Curve tracer: Design of a clone of the Heathkit IT-1121 / IT-3121

 By Dave Erickson

it1121

Intro:

As we all know, everyone wants a curve tracer. Building a trivial low power one is pretty easy. Trivial ones have currents in the low mA and low voltage, usable for small signal transistors. Even simpler is the Octopus or other simple AC curve tracers for generating V-I plots. My Analog Discovery 3 has a simple one (below). Designing or building a meaningful one is hard. It should have voltages in the hundreds or up to 1000 or so, current in the amps, be bipolar, have decent DUT connections. And ideally it should be both analog and digital.

I've been looking at the HeathKit IT-3121. This has decent specs, suitable for a DIY'er: Collector voltage up to +/- 200V, currents up to 1A, simple controls, lots of ranges. I see one old unit on Ebay for about $300. It uses an external oscilloscope for the X/Y display. It was designed for analog scopes, but can drive a digital scope in X/Y mode. The IT-1121 and IT-3121 are basically identical units, but the '3121 is a bit newer.  I don't see anything on the current market that is even close.

There is also the old B&K Precision 501A which is similar to the Heathkit. Only one V range of 100V, and it expects you will use the 'scope controls for H (voltage) scaling. Seems reasonable.

Ideally, for a DIY'er, it should be small. Requiring a 100 pound instrument on a rolling cart may be fine for a large work space, but not for a home DIY'er. Roll-around or a large bench instrument is OK if you use it every day. For occasional use, small is better. I want a small box that can live on the shelf until needed. Then 2 wires to connect to the scope, which I set up with a stored configuration for X/Y curve tracing. Easy peasy.

Of course you could buy two SMUs for about $4K to $10K. That's what Keithley and Agilent suggest. 

I did this with my DIY-SMU and Quad SMU. But transistor curve tracing takes 2 SMUs, external software, and a DUT test fixture with cables.

Tektronix has built the gold standard in stand-alone curve tracers since the 60's. If you are willing to commit the $$$, the space, and the effort to maintain an old one, go for it. However most are no longer supported and available only on the used market. Tek and Keithley expect you to buy 2 channels of SMUs to do curve tracing. SMUs have the advantage of being available in many voltage and currents. But expensive.

In my recent Analog Discovery 3 play, I tried the curve tracer. Digilent makes a specific curve tracer adapter with relays to set the base and collector current ranges, but you can do it with a small transistor socket, two resistors, and some wiring patience. I built a simple small hand-wired board to handle the details. It works well, but has limitations:

Here is the IT-1121 / 3121:

The Product build manual and Schematic.


Curve Tracer block diagram

Here is a block of the Heathkit curve tracer. 

block

Step Generator block diagram

Here is the Heathkit step generator. Tek uses a similar approach.

step

Heathkit IT-1121 Schematic

Here is the IT1131 Schematic. Pretty straightforward 70's analog design. 18 transistors, five '741 Opamps, one TTL counter. The magic is in the transformer and the switching. This represents all the features of a full-featured curve tracer, so is a good example. It is a good starting point for a DIY curve tracer.
it3121

And the Heathkit PCB  

pcb

Improvements to IT-3121

Design approach discussion

So to build an exact or a functional exact clone of the IT1131 there are a few challenges. The design is hardware-only.There are advantages and challenges to building a software controlled version.

Possible new implementations

How does Tek do Vc control? Check out the 577 service manual. It uses an AC variac, followed by a step up / down transformer with 5 taps from 6V to 1500V. The Tek 576 uses a similar approach.

Switches

The original design uses a few 12 and 9 position rotary switches. Several of these are 2 decks. Multiple-deck rotary switches are quite expensive, $40 up to $130, and are not PCB mounted. Some switches switch high-ish currents and voltages. Some, such as the base / gate step size are low voltage only. The low voltage ones could be replaced by solid-state switching.

The 2 position rocker switches are either 2PDT or 4PDT. Some of these switch high voltages and currents. Also hard to come by. Heathkit built a custom sheet metal bracket for these. I'd like to avoid this.

All these switches required a gnarly wiring harness to connect them all. The original used one deck of the rotary switches mounted to the PCB (see PCB above). That reduced the hand-wiring a bunch. I'd like to avoid this by using PC board mounted components. If it needs a second front-panel PCB, so be it. Hand wiring is the devil.

To use PCB mounted rotary switches, you are limited to one deck. Most of the high voltage / current switching can be done with one deck, at the cost of a few more resistors. Resistors are cheap and easy.

Voltage Control

The voltage control of the IT11131 is pretty elegant. It uses a simple 200K linear pot and a voltage buffer. The buffer is a simple 3 transistor power Darlington with a simple current limiter. It is driven from the raw 200V or 40V This is a simple and elegant way to control a variable 0 to 200V pulsed DC source.

Making this software and low-voltage knob controlled presents a design challenge. See below.

Custom Transformer

The AC transformer provides about 350VCT at 200mA (70 VA) or 70VCT at 1A  (70VA). These supplies are semi-isolated since the current sensing is in the return leg. In addition there is a low voltage CT winding for the +/- 15V and +5V supplies. First time I've seen a +5V logic supply with no regulator, just a dropping resistor from +15V.

It may take as many as 3 off-the-shelf transformers to do this. And like most commercial curve-tracers, the Heathkit transformers are shielded. Medical grade would help.

Step DAC and low voltage stuff

The step DAC ouptuts 10 steps. It uses 4 discrete transistors as current sources. DACs were expensive back then. These feed an inverting op-amp which nominally generates 1V / step. The feedback resistor can be varied to reduce the step sizes.

A nice precision current source converts the 1V/Step to base currents, using a single scaling resistor.

Software or Hardware?

The big questions. What would it take to make it software controlled? Is it worth it? Here are the pros and cons to software:

Mostly, software control is simply replacing switches with relays, But... building a software controlled 200V (or more) Vc generator is a significant design task. The Tek 370 does this by using a 50/60 Hz sinewave generator, an expensive analog multiplier, a high power (~100W) amplifier, and a tapped transformer. I have come up with a few crazy ideas for this, (motor driven pot, 200V R-2R DAC...) but have no final designs or even simulations yet.  So I plan to build the first version with a simple Heathkit-like 200V potentiometer. If this project takes off, and others are interested in a fully programmable system....

Of course if you're going this far, how about a scope-like display? This needs a decent size and resolution LCD (~VGA or higher) and more powerful CPU.

Tek Curve Tracers

I investigated several Tek curve tracers as part of this project, including the 575, 576, 577, 370A and the 7000 series plug-in, 7CT1N.  There is a lot to be learned from their service manuals and schematics.

575, 576, 577

These venerable CRT-Based curve tracers all use a small-ish AC variac (variable transformer) followed by a tapped transformer to control Vc. This approach has the advantage of efficiently providing maximum power to all voltage ranges, higher current at low voltage, and vice-versa.

The 575 is from the early 60s, is tube-based and has a round CRT.

The 576 and 577 are solid-state, and use a rectangular CRT. The 576 has incandescent, back-lit displays so a photo of the CRT includes the instrument settings.

370A/B

The 370A from the mid 80's, is a fully programmable curve tracer. It uses a MC68000 CPU, GPIB, and dozens of relays for control. It can digitize and store the waveforms. Very large and very expensive, it was intended for semiconductor companies.

7CT1N

This plug-in for the 7000 series oscilloscopes was designed in the 70's. It is an all analog (transistors and op-amp) design, including the step generator. It's Vc supply uses a triangle waveform generator, a class-B amplifier driving a tapped transformer. It can deliver just a few watts to the DUT so is intended for small-ish transistors. It will deliver up to 400V at 5mA, so about 2W.

Curve tracer build Notes:

I began to build up a prototype. I started with a 42VCT transformer I had laying around, and bought a 240V transformer. I bought a 200K pot on Amazon, hoping it will do 240VDC. As a hack, am using an IGBT that I had lying around, instead of the triple darlington transistors. It has about 5V of Vgs vs. the 2.1V Vbe of a darlington.  Since this has potentially dangerous voltages, a solid chassis is needed to hold all the parts.

The voltage range and voltage controls are working well. Waveform is 220V into a 5K resistor.

protoSch

And a very simple breadboard. Next step is to add a Vc range switch, DUT jacks, polarity switch, and current limiter.

proto


 coll

The next step is to add the current limiter, polarity switch, a couple of current measuring resistors, a simple collector resistor, and banana jacks for the DUT connection.

Once I get the basic collector supply working, with fixed resistors for the collector R and Current ranging, I'll probably lay out a PCB for the real thing.

Here is the first curve: the breakdown voltage characteristics of a 40V 1A schotttky diode.

first


New Base drive circuit:

There are a bunch of controls for the base drive circuit.
Doing this all in hardware takes a bunch of switches. Using two DACs makes them all doable in software, and minimizes the hardware complexity. Of course it adds some software complexity. My idea for the base circuit is to use two DACs, one driving the positive and one driving the negative amplifier inputs. This obviates the need for the polarity switch. You simply drive the positive DAC for positive voltage and the negative drive for negative voltage. The other advantage is that it makes the offset voltage simple for driving MOSFETs and other devices. A single bipolar DAC (Unipolar DAC with offset) could also do the job. The two 12b DACs sorta give better resolution, like having a 13 bit DAC.

The zero-crossing detect drives a CPU interrupt input, and then the DAC values for that step are output. Pretty simple software. The menu and controls are more complex.

A switch is still needed for voltage/current mode selection. This can be a CMOS switch if the output voltage is less than +/-15V but needs to be a relay or SSR if the voltage is greater than 15 volts. In the good-old-days, using two DACs for this circuit would be crazy but using a Microchip dual 12 bit DAC is pretty low cost and straightforward.

In addition, I plan to allow two different voltage ranges for the base circuit. MOSFETs do pretty well up to +/- 10 or 12 volts of gate voltage. But I can imagine needing more voltage for MOSFETs, and certainly more negative voltage for grid drive on vacuum tubes. The buffer could be replaced by a x2 amplifier such as one of the high voltage opamps like a OPA541. This could be powered from the unregulated plus or minus 15 which would give as high as plus minus 20 or 24 volts. Some fancier current limit would be needed for the current feedback amplifier.

Another advantage to using 12b DACs, is that the fine base current range selection (x1.0 x 0.5, x0.2) can be done with the DAC in software. The voltage range selection can also be done in the DAC. Then only 4 current limit resistors are needed for the decade selection, and relay selection of 4 resistors is much simpler. For +/- 15V, CMOS switches could be used for all but the 100 ohm resistor, which needs less than a 1 ohm switch to maintain better than 1% accuracy. That would need either a relay or a PhotoMos switch. Even with the 1K ohm resistor, less than 10 ohms for a +/- 15V CMOS switch is tricky (meaning pricey). Maybe multiple switches in parallel?

Implementing the base circuit with DACs and a microprocessor pushed the decision to use a microprocessor over the threshold. Yes, firmware and relays add complexity to the other switches. But it also allows programmability. Not sure if DIY's will use this capability.

DacStep

Proto Build

Design Tasks:

Notes:

Downsides of the IT-3121
    No separate Gate polarity switch or offset for gate voltages or Vacuum tubes (JFets)
    No AC mode, just + or - DC
    200V is useful, but 500-1000V better

Here's what I'm thinking:
    Similar specs to the IT-3121
    Similar form factor, smaller size if possible
    Encoders and buttons instead of rotary switches
    Digital interface via USB
    About $150 parts cost
    Need to find a transformer
    May need 3 transformers: +200V, +40V, low V
   
   
IT-3121
    Same as IT-1121, but newer.
    Pretty rare, one crusty one on Ebay for $300
    2 DUT connectors and L/R switch
        Bananas and CBE Sockets
   
 
   
Collector Voltage
    Good to vary by hand to sneak up on breakdown and high power
    Need > 100V, the higher the better.
    Tek does 1500V, 4 V ranges
    Old curve tracers use an AC variac + tapped transformer
    Use an AC isolation transformer and 1-turn V pot
    Need a high voltage amplifier who's output is proportional to the pulsating (0 to 200V) supply.
    Or an amplifier / waveform + generator

How about Digital interface?
    Controls: Need muxes and relays
        Still need rotary switches or up-down interface.
        Voltage control: encoder and High-V multiplying DAC??
    ADCs for H and V
        USB DAQ thing
        Use DSO
        Use analog scope
       




Back to Dave's Home Page

Last updated 6/23/2025