HP3458A FLATNESS DAC CONVERGENCE 195 error

[chilternview]

Just started seeing the following error on a HP3458A (old one, 1989 vintage) after an ACAL AC:

"ERR 204 HARDWARE FAILURE - FLATNESS DAC CONVERGENCE 195".

It started a couple of days ago after the meter had been left running for a week; previously it's been fine. Any tips on what to look for?

[PartialDischarge]

Check if is it running in a too warm environment

[chilternview]

Room temp is 24C, TEMP? gives 33.0

[Dr. Frank]

Room temp is 24C, TEMP? gives 33.0

That's not too hot a room temperature, but....

If your instrument is running for at least 2 hours under these conditions, then TEMP? should read at least 12°C higher than R.T., provided the fan filter is clean, the instrument stands free on a table, i.e. nor in a rack, neither any air blocking of fan inlet.

Otherwise, something is not correct with the TEMP? measurement.
Check all power supplies, please.

The error is related to the AC board A2.
It sounds, as if it doesn't happen all the time.
What's the status when the 3458A is just powered up (i.e. still "cool")?

What's the result of the extensive Self Test?

The assumed thermal error might affect the 12bit DAC, U301, or the OpAmp behind.
If the error pops up after warming up, you might specifically cool the ICs, as well U404, U406.


Frank

[chilternview]

Sorry, I should have been more accurate.

Room temp 10cm in front of the 3458A is 20.5C according to a Fluke 87V. The thermal probe reading 24C was between two stacked 3458A's - the lower good one has TEMP?=34.8C. The one with the fault described is on the top with TEMP?=33.7C As far as I can tell both the fan and filter are OK with no airflow blockage.

The error message occurs every time I run a reset or full test or an ACAL AC. As far as I am aware it started yesterday (Tuesday) morning; the unit had been powered down Sat evening and then powered on and left on Monday morning.

It seems to read OK on ACV at least on the 1v/10 ranges and at freqs 10Hz-1MHz. I wanted to keep it powered up for a while as I'm running the drift test, so far  (subject to the power cycle mentioned above, and after only 1 week) it's showing 0.09ppm/week drift compared to the other known good 3458A with 0.03ppm/week.

I'll try some freezer spray on those ICs as soon as its powered down again.

[martinr33]

Good suspect here isthe Elantec comparator, EL2039, I'd start wthe the freeze spray there.

These are super high speed, sensitive parts and somewhat prone to failure. They have not been made for many years, and there are more counterfeit than genuine sources out there.

[MiDi]

Good suspect here isthe Elantec comparator, EL2039, I'd start wthe the freeze spray there.

These are super high speed, sensitive parts and somewhat prone to failure. They have not been made for many years, and there are more counterfeit than genuine sources out there.

TiN seconds that: Flatness DAC convergence: 198

Good luck on finding genuine replacement 

[chilternview]

So I saved the contents of the CALRAM and settings RAMs using WinGPIB to dump to a bin file. Then replaced first the CALRAM after flashing the saved bin file and verified that was OK. Then did same for settings RAMs.

All looked fine although I did see a message about a settings ram checksum error initially. But running a full self test was OK apart from the 195 error. And the meter seemed to be working OK.

Until I tried to query CAL? 72. And found that there was no menu entry CAL?... Half the commands seem to have disappeared.

Question: Is this likely to be due to the settings RAM change? Which of U121/122 is the low address (0x120000-0x127fff) and which the high address (0x128000-0x12ffff)?

[Dr. Frank]

You obviously successfully wrote back the CALRAM content, but yes, you've lost the content of the setup RAM, maybe by confusing Hi and Lo Byte of U121/122.
That's no big deal, I would just manually re program any Keys and settings as before.

For accessing the full menu call: MENU FULL from the key pad.

Frank 
 

[chilternview]

Dr Frank, you're a hero! I was seriously pooping myself I'd done something irreversible...

Are the settings RAMs just that - settings that can be changed by the user and are non-volatile? I'll read the manual on changing the key settings, but in the meantime MENU FULL works just fine.

R.e the flatness error, that's next on the list to check U404, 406 etc. when I can find some freezer spray.

[Dr. Frank]

R.e the flatness error, that's next on the list to check U404, 406 etc. when I can find some freezer spray.

Well, I assume that you did not answer the question yet, if this error occurs on a warm instrument only, or if it vanishes if the instrument is cold.

If it occurs also on a cold instrument, this cooling method makes no sense.
So please clarify this effect first.

I also was a bit reluctant to directly point to any of the ELANTEC ICs, like MiDi, as I hope the best for you, that only this DAC is the root cause, which might also be difficult to find.

Frank

[chilternview]

Sorry Dr Frank, so far I've just tried powering off for a short while (10-30mins). On power up, there is no initial error. If a full self-test is run (or an ACAL AC) it shows the error though. I'll try leaving it powered down for longer - e.g. overnight - and see if it passes self test then.

[chilternview]

Well, I assume that you did not answer the question yet, if this error occurs on a warm instrument only, or if it vanishes if the instrument is cold.

Unfortunately with the instrument cold, then powered on this morning (room temp 22.1C), self test failed with the flatness error. So does not appear to be a temperature issue.

Does it make sense checking voltages or waveforms around the flatness DAC? And if so under what conditions?

[Dr. Frank]

Fortunately, I did not experience such a fault on my 3458A yet. So unfortunately for you, I can not give any specific advice on this problem.

You might check whether the DAC receives proper digital input (problem with the ring-shift-registers?), and if it outputs a proper analogue signal, which I assume to propagate across the AC amplifier. I wonder how to analyze the signal before and after the ELANTEC amplifier, maybe apply an AC signal in the correct AC mode and observe what signal comes out of it.

Sorry, I don't want to speculate any further, maybe TiN has some detailed hints on his xdevs pages, or you might contact him directly, how he identified the ELANTEC as the root cause.

Frank

[chilternview]

Thank you Dr Frank, I'll check around U401B/U301/U302 the components around the flatness DAC. I seem to remember seeing someone else had a problem which was traced to bad opamp U401B.

[chilternview]

Some probing voltages on the pins of the flatness ICs on board A2 (marked 03458-69502, ERC4312, 09-9226). All measured with a 34401A with input resistance set to 10Gohm.

U401 (MC34072P)
1: +11.7mV    8: +14.0V
2: +11.7mV    7: +2.0mV
3: +10.1mV    6: +2.0mV
4: -17.7V        5: +1.2mV

The above I believe look OK.

U301 (AD7240JN)
1: +2.0mV    18: -22.0mV
2:  0.0V        17: -0.6mV
3: -1.27V      16: +13.9V
4: -1.27V      15: -1.27V
5: -1.27V      14: +5.0V
6: +5.0V       13: +5.0V
7: +5.0V       12: -1.27V
8: +5.0V       11: +5.0V
9: +5.0V       10: -1.27V

Schematic says pin3 of U301 should be -1.4V but -1.27V is probably OK.

U302 (LT 9150)
1: +11.16V    8: -16.5V
2: +27mV      7: 14.0V
3: -8.3mV      6: +8.5mV
4: -17.7V       5: 11.18V

I'm wondering about these - the opamp output (pin 6) is +8.5mV, the +input (pin 3) is -8.3mV, the -input (pin 2) is +27mV.

[Dr. Frank]

Well, this A302 (wideband amp) probably is defect. Pin 2 should be -8.3mV as well, and pin 6 =  -83mV.
I assume that all voltages are measure relative to AGND.

That ain't no good message either, as this is a special part, maybe HP custom specific, as I don't even find a datasheet.

Let's see what's used on the new A2 board (black edition).
Seems still to be the original one, an LT318A, as denoted in the schematic. Still in production.
The 12bit DAC, U301 = AD7541

Frank

[chilternview]

Well, this A302 (wideband amp) probably is defect.
I assume that all voltages are measure relative to AGND.

Yes measured from TP402 (analog ground).

That ain't no good message either, as this is a special part, maybe HP custom specific, as I don't even find a datasheet.

Let's see what's used on the new A2 board (black edition)

Frank

It's shown as a LT318A on the CLIP schematic, which broadly fits the description.

[chilternview]

Iteresting... I removed the A2 board and took a closer look. It's interesting that the date codes on this board are mostly '92, whereas the other boards are datecode 88. So it may have been replaced early on in the life of this 3458A.

A closer look shows the EL2039CN has a datecode of 99 - and look at the back of the board, it's been replaced with the flux left as soldered!

I'll still swap U302 for a new LT318A, which should arrive today, but maybe that Elantec opamp is worth checking too.

[Dr. Frank]

All these Elantec ICs are the weak spot of the 3458A, over long time and / or long usage.
Therefore, you're lucky that it failed already and had been replaced.. I would leave it alone, as it probably is not the root cause in this case.

Would be really great if U302 was the only culprit.

Btw.: Have you already updated firmware (to 8 or 9) and refreshed the nvSRAMs?

Frank 

[chilternview]

All these Elantec ICs are the weak spot of the 3458A, over long time and / or long usage.
Therefore, you're lucky that it failed already and had been replaced.. I would leave it alone, as it probably is not the root cause in this case.

Would be really great if U302 was the only culprit.

I hope so!

Btw.: Have you already updated firmware (to 8 or 9) and refreshed the nvSRAMs?

Yes the firmware was updated to rev 8 and new(ish) DS1220Y-150 and DS1230Y-150 in. The calram contents were copied OK.

[chilternview]

Well some LT318A arrived in the post this morning and so U302 duly replaced. Alas it did not make any difference, self test gave the flatness dac error. So I guess the next to try replacing are U301 and U401?

[Dr. Frank]

Hello,
you had measured the voltages of Input + and Input - on this U302... which showed an offset. Have you repeated this measurement, and did that improve?
Do you see any oscillation on this OpAmp?
To replace U301 would be another action, although it makes sense only, if that offset is still present.
Why U401? Have you seen any strange voltages on these?

TiN made similar measurements on U404, and found a big Offset as well, so that would be the next test to do.

Let me find the link: https://xdevs.com/fix/hp3458a_u4/

He applied a DC signal in the ACDC mode, obviously, and measured across the signal path in the AC/DC amplifier, including U404

Frank

[chilternview]

Hello,
you had measured the voltages of Input + and Input - on this U302... which showed an offset. Have you repeated this measurement, and did that improve?

No, with the replacement U302 the voltage on pin 2 (-ve input) is +110mV, pin 3 (+ve input) is -0.96mV, pin 6 (output) is -8.5mV.
I'm still unclear why the + and - inputs are different, it seems odd to me. The voltage of the junction of R303/R304 is +0.04mV which is consistent with the output, but I don't see why the inverting input has quite an offset.

Do you see any oscillation on this OpAmp?

No, no oscillation.

To replace U301 would be another action, although it makes sense only, if that offset is still present.
Why U401? Have you seen any strange voltages on these?

It was a thought as U401B drives the reference input of U301, the DAC. Although the voltages on it seem fine.

TiN made similar measurements on U404, and found a big Offset as well, so that would be the next test to do.

I checked U404, the +ve and -ve inputs both read +6.970V which is as expected. This is the EL2039 that appears to have been replaced in the past.

[Dr. Frank]

No, with the replacement U302 the voltage on pin 2 (-ve input) is +110mV, pin 3 (+ve input) is -0.96mV, pin 6 (output) is -8.5mV.
I'm still unclear why the + and - inputs are different, it seems odd to me. The voltage of the junction of R303/R304 is +0.04mV which is consistent with the output, but I don't see why the inverting input has quite an offset.

No, there might be something wrong with one of the 3 resistors, R302, 303, 304.
If pin2 is more positive than pin3, the output should be - V_supply.
If you measure with your DMM (10M input) on pin2, and assumed, that the feedback path via R304 is open, then you will measure about zero Volt, or 10M times the bias current of the OpAmp, +/- 6nA * 10MOhm ~ +/- 60mV, which is about the 27mV you're measuring. These 40mV between R302 and R303 should be exactly 1/10 of V_out, therefore this does not fit as well.
I suspect that R304 is broken, or has an open solder joint, or maybe any of the other resistors.
Please measure continuity (i.e. the resistances) of all 3 resistors, by probing between the affected pins of the OpAmp, ground and the common junction of all 3 resistors.

That would be a very uncommon error, but I also once encountered defective Thin Film resistors in our applications.
The protection film on top of the resistors had pinholes, and humidity plus voltage across the resistor oxidized the NiCr thin film layer until the resistor was open.

You have an old unit, so who knows.

Frank

[Dr. Frank]

Wait a moment, I had a look on your picture.
You can see R303 and R304, right to U302, and both their soldering joints seem to be broken. Compare this to the solder joint of the blue capacitor to the left..
Or is it just a bad lighting in your picture?

Frank

[chilternview]

These 40mV between R302 and R303 should be exactly 1/10 of V_out, therefore this does not fit as well.

The junction of R302/3/4 was actually 0.040mV i.e. 40uV.

I measured the resistors with a 34401A:
R302 (1K) was 921ohms
R304 (10K) was 9.89Kohms
R303 (9K) was 2.7Kohms

Given that the output of U302 is shunted to earth by other resistors all of those look OK.

I also checked the readings with the DVM set to 10M and 10G input resistance, pin 2 read +95mV with 10Gohm input resistance, and +110mV with 10Mohm input resistance. It looks like R304 is OK.

About the solder joints, I think that's the light. I did try re-tinning the joints of the 3 resistors, no effect.

I can try replacing U302, not sure what else to try.

[Kleinstein]

One should be able to follow the signal path with a scope. So apply an AC signal to the 3458 and check the signal at U401 , U301 and U302. Unless the DAC is set very close to zero a copy of the input signal should be there.

Is the supply to U302 OK ?

[Dr. Frank]

Maybe you could check, if the supply voltages are ok.. or creating spikes..
Another U302 would mean, that the new one was damaged also, by overvoltage?
Does the old OpAmp still work properly on a breadboard, maybe at +/12V supply?

I'm baffled, meanwhile.

Frank

[Dr. Frank]

If the output is at -8.5mV, then 1/10 is about -850µV, but not +40µV, btw. That's not consistent, even on such low levels.

There's still something fishy in the resistor network, or the OpAmp draws too much current to its inputs.
What's the voltage across this 10k resistor?

Frank

[Kleinstein]

The LT318 can have quite some input bias current with typical 120 nA. So some 0.9 mV drop over the 10 K resistor is about normal - more on the low side.

Something like supply spikes are maybe a poor turn on sequencing possibly damaging the OP is a good point.

Parts of the amplifier are very fast and may oscillate at a frequency well above 100 MHz to fast to see with a slow scope. It is not very likely but can make parts behave strange.
Spikes on the -17 V supply could also cause some confusion.

[chilternview]

Thanks for the suggestions, so I tried applying a 10V 1KHz sinewave and setting the 3458A to DCACV. The value it measured was in the right ballpark (~3.5v).

A scope on pins 2 (inverting i/p), pin 3 (noninverting i/p) and pin6 (o/p) is attached. It looks like there is 10x gain from U302, with some spikes on the signal - these appear to be on the power lines.

Voltages measures were:

pin1 +11.71V      pin8 -16.5V
pin2 +114mV      pin7  +14.0VV
pin3 -0.4mV        pin6 -3.7mV
pin4 -17.75V       pin5 +11.73V

So the supplies are -17.75V and +14.0V. Are those OK?

[Dr. Frank]

If you also have bought a replacement for the 12bit DAC, then I would try it out.

Frank

[chilternview]

I am waiting for the DAC to try and see if that helps. In the meantime I tried measuring the voltage on pin2 of U302 for 4 different new LT318A's from Digikey, plus the original, HP-part number marked one.

The new ones measured from +120mV to +126mV. The original measured +65mV. Maybe the HP marked one was selected for low input bias current?

With R304 being 10K (I double checked by taking U302 out of its socket so R304 was floating under measurement)  then the input bias current can be calculated to be of the order of 12uA (or 6.5uA for the original part). Both are way over the datasheet spec of 150nA typical, 250nA max.

So it this high reading on pin 2 still makes no sense to me. And I'm not clear what the purpose of R304 is, either.

[Dr. Frank]

If the DAC has an output impedance of about 10k, then R304 would serve as an offset compensation for the bias current.

I still suggest to check the disassembled OpAmp on a breadboard if it works correctly.

Maybe you de-solder all 3 resistors, check them out of circuit and solder them again in place. I still do not trust these.

The bias current as well as the offsets cannot explained otherwise.

Frank

PS: the obsolete AD7240 has an output resistance between 7  to 15k, typ. 12k, so that's obviously the correct explanation about bias compensation.

I don't know, which replacement you have ordered, but the recommended AD7541 has a completely different pin out, requiring an adaptor.

[Kleinstein]

The DC readings at the U302 inputs is still very odd and can not be explained by the bias. Especially the +120 mV at pin 2 are strange.

A possible explaination would be that the meter used to measure the votlages effects the circuit: Additional capacitance (from the meter) at pin2 can cause oscillation of the OP amp. So when probing Pin2 one should have something like 100 K in series directly at the OP side to reduce the effect of the meter capacitance.

120 mV would also be enough to see with the scope. The scope trace from reply 32 shows much less than 120 mV.
The AC waveform shown from the scope suggest that U302 is working as intended at least for the AC part and also the DC part look OK with the scope.

[chilternview]

120 mV would also be enough to see with the scope. The scope trace from reply 32 shows much less than 120 mV.
The AC waveform shown from the scope suggest that U302 is working as intended at least for the AC part and also the DC part look OK with the scope.

Kleinstein, well spotted, why didn't I notice that!

With a 10X scope probe, DC coupled, there was no offset on pin2 (< 2mV) and no oscillation as in the pictures. When I connected the 34401A probe though, there was a 1V p/p ~500kHz oscillation! And of course the meter read a DC voltage in the tens of mV.

So U302 is OK and performing as expected. And I've learned a valuable lesson: always look with a scope if a reading is odd...

[chilternview]

One other question while I'm waiting for the replacement DAC:

Anyone know what the code at the end of the 'FLATNESS DAC CONVERGENCE 195' means?

I have seen others (e.g. TiN) report the error with number 198. Even someone with 197. But what does the code mean?

Has anyone disassembled the 3458 firmware so one could try and figure out what it's doing in that test?

[Dr. Frank]

that code is nowhere specified, unfortunately.
Even if there would be people who reverse engineered the code, they won't tell, to obvious reasons.

[Kleinstein]

The extra number probably is a slightly more specific number on which test failed. So 195 is likely an earlier failure in the flatness testing. This could be as little as a different range detecting the error, as the same DAC is used for different dividers.

I see 3 different types of error causing such an errors:
1) a problem with the DAC (it still looks like it works somewhat) or maybe the shift register
2) a problem with the test signal generation or maybe the measurement of the test signal (though this part should be tested in a different test, though that error may come later and may not show)
3) a problem with one of the dividers (drifting resistor or capacitor, bad JFET (or LM339) for switching), so that the DAC range is insufficient to get a good waveform.

If one has a good AC source or other meter one may be able to check the error in the AC ranges - chances are 1 or more of the AC ranges are out of spec, though not sure how much (could be just a few ppm or a few percent), either at low frequencies or higher frequencies (e.g. 10 kHz) or both. A gross error, like from a not working LM399 / JFET should be visible.

Another point to test could be to see how good the AC amplifier actually is, e.g. with a 1 kHz or similar rectangular signal, a bit like in the scope probe adjustment. Ideally this should give a relatively clean square wave also at the AC amplifer output. By tying the DAC output to ground one would get an intentionally bad setting.

[Traceless]

@chilternview:
Could you locate test point 901 on the A1 board, monitor it with your oscilloscope during ACAL AC and post a screenshot here?

[chilternview]

@chilternview:
Could you locate test point 901 on the A1 board, monitor it with your oscilloscope during ACAL AC and post a screenshot here?

I'm not sure if it helps, as there is a *lot* going on during an ACAL AC. Here is a bit caught just after the ratio tests.

[chilternview]

If one has a good AC source or other meter one may be able to check the error in the AC ranges - chances are 1 or more of the AC ranges are out of spec, though not sure how much (could be just a few ppm or a few percent), either at low frequencies or higher frequencies (e.g. 10 kHz) or both. A gross error, like from a not working LM399 / JFET should be visible.

I ran the output of a HP3245A set to 10VP/P AC sinewave via coax cable to the 3458A (HP3458A-2 below) and another, Keysight-calibrated 3458A (HP3458A-1 below). Here are some measurements. The range was set manually on the ACV reading.


Freq 1KHz           HP3458A-1        HP3458A-2
---------------------------------------------------
Range: 1KV         0003.687          0003.811
Range 100V         003.5366          003.5779
Range: 10V         03.53392          03.53555
Range: 1V           OVLD                3.421325

Freq 10KHz
----------------------------------------------------
Range: 1KV         0003.690          0003.781
Range: 100V        003.5411          003.5827
Range: 10V         03.53978          03.55297
Range: 1V           OVLD                3.575434

Freq 100KHz
----------------------------------------------------
Range: 1KV         0002.461          0003.157
Range: 100V        003.4822          003.5705
Range: 10V         03.53122          03.57405
Range: 1V           OVLD                3.593496

Freq 1MHz
----------------------------------------------------
Range: 1KV         0001.421          0001.361
Range: 100V       001.7199          002.2459
Range: 10V         03.0461           03.11238
Range: 1V           3.148369          3.169615

[Traceless]

@chilternview:
Could you locate test point 901 on the A1 board, monitor it with your oscilloscope during ACAL AC and post a screenshot here?

I'm not sure if it helps, as there is a *lot* going on during an ACAL AC. Here is a bit caught just after the ratio tests.

Hello chilternview,

thanks for posting the scope screenshot. What we are looking for is a square wave output with peaks > 10V and < -10V (which is outside the ADC operating range) immediately before the meter outputs the "error beep sound". If that happens *thoroughly* check Q802 (don't rely on one of the cheap IC testers). That should be an 2N4392-N-channel JFET.

That being said while the signal on your screenshot is still within the +/-10V range the part I marked red on your screenshot does look suspicious to me. I'm not sure what is happening at that point in time but, the -5.5 - -6V peaks look odd. It would probably be helpful if someone with a working 3458a could provide an ACAL log of TP901 for reference.

Cheers Traceless

[chilternview]

Thanks traceless,

The bit captured in the following was by single shot triggering on +9V rising edge. At approx t=6s after triggering, the beep indicating the error occurs.

There are +12 / -15v peaks then +3/-3v levels (approx 31 of each) - with spikes of varying amplitude. Not sure if this is revealing anything useful.

I have seen someone else refer to an issue with Q801/2 elsewhere, can't think where. I do have a couple of spares so it may be worth checking them as you say - I believe that part of the circuitry is generating the cal voltage input?

[chilternview]

Tried replacing both Q801 and Q802, alas no difference...

[Kleinstein]

The TP901 signal is the signal going to the ADC in fast mode. For the compensation test this would likely be the output of the S&H stage.
The rather large overshoot does not look very good, but is quite short and may be OK. It is not so clear at what time the actual ADC conversion is done.
For comparison one could looke at the AC main amplifier output (TP401) to see if the large overshoot is an effect of the AC amplifier or the S&H path. For the main amplifier that much overshoot would be suspicious. For the S&H part it may be OK, if only short.

What is a bit strage is that the amplitude for the +3-4 V signal varries quite a bit between the pulses, while the negative side look more or less stable. This suggest that there is quite some effect of the compensation settings for the positive side only. The neg. side may be fixed just to test the other sign too and no real search for the best solution there.

The AC test shows surprisingly large errors even for the "good" meter. Chances are the 3.5 V eff are borderline for the 1 V range, so the overload result is not a significant result. To test the smaller ranges it would need a smaller amplitude (e.g. some 100% of nominal range). The question would be if the low frequency (e.g. 100 Hz) and higher frequency (e.g. 10-50 kHz) readings are similar in ratio. The DAC for the flatness should effect the low frequencies below some 2 or 8 kHz.
The test in the 1000 V range and a bit already the 100 V range are on the low side for the amplitude, especially for the analog RMS case. So the difference there could be an effect of the RMS chip.

[chilternview]

Sorry Kleinstein, that *was* TP401 that the scope was connected to i.e. the amp output.

[Kleinstein]

For the amplifier output the levels at some +12 and -15 V are about the limits on what the amplifier can output in saturation. The test signal is somehow generated idurectly via flip flops and that may cause the extra phases at the start. So no problem with the asymmetry there and the pulses itself may also be no problem.

When reaching the slew rate limit the amplifier may be a little asymmetric can slower in one direction than the other. So asymmery in the overshoot is not a big surprize.
A possible reason for higher than normal ringing could be an error in the amplifier compensation adjustment, the part that goes via CR402 and seems to be not part of the normal ACAL. So a wrong value in the CAL RAM could cause this. A rather small voltage at the cathode of CR402 / output of U402B would be a bit suspicous.

The ringing may as well come from the test signal and may not be an issue at all.  One could check this with a a good square wave at the input.

[chilternview]

For the amplifier output the levels at some +12 and -15 V are about the limits on what the amplifier can output in saturation. The test signal is somehow generated idurectly via flip flops and that may cause the extra phases at the start. So no problem with the asymmetry there and the pulses itself may also be no problem.

When reaching the slew rate limit the amplifier may be a little asymmetric can slower in one direction than the other. So asymmery in the overshoot is not a big surprize.
A possible reason for higher than normal ringing could be an error in the amplifier compensation adjustment, the part that goes via CR402 and seems to be not part of the normal ACAL. So a wrong value in the CAL RAM could cause this. A rather small voltage at the cathode of CR402 / output of U402B would be a bit suspicous.

Thanks, I'll take a look.

The ringing may as well come from the test signal and may not be an issue at all.  One could check this with a a good square wave at the input.

I have tried running a square wave of amplitude 0.1/1/10v and frequencies of 100Hz-1MHz and the waveform at TP401 looks pretty good - some sloping at 100Hz from LF rollof, very good from 1KHz to 100KHz and some HF degradation above 100KHz.

I don't see much signal amplitude on the REF input of the DAC U301, what sort of voltage (DC or AC?) should be expected here?

[Kleinstein]

The schematics (from TiN's website) has some missing parts in where the ref signal is comming from.  It looks like there are 2 passible paths to choose from.
One looks like a 1/10 divider (R202,R201 = 9K and 1 K)  from the guard signal and thus 1/10 of the input to the amplifier. This would also be relatively small: 1/200 or 1/2000 of the amplifier output.

The lower ranges (10 mV,100mV and 1 V) could also use the gain 0.001 signal and thus 1/1000 the input signal and thus also a rather low level.

There likely is a reason to use such a small signal and than have a gain of 10 after the DAC.

[Traceless]

Thanks traceless,

The bit captured in the following was by single shot triggering on +9V rising edge. At approx t=6s after triggering, the beep indicating the error occurs.

There are +12 / -15v peaks then +3/-3v levels (approx 31 of each) - with spikes of varying amplitude. Not sure if this is revealing anything useful.

I have seen someone else refer to an issue with Q801/2 elsewhere, can't think where. I do have a couple of spares so it may be worth checking them as you say

I'm not sure, the peaks you recorded are extremely short and a lot less frequent compared to the case I had in mind (see video link below, the interesting part for you starts at around 25 minutes into the video). Those transient spikes may or may not be okay.

- I believe that part of the circuitry is generating the cal voltage input?

Yes you can see it on this schematic* the CAL line from the "Calibration" part of the circuit where Q801 and Q802 are located (page 3) is connected to the input of K203 which is part of the "High Voltage Attenuators and Input Switching circuit" on page 1.


(*While the schematic linked above is a lot easier to read than this one Q802 seems to be mislabled as Q803 compared to the original documents (p. 53)).

[chilternview]

Finally a AD7420CJ DAC arrived in the post. With the new part I eagerly awaited a self test... and it failed again with the same error.

[seebeck]

Hi Folks,
My 3458A turned out to have the same problem ，“Self test failed” --> "Flatness DAC Convergence:198" and showed 3xxx.xxxV on ACV with shorted input such as MiDi's plot# https://www.eevblog.com/forum/metrology/3458a-worklog/msg3627094/#msg3627094.
According to https://xdevs.com/fix/hp3458a_u4/, measured TP401 -7.89 VDC, then to checked U404 EL2039, inverting input had +6.892 VDC, non-inverting input +7.256 VDC.
I have been removed U404 and measured the pin1 non-inverting input and pin14 inverting input, voltages are 7.258 VDC and 7.093 V. Normally，TP401 should be equal to ~1.65 VDC if U404 is working well. It's a common issue for old 3458A.
U404 Elantec EL2039 is EOL, and Elantec as been bought up, Intersil... HA-2839（ref to https://www.alldatasheetcn.com/datasheet-pdf/pdf/91712/INTERSIL/HA-2839.html） is a p2p for EL2039. Currently, Renesas Electronics. neither of them had a great substitute for it.
EL2039/HA-2039/HA-2839，600MHz GBW and 600V/us high slew rate.. In the open market，you can't get any good one these days.
Q: How to replace this IC? Do you have any suggestions?

BR.
Seebeck

[syau]

Possible reliable source for EL2039 is utsource

[seebeck]

Hi Dr. Frank,

I found ADI, NXP and Signetics ever had a p2p device and their basic spec are almost same as EL2039.
The model is ADI AD5539 / NXP(early produced by Philips) or Signetics (currently, TI) NE5539, but this device are also being discontinued. 
AD5539's TDR is provided in attachment, please have a review.
Thanks a lot!

BR.
Seebeck

[Dr. Frank]

Supply voltage is +/- 10V only, so this seems to be the culprit, same goes for the other ELANTEC comparators.
You should have a look on the PCBs, how these parts have been replaced in the 3458A Black Edition.

Somebody has high-res photos on his site 

For the fast comparators, +/-5V types from T.I. I tink were used, and individual +/-5V supplies with small 78/79L05 were added.

Frank

[seebeck]

Well, really!
EL2039's supply voltage is +/-15 VDC, 10V is too low. AD5539/NE5539 was mainly used in audio circuit.

Somebody I'll check it immediately.
Thanks a lot.

With best regards.
Seebeck.

[Kleinstein]

The clipping detection is at +-10 V for the output. One might get away with a slightly lower supply, as long as the amplifier can drive up to +-10 V. So from the supply a AD8021 (+-12 V max) may be boarderline working, though a little slower.

The new black edition may give a hint on what they use.

A different amplifier may want some extra supply regulation and possibly slightly different capacitors for the compensation. The higher frequency end of interest can already be effected by the input capacitance and details of the frequency response. There are more parameters than just the GBW. Just the amplifiaction would get away with a little less.