Is it worth going for a all through-hole construction on a new design?

[technix]

I am still pondering on the idea of making my own 4.5-digit multimeter but a few changes occurred: out goes the micro.

I found an Intersil application note on building an autoranging volt-ohm meter using ICL7103/ICL8052 pair (precursor to ICL7135 which combined the two chips into one but the built-in reference in ICL8052 is lost) and I am thinking about expanding it into a proper volt-amp-ohm meter.

For now I have almost all of the supporting parts but they are all through-hole. (If you are interested, digital section is mostly 4000-series CMOS and I can go 4000-series only if I want to. The only two chips not 4000-series is ULN2003 and GAL16V8) Should I go ahead with it and build a brand-new design that end up being assembled using NOS through-hole chips? (a few chips in the chipset, namely the AD736 RMS-to-DC converter and the REF03 reference chip are not cheap bought new, so I can only afford NOS or even used) Or should I just scratch that design and go back to the microcontroller-based design and keep hunting for good-enough surface-mount ADC at reasonable price?

[Fungus]

Should I go ahead with it and build a brand-new design that end up being assembled using NOS through-hole chips?

Sure, why not?

[Kleinstein]

The ref03 is not that special - there might be more economic references if it comes down to price.
Doing autoranging with standrad logic / GAL is outdated - even a simple µC is much better suited for that, even if combined with a display oriented ADC like the 7135.  If you are familiar with µC's, I would go with a µC based solution. If not a I would skip autoranging. Autoranging with the current ranges is problematic anyway.

Going through hole is not bad for a one of a kind project. At that level changes / fixes are easier than with a smaller all SMD design. It's also possible to combine THT chips the SMD resistors and capacitors (epsecially decoupling). I would consider the TH form of the ICL7135 an advantage. Also TH chips in sockets are easy to replace / reuse.

I have not seen the ICL7103/8052 - sound's like antique and hard to get a replacement, even I you still find a set. So I would definitely not use this - especially if it's just for the reference.  The ICL7135 is much more comon and chances are you still get one in 5 years.

If you want a really low cost solution, I would go for µC + sigma delta ADC and RMS in software. With just 4.5 digits, something like a MCP3911 and a µC (RMS would profit from a faster µC, like a small ARM) can do most of the job. As a benefit one might have the option to do things like a PC interface (e.g RS232).

[nowlan]

FYI, might be interested in this.
Hameg HM8012 Multimeter - Service manual

[T3sl4co1l]

Do you expect to sell anything, like, in production?

If not, nope, doesn't matter, follow your heart.

Prototypes of modest complexity are kind of-sort of-tempting to breadboard (either with actual breadboard of the solder or -less varieties, or more free-form methods such as dead-bug), but if you have a serious number of traces to route, and pins to connect, yeah, it can be worthwhile making a PCB, even if you only wanted one or two, not 10 or 20 of them!

Tim

[Ian.M]

One-off or very small volume (<10)?  Make it modular.  Its easier to test and you can reuse/repurpose bits of it when you come up with a better design.

[technix]

The ref03 is not that special - there might be more economic references if it comes down to price.
Doing autoranging with standrad logic / GAL is outdated - even a simple µC is much better suited for that, even if combined with a display oriented ADC like the 7135.  If you are familiar with µC's, I would go with a µC based solution. If not a I would skip autoranging. Autoranging with the current ranges is problematic anyway.

Going through hole is not bad for a one of a kind project. At that level changes / fixes are easier than with a smaller all SMD design. It's also possible to combine THT chips the SMD resistors and capacitors (epsecially decoupling). I would consider the TH form of the ICL7135 an advantage. Also TH chips in sockets are easy to replace / reuse.

I have not seen the ICL7103/8052 - sound's like antique and hard to get a replacement, even I you still find a set. So I would definitely not use this - especially if it's just for the reference.  The ICL7135 is much more comon and chances are you still get one in 5 years.

If you want a really low cost solution, I would go for µC + sigma delta ADC and RMS in software. With just 4.5 digits, something like a MCP3911 and a µC (RMS would profit from a faster µC, like a small ARM) can do most of the job. As a benefit one might have the option to do things like a PC interface (e.g RS232).

I am using ICL7135 from the get-go, since it have both ICL7103 & ICL8052's circuitry combined into one chip except the reference (which is now done using a REF03) and the autoranging logic, using the CD4017/CD4098/GAL16V8 chipset, comes right off the Intersil datasheet.

If you are suggesting MCP3911+ARM, I happen to know a good MCU to use - a proudly Made in China CH563 which is not expensive but have all the bells and whistle - USB (device and host) as well as 10/100 Ethernet (no PHY chip needed.) Although I will still keep the AD736 and REF03 (although this time I may be using the surface-mount version ADR03 and a surface-mount RMS-to-DC converter.)

[technix]

One-off or very small volume (<10)?  Make it modular.  Its easier to test and you can reuse/repurpose bits of it when you come up with a better design.

Oh definitely. The project will go with three boards: display + switch, main processor and power + input protection.

[Ian.M]

That's *NOT* how I would break it down.

Input protection, range switching, attenuators etc. need to go together on the first board.

Then you need the mixed signal stuff on another board with optosolators on the digital lines

Then there's the MCU, which could easily be an off-the-shelf development board.

Display and front  panel controls - probably on a serial bus of some sort to keep the cabling and MCU pin count reasonable.

Finally there's the PSU, which must provide an isolated supply for the front-end + power all the other modules.   

[sarepairman2]

you should upgrade it to a high voltage 4.5 digit source meter and add a transistor assisted vacuum tube to it.

that would actually be a really fucking cool project. multimeters up to 5.5 digits are so cheap now adays (90-150$).. but if you made a HV source meter it would be a unique project and you would probobly be remembered.

im serious about the vacuum tube btw, it would work better for a one off prototype then a bunch of blown transistors.

[technix]

That's *NOT* how I would break it down.

Input protection, range switching, attenuators etc. need to go together on the first board.

Then you need the mixed signal stuff on another board with optosolators on the digital lines

Then there's the MCU, which could easily be an off-the-shelf development board.

Display and front  panel controls - probably on a serial bus of some sort to keep the cabling and MCU pin count reasonable.

Finally there's the PSU, which must provide an isolated supply for the front-end + power all the other modules.

For the ICL7135-based design it is going to be battery operated so no internal isolation required. If I go with MCP3911 only the analog section, up to the ADC, will be isolated. Equipment still will be battery powered but there will be USB connectivity and rechargeable batteries inside.

[technix]

you should upgrade it to a high voltage 4.5 digit source meter and add a transistor assisted vacuum tube to it.

that would actually be a really fucking cool project. multimeters up to 5.5 digits are so cheap now adays (90-150$).. but if you made a HV source meter it would be a unique project and you would probobly be remembered.

im serious about the vacuum tube btw, it would work better for a one off prototype then a bunch of blown transistors.

If I know how to roll that I may give it some consideration, as an optional accessory of the meter. However powering the valve with 9V is a significant challenge as I don't think any SMPS can do that high a voltage.

[Kleinstein]

You only need a powerfull µC if you want to do true RMS in software, as an alternative to the old analog RMS converter. The ADC/µC RMS version might be lower cost and can have some advantages (better linearity, stability) but usually will have a limited bandwidth.  With analog true RMS you can uses just a low power µC (PIC,AVR,8051,MSP430 ...) - even integrated USB is of very limited use, as you still need isolation, which is easier at UART level.

In a small design you may not need optoisolation other than for the interface. The easy way is having a optoisolated UART and than a UART to USB chip powered from the USB. The USB side needs power only when used, so no need for a isolated powersupply, when battery operated. Though having the option to power the whole thing from USB is likely a good idea.

The old GALs are rather power hungry and thus no good option for batteriy operation.

 

[technix]

You only need a powerfull µC if you want to do true RMS in software, as an alternative to the old analog RMS converter. The ADC/µC RMS version might be lower cost and can have some advantages (better linearity, stability) but usually will have a limited bandwidth.  With analog true RMS you can uses just a low power µC (PIC,AVR,8051,MSP430 ...) - even integrated USB is of very limited use, as you still need isolation, which is easier at UART level.

In a small design you may not need optoisolation other than for the interface. The easy way is having a optoisolated UART and than a UART to USB chip powered from the USB. The USB side needs power only when used, so no need for a isolated powersupply, when battery operated. Though having the option to power the whole thing from USB is likely a good idea.

The old GALs are rather power hungry and thus no good option for batteriy operation.

 

If I go with isolated USB route I would run the uC on non-isolated side and isolate-power the ADC. Autoranging is done using relays so that are already isolated

[kripton2035]

I think it's only a matter of production quantity
if you're making a prototype, let's go for throught hole, or what you like
if you intend to sell it later, you've better go with a modern smd design
I always use dip adapters for all smd chips that I put in my prototypes
you will surely have problem finding throught hole chips in big quantities.

[Kleinstein]

with just 4-5 digits, there is no real need to isolate the analog input from the µC and display. Especially if the µC is not doing much, it does not produce much noise. Modern single chip solutions are much better in this respect than the old way with a CPU / RAM and EPROM in separate chips.

[technix]

with just 4-5 digits, there is no real need to isolate the analog input from the µC and display. Especially if the µC is not doing much, it does not produce much noise. Modern single chip solutions are much better in this respect than the old way with a CPU / RAM and EPROM in separate chips.

The reason of isolating the analog section is that I can use unisolated USB and internal rechargeable Li-ion batteries. If I want to do software RMS I might as well go with a more powerful MCU like CH563 (based on ARM9, proudly Made in China) which also comes with built-in Ethernet MAC/PHY (something I definitely will put into use) and USB (definitely used too). Also unisolated USB means I can add an Li-ion internal battery bank, recharged using power from USB bus power.

[technix]

I think it's only a matter of production quantity
if you're making a prototype, let's go for throught hole, or what you like
if you intend to sell it later, you've better go with a modern smd design
I always use dip adapters for all smd chips that I put in my prototypes
you will surely have problem finding throught hole chips in big quantities.

A few components in the ICL7135-based design don't have SMD alternatives (or hard to find): ICL7135 itself is one, and the SMD version of REF03 is several times more expensive as there is no NOS stock for either REF03 or ADR03, and the same for AD736. Precision metal film resistors in SMD packages are also a few times more expensive than their THT counterparts.

[Kleinstein]

You can also have the isolation between analog and digital part. However this may need a few more isolators for the ADC and some range switching that does not use relays. Relay are still buky and need quite some power if not bistable. You also need a isolated power supply, whith the benefit of having free choice with the voltages.

In a hand soldered low volume projekt mixing THT and SMT is no problem.

[SeanB]

If you ignore USB ( or use a USB isolator instead) and only use ethernet then isolation is not needed, as the magnetics provide 2kV of isolation, though you will probably want a separate magnetics block and socket, so you can connect the capacitors needed to the cable shield only, and have a very low value capacitor across the isolation barrier and a high value ( 22M or higher) bleed resistor for the isolation gap. If you want to run with batteries and charge at the same time a DC-Dc converter will be needed, though there you can use a POE unit with isolated output to give you a 5V rail to charge the internal cells.

Interesting to have a DMM which connects over POE for charging or just works alone.

[technix]

If you ignore USB ( or use a USB isolator instead) and only use ethernet then isolation is not needed, as the magnetics provide 2kV of isolation, though you will probably want a separate magnetics block and socket, so you can connect the capacitors needed to the cable shield only, and have a very low value capacitor across the isolation barrier and a high value ( 22M or higher) bleed resistor for the isolation gap. If you want to run with batteries and charge at the same time a DC-Dc converter will be needed, though there you can use a POE unit with isolated output to give you a 5V rail to charge the internal cells.

Interesting to have a DMM which connects over POE for charging or just works alone.

If I use Ethernet I will make sure the DMM can talk to computers (or better, iPhones). However both PoE and USB isolation are expensive so still prefer unisolated MCU with isolated analog section.

[dardosordi]

Hi,

I know its an old thread, but Im thinking of building a bench meter with ICL7135, do you mind sharing some schematics ?

thanks,
Dardo.

[technix]

Just search for "ICL7135 autoranging" and follow that application note. I also squished the entire autoranging logic into one programmable logic chip, and used the CD4511/ULN2003 instead of 74LS47/discrete NPN as the display driver chipset.