input signal over a extended period of time and combine the samples together to produce the waveform on the display. The following screenshots demonstrate the displays of a 20nS pulse that my home made comb generator outputs. Note great improvement in the waveform details and rise/fall time when REPET mode is on vs. real time sampling.

Real time sampling of a 20nS pulse (6nS rise time)with sinx/x interpolation.		Equivalent time sampling of the same signal in REPET mode. Much finer details can be seen.

Conclusion

After the repair and adjustment the scope is functional with the caveats mentioned above and is back on the bench. Hoping to get a few more years of life out of it, though indications are the Hybrid ICs in 2430 series and other Tek scopes from the era may develop internal failures with time and no replacement parts will be available. It is sad because it is a nice machine overall.

References

1. Tektronix 2430 Service Manual

2. Tektronix training manual for service technicians "Troubleshooting Your Oscilloscope".

3. Agilent Application Note "Evaluating Oscilloscope Sample Rates vs. Sampling Fidelity"

In case you have questions, please feel free to email me at the address below.

Contact: miv@makarov.ca