DIY calibration source

[james_s]

I'm toying with the idea of making a simple voltage calibration source specifically aimed at calibrating some scopes I've been fixing, I have a proper DC calibrator I can use to check it but I want something small and portable with a simple rotary switch I use to select just the voltages I need, and inexpensive enough that I can send one to a friend too. I need 10V, 1.3V, 0.8V, 300mV, 80mV and 15mV as I recall, + and - of each. These scopes seem to be exceptionally sensitive to drift and noise, they must have a very narrow adjustment range because occasionally calibration will fail even with the real 0.2% source. I was thinking it a reasonable approach might be a 10.0V reference IC with precision dividers for each voltage I need with a pot to tweak each, followed by a DPDT switch to set the polarity then a unity gain buffer so it can drive a 50 ohm termination. Is this a reasonable approach or am I going about this the wrong way? I've never tried to buffer a precision voltage source so that's the part I'm most concerned about.

[feedback.loop]

What is your budget for such a thing?
How about this, for instance:
http://www.ianjohnston.com/index.php/onlineshop/handheld-precision-digital-voltage-source-v2-detail

[james_s]

That's at least an order of magnitude over the budget for this project. Looks like a nice instrument, but it's well over twice what I paid for the perfectly functional proper calibration standard I already have. What I'm hoping for is something in the ~$20 range, a secondary source that I can calibrate against my standard that will hold a nice stable low noise voltage with good short term accuracy. I'm not particularly worried about long term accuracy since it can easily be tweaked to better than the necessary absolute accuracy using a decent multimeter. If I can tweak each range and have it hold steady for an hour or two that's good enough, if it performs better than that, that's a nice bonus.

Also buying a ready made instrument would defeat the purpose of learning something.

[feedback.loop]

$20 seems to be too low for the requirements you list. Perhaps, something like this?
http://www.ebay.com/itm/Datel-DVC-8500-Voltage-Calibrator/332191099374

[james_s]

Perhaps it is, the project may be a non-starter, that's why I'm asking. Either way while I appreciate the suggestions, I'm not looking to purchase a ready made instrument at any price, I already have a calibration standard that allows me to set any voltage I want from 0 to 9.9999V, it works just fine and I don't need to buy another one. I'd like to build one as much as a learning experience as anything else, and have something cheap and reproducible that I can toss in my toolbox and not worry about.

[feedback.loop]

OK, fair enough. The Datel thing is quite compact, so I thought it might be handy.

[cellularmitosis]

Hey James, it sounds like I am working on something similar (a multi-voltage transfer standard).  Mine is based on the circuit in the big LM399 thread.   The BOM cost would be over your budget, but you could probably use cheaper parts and maybe ovenize the enclosure to get "good enough" performance.

[cellularmitosis]

(Note: there is an error in that schematic: I think the 2057 SD pin can't handle 36v)

[james_s]

Well the budget is not cast in stone, I figured whatever reference source I use will be the most expensive part but there is also the option of using a surplus device. Since I already have a 0.02% DC calibrator and multiple meters that have been checked against it, absolute accuracy is not terribly important, I want a secondary source.

The Metrology section may not be the best place for this since my question is more about other aspects of the circuit. Primarily the buffering part, I've used an op amp as a unity gain buffer before but I've never explored the ultimate precision that can be gained from such a setup. Let's say for the sake of discussion that I have a *perfectly* accurate 5.000000V source capable of delivering a few tens of microamps and I wanted to buffer it so that it can deliver a reasonable current, I'll just pull a number out of my backside here and say 100mA. An ideal op amp configured as a buffer would be able to do this no problem, but since ideal op amps don't exist outside of a simulation I'd like to explore the real world approaches that could be used.

The circuit above is essentially what I had originally thought of, but then I started thinking that seemed too simple and I must be missing something.

[Kleinstein]

A suitable OP and maybe an additional buffer (e.g. transistor) for higher current  can make a rather good buffer. The errors of the OP are usually not the biggest concern. The more tricky part is getting the ground right, to compensate for the drop on the wires / connectors.

With the choice of OP, there is a choice between classical based OPs like the LT1001 / OP177 that offer low noise at higher frequencies and AZ OPs with a little more noise, but essentially no drift at all and very high DC gain. Except for noise (especially input current spikes), the AZ OPs come rather close to an idea OP.

The tricky part with a DIY calibration source is more the stability of the voltage setting dividers (with a few fixed outputs) or the DAC circuit with an adjustable source. Today one might consider a feedback with ADC reading directly from the output as an option, to compensate for long time drift and DAC nonlinearity. A relatively cheap SD ADC like the LTC2410 has a surprisingly stable and accurate gain. One might need to spend more on two resistors to do it the old way.

@cellularmitosis:
The stability of the LM399 may not absolutely need an AZ buffer. The buffers for the 30 mV and maybe 300 mV would profit from a low offset buffer. So the AZ OP may be better placed at these outputs.

[chris_11]

For those 10:1 dividers the Caddock 1776 series might be sufficient. Is used a lot in Fluke Handheld meters and might be cheaper than individual precision resistors. Other alternative is a 16bit D/A and calibration if you are in uP and software. If not switches and banana jackets will do.

[MisterDiodes]

James_s, cellular:  Head's up on '2057 (and AZ amps in general) - those typically have very little drive current at their output, and high frequency noise at chopper freq  (~100kHz) splattering everywhere you maybe don't want.  LT Spice will never show you that, nor does it show up on the '2057 datasheet.  That might be important to you, maybe not.  Those are NOT anywhere near close to an ideal op-amp if you want a -very- quiet circuit.

You could follow the '2057 with an LT1010 driver, but at the cost of a lot of noise density.

For this box I'd suggest something like a good low-noise current-limited transistor output buffer.  Those will offer a much lower noise output at the expense that the buffer would be current -source- only. For some applications this would be OK, and you don't need exotic transistors (like how 732b does it).  Sometimes you really want that output buffer to have true source + sink capability, and that raises complexity when you need bipolar current flow on your box output terminals. 

Depends on your application - and do you need a bipolar current flow?

If you're able to calibrate this box often, I wouldn't worry about using a voltage ref with good divider resistors, quiet bipolar amps and a trimmer pot/s as required.

You will probably get what you pay for also...so you might have to adjust your budget to get whatever more pizzazz as required.

[Andreas]

those typically have very little drive current at their output,

Hello the typ 30mA of the LT2057  is more than a Fluke 732 is specced.
I think you are messing that up.

With best regards

Andreas

[chris_11]

The chopper noise can be filtered out for a DC output. The bigger problem might be driving capacitive loads i.e. from cables. Do you need to drive cables ( Ballpark 100pF/m) direct into high impedance like 1MOhm or are the inputs 50 Ohm? There are ways to compensate for driving almost infinitive capacitance but that will require a bit more complex drive circuitry. Those Opamps, which are made to drive any capacitive load are usually higher offset.

Christian

[Kleinstein]

Even if the low offset OP can drive a significant current one usually would not want this: driving a high current adds to the power loss and this can add to offsets.

The other point is capacitive drive and the high frequency noise. The high frequency noise can be filtered, but this would need another buffer / amplifier. So a more complicated driver circuit, that might include filtering would in deed be needed. More like a compound amplifier, with an AZ type OP for the DC part and a more normal, maybe higher current OP for the higher frequencies and as a current driver. With filtering and no need to buffer a high frequency signal a simple transistor buffer (though likely more than just one transistor) might be good enough for this.

[james_s]

Ideally the scopes I'm working with are supposed to be fed from a 50 ohm terminated line from the DC calibrator. In practice I've found it makes no difference I can see in the final result, so long as the voltage is correct. The highest range used is +/-10V used for calibrating the trigger and into 50 ohms that results in a significant (for a calibration source) current draw.

[MisterDiodes]

those typically have very little drive current at their output,

Hello the typ 30mA of the LT2057  is more than a Fluke 732 is specced.
I think you are messing that up.

With best regards

Andreas

Check your '2057 datasheet, specs are for 5mA max for good rail output swing - and LT apps engineering wants that even -less- output current for lower output chopper noise, say down around 2mA max, source or sink.  Any more than that and '2057 really wants some sort of buffer follower.  Typically, if you pull current from an AZ amp, the chopper noise starts to be an issue on output signal.  It just depends on application and how much you want to add filters.  AZ amps in general are not known for being a great current drivers - if you're trying to keep things quiet.

For this application, since James_S says that he can calibrate often, then there is probably no need for AZ / zero-drift amp and a quiet bipolar amp would be simpler, maybe.  You can get more drive current -quietly- out of a good bipolar amp, too.

732b does a current source only, current limited output buffer with typically no more than a couple 2N3904's...although some versions use slightly different parts.  If you only need a buffered -sourcing- output, even cheap jellybean bipolar transistors can work better (just a few nV/rtHz) than an op-amp buffer  - just make sure to feedback your final output back to the amp, and then any voltage drop across the buffer transistors is accounted for.

It works well and is a cheap & simple buffer that gives you some decent sourcing drive current plus current limit to boot.

EDIT: Especially applies if you're driving 50 ohm load:  I would strongly suggest a simple approach with your Vref chip / divider resistors as required, and the output driver I'd look at a quiet bipolar amp with optional jellybean transistor buffer for current limit, and use trimmer pot /s as required.  For voltage reversal a simple DPDT switch arrangement is probably easiest & cheapest - or just swap the plugs around.

[Edwin G. Pettis]

That current output Andreas quoted is for short circuit current, 30mA typical and 19mA minimum.....that is not a useful current specification for active outputs, that is into a short circuit.  Indeed, drawing very much current out of an AZ amp is going to degrade the accuracy of the output and increase noise significantly, drawing no more than necessary, 1mA or 2 mA at most keeps things accurate and quiet.

[chris_11]

With 10V @ 50 Ohms you are clearly in LT1010 or LT1970 current regions. You can make this stage discrete with over current protection like an audio amp, but for N=1 I would go for the version with the least amount of components and some overload/short circuit protection.

[james_s]

In an ideal world I'd like to be able to drive 50 ohms, but in practice my DC calibrator that I have can't drive that so I used it without the terminator and it worked fine.

On that note, anyone know why they specify hooking it up with a 50 ohm terminator? It's DC calibration, what difference does the terminator make? Reducing noise?

[MisterDiodes]

james_s:  Do these 'scopes have a switchable "high impedance" function on the inputs?  Usually 50 ohms impedance Z is for an AC matched impedance source, like RF generator, etc.  For an accurate low-level AC signal represented on the 'scope, you'll need that matched 50 termination on the 'scope input - otherwise you just get reflections back and forth along the test cable, and you won't know what the signal is you're looking at on the 'scope.

But for DC: 10VDC into 50Ohms is going to be around 200mA current flow, and that means 2 watts power across that 'scope input terminator.  Which is a pretty tall order for a "calibrator" and the scope input system.  Do you really want that or did I read something wrong?.  I know all the 'scopes we have here (LeCroy / Keysight / Tek) have major warnings on the front and in the software for an absolute  5V max level on the probe / trigger inputs if they are switched to 50 Ohm mode.  In general, that 50 Ohm input mode is a special function you want enabled only when necessary.  OR you do a 50 ohm terminator external to the 'scope input via a coax "Tee" if you need to handle a higher level signal.

Can you calibrate this equipment if it is switched into High Impedance mode?  (Z on 'scope inputs is usually 1MOhm // 10pF or similar).  That would make your calibrator box not have to supply so much current - and that would be a lot easier.  For DC calibration it -shouldn't- need a specific terminator, unless they are talking about accurate DC level offsets riding along with an AC signal??  I have seen that before.



 

[james_s]

The TDS300 scopes do not have switchable input impedance, I think the TDS400 scopes do but I haven't fixed the PSU for either of those yet. The service manual for the TDS300 lays out the procedure pretty clearly. When I calibrated one of them I omitted the terminator because my calibrator can't drive such a low impedance. Below is an excerpt straight from the service manual.


Equipment Required: One DC calibration generator (Item 4), one precision
coaxial cable (Item 3), one fast–rise step generator (Item 5), and one 50 ?
termination (Item 2).
1. Press UTILITY.
2. Press the leftmost main menu button until the pop-up menu shows the Cal
selection. This calls up the system calibration menu, shown in Figure 5–1.
Figure 5–1: The system calibration menu
3. Remove all input signals from the front panel BNC connectors.
4. Press the main menu button Voltage Reference. Read the on-screen text
before continuing.
Adjustment Procedures
TDS 340A, TDS 360 & TDS 380 Technical Reference
5–3
5. Press the side menu button OK Calibrate Voltage Ref. Connect a DC
calibration generator (Item 4) to the CH 1 input through a 50 ? coaxial
cable (Item 3) and follow the instructions on the screen.
6. Press UTILITY.
7. Press the main menu button Signal Path. Read the on-screen text before
continuing.
8. Press the side menu button OK Compensate Signal Paths.
9. Wait. The signal path compensation routine takes about four minutes to run.
10. Move the DC calibration generator (Item 4) from the CH 1 input to the
EXT TRIG input.
11. Press UTILITY.
12. Press the main menu button Ext Trig. Read the on-screen text before
continuing.
13. Press the side menu button OK Calibrate External Trig. Follow the
instructions on the screen.

[Vgkid]

Input impedance of the TDS-3X0 series is 1Megohm, were safe.

[jh15]

What about the horizontal calibration?

I used Tek stuff, usually tm503 rack plugins that gave you a differentiated pulse from a small cap derived from a precision square wave.

the sharp peaks made it easier to check for non-lineararities at the screen tic points.

(I repaired the calibrators, not used them much, but as all my young life was impressed with Tektronix engineering. even my first smd repairs were done in these).

[james_s]

I've read that portion of the manual a bunch of times and just now realized it doesn't actually say to use a terminator with the DC voltage calibration, it only mentions the 50 ohm coax, so I guess I was ok calibrating that the way I did without the terminator.

The timing calibration I'm still working on. It wants a 1ms pulse with a rise time of less than 10ns. I have a HP 8013B but it wasn't satisfied with the pulses from that. I was going to try triggering it externally to reduce the jitter, otherwise it's back to the drawing board. I've gotten reasonably proficient at working on these scopes so I'd like to have a way to calibrate them.