Advice on determining a logic analyzer for use with 1980s synthesizers. TLA715?

[sku_u]

I am looking for a logic analyzer that will be used for poking around inside vintage synthesizers and digital delays primarily from the 1980s. These devices are chock full of CMOS and TTL logic and contain processors such as the 6809, 8049, 8051, and Z80 derivatives. The delays use discrete logic and 4164 DRAM to playback a delayed audio signal. Chips like the 8255 and 8253 can be found inside of these devices as well. I want a logic analyzer to help troubleshoot faults and also to learn more about how these devices work.

I am aware of the maintenance that older equipment requires so I have no problem with purchasing a logic analyzer that is 2 or 3 decades old. I am also familiar with HP-UX from HP's PA-RISC workstations and come from a background of vintage computing so I am well aware of what I am getting myself into. I am wanting a standalone device so an USB LA is off of the table for me.


I missed out on an HP 16500C earlier this year which has accelerated my desire of obtaining an LA. I have been researching other options but I still am not sure what would be the most suitable device for me.

A Tektronix TLA715 is available locally and I am considering this unit but I basically know nothing when it comes to the probes and adapters I need. It has two of the TLA7N4 (2GHz timing, 64k) modules installed and comes with several of the P6434 probes. Of course these probes are the ones with the Mictor connector and are of no use to me unless I buy or make a Mictor to 2.54mm adapter. I need help determining which probes would be most suitable for my application and I have allocated several hundred dollars for obtaining these probes. According to the pdf, "Tektronix P6400 P6800 P6900 Logic Analyzer Probes" from https://w140.com/tekwiki/wiki/P6434, it looks like most of the probes are compatible with TTL & CMOS levels. I need some sort of flying lead that can be attached to the pins of chips. I have made a little DIP-14/16 breakout PCB with 2.54mm pin headers that I can attach a probe to as well. The P6417 and P6418 appear to be something I can use but I would like to confirm.

I am also making this post to see if there is a better LA for my use case. I have come across other units like the HP 16702B but I am not sure what features and specifications make one logic analyzer better or more preferable for my intended use than another. What about the 16500C? Are there any other analyzers that would be better than the units I've mentioned? Is the TLA715 even a good fit for my needs? I apologize for the very noobie questions but I figured this would be the best place to ask. I greatly appreciate any and all advice and help.

[tggzzz]

When debugging/repairing equipment of that age:

• measure Vcc voltage and ripple, using voltmeter and scope
• use scope to verify signal integrity. 100MHz is sufficient, but be aware of the ringing caused by a 6" ground lead
• thereafter use a logic analyser

The LA width is the key spec. Ideally that should be 32 bits: 16 address, 8 data, Clk,CS/RW/etc.

Speed isn't likely to be an issue. Even an early 80s LA is fast enough.

Capture depth is more likely to be an issue. Early 80s LAs were just sufficient, with skill and imagination.

Arm/trigger/filter facilities are vital. Modern cheap USB LAs are hopeless for that.

The older HP LAs can be very cheap, but with all LAs be aware of the price of missing probe clips. Replacements might cost more than the LA!

Summary: almost any working old "wide" LA will be sufficient. I'd look for any LA >HP1630 series with state machines >=25MHz.

[artag]

I like the slightly later 16700, which seems to have all the best parts of the 16500 with some shortcomings fixed but without the high cost of the latest models.

I started writing a more detailed answer but better descriptions have been written here in the past by more knowledgable people than me. Hunt them down ! Tggzzz is one of them

There are lots of options in terms of digital and analogue capture. Get the deepest memory you can afford but be aware you'll sometimes limit the capture to keep the display time down. On these parallel bus processors you'll be using it in state mode though, which is far more efficient for memory use. There is an analogue scope option. It's often pricy and is very poor compared with modern scopes, but having it in the same box with time-aligned displays and cross triggering is very powerful on a mixed-signal device like a synth.

I know nothing aboiut the Tek analysers so I can't compare them. Broadly though, from what I've seen on ebay, you'll pay less for a well-configured instrument but struggle to find upgrades later. I suspect the UI, like scopes, is a matter of personal opinion and depends what you grew up with.

Some people describe the 16500/700 UI as complicated and hard to understand. I don't think this is true, but if, like me, you started with a 1630 and updated over time it will all be very familiar.

If you have no prior experience of logic analysers I'd actually recommend doing the same thing. A 1630 (1630D for mixed signal, 1630G for software development) is very cheap and is explained in terms that makes it very comprehensible to a hardware engineer. Later models lacked that straightforward explanation and wrote the manuals in terms of usage examples. IMHO this denies you the basic understanding of what's going on that proper understanding requires. Once you start to suffer from it's limitations, replace it with a more modern one. But your synths were designed by people using that class of analyser - upgrading to a '90s machine is the lap of luxury, adding small probes, deep memories, bigger  / better displays.

[xtech]

Signature analyzer would be much more useful than ordinary logic analyzer. I have few logic analyzer including TLA715. I'm always doing everything not to start that monster but rather use handy DSlogic or Salaae Logic 16 (clone) - rarely you need to monitor all the buses in the serviced equipment. Buy good logic probe and logic pulser. Old HP current tracer and logic clip are also not much expensive but very useful. Let TLA715 rest

[tggzzz]

Signature analyzer would be much more useful than ordinary logic analyzer. I have few logic analyzer including TLA715. I'm always doing everything not to start that monster but rather use handy DSlogic or Salaae Logic 16 (clone) - rarely you need to monitor all the buses in the serviced equipment. Buy good logic probe and logic pulser. Old HP current tracer and logic clip are also not much expensive but very useful. Let TLA715 rest

A signature analyser is a very different beast, only useful for a pass/fail tests. It takes a single bit input stream, and uses an LFSR to generate a hash. If the hash is the same as the predefined hash, the module is working.

The OP might like to consider whether a simple protocol analyser would help. A BusPirate5 is a starting point.

I started writing a more detailed answer but better descriptions have been written here in the past by more knowledgable people than me. Hunt them down ! Tggzzz is one of them

Thamks, but don't regard me as an oracle!

[MarkL]

The Agilent 16702B is a modular LA with 5 slots and lots of parallel capacity (typically 68 channels per slot) and captures in state or timing mode.  You can also get scope cards with limited points (32k) that allows correlated capture with the digital inputs which is useful for checking signal integrity.  It runs HP-UX, which you might find comfortable.

However it's BIG and bulky (and loud, if you care about that).  And the capture cards have a propensity for going bad due to corrosion issues, so you need to purchase with a right to return.

There's no lack of used accessories for it on ebay (grabbers, flying lead pods, cables, etc.), but it's always better to find one that's already outfitted since post-purchase of these can add up.  Capture cards using the "40-pin probing system" are available with anywhere from 512k to 64M points, amd hacks have been discovered to configure a card to its maximum capabilities.  TTL, CMOS, ECL, and custom -6V to +6V threshold levels are supported.

Z80 and other processor decoders are available.  Support for modern serial protocols, such as SPI and I2C, is non-existent, although of course it will show you the waveform.  Included in the OS is development software, so you can write your own decoders and disassemblers, if so inclined.

There's also the 16700B which is the same as the 16702B, but uses an external monitor which might allow you save on bench space.  And in either case, get option 003 which maxes out the display memory (difficult to add post-purchase).

If you want something a little smaller, the 1670G and friends is a good compromise, but doesn't provide an accessible OS like the 16702B.  It uses the same 40-pin probing system, as does many other from HP/Agilent units from this era.

The 1630 series was mentioned above.  I had a 1631D for years, and the triggering was very sophisticated and easy to use.  But I found its capture memory of only 1024 points to be extremely limiting.

[TomKatt]

What frequency do these vintage instruments run at?  I’d guess many are Z80 or similar running around 4MHz or so.

I have zero experience working on equipment like this at the logic level, but if I had to make a blind start I’d probably consider a DSlogic 32 - at $400 that must be the price you’d pay (or less) for capable vintage equipment outfitted with all the probes and accessories you’d need for them.  And you’d eliminate concerns about faulty hardware or corroded connections.  Sampling at 10X frequency of 50MHz or less, you could stream nearly 13 minutes of data over usb to a pc for examination of a full address / data bus.  I haven’t used DSview for parallel logic, but it seems like it’s capable of triggering on parallel logic so you could monitor for specific addresses or data.  I can’t imagine vintage equipment has that deep memory.

That said, I love vintage equipment both from its build quality and value.  Though I have a few modern DSO’s, I still use my old HP180A all the time for analog signals.  Plus it’s fun.

Edit - I’m also frequently impressed watching people work on gear like this with almost simple test gear such as a basic logic probe…. And I wish I could afford the HP 547A current tracer mentioned earlier - its amazing that something like that isn’t available today at some affordable price.  There must be a reason those things go for $500+ in good condition…

[nctnico]

A Tektronix TLA715 is available locally and I am considering this unit but I basically know nothing when it comes to the probes and adapters I need. It has two of the TLA7N4 (2GHz timing, 64k) modules installed and comes with several of the P6434 probes. Of course these probes are the ones with the Mictor connector and are of no use to me unless I buy or make a Mictor to 2.54mm adapter. I need help determining which probes would be most suitable for my application and I have allocated several hundred dollars for obtaining these probes. According to the pdf, "Tektronix P6400 P6800 P6900 Logic Analyzer Probes" from https://w140.com/tekwiki/wiki/P6434, it looks like most of the probes are compatible with TTL & CMOS levels. I need some sort of flying lead that can be attached to the pins of chips. I have made a little DIP-14/16 breakout PCB with 2.54mm pin headers that I can attach a probe to as well. The P6417 and P6418 appear to be something I can use but I would like to confirm.

I have a TLA715 myself and so far I've been very happy with it (its my 4th logic analyser). A newer model is the TLA7000 series which can use the same modules.

Having a mictor to flying leads adapter is one way to go. I have made something like that a long time ago.
If you want to hunt for flying lead probes, make sure to get the P6417. These are the older model which has tongue shaped contacts which will grab on thin wires as well.

I see some mentioning getting USB logic analysers but I don't think these are very handy as these don't have the elaborate trigger and storage qualifiers a real logic analyser has. What the TLA715 can do is store only when a signal changes. This means A) you can use read/write strobes, chip selects, etc to get a list with addresses a microprocessor is accessing and thus trace which memory locations are accessed B) you need far less memory to store a long sequence of accesses and still you retain full samplerate time resolution. This is a very powerful feature when dealing with microprocessor circuits like the ones you are describing. Another possibility is to take a snapshot of samples around an event which can help to find what the signals look like before and after an event (for example: see what happens in surrounding logic when a setting in hardware is changed).

The TLA7N4 module can be hacked to support 8Mpts per channel. You have to send a few simple commands over the TekVISA control panel to do this. More info can be found on this forum.

[artag]

And I wish I could afford the HP 547A current tracer mentioned earlier - its amazing that something like that isn’t available today at some affordable price.  There must be a reason those things go for $500+ in good condition…

Vanishingly rare, mostly, rather than because they were excellent and unreproducible. I have one and don't really use it because modern data rates and board densities reduce it's usefulness. But there are modern equivalents. See EEVBLOG #296 and Mikes's Electric Stuff's very creative review of the same product (making a 3d magnetic map of the board).

There was a project on here to build your own using a honeywell magnetic sensor. I bought a sensor but haven't got around to trying it.

Both these are a bit niche as is evidenced by there not being a cheap, widely available equivalent despite being ancient designs.
 They might be useful to sku_u but he's looking for solutions to microprocessor-based instruments with external ram and rom, which often fail because of address decoding or data bit errors. A wide logic analyser is kind of overkill but does actually do a really good job as long as you have a decent probing solution.

For machines with a 40-pin DIP CPU (a lot of them !) I recommend a 40 pin dip clip with the LA probes connected to address, data and control signals - enough to get an instruction trace. Maybe add a few external signals such as chip selects on rom and ram chips. Capture the first few hundred cycles out of reset. This is long enough to do plenty of instruction fetches and, critically, some stack access,

Examine the trace, verifying :
1. What happens at reset ? Typically it should fetch a reset vector, access a ROM and then start executing there. Is it sensible ?
2. Are all the data bits active ? any appear to be shorted together or stuck high / low ?
3. Watch a call occurring (the PC should get pushed onto the stack, maybe with other things)
4. Find the corresponding return. Does the same value get pulled from the stack or has it been corrupted ?

These will show up most of the common problems around a CPU/RAM/ROM core. You might be able to extend it to some I/O and a larger range of address bits but I have used this heavily when training production technicians to debug new builds that pass ATE but don't work. Both 8086 and ARM2/3 machines. These are especially hard because they often have multiple faults - they have NEVER worked. Most repairs are on bits of equipment that worked until they got a fault and then were sent for repair - so they only have one fault. Old equipment thats been rusting for years may have more but they're often in very different areas (logic, power supply, hard discs).

This procedure feeds back into the initial question slightly.
You want an LA that's wide enough for all address, data and control buses and a few extra too. This means 32 bit for 8 bit machines and more for 16 bit machines. Not many 32 bit busses in this era. but Acorn Archimedes is an exception - though the conditional instructions makes it a bugger tom determine execution path other than through the address bus.
Some machines (8086, Z80) have a multiplexed address/data bus. Thius reduces the nuber of pins you need to moinitor but requires a state clocking mechanism that can record both data and address phases. I think this needed an external hardware bus processor for the 8086. 
Huge depth is not necessary (though it's nice). Even with 1k depth you'll usually see a stack call & return, and if necessary you can trigger on ram accesses over a small address range. This is the advantage of the complex state triggers these LAs have.
A disassembler (possibly including special hardware) is also very helpful to make the instruction stream more readable.

[artag]

Another device that has  a bit of a premium price because of rarity and the appreciation of it by, especially, arcade techs, is the Fluke Troubleshooter. This is a simplified in-circuit emulator with personality pods to suit popular processors. The processor is removed and the pod end plugged in its place (it may be possible to clip over a soldered-in device).

The troubleshooter then runs a bunch of canned tests such as ram, rom checksum, address and data  bus etc. by generating bus cycles at the processor that exercise those things without requiring the real roms and rams to work. I think it has an actual processor in the pod, it's not just a digital vector generator, but that processor is using its own rom and ram and then steering the bus access internally or externally to suit. This can give very rapid detection of shorted or open data and address busses.

I'm not sure if there are enough machine of this era needing repairs to justify the development effort, but it could be reproduced using something like a Pico. Maybe worthwhile for fixing pinball machines which have large numbers of different designs having a  common CPU board.

[tggzzz]

There are lots of options in terms of digital and analogue capture. Get the deepest memory you can afford but be aware you'll sometimes limit the capture to keep the display time down. On these parallel bus processors you'll be using it in state mode though, which is far more efficient for memory use.

Memory efficiency is relatively unimportant, especially now memory is so cheap.

What is important is being able to ignore rubbish, so you can concentrate on what is important. That's why state-based arm/trigger/filter is so important: it saves your time. If you don't capture the crap, there's nothing to ignore.

Alternatively: good luck finding the "wrong" bit in 10MS of 32 bit captures.

[nctnico]

There are lots of options in terms of digital and analogue capture. Get the deepest memory you can afford but be aware you'll sometimes limit the capture to keep the display time down. On these parallel bus processors you'll be using it in state mode though, which is far more efficient for memory use.

Memory efficiency is relatively unimportant, especially now memory is so cheap.

What is important is being able to ignore rubbish, so you can concentrate on what is important. That's why state-based arm/trigger/filter is so important: it saves your time. If you don't capture the crap, there's nothing to ignore.

Alternatively: good luck finding the "wrong" bit in 10MS of 32 bit captures.

OTOH: if you don't know what to look for, doing offline analyses on a long capture using advanced search methods (which modern logic analysers do provide) is way more efficient compared to doing lots of captures while trying to kind of narrow down the correct trigger condition. On top of that, a long capture also allows to do other analysis / sanity checks on the data being captured which is simply impossible to do with limited memory depth. Quite often you can find interesting information that way you would have missed otherwise.

[TomKatt]

But if you know what makes it “wrong”, you could always filter out the noise.

If you don’t know what makes it “wrong”, you can’t set up a trigger to begin with.

And sure - I’d take a hardware trigger over not, all other things being equal.

[tggzzz]

But if you know what makes it “wrong”, you could always filter out the noise.

If you don’t know what makes it “wrong”, you can’t set up a trigger to begin with.

And sure - I’d take a hardware trigger over not, all other things being equal.

That is true whenever/wherever the analysis is done.

Blindly capturing everything is suboptimal and is rarely necessary. More often - and more profitably - you have a concept of the problem's cause and consequences, and filter based to observe your hypothesis

It is much more than just a trigger, of course. With appropriate filtering you can capture only the sequence of instructions fetched, or read/write to an address range, or latencies between an interrupt occurring and the first instruction of the ISR being executed, etc, etc, etc. That's made easier with old processors since all that information appears on external pins.

[tggzzz]

There are lots of options in terms of digital and analogue capture. Get the deepest memory you can afford but be aware you'll sometimes limit the capture to keep the display time down. On these parallel bus processors you'll be using it in state mode though, which is far more efficient for memory use.

Memory efficiency is relatively unimportant, especially now memory is so cheap.

What is important is being able to ignore rubbish, so you can concentrate on what is important. That's why state-based arm/trigger/filter is so important: it saves your time. If you don't capture the crap, there's nothing to ignore.

Alternatively: good luck finding the "wrong" bit in 10MS of 32 bit captures.

OTOH: if you don't know what to look for, doing offline analyses on a long capture using advanced search methods (which modern logic analysers do provide) is way more efficient compared to doing lots of captures while trying to kind of narrow down the correct trigger condition. On top of that, a long capture also allows to do other analysis / sanity checks on the data being captured which is simply impossible to do with limited memory depth. Quite often you can find interesting information that way you would have missed otherwise.

The low end modem LAs have very restricted functionality w.r.t. filtering and triggering. Decent modern LAs are fine, as are old LAs from that period, e.g. >HP163x.

[TomKatt]

The low end modem LAs have very restricted functionality w.r.t. filtering and triggering. Decent modern LAs are fine, as are old LAs from that period, e.g. >HP163x.

That would seem to rule out the LA features included in entry level oscilloscopes…. You’re definitely going to need a good dedicated device for this kind of analysis.

[artag]

The low end modem LAs have very restricted functionality w.r.t. filtering and triggering. Decent modern LAs are fine, as are old LAs from that period, e.g. >HP163x.

That would seem to rule out the LA features included in entry level oscilloscopes…. You’re definitely going to need a good dedicated device for this kind of analysis.

That's right. I've only ever seen timing analysers included in oscilloscopes. This is not really an LA, it's a bunch of very low resolution oscilloscope channels for use in systems where 2 or 4 analog channels aren't very useful. That's why they're called 'mixed signal oscilloscopes' rather than 'mixed domain analysers'.

The better ones offer SPI/RS232/I2C protocol analysis via oversampling which helps make sense of serial data but to usefully display an instruction stream you want clocked capture and possibly demultiplexing clocks.

[artag]

Memory efficiency is relatively unimportant, especially now memory is so cheap.

What is important is being able to ignore rubbish, so you can concentrate on what is important. That's why state-based arm/trigger/filter is so important: it saves your time. If you don't capture the crap, there's nothing to ignore.

Alternatively: good luck finding the "wrong" bit in 10MS of 32 bit captures.

Sure, for modern analysers, but I was referring to getting remaining value out of old machines like a 1630 or 16510 card.  If operating in state mode you're only using one word per instruction, rather than oversampling, and per my description elsewhere you can get a lot of useful repair information out of the first few K of execution cycles (or a few K around the right trigger point). IIRC even the 1630 had a couple of levels of trigger, not just an address comparator.

Totally agree about using the trigger to reduce the amount of unwanted data though.

[tggzzz]

Memory efficiency is relatively unimportant, especially now memory is so cheap.

What is important is being able to ignore rubbish, so you can concentrate on what is important. That's why state-based arm/trigger/filter is so important: it saves your time. If you don't capture the crap, there's nothing to ignore.

Alternatively: good luck finding the "wrong" bit in 10MS of 32 bit captures.

Sure, for modern analysers, but I was referring to getting remaining value out of old machines like a 1630 or 16510 card.  If operating in state mode you're only using one word per instruction, rather than oversampling, and per my description elsewhere you can get a lot of useful repair information out of the first few K of execution cycles (or a few K around the right trigger point). IIRC even the 1630 had a couple of levels of trigger, not just an address comparator.

Totally agree about using the trigger to reduce the amount of unwanted data though.

We are in violent agreement

With skill and imagination and the right task, old working LAs can still be useful.

[tggzzz]

The low end modem LAs have very restricted functionality w.r.t. filtering and triggering. Decent modern LAs are fine, as are old LAs from that period, e.g. >HP163x.

That would seem to rule out the LA features included in entry level oscilloscopes…. You’re definitely going to need a good dedicated device for this kind of analysis.

That's right. I've only ever seen timing analysers included in oscilloscopes. This is not really an LA, it's a bunch of very low resolution oscilloscope channels for use in systems where 2 or 4 analog channels aren't very useful. That's why they're called 'mixed signal oscilloscopes' rather than 'mixed domain analysers'.

The better ones offer SPI/RS232/I2C protocol analysis via oversampling which helps make sense of serial data but to usefully display an instruction stream you want clocked capture and possibly demultiplexing clocks.

For decoding SPI/UART/I2C protocols, if I didn't already own something suitable then I would evaluate a BusPirate5 for ~$45.

BusPirate3 was becoming long-in-the-tooth, BuisPirate4 was a misfire, but BusPirate5 is better.

[TomKatt]

FWIW I was playing around with my DSlogic LA today and noticed that it has a full Z80 decoder available.  You might have to play around with triggering options as discussed above, but you can capture and filter addresses, data R/W etc pretty easily.   With the large memory ability provided by streaming capture, you could easily analyze the startup process to verify proper operation.

Just figured I’d mention that as it seems like it would be helpful for working on that type of gear.  I see that the OP wants a stand-alone device, so prefers a non usb LA…. But if the tool works for the task I don’t see how that applies to the issue?  A stand-alone device is basically the logic hardware with a self contained processor running the box, so what difference would it make if the logic hardware and processor were separate devices?   Surely there’s no lack of laptops available to make a dedicated usb LA setup?  That’s basically what I do - a cheap older tough Thinkbook makes a good addition to my bench.  In addition to running my LA, it also connects to my power supplies, scope, and meters for data logging.  It’s a tool like everything else.

Edit - looked up the HP 16500.. very impressive.  But obviously bench space is not the limiting factor if you’re looking at gear that size lol.

[sku_u]

Thank you everyone for replying. There is a lot of invaluable information in here and a lot of points that I would have not considered if it wasn't for your comments.

Some of the key points I've taken away
-LA should be wide enough for all of the address, data, & control lines
-Triggering options
-Capture depth

I am going to inspect the TLA715 this week. I may be able to get this unit fairly cheaply and I may pull the trigger on it if the price is right and it's in good nick unless anyone thinks it would not be a good fit

When debugging/repairing equipment of that age:

• measure Vcc voltage and ripple, using voltmeter and scope
• use scope to verify signal integrity. 100MHz is sufficient, but be aware of the ringing caused by a 6" ground lead
• thereafter use a logic analyser

Some excellent pointers there along with other fantastic advice in your post!

There are lots of options in terms of digital and analogue capture. Get the deepest memory you can afford but be aware you'll sometimes limit the capture to keep the display time down. On these parallel bus processors you'll be using it in state mode though, which is far more efficient for memory use. There is an analogue scope option. It's often pricy and is very poor compared with modern scopes, but having it in the same box with time-aligned displays and cross triggering is very powerful on a mixed-signal device like a synth.

If you have no prior experience of logic analysers I'd actually recommend doing the same thing. A 1630 (1630D for mixed signal, 1630G for software development) is very cheap and is explained in terms that makes it very comprehensible to a hardware engineer. Later models lacked that straightforward explanation and wrote the manuals in terms of usage examples. IMHO this denies you the basic understanding of what's going on that proper understanding requires. Once you start to suffer from it's limitations, replace it with a more modern one. But your synths were designed by people using that class of analyser - upgrading to a '90s machine is the lap of luxury, adding small probes, deep memories, bigger  / better displays.

Thank you very much. I had not considered cross-triggering but it seems like a very useful tool to have available. Thank you for mentioning the HP 1630. I really enjoy learning about the kinds of tools and developmental processes the engineers of these synthesizers would have used back in the day. They appear to be fairly inexpensive so I will definitely pick one up in the future!

The Agilent 16702B is a modular LA with 5 slots and lots of parallel capacity (typically 68 channels per slot) and captures in state or timing mode.  You can also get scope cards with limited points (32k) that allows correlated capture with the digital inputs which is useful for checking signal integrity.

Capture cards using the "40-pin probing system" are available with anywhere from 512k to 64M points, amd hacks have been discovered to configure a card to its maximum capabilities.  TTL, CMOS, ECL, and custom -6V to +6V threshold levels are supported.

Z80 and other processor decoders are available.  Support for modern serial protocols, such as SPI and I2C, is non-existent, although of course it will show you the waveform.  Included in the OS is development software, so you can write your own decoders and disassemblers, if so inclined.

If you want something a little smaller, the 1670G and friends is a good compromise, but doesn't provide an accessible OS like the 16702B.  It uses the same 40-pin probing system, as does many other from HP/Agilent units from this era.

The 1630 series was mentioned above.  I had a 1631D for years, and the triggering was very sophisticated and easy to use.  But I found its capture memory of only 1024 points to be extremely limiting.

Thank you for chiming in about the 16702B. I wasn't aware of the Z80 and other processor decoders out there!

What frequency do these vintage instruments run at?  I’d guess many are Z80 or similar running around 4MHz or so.

Usually these early 80s synthesizers have 2 processors; one is the key assigner that handles scanning the keybed and the other is the programmer which scans the front panel I/O and is in charge of saving and recalling patches (sounds). Usually they work independently of each other. The Juno-106 uses 2 NEC 7811s, which is a Z80 derivative, as its programmer & key assigner running at 12MHz each. Later Roland synths like the JX3-P use an 8031. The Korg PolySix uses a 8048 & 8049 both @ 6 MHz and the Poly-61 uses 2 8049s at 6 MHz. Interestingly the Korg SP-80's (1983) oscillator design is very similar to the Poly-61 (1982) using 8255s and 8253s to generate square waves, but the SP-80 uses an NEC D780C, a Z80 derivative. There is an 8MHz crystal on the SP-80 board but I believe it uses some circuitry to divide that clock down, but it's been a while since I've looked at it. The SP-80 looks like it was designed in 1980 based on the design of some of the circuit boards inside of it. Ensoniq, an American synth manufacturer, used 6800 derivatives almost exclusively in all of their synthesizers. The ESQ-1 uses 2 6809s with its main programmer running @ 8MHz.

A Tektronix TLA715 is available locally and I am considering this unit but I basically know nothing when it comes to the probes and adapters I need. It has two of the TLA7N4 (2GHz timing, 64k) modules installed and comes with several of the P6434 probes. Of course these probes are the ones with the Mictor connector and are of no use to me unless I buy or make a Mictor to 2.54mm adapter. I need help determining which probes would be most suitable for my application and I have allocated several hundred dollars for obtaining these probes. According to the pdf, "Tektronix P6400 P6800 P6900 Logic Analyzer Probes" from https://w140.com/tekwiki/wiki/P6434, it looks like most of the probes are compatible with TTL & CMOS levels. I need some sort of flying lead that can be attached to the pins of chips. I have made a little DIP-14/16 breakout PCB with 2.54mm pin headers that I can attach a probe to as well. The P6417 and P6418 appear to be something I can use but I would like to confirm.

I have a TLA715 myself and so far I've been very happy with it (its my 4th logic analyser). A newer model is the TLA7000 series which can use the same modules.

Having a mictor to flying leads adapter is one way to go. I have made something like that a long time ago.
If you want to hunt for flying lead probes, make sure to get the P6417. These are the older model which has tongue shaped contacts which will grab on thin wires as well.

I see some mentioning getting USB logic analysers but I don't think these are very handy as these don't have the elaborate trigger and storage qualifiers a real logic analyser has. What the TLA715 can do is store only when a signal changes. This means A) you can use read/write strobes, chip selects, etc to get a list with addresses a microprocessor is accessing and thus trace which memory locations are accessed B) you need far less memory to store a long sequence of accesses and still you retain full samplerate time resolution. This is a very powerful feature when dealing with microprocessor circuits like the ones you are describing. Another possibility is to take a snapshot of samples around an event which can help to find what the signals look like before and after an event (for example: see what happens in surrounding logic when a setting in hardware is changed).

The TLA7N4 module can be hacked to support 8Mpts per channel. You have to send a few simple commands over the TekVISA control panel to do this. More info can be found on this forum.

Thank you very much for chiming in about the TLA715! There is a lot of great knowledge in this forum about this unit.  I am going to check out the one local to me this week. I don't have anything with Mictor connectors that I can test the unit with. Is there any other way I can test the logic analyzing capability of the unit without buying some probes I can test with? Of course I am going to do the obvious inspecting and testing that I can do. Any cons or things you wish it did that it doesn't?

They might be useful to sku_u but he's looking for solutions to microprocessor-based instruments with external ram and rom, which often fail because of address decoding or data bit errors. A wide logic analyser is kind of overkill but does actually do a really good job as long as you have a decent probing solution.

For machines with a 40-pin DIP CPU (a lot of them !) I recommend a 40 pin dip clip with the LA probes connected to address, data and control signals - enough to get an instruction trace. Maybe add a few external signals such as chip selects on rom and ram chips. Capture the first few hundred cycles out of reset. This is long enough to do plenty of instruction fetches and, critically, some stack access,

Examine the trace, verifying :
1. What happens at reset ? Typically it should fetch a reset vector, access a ROM and then start executing there. Is it sensible ?
2. Are all the data bits active ? any appear to be shorted together or stuck high / low ?
3. Watch a call occurring (the PC should get pushed onto the stack, maybe with other things)
4. Find the corresponding return. Does the same value get pulled from the stack or has it been corrupted ?

These will show up most of the common problems around a CPU/RAM/ROM core. You might be able to extend it to some I/O and a larger range of address bits but I have used this heavily when training production technicians to debug new builds that pass ATE but don't work. Both 8086 and ARM2/3 machines. These are especially hard because they often have multiple faults - they have NEVER worked. Most repairs are on bits of equipment that worked until they got a fault and then were sent for repair - so they only have one fault. Old equipment thats been rusting for years may have more but they're often in very different areas (logic, power supply, hard discs).

This procedure feeds back into the initial question slightly.
You want an LA that's wide enough for all address, data and control buses and a few extra too. This means 32 bit for 8 bit machines and more for 16 bit machines. Not many 32 bit busses in this era. but Acorn Archimedes is an exception - though the conditional instructions makes it a bugger tom determine execution path other than through the address bus.
Some machines (8086, Z80) have a multiplexed address/data bus. Thius reduces the nuber of pins you need to moinitor but requires a state clocking mechanism that can record both data and address phases. I think this needed an external hardware bus processor for the 8086. 
Huge depth is not necessary (though it's nice). Even with 1k depth you'll usually see a stack call & return, and if necessary you can trigger on ram accesses over a small address range. This is the advantage of the complex state triggers these LAs have.
A disassembler (possibly including special hardware) is also very helpful to make the instruction stream more readable.

Some fantastic information here!!!

Another device that has  a bit of a premium price because of rarity and the appreciation of it by, especially, arcade techs, is the Fluke Troubleshooter. This is a simplified in-circuit emulator with personality pods to suit popular processors. The processor is removed and the pod end plugged in its place (it may be possible to clip over a soldered-in device).

The troubleshooter then runs a bunch of canned tests such as ram, rom checksum, address and data  bus etc. by generating bus cycles at the processor that exercise those things without requiring the real roms and rams to work. I think it has an actual processor in the pod, it's not just a digital vector generator, but that processor is using its own rom and ram and then steering the bus access internally or externally to suit. This can give very rapid detection of shorted or open data and address busses.

I'm not sure if there are enough machine of this era needing repairs to justify the development effort, but it could be reproduced using something like a Pico. Maybe worthwhile for fixing pinball machines which have large numbers of different designs having a  common CPU board.

Fluke 9010A? This looks like such a cool and interesting unit!

FWIW I was playing around with my DSlogic LA today and noticed that it has a full Z80 decoder available.  You might have to play around with triggering options as discussed above, but you can capture and filter addresses, data R/W etc pretty easily.   With the large memory ability provided by streaming capture, you could easily analyze the startup process to verify proper operation.

Just figured I’d mention that as it seems like it would be helpful for working on that type of gear.  I see that the OP wants a stand-alone device, so prefers a non usb LA…. But if the tool works for the task I don’t see how that applies to the issue?  A stand-alone device is basically the logic hardware with a self contained processor running the box, so what difference would it make if the logic hardware and processor were separate devices?   Surely there’s no lack of laptops available to make a dedicated usb LA setup?  That’s basically what I do - a cheap older tough Thinkbook makes a good addition to my bench.  In addition to running my LA, it also connects to my power supplies, scope, and meters for data logging.  It’s a tool like everything else.

Edit - looked up the HP 16500.. very impressive.  But obviously bench space is not the limiting factor if you’re looking at gear that size lol.

Thank you for chiming in about the Z80 decoder.... this definitely makes it an interesting option. I haven't completely written off an USB based analyzer and I do have an older Thinkpad laying around...

[tggzzz]

Thank you everyone for replying. There is a lot of invaluable information in here and a lot of points that I would have not considered if it wasn't for your comments.

Some of the key points I've taken away
-LA should be wide enough for all of the address, data, & control lines
-Triggering options
-Capture depth

And filtering operations.

With filtering operations you can avoid capturing irrelevant stuff, e.g.

• irrelevant stuff that will be ignored by the ICs because it is between the ICs' clocks. If the IC ignores it, you shouldn't have to look at it!
• stuff unrelated to a particular suspect peripheral or memory range, or I/O or interrupt line

Those may or may not be inherent in a "personality module" for a specific processor. You may still need them for your system's peripherals.

Good to see someone listening to responses and thinking and replying. That doesn't always happen 

[nctnico]

A Tektronix TLA715 is available locally and I am considering this unit but I basically know nothing when it comes to the probes and adapters I need. It has two of the TLA7N4 (2GHz timing, 64k) modules installed and comes with several of the P6434 probes. Of course these probes are the ones with the Mictor connector and are of no use to me unless I buy or make a Mictor to 2.54mm adapter. I need help determining which probes would be most suitable for my application and I have allocated several hundred dollars for obtaining these probes. According to the pdf, "Tektronix P6400 P6800 P6900 Logic Analyzer Probes" from https://w140.com/tekwiki/wiki/P6434, it looks like most of the probes are compatible with TTL & CMOS levels. I need some sort of flying lead that can be attached to the pins of chips. I have made a little DIP-14/16 breakout PCB with 2.54mm pin headers that I can attach a probe to as well. The P6417 and P6418 appear to be something I can use but I would like to confirm.

I have a TLA715 myself and so far I've been very happy with it (its my 4th logic analyser). A newer model is the TLA7000 series which can use the same modules.

Having a mictor to flying leads adapter is one way to go. I have made something like that a long time ago.
If you want to hunt for flying lead probes, make sure to get the P6417. These are the older model which has tongue shaped contacts which will grab on thin wires as well.

I see some mentioning getting USB logic analysers but I don't think these are very handy as these don't have the elaborate trigger and storage qualifiers a real logic analyser has. What the TLA715 can do is store only when a signal changes. This means A) you can use read/write strobes, chip selects, etc to get a list with addresses a microprocessor is accessing and thus trace which memory locations are accessed B) you need far less memory to store a long sequence of accesses and still you retain full samplerate time resolution. This is a very powerful feature when dealing with microprocessor circuits like the ones you are describing. Another possibility is to take a snapshot of samples around an event which can help to find what the signals look like before and after an event (for example: see what happens in surrounding logic when a setting in hardware is changed).

The TLA7N4 module can be hacked to support 8Mpts per channel. You have to send a few simple commands over the TekVISA control panel to do this. More info can be found on this forum.

Thank you very much for chiming in about the TLA715! There is a lot of great knowledge in this forum about this unit.  I am going to check out the one local to me this week. I don't have anything with Mictor connectors that I can test the unit with. Is there any other way I can test the logic analyzing capability of the unit without buying some probes I can test with? Of course I am going to do the obvious inspecting and testing that I can do. Any cons or things you wish it did that it doesn't?

The acquisition module can run a diagnostic self-test which should tell you whether they work (or not).

The only wish I have is that the software should run a bit faster. I'm using my TLA715 remotely from my PC (using the same software as runs on the logic analyser) which is great because the bigger monitors and mouse + keyboard make life a lot easier. Nevertheless the software remains quite slow; it doesn't load the processor anywhere near 100% so the slowness seems to be on purpose for some reason.

[TomKatt]

Thank you for chiming in about the Z80 decoder.... this definitely makes it an interesting option. I haven't completely written off an USB based analyzer and I do have an older Thinkpad laying around...

I suspect the Tek TLA715 is in an entirely different class compared to my DSlogic LA lol.  Nonetheless, for informational purposes here is what the Z80 decoder menu looks like...