Hi
I know this question has been asked numerous times on the internet and it has been discussed here before, but does anyone know the part number for the Schottky diodes in the HP 8405A sampling probes?

The black diodes in the probes of this particular vector voltmeter show three lines of text on the diode body.

HP 1

517

152

I assume this means HP 1517 followed by 152

I suspect the HP part number above will be for a matched set of diodes so this part number is for four diodes perhaps? So if this is the case, I would need to know the part number for a single (unmatched) diode.

Looking online I've found HP 1901-1517 and this cross references to 5082-5414.
I have another much older HP 8405A here and it has diodes with coloured bands.?
Orange White Orange? (393?)
Orange White Red (392?)
All research draws a blank in terms of finding a part number.


I'm aware of the Richard Carey W6RLC pdf document that suggest using 1N6263 diodes but looking at the datasheet, I think the 1N6263 is not going to be the same as the original diodes. I really want to restore this VVM to full working order and it has some issues with the probes. It looks like someone has fitted 1N6263 diodes to probe B to repair it but I want to do a proper job of restoring the performance of the probes. The P JFET has also been replaced as per the Richard Carey W6RLC pdf document. At some point I'll need to see if this JFET is OK to use. However, it's the diodes I'm most concerned about.

This HP 8405A is in excellent unmarked condition otherwise, and it's a late model made in 1986. It's not mine, but if the proper diodes can be found then I will probably buy it to replace my much older HP 8405A.
Thanks!

Thanks. I still have a large bag of HP 5082-2800 glass diodes here in a little HP envelope dated 1974. There were originally 250 diodes in it but it's down to about 50 now.?
Once I know the HP 8405A is mine (I've made an offer) I plan to remove a good diode from the good probe and measure it on a curve tracer and also measure it on a VNA.

I've measured the input impedance of the probes across 2-1000MHz and it's easy to spot that the repaired probe has too much capacitance and the Rp is much lower up at VHF and into UHF.

I need to find diodes that give the correct input Cp capacitance and Rp resistance vs frequency when fitted to the probe.

I also need to check out the good P JFET on a curve tracer and measure its s-parameters using a VNA up to 1GHz or so. I've done this many times for N channel JFETs but not P channel.

The HP part number for the diodes might actually be 1901-0517 rather than 1901-1517. I really don't know how the HP system works.

See below for a measurement of the probes from my really old HP 8405A for channel A and B. I didn't bother using a narrow RBW or averaging so the plot is a bit noisy once Rp is above about 40k ohms at lower frequencies.?

I think these probes contain the original diodes. You can see that channel A and B both have an input capacitance of just under 2.5pF.
The Rp is 100k as expected at low frequencies and this gradually reduces with increasing frequency (as expected). Both probes agree very closely for Cp and Rp vs frequency.

Sadly, the 'repaired' channel B probe in the newer HP 8405A VVM shows about 4pF input capacitance and the Rp resistance is a fair bit worse up at VHF and UHF.?

Those diodes should work fine, as long as they're matched (for both forward voltage and capacitance -- and be sure to tweak the adjustments for balance, as described in the manual). What problems are you seeing with them?

You'll want to replace them in matched quads for best performance. I once repaired a probe that had "never worked right", according to its most recent owner. It had been repaired before by someone who had replaced just one of the diodes. Replacing all four with 1N5711s (pretty much the same as 1N6263) that came from consecutive positions on a tape worked fine. If you can find some HSMS-280x surface-mount parts, they should work very well (these contain 1N6263-equivalent devices). There are even some in that series that are in the bridge configuration (though rare; I've never found them for a reasonable price).

--Tom

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
420 Via Palou Mall
Stanford University
Stanford, CA 94305-4070

The main problem is that the input capacitance of the probe is too high and the Rp is lower than it should be up at VHF when compared to a good probe. I think this could be caused by the wrong choice of diodes (1N6263). I won't know for sure until I'm able to take the diodes out of the probes. It's not my VVM yet so I can't do any further internal testing.

This VVM is in such good condition it deserves to be repaired properly and if I do buy it I will service/adjust it to try and get as close to the original performance as possible.
I've got lots of the HP accessories to go with these meters plus I made a few of my own plus I have various directional couplers to use with it.

It's all a bit nostalgic because I also have a couple of modern lab VNAs here, but the HP 8405A is a classic instrument and it's also quite educational to use. I usually use the DMM recorder outputs at the back for voltage and phase, but I might also experiment by coupling up an Analog Discovery 2 to the 20kHz outputs and code it up to measure voltage and phase at 20kHz.?

What's your measurement setup, and what capacitance (and resistance) values are you seeing?

If you're seeing significantly higher capacitances, then I wonder about fixturing. Or maybe whoever did the repair wasn't careful about parasitics.

If you're seeing just somewhat higher C and lower R, then perhaps you could try some 1N5711s. These are often double-marked as 5711/6263, but the original 6263 had 10-15% or so higher capacitance, IIRC.

That is a fine instrument, for sure, and I hope you will get it back to like-new performance. It's also fun to use as a front-end for a 1GHz ETS sampling scope.

--Tom

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
420 Via Palou Mall
Stanford University
Stanford, CA 94305-4070

开云体育

I'm using an Agilent E5071B VNA and a decent cal kit and I use the fixture simulator feature in the VNA to correct for the fixture. I normally do this stuff up to many GHz so I think the fixture is fine.
The diode repair/fitment looks fine apart from the fact that the diodes are different.

I'm also measuring the relative differences between healthy probes (they all agree) and the modified probe (shows higher capacitance and poor Rp).

You can see the difference below. Channel B obviously has a problem with input capacitance and the Rp is also degraded up at VHF.

Also, if I do a flatness test across 10MHz through to 300MHz, my old HP 8405A meter and probes agree within <0.3dB up to about 300MHz. This meter hasn't been adjusted in about 10 years so this is a good result I think.

The newer HP 8405A is obviously poorly because the response isn't as flat and the two probes have up to a 1.5dB difference between them across this frequency range.?

So something is obviously wrong with it and I think it will be the repair on probe B. It might be possible to improve things with internal compensation but probe B is obviously not right. The capacitance is too high for one thing. These diodes need to be replaced with something closer to the correct part.

I think the official HP 'tee' in the HP 11570A kit will be designed for a probe with 2.5pF input capacitance so it's obviously going to cause some issues if used with a probe that now has 4pF input capacitance. I don't have an 11570A kit here but I do plan to buy one at some point. I have a homebrew versions of the tee here but they are ugly lashups and I rarely use them. They are designed for a probe with 2.5pF capacitance though.

I agree that your measurements are trustworthy and so the diodes are indeed the main suspect.

Dave Wise’s suggestion of a BAT62 is a great one if you don’t mind a breakdown voltage that is a bit less than the spec’d allowable input voltage.

And they’re more readily available than the original through-hole parts.

Tom
Sent from an iThing; please forgive the typos and brevity

Thanks, I'm going to consider all diode types from all manufacturers so I'd be happy to use BAT parts if they work OK. I'd welcome any other suggestions as well. I do have an old Avago diode kit here with their whole Schottky diode range in it but this is all in SMD.

I did another test using a decent ESGD sig gen as the source. I used my (SMA) homebrew compensated tee with the far end terminated in a decent 18GHz SMA load. Normally, I would externally level the generator using a power meter and external ALC but in this case, I'm not after great precision (yet). This sig gen is usually well within +/- 0.1dB across this frequency range when using its own internal levelling as long as I pad the output and don't turn the level up too high.

If I test the channel A probe for flatness with my homebrew tee or a homebrew compensated termination I see a gradual upwards slope up to about 500MHz. However, this slope is only about +0.25dB in total. I don't think this is in the sig gen. Some of it will be in my tee and some of it in the probe itself. However, this agrees with my old HP 8405A when I do the same test on its channel A probe.

If I then plug the channel B probe PCB into the channel A probe on the newer HP 8405A, the performance degrades a lot in terms of flatness. It was so disappointing I didn't write down the results but it exceeded 1dB by about 150MHz and hit +2dB by 250MHz.?

So something is clearly wrong with this probe element. I have a fair bit of design experience designing Schottky detectors up to a few GHz and usually when something like this upwards slope happens it will be a function of the input capacitance and any stray package inductance in the overall detector design. This typically causes some peaking up towards UHF. Adding damping resistance can help but this can spoil the input Rp.

I think it could take me a while to find the right diodes to use, but I'm in no hurry as I have the other HP 8405A here and I rarely use it anyway. I want to get this newer HP 8405A working really well, hopefully as good as the older one in terms of probe performance :)

Note that even in its current state I think it would be fine to use up to about 100MHz even with the wrong diodes in the probe. I suspect that the majority of 8405A uses will typically use these meters below 150MHz.

I've been told I can have the 8405A meter, so I can now start examining the B probe more closely. It may be that something else is damaged in it and this is causing the poor frequency response and increased input capacitance. All the diodes check out OK on a meter but maybe one of the surviving diodes is subtly damaged in some way.

I've also simulated the entire sampler circuit in the time domain on a computer and this was quite educational for me. I'm now less concerned about spreads in the JFET characteristics.
However, I will try and reverse engineer the original type of JFET used in the good probe. I can do this with a VNA and also a curve tracer. The suggested alternative JFET is the MMBFJ176 and this is a process 88 part. Looking at the datasheet, this does look to be the right type of JFET to use so I'm hoping this will turn out OK. The JFET analysis might not happen for a while though.?

I'm a bit poorly at the moment so I can't do much, but I'll try and look at the original diodes today.?

I've just completed the curve tracer comparison of the HP8405A diode against the 1N6263 diode. The diodes are quite similar, but the HP8405A diode shows much lower dynamic resistance when biased fairly hard. It gets as low as about 11 ohms at about 10mA forward bias. This is probably good when used in a sampling gate? I'm just guessing though...

I use a couple of 6.5 digit Keithley DMMs and a precision programmable PSU and a precision current sensor to measure the diode down to about 0.01uA forward bias. I usually use lots of averaging to get smooth data but I've not bothered in this case. So there is some wiggle in the curves down below about 1uA.

The VNA tests are going to be much more interesting as I test each diode as a two port device (across 300kHz to about 2GHz) and then vary the bias conditions for both the on and the off states.
I can then post process the data to get information about resistance and capacitance for each bias condition. This will create lots of data for these diodes.

I think it's probably best I post up most of my results and data on Eevblog as there are going to be lots of plots and simulations. I think there will be too much traffic for some users on the HP groups.
There's not much danger of me overloading Eevblog :)

I've just done some basic 1 port tests of the 1N6263 and the 8405A probe diode in various forward and reverse bias conditions and the 8405A is clearly a different diode to the 1N6263. The capacitance is much lower in the 8405A diode in the reverse bias condition. The dynamic resistance at HF and VHF is very close to the dynamic resistance predicted by the IV tests done at DC. This again shows a clear difference between these diodes.

However, it's probably fine to use the 1N6263 in these probes if you don't want to (critically) use the meter above about 100MHz, but I think the performance above this with begin to degrade. I've just ordered some P JFETs and several Schottky diodes that may be contenders (including the BAT-62). I've already got quite a few Schottky diodes here that I can also test.

I'm not yet sure about how to test the switching performance when used in a sampling gate. I could rig something up, but it might be easier and more relevant to simply test each diode type in the probe and look at the efficiency.

The old Philips (NXP) 1PS70 diode looks quite good on the datasheet. The IV curve isn't quite the same as the 8405A but I think it will have similar capacitance when reverse biased and the dynamic resistance will be quite low when forward biased. I think this would make this diode suitable for use as a sampling gate.

I've ordered some of these. Sadly, they are in a tiny SOT323 package. Hopefully, I can eventually find something in a classic glass axial package.

?It would certainly be nice to use HP glass diodes for the repair and I only have to find two of them :)

I think the capacitance of the classic 70V 5082-2800 will be too high. Also, I've already got a curve trace result for the 1N5711 (2800) from several years ago and it doesn't match the 8405A diodes either.

I looked at the (datasheet) IV curves of the other 5082-28xx diodes and some of them look to be quite promising. I have these here in SMD but not in the glass package.

I also have an old curve trace for a 5082-2826 taken from a Marconi 2022 sig gen and this looks to be quite close to the 8405A diode. The dynamic resistance at 10mA is very similar and the overall curve is quite similar. Back then I didn't measure the capacitance of this diode though. It is back in the sig gen (it is the ALC diode) so I don't want to remove it again. This is a matched diode and I think a matched pair of 5082-2811 diodes are used in the 2022 ALC system and this has the 5082-2826 part number. Greg kindly sent me a datasheet that shows how the matched parts are related to the unmatched parts. So I think I have this correct...

So the 5082-2811 is definitely a contender :)

Just be aware that the -281x series is rated for 20V, so you get a reduced capacitance at the expense of a reduced voltage rating for the probe. The BAT62 that Dave Wise mentioned has a lower capacitance and double the breakdown voltage, so it may be worth a close look.

And yes, the -2800 series and the 1N5711 are based on the same device, so if the latter is unsuitable, then so will be the former.

-- Cheers,
Tom

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
420 Via Palou Mall
Stanford University
Stanford, CA 94305-4070

开云体育

Thanks. I'm not sure what device will be most vulnerable when connecting the probe tip to +/- 50V DC. There will be a large transient generated via the 150pF input blocking capacitor and it might damage the PN junction of the P JFET. It might be able to damage the diodes but the transient will be quite narrow and Schottky diodes can avalanche for long periods without failure as long as the current is limited. So I'm willing to take the risk if i can get the performance from >= 15V rated diodes.

I had a rummage through my stash of Schottky diodes and managed to find some of the NXP 1PS70 82 diodes.

I fitted these to the B probe and also refitted the 8405A diode to it. I re-tested it up to 300MHz and it was much improved compared to the 1N6263 diodes.

There was a gradual slope with increasing frequency that raised the level about +0.25dB by 200MHz and +0.4dB by 300MHz. This compares to +2dB by 250MHz with the 1N6263 diodes.?

The datasheet for the NXP diode does show the capacitance is a little bit higher than my measurement of the the 8405A diodes (both at 3V reverse bias) so maybe this explains the fact that these diodes are still not giving quite the same frequency response as the originals. It's much closer, but still not right.