what is a good dual matched BJT for DC servo (slow+prec.)?

[mikep]

the subject pretty much says it all. I am trying to make a discrete DC servo, want to find a dual monolithic BJT transistor, like LM-394 or MAT-0x, but optimised for DC precision. low noise, low ft. cheap is good too. I don't know where to look. is this a common part?

mike

 

[Svart]

not common at all. are you willing to match them yourself or are you looking specifically for a set on a common die?

 

[mikep]

[quote author="Svart"]not common at all. are you willing to match them yourself or are you looking specifically for a set on a common die?[/quote]

was looking for something on a common die. don't want to match. maybe I can come up with a way to trim the offset of an unmatched pair? thing is, it really should remain trimmed after something like 10% PSU voltage variation.

the THAT 300 looks interesting.

 

[Samuel Groner]

I am trying to make a discrete DC servo, want to find a dual monolithic BJT transistor, like LM-394 or MAT-0x, but optimised for DC precision.

Pardon me, but to my best knowledge both transistor arrays you mention are optimised for DC precision... And in fact they are the most precise ones I know of--the THAT stuff is miles behinde.

Making a discrete precision opamp is not an easy and even less so a cheap thing to do. There are ways to get it close to what you can expect from the best precision ICs. Let us know about your way!

Samuel

 

[mikep]

[quote author="Samuel Groner"]
Pardon me, but to my best knowledge both transistor arrays you mention are optimised for DC precision... And in fact they are the most precise ones I know of--the THAT stuff is miles behinde.

Making a discrete precision opamp is not an easy and even less so a cheap thing to do. There are ways to get it close to what you can expect from the best precision ICs. Let us know about your way!

Samuel[/quote]

I just thought those arrays were overkill for a slow-ass DC servo app. I may try it anyway.

My way? I don't know! I do know how hard it is to make a precision discrete opamp, but that is not what I am trying to do. I want to make a DC servo that will compensate for very slow thermal drift (+/- 2 mV over several minutes) I was just trying to see if there was a part out there (or a class of parts) that would allow me to make a low gain DC-diff amp with little more than a cheap matched BJT pair. FWIW I was thinking about coupling the servo output into the amp with a vactrol. an IC servo is slightly unattractive because eventually I would like to run high rails, and want to avoid multiple supply voltages if possible.

mike

 

[Samuel Groner]

I never said they were better.

I don't think I said that you said? :grin:

I was just trying to see if there was a part out there (or a class of parts) that would allow me to make a low gain DC-diff amp with little more than a cheap matched BJT pair.

Honestly I don't see how a differential pair alone would help you much--can you elaborate or show a schematic? Just curious!

An IC servo is slightly unattractive because eventually I would like to run high rails.

Note that most transistor arrays do not eat more voltage than an IC. The only exeption I know of is the obsolete 2SA1349/2SC3381 complementary pair.

Samuel

 

[mikep]

Honestly I don't see how a differential pair alone would help you much--can you elaborate or show a schematic? Just curious!

not really... I haven't got that far yet. If the parts I was dreaming of actually existed, maybe I could go further. I was thinking something along the lines of a diff pair directly driving a vactrol. or a pair of vactrols, with one LED in each leg. I try all sorts of stuff that never ends up working, so I guess this might be one of those half baked ideas.

mike

 

[mikep]

[quote author="mediatechnology"]
I find the concept of using LEDs in a diff pair to drive Vactrols interesting. But I don't know why. May come in useful someday...[/quote]

thanks for the encouragement.

D3 and D4 would actually be vactrol LEDs The "current mirror with gain" gets about 20mA standing current in the diodes(edit: which is too much, should be 5ma), putting it roughly in the range for a VTL5C4 and amplifying the current difference. as shown you get about 3ma difference between the diodes for a 1mV offset.

In real life, do you think I could simply trim R4 to zero the offset?

 

[jdbakker]

[quote author="mikep"]I was thinking something along the lines of a diff pair directly driving a vactrol. or a pair of vactrols, with one LED in each leg.[/quote]
Modern LEDs have a huge intensity spread. We humans don't notice, since the eye is not very sensitive to intensity variations. Oh, you can get LEDs sorted (binned) by intensity, but it's not uncommon for the intensity of a LED in the top of a single bin to be ~150% of the intensity of a LED in the bottom of the same bin. And the temperature coefficient of a LED's efficiency can be up to 1% per degree C. So you might end up needing a DC servo for your DC servo :grin:
[quote author="mikep"]I try all sorts of stuff that never ends up working, so I guess this might be one of those half baked ideas.[/quote]
What Wayne said. Just be aware of the fundamental limitations.

JDB.
[who is looking at a microcontroller + a simple discrete DAC for a servo]

 

[mikep]

[quote author="jdbakker"]looking at a microcontroller + a simple discrete DAC for a servo[/quote]

that sounds like a cool idea.

not really the same thing but it reminds me of something I was pondering... I was thinking about using a type of switched-capacitor filter in a servo circuit. as it comes close to null, the switching frequency can automatically slow down, lowering the cuttoff freq. depending on how you do it, could have some cool advantages. for instance the servo action can "freeze" for periods when the offset is within some tolerated range. also, if you use a 2 cap circuit that toggles between the two, you could achieve a "never directly connected" scenario. I think it could be done with a few FETs, caps and a precision opamp, but you could program it in DSP as well.

mike p

 

[Dan Kennedy]

Take some of the heat off, reduce the current in the diode part of the mirror, put it in the output. So a resistor in each emitter leg of the drive transistors. Will reduce drift a lot, maybe enough.

 

[mikep]

Dan, Im not sure I understand. are you saying to reduce the current thru D3 and D4 by putting resistors in the emitters of Q5 and Q7? those currents ARE the outputs.

I tweaked it to reduce the currents virtually everywhere else:

 

[Dan Kennedy]

No, put some resistors in the emitter legs of Q5 and Q7.

You don't need all the gain in them, and by putting the resistors in you'll reduce the temperature sensitivity.

Somebody handy with spice can give you the best numbers, but start with 100 ohms in each.