Repair of a 7S14


 

I recently acquired a 7S14 dual channel sampler plug-in advertised as needing repair. I have sort of verified the symptoms listed by the seller (noisy Ch 1 and Ch 2, no trigger on Ch 1, dead mercury cells). I've created an album with some pictures (three pictures of the plug-in running in my 7633, and one of the sampling boards with the cover off) here https://groups.io/g/TekScopes/album?id=267375

I'm hoping that the lack of trigger is going to be due to a dirty switch (optimistic, I know, but I don't see anything else in the schematic that could cause Ch 1 not to trigger, but would let you trigger on an external signal. Also, "optimistic" is a relative term: getting to the switch is going to be a chore), and I'm planning to use Ed Breya's solution to replace the mercury cells, but that leaves the noisy inputs.

I finally worked up the courage to open the sampler box (would it have killed Tek to use retained screws?), just to have a look at how much work it would be to replace the mercury cells, but while I had it open I figured I would check the sampler diodes. My results are, I think, not encouraging. I just checked the forward and reverse resistance on the diodes using a multi-meter. Here are the results

Component Forward R Reverse R
Ch 1 CR1A 1.35 MΩ 27 MΩ
Ch 1 CR1B 1.57 MΩ 24 MΩ
Ch 2 CR1A shorted open
Ch 3 CR1B 3.9 MΩ 1.64 MΩ

only Ch 2 CR1A looks like a diode to me, and an ideal one at that (which is not what I expect: I expect a small forward resistance, and a very large reverse resistance). Ch 2 CR1B looks like it's been destroyed. Ch 1 looks bad, but at least it's got a 1:10 difference forward:reverse, which is at least marginally diode-like behavior. Maybe that's even the correct behavior, since I know that there is enough reverse leakage that Tek designed a circuit to cancel it out?

I tried to find specs for the the dual diode (152-0572-00) in the CDPC, but came up empty. I do see the dual diode listed in the RPR, but that document doesn't list specs (or not where I looked, at least). There are lots of dual diodes in current production, many listed as fast switching diodes, all in surface mount packages (which is fine, considering that the diodes Tek used are also surface mount devices, or at least SMD-like), but I can't tell if they would work without knowing the specs I'm trying to match. I assume that the Trr will need to be very fast, and that the junction capacitance will need to be correspondingly tiny.

All it says in the RPR is "SEMICOND DVC, DI" and "BRIDGE, SAMPLING GATE" and that it cost $75 in the mid-90s.

So, my questions are:

1. am I drawing the correct conclusions from my measurements?
2. what are the specs for the dual diode (152-0572-00) supposed to be?
3. what would be a suitable replacement, and where can I get one?

All help is greatly appreciated.

-- Jeff Dutky


Tom Lee
 

Jeff,

Those are not junction diodes, so trr is not a relevant parameter. You'll be using Schottkys. Focus on capacitance, reverse leakage and breakdown voltage.  Matching is helpful as well (both static and dynamic). The good news is that 1GHz is not particularly ambitious, so you have many options. HSMS-280x should work fine, for example. They're obsolete, but still readily available at low cost from many sources. There are faster, lower capacitance diodes out there, but they are also more delicate. For 1GHz samplers, the 280x series are my standard go-to choices.

You're going to have fun bringing that back to life!

-- Tom

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
http://www-smirc.stanford.edu

On 8/23/2021 02:40, Jeff Dutky wrote:
I recently acquired a 7S14 dual channel sampler plug-in advertised as needing repair. I have sort of verified the symptoms listed by the seller (noisy Ch 1 and Ch 2, no trigger on Ch 1, dead mercury cells). I've created an album with some pictures (three pictures of the plug-in running in my 7633, and one of the sampling boards with the cover off) here https://groups.io/g/TekScopes/album?id=267375

I'm hoping that the lack of trigger is going to be due to a dirty switch (optimistic, I know, but I don't see anything else in the schematic that could cause Ch 1 not to trigger, but would let you trigger on an external signal. Also, "optimistic" is a relative term: getting to the switch is going to be a chore), and I'm planning to use Ed Breya's solution to replace the mercury cells, but that leaves the noisy inputs.

I finally worked up the courage to open the sampler box (would it have killed Tek to use retained screws?), just to have a look at how much work it would be to replace the mercury cells, but while I had it open I figured I would check the sampler diodes. My results are, I think, not encouraging. I just checked the forward and reverse resistance on the diodes using a multi-meter. Here are the results

Component Forward R Reverse R
Ch 1 CR1A 1.35 MΩ 27 MΩ
Ch 1 CR1B 1.57 MΩ 24 MΩ
Ch 2 CR1A shorted open
Ch 3 CR1B 3.9 MΩ 1.64 MΩ

only Ch 2 CR1A looks like a diode to me, and an ideal one at that (which is not what I expect: I expect a small forward resistance, and a very large reverse resistance). Ch 2 CR1B looks like it's been destroyed. Ch 1 looks bad, but at least it's got a 1:10 difference forward:reverse, which is at least marginally diode-like behavior. Maybe that's even the correct behavior, since I know that there is enough reverse leakage that Tek designed a circuit to cancel it out?

I tried to find specs for the the dual diode (152-0572-00) in the CDPC, but came up empty. I do see the dual diode listed in the RPR, but that document doesn't list specs (or not where I looked, at least). There are lots of dual diodes in current production, many listed as fast switching diodes, all in surface mount packages (which is fine, considering that the diodes Tek used are also surface mount devices, or at least SMD-like), but I can't tell if they would work without knowing the specs I'm trying to match. I assume that the Trr will need to be very fast, and that the junction capacitance will need to be correspondingly tiny.

All it says in the RPR is "SEMICOND DVC, DI" and "BRIDGE, SAMPLING GATE" and that it cost $75 in the mid-90s.

So, my questions are:

1. am I drawing the correct conclusions from my measurements?
2. what are the specs for the dual diode (152-0572-00) supposed to be?
3. what would be a suitable replacement, and where can I get one?

All help is greatly appreciated.

-- Jeff Dutky




 

Tom,

I'm already having fun, both because I'm getting a small taste (despite the fact that the instrument isn't working) of what it's like to use a sampling scope (quite different from my experiences so far), and because it's giving me an opportunity to use the storage features of my 7633, which is also new.

Any idea what parameters I'm aiming for? I'm afraid that I'm not familiar with the spice model parameters, which is how the HSMS-2800 data sheet appears to want to impart information to me. I see that the capacitance is 1.6 pF, and that the breakdown voltage is 75 V, but I don't know which parameter is reverse leakage.

-- Jeff Dutky


Tom Lee
 

Hi Jeff,

Sorry for the typo—I meant HSMS-282x. That’s the series that hits the sweet spot for 1GHz samplers. The 15V breakdown makes it harder to kill than some lower-capacitance alternatives.

As for leakage, the spice model won’t help you much. Just look at the data sheet for small clues. A bigger challenge is to discover what leakage that sampler can tolerate. In the end you might be forced to cross your fingers. I suspect that the 282x diodes will be fine.

Cheers
Tom



Sent from my iThing, so please forgive brevity and typos

On Aug 23, 2021, at 04:00, Jeff Dutky <jeff.dutky@gmail.com> wrote:

Tom,

I'm already having fun, both because I'm getting a small taste (despite the fact that the instrument isn't working) of what it's like to use a sampling scope (quite different from my experiences so far), and because it's giving me an opportunity to use the storage features of my 7633, which is also new.

Any idea what parameters I'm aiming for? I'm afraid that I'm not familiar with the spice model parameters, which is how the HSMS-2800 data sheet appears to want to impart information to me. I see that the capacitance is 1.6 pF, and that the breakdown voltage is 75 V, but I don't know which parameter is reverse leakage.

-- Jeff Dutky





 

Tom,

As always, I’m uncertain about how to read the data sheets, and how to make inferences from the (nearly random) specs provided therein. Most of that hesitance is just me being insecure about my limited understanding of the engineering and science.

I’ve found the HSMS-2323 and HSMS-232C on eBay for what might be reasonable prices, assuming that the parts are even close to what they claim to be, so I’m going to order some and see what I get. I can’t very well make the instrument LESS functional than it currently is, and I don’t think that there’s anything else in the sampler board that is nearly as unobtainable as the dual diodes that I’m replacing (the other bits of unobtainium are in the trigger circuit on the other side of the instrument), so I feel as if I can dig a much deeper hole than I’m already in.

Do my measurements bear out the diagnosis that both sampling bridges are dead, or might the Ch 1 diodes still be good?

— Jeff Dutky


 

Hmmm I've not heard of HSMS232x - did you mean HSMS282x?

I have quite a few HSMS282R (uncommited ring quad) which I used in my SG504 levelling head replacement, and it might be possible to press those service with a bit of creativity. However I suspect the best match for a sampler would be the HSMS282P uncommitted bridge quad.

If you need one (after checking the datasheet to see if they would work for you), please contact me off list.

Cheers
David

-----Original Message-----
From: TekScopes@groups.io <TekScopes@groups.io> On Behalf Of Jeff Dutky
Sent: 23 August 2021 15:50
To: TekScopes@groups.io
Subject: Re: [TekScopes] Repair of a 7S14

Tom,

As always, I’m uncertain about how to read the data sheets, and how to make inferences from the (nearly random) specs provided therein. Most of that hesitance is just me being insecure about my limited understanding of the engineering and science.

I’ve found the HSMS-2323 and HSMS-232C on eBay for what might be reasonable prices, assuming that the parts are even close to what they claim to be, so I’m going to order some and see what I get. I can’t very well make the instrument LESS functional than it currently is, and I don’t think that there’s anything else in the sampler board that is nearly as unobtainable as the dual diodes that I’m replacing (the other bits of unobtainium are in the trigger circuit on the other side of the instrument), so I feel as if I can dig a much deeper hole than I’m already in.

Do my measurements bear out the diagnosis that both sampling bridges are dead, or might the Ch 1 diodes still be good?

— Jeff Dutky


 

David,

Yes, 282, not 232, that was my repeated typo. I have ordered the HSMS-2822 and HSMS-282C I found on eBay (total cost is only $12, so I'm not out much if they turn out to be fakes, but who would be making fake Schottky diodes at these prices?).

It looks like I will need to install the HSMS-2822/282C upside down, if I've translated the schematic correctly onto the board layout, but that's hardly a serious problem. Also, would it have killed Tek to put a Schottky diode symbol in the schematic for CR1A and CR1B?

I take it, from everyone's silence on the question, that my measurements do, indeed, indicate that both sampling bridges are dead. I tried measuring a 1N5817 and a 1N5819 but I'm still not sure I understand what I'm seeing. With my multimeter on the 1 MΩ range I see a forward resistance of 38 kΩ and a reverse resistance of 100 kΩ. That's not what I expect from a diode (I expect the resistance in reverse to be at least an order of magnitude higher), but I don't actually know what to expect from a Schottky diode. It certainly looks a lot like what I measured for the Ch 1 diode bridge. Does that suggest that the Ch 1 diode bridge may actually be okay?

In any case I will find out shortly.

-- Jeff Dutky


Tom Lee
 

Hi Jeff,

Yes, manufacturers do seem to work overtime to encrypt the information on datasheets. I think they hire steganographers to write them.

In samplers, the diodes are supposed to act as switches, so what matters most are off state impedance (dominated by capacitance at high frequencies), and on-state conductance (largely a function of forward current). At 1GHz, a 1pF capacitance has a reactance of about 160 ohms. For the 282x, the capacitance at a couple of volts of reverse bias is below 1pF. In the forward direction, the on-state conductance will typically be smaller than this reactance by more than an order of magnitude. An off/on impedance ratio exceeding 10:1 (and typically approaching 20:1 here) is conventionally the threshold for "good enough".

There are other specs to watch out for, such as package parasitics, maximum ratings, and the like. The SOT-23 package is plenty good enough at 1GHz. Just don't add lengthy runs of wire hooking up to them and you'll be fine.

You should feel encouraged by the tekwiki entry on sampling diodes (https://w140.com/tekwiki/wiki/Sampling_diodes), where 282x diodes have been used successfully to repair 1GHz samplers.

-- Cheers,
Tom


--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
http://www-smirc.stanford.edu

On 8/23/2021 07:50, Jeff Dutky wrote:
Tom,

As always, I’m uncertain about how to read the data sheets, and how to make inferences from the (nearly random) specs provided therein. Most of that hesitance is just me being insecure about my limited understanding of the engineering and science.

I’ve found the HSMS-2323 and HSMS-232C on eBay for what might be reasonable prices, assuming that the parts are even close to what they claim to be, so I’m going to order some and see what I get. I can’t very well make the instrument LESS functional than it currently is, and I don’t think that there’s anything else in the sampler board that is nearly as unobtainable as the dual diodes that I’m replacing (the other bits of unobtainium are in the trigger circuit on the other side of the instrument), so I feel as if I can dig a much deeper hole than I’m already in.

Do my measurements bear out the diagnosis that both sampling bridges are dead, or might the Ch 1 diodes still be good?

— Jeff Dutky




Tom Lee
 

When using a DMM to test diodes, you must use the special "diode test" mode (often indicated by a diode symbol). Ordinary resistance mode does not apply sufficient forward bias to provoke diode behavior. Your measurements seem to have been made in ordinary resistance mode, so you'll have to re-do the measurements.

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
http://www-smirc.stanford.edu

On 8/23/2021 13:53, Jeff Dutky wrote:
David,

Yes, 282, not 232, that was my repeated typo. I have ordered the HSMS-2822 and HSMS-282C I found on eBay (total cost is only $12, so I'm not out much if they turn out to be fakes, but who would be making fake Schottky diodes at these prices?).

It looks like I will need to install the HSMS-2822/282C upside down, if I've translated the schematic correctly onto the board layout, but that's hardly a serious problem. Also, would it have killed Tek to put a Schottky diode symbol in the schematic for CR1A and CR1B?

I take it, from everyone's silence on the question, that my measurements do, indeed, indicate that both sampling bridges are dead. I tried measuring a 1N5817 and a 1N5819 but I'm still not sure I understand what I'm seeing. With my multimeter on the 1 MΩ range I see a forward resistance of 38 kΩ and a reverse resistance of 100 kΩ. That's not what I expect from a diode (I expect the resistance in reverse to be at least an order of magnitude higher), but I don't actually know what to expect from a Schottky diode. It certainly looks a lot like what I measured for the Ch 1 diode bridge. Does that suggest that the Ch 1 diode bridge may actually be okay?

In any case I will find out shortly.

-- Jeff Dutky




Dave Daniel
 

Bob Pease wrote an article entitled “How to Read a Datasheet”. I think it was published in 1989 in one of National Semiconductor’s application note compendiums. It is also included as an appendix (F, I think) in Bob’s book “Troubleshooting Analog Circuits” (well worth reading!). The article is an insightful review of how to understand that which a datasheet tries to communicate.

DaveD

On Aug 23, 2021, at 17:00, Tom Lee <tomlee@ee.stanford.edu> wrote:

Hi Jeff,

Yes, manufacturers do seem to work overtime to encrypt the information on datasheets. I think they hire steganographers to write them.

In samplers, the diodes are supposed to act as switches, so what matters most are off state impedance (dominated by capacitance at high frequencies), and on-state conductance (largely a function of forward current). At 1GHz, a 1pF capacitance has a reactance of about 160 ohms. For the 282x, the capacitance at a couple of volts of reverse bias is below 1pF. In the forward direction, the on-state conductance will typically be smaller than this reactance by more than an order of magnitude. An off/on impedance ratio exceeding 10:1 (and typically approaching 20:1 here) is conventionally the threshold for "good enough".

There are other specs to watch out for, such as package parasitics, maximum ratings, and the like. The SOT-23 package is plenty good enough at 1GHz. Just don't add lengthy runs of wire hooking up to them and you'll be fine.

You should feel encouraged by the tekwiki entry on sampling diodes (https://w140.com/tekwiki/wiki/Sampling_diodes), where 282x diodes have been used successfully to repair 1GHz samplers.

-- Cheers,
Tom


--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
http://www-smirc.stanford.edu

On 8/23/2021 07:50, Jeff Dutky wrote:
Tom,

As always, I’m uncertain about how to read the data sheets, and how to make inferences from the (nearly random) specs provided therein. Most of that hesitance is just me being insecure about my limited understanding of the engineering and science.

I’ve found the HSMS-2323 and HSMS-232C on eBay for what might be reasonable prices, assuming that the parts are even close to what they claim to be, so I’m going to order some and see what I get. I can’t very well make the instrument LESS functional than it currently is, and I don’t think that there’s anything else in the sampler board that is nearly as unobtainable as the dual diodes that I’m replacing (the other bits of unobtainium are in the trigger circuit on the other side of the instrument), so I feel as if I can dig a much deeper hole than I’m already in.

Do my measurements bear out the diagnosis that both sampling bridges are dead, or might the Ch 1 diodes still be good?

— Jeff Dutky








 

Tom,

I re-did the measurements using the diode test function of my Tek multimeter (I took a little detour investigating what voltages were used by different meters on different modes and ranges, and the results of that were also fascinating: The Tek meter uses a lot higher voltages for all modes and ranges than my UNI-T meter does) and here are the results:

Component Forward V Backward V
Ch 1 CR1A 0.42 V 0.L
Ch 1 CR1B 2.1 V 2.1 V
Ch 2 CR1A 0.45 V 0.L
Ch 2 CR1B 0.46 V 0.L

These results seem to make a lot more sense, but this seems to say that the diode bridge on Ch 2 is good, and that only half of the diode bridge (CR1B) on Ch 1 is damaged. I'm not sure how to reconcile this with the strange behavior of Ch 2 (see the photo album: Ch 2 is showing what looks like three traces with at least 8 divisions of noise, while Ch 1 shows one trace with well around one minor division of noise), but I'm also expecting that there are other faults that I have not tried to investigate yet.

Among those other failures are, obviously, the mercury cells. With the mercury cells dead shouldn't the diode bridge basically not work at all? Or am I misunderstanding what the bias voltage provided by the mercury cells is supposed to do? Scratch that, I'm certain that I don't understand how the sampling diode bridge works.

I'm still working on building the LED replacement for the mercury cells, though I've considered just buying some zinc-air batteries instead. I'm not sure how I would mount them: I don't have a spot welder, though I see that they can be had very inexpensively, but I'm also kind of scared a $50 spot welder. Until I have replaced the mercury cells I don't really expect the sampling to work at all.

-- Jeff Dutky


 

On Tue, Aug 24, 2021 at 01:18 AM, Jeff Dutky wrote:


I'm still working on building the LED replacement for the mercury cells,
though I've considered just buying some zinc-air batteries instead. I'm not
sure how I would mount them: I don't have a spot welder, though I see that
they can be had very inexpensively, but I'm also kind of scared a $50 spot
welder. Until I have replaced the mercury cells I don't really expect the
sampling to work at all.
Hi Jeff,
Simple primary (i.e. non-rechargeable) hearing-aid zinc-air batteries are useless in this application: Once activated, they will discharge and be dead in a relatively short while (months at most I guess), whether they have to supply any current or not.

Around the same time that Ed Breya published his article on using a photo-voltaic bias solution on TekWiki, I modified my first 7S14 using head-to-head glued LEDs, using cyanoacrylate glue. So, basically, same as Ed's solution.
My modified 7S14 samplers are behaving quite well in three 7S14's. Obviously, hardly any current is drawn. The bias voltage is not at all as critical as was thought earlier. Just make sure that about 1.3V - 1.5V is generated in-circuit by the receiving LED and that each pair (the two in each sampler channel) matches within less than 0.1V.

Raymond


Tom Lee
 

Hi Jeff,

Those values make much more sense. CR1B is open. The other three devices are acting like diodes. That said, it is advisable to replace these diodes in pairs (e.g., with a 2822 or 2825).

Without proper bias, it's hard to say whether what you are seeing points to any problems besides a dead diode.  The dead batteries will act as some random impedance, so noisy output is perhaps not a surprise.  Certainly, check what you can conveniently get to, but I wouldn't expend too much additional troubleshooting effort until after you have healthy diodes and get the battery substitution circuit up and running. Then you'll have a much better idea of where you stand.

Cheers
Tom


--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
http://www-smirc.stanford.edu

On 8/23/2021 16:18, Jeff Dutky wrote:
Tom,

I re-did the measurements using the diode test function of my Tek multimeter (I took a little detour investigating what voltages were used by different meters on different modes and ranges, and the results of that were also fascinating: The Tek meter uses a lot higher voltages for all modes and ranges than my UNI-T meter does) and here are the results:

Component Forward V Backward V
Ch 1 CR1A 0.42 V 0.L
Ch 1 CR1B 2.1 V 2.1 V
Ch 2 CR1A 0.45 V 0.L
Ch 2 CR1B 0.46 V 0.L

These results seem to make a lot more sense, but this seems to say that the diode bridge on Ch 2 is good, and that only half of the diode bridge (CR1B) on Ch 1 is damaged. I'm not sure how to reconcile this with the strange behavior of Ch 2 (see the photo album: Ch 2 is showing what looks like three traces with at least 8 divisions of noise, while Ch 1 shows one trace with well around one minor division of noise), but I'm also expecting that there are other faults that I have not tried to investigate yet.

Among those other failures are, obviously, the mercury cells. With the mercury cells dead shouldn't the diode bridge basically not work at all? Or am I misunderstanding what the bias voltage provided by the mercury cells is supposed to do? Scratch that, I'm certain that I don't understand how the sampling diode bridge works.

I'm still working on building the LED replacement for the mercury cells, though I've considered just buying some zinc-air batteries instead. I'm not sure how I would mount them: I don't have a spot welder, though I see that they can be had very inexpensively, but I'm also kind of scared a $50 spot welder. Until I have replaced the mercury cells I don't really expect the sampling to work at all.

-- Jeff Dutky




 

Raymond,

Cool! So, to be clear, each mercury cell is replaced by a single pair of LEDs, one LED illuminating the other which generates the bias voltage.

Where did you tap the power to drive the illuminating LED from? Ed's solution seemed to require an external power supply, as well as modification to the PCB, both of which I would like to avoid. I was going to reduce the size of the assembly by using 3mm diodes and filing off most of the lens, allowing me to glue the two dice nose-to-nose, but I wasn't clear on why he used an external power supply, rather than tap a power rail in the plug-in. I see that there is a +15 V rail on the sampler board, and my plan was to tap into that.

Ed's solution also seemed to have a voltage divider network for some reason, even though my impression was that the voltage coming out of the red LED was very close to the 1.35 V produced by the mercury cells. I was planning to dispense with the voltage divider in my construction, and just have filter caps on both LEDs, and a single current limiting resistor on the white, illuminating LED.

I'm not very far along with this part of the repair: I haven't even calculated what the value of the current limiting resistor should be; I'm still experimenting with filing down the LED lenses.

-- Jeff Dutky


 

On Tue, Aug 24, 2021 at 02:11 AM, Jeff Dutky wrote:


Cool! So, to be clear, each mercury cell is replaced by a single pair of LEDs,
one LED illuminating the other which generates the bias voltage.
That's correct.

You can find a picture of my solution here:
https://w140.com/tekwiki/wiki/7S14#/media/File:Modified_sampler_with_head-to-head_leds.jpeg
The black shrink-tube material covers a set of two head-to-head LEDs.
You can see two cylindrical tantalum caps of 1.5uF vertically, one across each receiving LED.
I do use the 15VDC supply on the sampler, like you plan.
I select the LED sets + resistors on a separate 15VDC power supply and mount them head-to-head *without* filing down. I just arrange them by hand and put a drop of cyanoacrylate glue between them.
The resistance value is established empirically. Somewhere from 2k2 up works. Some white LEDs work best as generators.
It's all much less sophisticated than you'd think and it works fine: I verify using sine waves (up to above 1 GHz) and fast edges: both around 40 ps and 200 ps rise times and flat tops plus rise and fall times of about 500 ps. No problem getting consistent results for and between each of these.

Raymond


Tom Lee
 

Hi Raymond,

Just out of curiosity, did you use clear-packaged LEDs or ones with diffuse lenses? When making homebrew optoisolators, I've always used clear ones, for higher CTR. Just wondering if that matters at all here.

-- Cheers,
Tom

--

Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
http://www-smirc.stanford.edu

On 8/23/2021 17:31, Raymond Domp Frank wrote:
On Tue, Aug 24, 2021 at 02:11 AM, Jeff Dutky wrote:

Cool! So, to be clear, each mercury cell is replaced by a single pair of LEDs,
one LED illuminating the other which generates the bias voltage.
That's correct.

You can find a picture of my solution here:
https://w140.com/tekwiki/wiki/7S14#/media/File:Modified_sampler_with_head-to-head_leds.jpeg
The black shrink-tube material covers a set of two head-to-head LEDs.
You can see two cylindrical tantalum caps of 1.5uF vertically, one across each receiving LED.
I do use the 15VDC supply on the sampler, like you plan.
I select the LED sets + resistors on a separate 15VDC power supply and mount them head-to-head *without* filing down. I just arrange them by hand and put a drop of cyanoacrylate glue between them.
The resistance value is established empirically. Somewhere from 2k2 up works. Some white LEDs work best as generators.
It's all much less sophisticated than you'd think and it works fine: I verify using sine waves (up to above 1 GHz) and fast edges: both around 40 ps and 200 ps rise times and flat tops plus rise and fall times of about 500 ps. No problem getting consistent results for and between each of these.

Raymond




 

On Tue, Aug 24, 2021 at 02:43 AM, Tom Lee wrote:


Just out of curiosity, did you use clear-packaged LEDs or ones with diffuse
lenses? When making homebrew optoisolators, I've always used clear ones, for
higher CTR. Just wondering if that matters at all here.
Hi Tom,
I never tried diffusing lenses. While adjusting and measuring output voltage with the clear-packaged LEDs, I can clearly (!) see the transmitter's spot illuminating the receiver's chip.
So, unfortunately, I can't tell you whether diffusing lenses perform worse. I chose clear-packaged LEDs thinking that the directivity of the emitted light results in a greater fraction of the light hitting the receiver chip.

Raymond


Tom Lee
 

Thanks for the fast reply, Raymond. That's what I found as well. I'll stick with the clear ones.

Cheers
Tom

--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
http://www-smirc.stanford.edu

On 8/23/2021 17:52, Raymond Domp Frank wrote:
On Tue, Aug 24, 2021 at 02:43 AM, Tom Lee wrote:

Just out of curiosity, did you use clear-packaged LEDs or ones with diffuse
lenses? When making homebrew optoisolators, I've always used clear ones, for
higher CTR. Just wondering if that matters at all here.
Hi Tom,
I never tried diffusing lenses. While adjusting and measuring output voltage with the clear-packaged LEDs, I can clearly (!) see the transmitter's spot illuminating the receiver's chip.
So, unfortunately, I can't tell you whether diffusing lenses perform worse. I chose clear-packaged LEDs thinking that the directivity of the emitted light results in a greater fraction of the light hitting the receiver chip.

Raymond




 

Raymond,

I've done some experiments with different color LEDs for both the illuminator and the driver, and found that modern white LEDs don't work at all as the driver, but are excellent illuminators. For the drivers I am going to use modern red LEDs, which produce almost exactly the target voltage of 1.35 V in my trials.

There will be something like a 12 V drop across the resistor, and the white diodes I plan to use have a 20 mA max current, so the resistor should be no less than 600 ohms, but probably more like 1k if I don't want to risk burning out the illuminating LED. Even at half max current they will still be mighty bright.

I may not bother to file down the LEDs in the final build, but I have a theory that the close proximity of the file surfaces will make alignment of the dies less important. I'm comparing the voltage output of the filed vs. unmodified LED pairs, and will go with whichever arrangement gives me the higher output voltage. And now that I've said that, I realize that I need to compare filed and un-filed versions of the same LEDs, so test before filing them down, and test again after, otherwise variation between LEDs will make my results meaningless.

-- Jeff Dutky


 

On Tue, Aug 24, 2021 at 03:03 AM, Jeff Dutky wrote:

Hi Jeff,

I've done some experiments with different color LEDs for both the illuminator
and the driver, and found that modern white LEDs don't work at all as the
driver, but are excellent illuminators. For the drivers I am going to use
modern red LEDs, which produce almost exactly the target voltage of 1.35 V in
my trials.
That matches what I remember: use white LEDs as emitters.

I've never needed to go near maximum current.

I align the opposed lenses in a rig, monitoring voltage output. If anything, the drop of glue seems to even slightly increase optical coupling. I've never filed the lenses after the very first set I made, still with ample voltage output.
Of course, it never harms to play with things, just to get some feel for contributing factors. I'd be interested in hearing your results with the different setups.

Signing off for the night: It's 3:19AM (GMT+1 + DST).

Raymond