Hi all:
How would I go about producing super precision ref Vs with 0.25% precision?
Max load is 50 Ohms.
Specifically: (0.5V, 0.2V, 0.1V, 50mV, 20mV, 1V and 10V)
Hi Mosaic,
Like many topics on ETO, I've been reading this one of yours with interest. MrAl's suggestion of using a digital to analog converter seems like a great idea and opens up a whole raft of possibilities, even adding a microcontroller in the future. After mulling over your requirement for a few days, I ended up doing the attached schematic outlining an additional approach.
As KeepItSimpleStupid and Tony Stewart imply, the requirement for 10V with a 50Ω load is a little challenging. I agree, so the design only provides a maximum output of 5V. crutschow makes some good points about calibrating scopes, but on rare occasions you may need to calibrate a 50 Ω input on a scope so this design provides a 50 Ω output.
Mikebits mentions about checking the frequency response of a scope. Many scopes have built-in HF calibrators typically providing a 100mv, 1 Khz square wave with fast edges and flat tops and bottoms. If using a scope probe you need a good square wave to set up the probe compensation. My suggested design should produce a reasonably good square wave, but it would be much better to have a separate circuit on the scope calibrator designed especially for this purpose. The same general approach could be used, but simplified with only one low output voltage and using fast components.
I don't suppose a negative output voltage would be acceptable, because if so, it would allow a much wider choice of the voltage reference and operational amplifier. Both greatly affect the accuracy.
Here is a high-level description of the suggested scope calibrator:
The essence of the design is that a precise voltage controls a precise constant current generator that feeds into a precise 50Ω resistor to generate a precise output voltage. All those 'pecises' may sound good but every one represents an error that detracts from the overall accuracy. The suggested approach is fairly common in traditional oscilloscope calibrators, certainly the ones that I have worked on. I have also used the general circuit for many other applications, from power supplies to scope time base generators and it seems to work quite well. I've even designed and built a scope voltage calibrator using this technique.
Two output connections are provided: '50Ω OUTPUT' and 'VOLTAGE OUTPUT'. The first has a constant impedance of 50Ω and the second has an impedance of pretty much 0Ω at DC for load currents up to 20mA, depending on the choice of operational amplifier, N2. Outputs from 10mv to 5V in standard 1:2:5 ratios are selectable by a 9 position switch. Other output voltages could be configured by simply changing the value of one resistor. Another switch selects an output of either DC or a square wave at 1 kHz.
The suggested scope calibrator requires a stabilised 12V supply and consumes 120 mA, worst case, when the 5V output is selected.
The circuit should meet your accuracy specification which I take to be +-0.25% without trimming. I haven't done an error budget, but worst case calculations would probably show around +- 0.5%. That would be if all the errors go the same way. Similarly I haven't calculated a thermal error budget but I don't expect thermal effects to be significant with adequate cooling and layout.
To answer the initial question in your opening post, I suggest that the approach shown in the schematic would be a good starting point for your project, and could be developed into a reliable scope calibrator. Please remember though that this is not a fully developed design and is certainly not optimised. I didn't spend any time researching the components shown on the schematic so no doubt these could be improved on. None of the components should be prohibitively expensive. The precision 0.1% high stability resistors cost about 85p but I haven't checked the cost of a high quality switch.
I hope this helps and if you spot any mistakes, there always are, or need any clarification please let me know. I have put some more information in the APPENDIX. As you may have guessed, I would love to have a go at this project myself. Do let us know how you get on if you decide to go ahead with it.
APPENDIX
DATA SHEETS
* OPAx92 operational amplifier family datasheet:
http://www.ti.com/lit/ds/symlink/opa192.pdf
* ADR8421 precision voltage reference family datasheet:
http://www.analog.com/media/en/technical-documentation/data-sheets/ADR5040_5041_5043_5044_5045.pdf
* LM555 timer datasheet:
http://www.fairchildsemi.com/datasheets/LM/LM555.pdf
* Welwyn precision metal film resistors:
**broken link removed**
DESIGN, DEVELOPMENT, AND CONSTRUCTION
The 50Ω output should produce a well-shaped square wave with fast edges and little ringing, and the well-defined 50Ω impedance helps minimises reflections, with 50Ω cables that is. Another important characteristic of this approach is that all the precision circuitry works at DC and is isolated from the outside world by the transistor in the current generator. This greatly simplifies things. Finally it provides a very robust output, essentially a resistor and the drain/collector of a transistor. The 'VOLTAGE OUTPUT' is a different story and would need comprehensive protection in a practical scope calibrator.
For even better absolute accuracy the value of R15 could be changed slightly; a trim potentiometer would not be suitable as the calibrator temperature stability may be ruined.
If a negative output voltage was acceptable a series type voltage reference could be used. These generally have a better performance than the shunt type shown in the present design. And to a lesser degree, the operational amplifier in the constant current generator would be operating with a better input voltage. Having said that, it may be possible to use a shunt voltage reference with the positive version of the calibrator, but this is another area I haven't investigated.
In case you're wondering about the parallel resistors, they are just there to dissipate power: high power, high frequency, precision, 50 Ω resistors are very expensive. The capacitors across the supplies etc are only notional decoupling; this is an important area to get right in a practical design. Tantalum for the electrolytics and ceramic for the non-polarised capacitors would be a good choice.
The LM555 is only there to provide a quick and easy way to get a square wave for development. Apologies for the spaghetti schematic layout around the LM555 but I just grabbed a part from the library and didn't bother to redraw it. In the final version I suggest a square wave generator with an xtal oscillator and accurate 1:1 mark-to-space ratio, all pretty straight forward and cheap to do. This would turn the calibrator into an accurate frequency generator as well. You could even provide a range of frequencies selectable by an additional switch. The highest frequency should probably be 100kHz to keep things simple though. You could go a lot higher if a separate frequency calibrator was built into the scope calibrator but it would be best not to run the main circuit at high frequencies.
There is absolutely no reason why, for initial development purposes, standard non-precision parts couldn't be used. The resistors could just be ordinary types and you could even start with a PNP bipolar transistor for the constant current generator. For prototyping you wouldn't need to worry too much about input currents and low offset voltage so a lesser operational amplifier could be used provided that it will operate with its inputs just 1V down from the 12V supply rail. The best approach may be to build a negative output version of the calibrator first to establish the fundamentals of the circuit operation. The decoupling components and layout etc must be the real deal though.
MOSFET, Q1, will need careful selection to ensure a decent square wave, and low leakage, especially as it will be dissipating quite a bit of power at the higher output voltage settings. Maybe another type of transistor would be more suitable. Of course, the big advantage with FET devices in this case is that the current flowing up the source is the same current that flows out of the drain. The big drawback with FET devices though is their leakage currents and capacitances and I have a feeling that in the end a couple of bipolar transistors will do a better job.
Another area to watch out for is stability: although the constant current generator is only working at DC it has a very high gain and will be prone to oscillations without a good layout and adequate decoupling. It may even be necessary to add some frequency compensation in the feedback loop. The operational amplifier is operating with its input at only 1V more negative than the 12V supply rail. This should not be a problem and the OPA192 shown should do the job well, but as the upper difference amplifier in the chip is operating it is not ideal (see my post on the OPA192).
Apologies if I'm preaching to the converted, but here are a few general comments. The devil is in the detail is the adage to keep in mind when developing accurate analogue designs like this. HiFi amplifiers are the same. A very good power supply is absolutely essential, certainly not a switch mode type, for initial development anyway and solid grounding is vital. The physical layout and screening would also be critical to achieve the performance. And heat is always your enemy. One last piece of cracker-barrel advice, you need to use 'four terminal connection' when the precision parts are wired into the circuit or when the printed circuit board is laid-out. Without this, the accuracy will be lost. Believe it or not, you also need to use solder that gives a good solid joint with low resistance and minimum galvanic voltages: some of the zero-lead types are very poor in this respect and they get worse as they age!
The switch for selecting the output voltages is in a critical place in the circuit as far as accuracy and frequency stability is concerned, so it should be a good quality low resistance type, ideally with gold contacts. Rather than mounting the switch on the front panel and having long leads to the precision resistors, I'd strongly advise mounting the switch to optimise the electrical performance and keeping the leads as short as possible. You can always use a mechanical method to operate the switch from the front panel.
Here is a list giving the resistances and currents for the range of output voltages
(1) 10mv 200μA 5k
(2) 20mv 400μA 2K5
(3) 50mV 1mA 1k
(4) 100mV 2mA 500R
(5) 200mV 4mA 250R
(6) 500mV 10mA 100R
(7) 1V 20mA 50R
(8) 2V 40mA 25R
THIS VERSION IS NOW OBSOLETE> PLEASE SEE REVISED CIRCUIT IN POST 34 of 05 November 2015.
ERRATA (items in red are important and affect the fundamental function of the circuit)
(1) R15 should be 15K not 25K (Typo) With R15 at 25K the circuit will still function ok but the output voltages will only be 70% of the values shown on the schematic.
(2) Add 100nf ceramic capacitor between the positive and negative supply pins of N1. (good practice to help keep N1 stable)