dual Jfets (LS844)as constant currrent sources?

[Ghostie]

I just got 4 dual jfets in a pdip package from Linear Systems(LS844 Ultra Low noise dual jfet) that I want to try to use as a constant current source for an amp I'm building. Does anyone have any experience using these? I saw them originally in Erixs' Kustom Amp thread, he then pointed me to a thread about constant current sources where PPL mentions them.
It came with test results of the idss of each chip. How far apart does the idss have to be? It seems that each chip have a very close idss. For example chip 1 has an idss of 7.123 on one side and 7.120 on the other, chip 2 has 5.983 and 5.982. Thanks
Ghostie

 

[ppl]

The Idss of a fet current source can be the same IMHO. the LS-840 being a dual matched monolithic transistor pair it's parameters are matched quite well. these transistors are about the best available for Biasing the Op-Amp in Op-Amp/Buffer Combo's.

 

[erix]

I threw in a pair of LS840's in my Kustom amp along with OPA637 and was pleasently surprised. I Don't know how closely they match but it seems that matching might be a good thing..

I wouldn't sweat it. The pair are in a 8-pin DIP and if you figured out that you must have the left hand side first on this channel but the right hand side first on the other channel, well - there goes your layout!

Pop them in and enjoy!

ok,
erix

 

[Ghostie]

PPL: So your saying it would be fine to have one chip per cascode?
Since I have't built it yet, I'm was trying to figure out which way to do it. To have one chip be the cascode for one op-amp (the same idss) or be half the cascode for each op-amp (different idss.)
I like that they sent me a sheet showing test results of each chip, then numbering the chips too!
Thanks for all your input!
Ghostie

 

[ppl]

I would use one chip per channel to avoid any interchannel interaction similar to using a dual op amp for both Channels vs. one on each. like eric sed throw them in and enjoy.

 

[Ghostie]

Great! Thanks so much for your help ppl and erix.
Ghostie

 

[jcx]

I might have made the quick and incorrect assumption that matched fets might be OK in a cascode connection before I criticized PPL’s “current source” op amp power supply decoupling in the PPA project ( the Q3 fets )

But at the moment the reason why matched fets are a bad idea in series in a cascode is clear (to me at least)

See the attached Pspice current source “tapas platter” (spice is really good at giving exactly matched devices no matter what the real world parameter distribution so all the fets in this simulation are exactly matched)

The single J0 fet current sink gives the Idss (shown in table of I_dc in the sense resistors)

J2/J3 cascode gives ~1/2 the current with higher output impedance (= less sine I_ac amplitude)

But J1 single fet with drain-gate resistor that sets I_dc the same as J2/J3 has only 10% worse output impedance as the “cascode” J2/J3

Which can be restated as: J2 is only a 10% better “current source “ than a 150 ohm resistor!

Why isn’t J2 (or PPA Q3s) regulating current? – because the voltage across the device (Vgd) is less than the Vp pinch off voltage (~= Vgs_off) for the device, in the cascode circuit is guaranteed by the use of matched fets J2/J3 where J3 Vgs must be < Vp when conducting ½ Idss

J4/J5 shows another attempt at getting the cascode to do something useful but only raises the output impedance 2X, hardly worth the expense of using 2 (matched) fets

This shows that matched fets are fundamentally inappropriate for series cascode use, the Idss and Vp of the upper fet must be selected to have a Vgs at the operating current that allows more than (the lower fet’s) Vp bias voltage for the lower fet to begin acting as a current regulator

I_dc
R_R0 -.0120
R_R1 -.0059
R_R2 -.0059
R_R3 -.0049
R_R4 -.0050
R_R10 -.0060

What’s that really flat line? – the output of a bipolar Wilson current source that can be designed and built to <5% tolerance without trimming or device selection, costs less too

I also snuk in an AC bootstrapped resistor current source for op amps used with unity gain buffers

 

[Ghostie]

Thanks for the input and Spice simulataion jcx.

Quote:

This shows that matched fets are fundamentally inappropriate for series cascode use, the Idss and Vp of the upper fet must be selected to have a Vgs at the operating current that allows more than (the lower fet’s) Vp bias voltage for the lower fet to begin acting as a current regulator

So using matched jfet are just counterproductive and expensive to use since it only is 10% better than a resistor ccs! So it is still best to have a dissimilar idss.
I'm interested in hearing more about this bipolar Wilson current source. Is that the first css shown in the simuation? Do you have a schematic or more information about it? I like cost savings and precision all in one!

If anyone else has a favorite CSS topology please chime in! Thanks once again for the discussion.

Ghostie

 

[erix]

John--> Welcome to the fray...

I've never heard a Pspice amplifier - are they new? Does Headroom sell them?

Or will somebody be making a circuitboard for them?

Quote:

(spice is really good at giving exactly matched devices no matter what the real world parameter distribution so all the fets in this simulation are exactly matched)

So this simulation is based on something that rarely, if ever happens - FETS that are matched exactly?

What other 'liberties' did you take to make your point?

Let's see..

-the FETS you used are not the same ones as shown in either the PPA schematic or the META42 schematic. You could have at least bothered to get the Spice model of the LS844 that was mentioned. www.linearsystems.com

- you are using an ideal 15V supply. Perfectly regulated, free of ripple or sag - a far cry from the battery supply that will be used in practice.

- you are feeding a 1K sine wave through the amp to take your measurements. I don't know about your musical preferences but I think sine waves are boring. Kinda like Philip Glass, ya know? I mean it's good, if you're into that kind of thing..

I don't trust Pspice for anything other than making sure I won't start a fire when I plug in. In fact, I think I trust George W. Bush more than Pspice.

This may make me the minority - people seem to care more about how their amp measures versus how it sounds.

Things I trust:
#1 My Ears
#2 My instinct.
#3 ppl, Gilmore, Peranders, ZVex.. Any designer who actually builds stuff first and measures it later.
.
.
.

#672,413 OJ Simpson
.
.
.
.
.
#1,000,006 Pspice simulations

I've built a few amps with cascodes made from mismatched 2N5484's (sorted and marked by the God of DIY himself) and they've all sounded flat and lifeless compared to the same amp without the cascode.

Amps I've built with LS840 (and LS844) cascodes sound much livlier to me than the same amp without.

That's my experience and I'm sticking with me original advice - pop 'em in and enjoy.

ok,
erix

P.S. John - I don't mean to **** on your numbers - I've no idea what they EVEN mean and I'm proud of it! My point is that if you are going to simulate something you should at least simulate the RIGHT thing.

 

[morsel]

Coincidentally, we have been talking about FET CCS issues lately, before JCX's post. Kurt is in the process of doing more detailed tests. We'll get back to you on this when Kurt is done.

 

[Voodoochile]

Quote:

In fact, I think I trust George W. Bush more than Pspice.

I bet you will see that again during the campaign. Target marketing at all the EE democrats who failed him the first time around!

Disclaimer: This is satire. I do not care, nor do I mean to insinuate any party preference among EEs. I merely had an actual LOL moment with this remark of erix's.

 

[L Daniel Rosa]

jcx,

Would you add my favorite to the simulation? I'd like to see how pspice renders it.

 

[Ghostie]

We'll I guess I'll have to try them all to see which I like best!

 

[ppl]

Jcx. we have discussed this issue offline however it appears you will not give it a rest. You lumped two separate issues into one. The jfets on the opamp rails of the PPA and the cascode current source used to Bias the Op-amp into Class A. This is confusing to say the least. since this post is about cascode current source to bias an op-amp into class A I just address that issue alone hear.

First and foremost you totally missed the point about the reason to cascode. It is not to achieve an ultra high output impedance although a High output impedance is a natural attribute of any cascode circuit. the reason to cascode anything is to reduce the effect of non linear capacitance upon the performance of the circuit. As you may Hopefully know cascodes have low miller capacitance and thus is why cascoding is often used when high frequency performance is desired. if one wanted to obtain an ultra high output impedance from a cascode current source is is easy to do however the dropout voltage is dramatically increased.

Sound Quality and technical perfection often do not go hand in hand as Eric pointed out. what dose a p-Spice amp sound like? I have heard one. It was a team project done some years ago. The Perfect Amp DC- light speed bandwidth, Perfect sq wave performance. THD into the 0.0001% range ect… ect.. ect.. Well it sounded Bright and hard no low end at all. Not that Technical competence is not required to obtain good sound because it is to a point. the best sounding Amps have never been the best performing on the Bench. This very fact is what gave the High end Audio Industry a reason for being.

Are my ideas perfect! Heck no, However one could poke holes in anything if they look in the right places. What is important is what a device sounds like in a real circuit in comparison to other similar methods. under no circumstances am I saying that anyone should blindly take my ideas on faith alone. just try it and if it works for you great! if not move on to something else.

When Tangent and morsel approached my about the PPA Project I wanted to keep it a closed design project and how it came out is what I intended it to be not a mix of everything from everybody. Now thay at that time convinced me that an open source design is a good thing. Open sourced design is not a concept I have been used to, I am a Microsoft baby so to speak. Now that Idea got Microsoft into a lot of Legal trouble by third Party competitors wanting to be able to have there products incorporated into Bill's operating system. Well after they somewhat got there way in court we now have all these other products available for our Windows operating system. Some work well others do not and still others will make you regret installing them at all. At least if the software to be installed is part of the family it will work and not wreck your system. This is also true of Electronic design when you get a lot of hands into something it can get real fragmented and end up not being what you intended at all.

"A Camel is a hoarse designed by a committee"

 

[morsel]

"Pay no attention to the man behind the curtain." -Oz

PPL, take a deep breath and relax. This is an open forum. Exchanging ideas is encouraged. Try not to take things so personally. This is a thread on current sources. You are taking it as a personal attack, and this stuff about PPA open source is way off base.

Gold Five to Gold Leader: "Stay on target."