What happened to Current Regulating Diodes?

[ezkcdude]

It was suggested to me that the output of my ezDAC could be improved by adding a constant current source in the form of a CRD (or JFET + R). So, not knowing much about this, I looked into it a little (even came upon Tangent's website - to no suprise of mine, of course). So, I figured the CRD might be a nice solution - save some board space, since it's only a single component, and not have to worry about fiddling around the with R values. Well, it seems these CRD's are very difficult to find these days. Some of them are in stock, but many are now obsolete or discontinued. This raises a few questions. 1) What happened to the CRD? (I'm just interested from a history standpoint). 2) Is there now a reliable (active) source for obtaining CRD's (preferably through-hole) and 3) If the answer to (2) is "NO", then is the JFET + R a good enough alternative to bother, or would I need to go with the JFET cascode to see any real improvement?

 

[MisterX]

http://www.mouser.com/catalog/631/374.pdf

Or page 410

http://www.mouser.com/catalog/631/410.pdf

 

[ezkcdude]

Thanks, MisterX. I hadn't seen those. Many of the J and CR series from V-S are obsolete, though.

 

[Pars]

Someone in one of the threads (might be on HW under the sigma22 thread) had an explanation regarding Vishay Silliconix' fab being sold and outsourcing FET production. They now apparently have yield/characterization problems which is why you don't see the J series available. That coupled with somewhat hideous prices (the 4.7mA 1N5314 part is $4.75 at Mouser, and I think $11 something at Newark

). Might be another instance of bean counters getting involved

.

 

[pmillett]

Just like JFETs (CRDs are really JFETs with internally strapped gates) and other small discretes, the demand has dropped low enough that the "big guys" like Vishay don't bother with them much. So smaller specialty houses - and low-cost Asian suppliers - pick them up.

Hint:

As I understand it these things are sorted at test into different current bins; process variations make them come out all over the map. In other words the "1mA" part is built no differently that the "0.91mA" part. So the parts with high demand (like 1mA) wind up in short supply. That's why you'll notice that the 1mA part in Mouser costs more than 2x the 0.91mA part. The manufacturer probably throws away 0.91mA parts in order to build enough to meet the demand on the 1mA binning.

So, design in an odd value and save money...

Pete

 

[tangent]

It's the mainstream world of electronics that drives what components we see, and discrete analog electronics is very much in decline. The need for constant current sources is still quite high, but they're quietly buried in monolithic analog components like op-amps and references.

As DIYers, our time is relatively cheap and we often aren't too concerned with compactness. So, we also tend to avoid CRDs because we can get better performance by hand-matching some JFETs and maybe adding a resistor or two. The cheapest thru-hole CRD I know of runs about $1.75 each, and that'll buy a lot of JFETs.

 

[ezkcdude]

Quote:

Originally Posted by tangent /img/forum/go_quote.gif So, we also tend to avoid CRDs because we can get better performance by hand-matching some JFETs and maybe adding a resistor or two.

What do you mean here by "performance"? Does this refer to current regulation, noise, conductance? Which of these factors will affect the audio signal? For example, if you want to do the Class A biasing, does it matter whether the current sink varies by 10 or 20%, as long as it is more than enough? Does the current variation itself somehow affect the signal?

 

[tangent]

Quote:

Originally Posted by ezkcdude /img/forum/go_quote.gif What do you mean here by "performance"?

I thought you read this article. CRD: 1% current variation with voltage change. JFET cascode: 0.5% variation.

Quote:

does it matter whether the current sink varies by 10 or 20%, as long as it is more than enough?

Yes, it matters. Keeping the output stage in class A is only part of it. We want to keep it in class A at a fixed bias point to avoid thermal variation.

 

[amb]

One thing attractive about the 1N5xxx and CRxxx series CRDs is theur 100V rating. Even the J5xx series were rated 50V. There are very few discrete JFETs with that sort of ratings, so in higher voltage apps (like in the β22 amp) where the parts count is already high and there are many points in the circuit that need current regulation, there is no simple and elegant alternatives to CRDs.

 

[Pars]

Could try these... a bit overkill (350 and 450V). TO220AB package (or SMD), and requires a resistor. I've seen these recommended for CCS in tube apps. before... bought one for my CP preamp, but haven't used it yet.

 

[ezkcdude]

I may end up adding pads for 1N53XX from Central Semi (1N5314 = 4.7 mA). It's a DO-35 package, and there are plenty available at Mouser. Leo K hardwired an SST511 and said it "sounds quite good".

 

[Seaside]

One good thing about CRD over JFET+R is, you don't have to worry about boring things like pair matching, current calculation etc. Bad thing about CRD is, you're gonna pay more than what you pay for JFET+R.

 

[ezkcdude]

Just to follow up, I re-read the two Walt Jung articles on current sources that were published several months ago in AE. Very useful. Looking at his measurements, I'm pretty satisfied that the CRD is a good choice for my application.

 

[applegd]

Quote:

Originally Posted by Seaside /img/forum/go_quote.gif One good thing about CRD over JFET+R is, you don't have to worry about boring things like pair matching, current calculation etc. Bad thing about CRD is, you're gonna pay more than what you pay for JFET+R.

I think it is the other way around. Look at the Vishay CRD datasheet, take SST511 as example, guaranteed tolerance is 20%, base on my experience, I have to buy more(for my case 20) than I actually need for better matching(<=1%) although it is expensive.

For JFET+VR, you have much less matching concern, just let VR do the job. So you actually save money for better performance.

Now it is pain in ass to source these SST5xx,mouser still has about 20 SST508-E3 for 2.4mA, soon it will run out......