I have been an advocate of using Constant current Sources to bias an Opamp driving a Buffer into Class A for over a decade. The Advantages is to remove the Possibility of Crossover Distortion From the Class b or Class ab Opamp output stage. Some Opamps like The AD-845 Have a Class A output stage and as sutch will not be improved by this technique. Some Opamps notably all the Most used Burr Brown Types used by Head-Fi'ers show a dramatic improvement with this method. My most often used method of suppling this Constant current is via one Of the Current regulator Diodes. or using a Jfet with the Gate and Source connected together and this connecting Between the Vee (Neg. Rail) and the Opamp's output. This works alot better than the simple resistor most folks use to do this as the CRD will not load the Opamp's output stage with a low resistence from a Resistor and thus removing the Dynamic Benifits of using a Buffer. This method while Great is still not the Best way. The Nonlinear capacitence of the CRD or j-FET will have a negitiive impact upon the sound quality of the Opamp. I have got around this in the Past by Placing a resistor in series with the CRD and the Opamps output to Buffer this capacitence. Since most of my recent amps have been battery powered The Possibility of the CRD droping out of regulation on large signal swings is a possibility. I have thus been testing Current sources made from cascoding Two J-FET's. A cascode circuit has a constant capacitence and Leakage current with voltage. This may at first seem a minor Point unless like Me you Think god is in the details. and if your search for the Holy grail in sound quality from an Opamp based Headphone amp I cay say without reservation The Cascode Current source is so mutch of an improvement over a Current regulator diode in sound quality that I was Most impressed. The most notable improvement is in the Midrange upward. My first Impresions using an OPA-637 Driving a stack of EL-2002's were staggering almost like another Otave was added to the Reproduction and this changed what is normaly a slightly dark sounding Opamp with a Normal CRD into a High and Mid frequency sonic Wonder. All BB OPA's without any Resistor or CRD sound Dark and vailed to me. using a 5K resistor strips all the dynamics away, while still only providing trivial improvements. replacing the resistor with a resistor Bufferd CRD brings the dynamics back while almost Compleatly removing the vail over the Sound image. Replacing This with a cascoded low capacitence jfet pair made this dark sounding opamp into a nice and detailed Performer. This same improvement happens with AD Opamps like the AD-825 & AD-823 also. Infact the sibelense sometimes present on the AD-825 is dramiticly reduced.

In conclusion If your going to get the best sound out of your Opamp Buffer Combo cascode a Pair of jfets like the MPF-103 or a Lower capacitence type and enjoy a sonic improvement You will find hard to resist.

I think this belongs in DIY, no? I'm gonna move it

ppl is at it again . Very nice article. I'll try it when I start experimenting with DAC/headphone amps again.

I'm sold! I'll get a few of those jfets with my next orders.

trust in the ppl, the ppl is good... hasn't steered me wrong once. that's just me, though... heh

oh incidently, where was this MacDEF, that you had to move it?

Great idea PPL. I am laying out an amp right now and will include this. I have examined some headphone circuits lately and all have extra transistor pairs like this.
I have two old Denon portable CD players. I repaired one with a digital out last week and disassembled another one and both had this type of circuit. The two designs wer very different although from the same era. These were very expensive players and considered the best portables ever made.
The second one which I have had for years is full of Elna Electrolytic caps. These things eat batteries but have the best headphone amps I've heard on any portable. No shock stability but they sound great.
Dan

Dan

I can see a commercial headphone amp having a push-pull transistor pair as an output stage, but i seriously doubt any consumer product would use this type of design to bias opamps. These transistors are not used in the signal path, they only bias the output stage transistors of the opamp into class-A.

I'm sure this design sounds amazing, but i wouldn't use it in a portable unless you have a monster 16AA battery pack.

PPL: Can this same technique be used in buffers like BUF 634 or EL2001? Seems like it should work, but i've never seen it done before...

Oh Darn! I forgot to Change Fourms before i posted this I was reading the Music Reviews and forgot to Change to the DIY fourm, Sorry. Gess that is what Moderators are for. Anyway moving along Regarding Thomas Idea of Biasing Buffers into Class A. Elantec Recomended placing a 100 ohm Resistor between V- and The Output of the EL-2008 for a Video Application in order to get the Best Possible Diff. Phase and gain. On an AC powered Headphone Amp Using the EL-2009 I at First used this Resistor and as expected The dynamics were somewhat Compressed so i changed That to a Current source using a Bipolar transistor with an LED for a Reference. This made a slight Improvement and was most noticeable using the Amp as a Line Level Preamp. Havent used a Jfet in that Application because of the amount of Current Needed. Using Elantec's recomended 100 ohms at 12 volts results in 120 Ma. I dont Know any Jfet with an Idss that High. As far as usiing My Cascode Current source Idea on the Opamp in a battery Powered amp. The Current source only Needs to be like 1-2 Ma. The LS-840 Dual J-Fet i used in my Present amp Design as discribed, supply 0.68 Ma in cascode Mode and 1.75 Ma per J-fet half if used in a Normal non cascoded Current source. However this was maintained from 1.2 -15 Volts. where as an 1N5298 CRD went from 0.7 to 1.1 Ma Under the same Voltage range and if a Buffering resistor is added then the Voltage could only Go down to 2.2 Volts while sourcing 0.7 Ma. This may seem unimportent untill you consider that most Opamps can swing to wihin 2 Volts of the rail and Some Like the OPA-132 series and The AD-823 Opamps can swing to within less than 1 volt of the rails. Now you may be Thinking that no one is going to Listen to There Headphones at that level and this would be Good Thinking but The Change in current within the Normal Operating ranges can Impact the Opamps Performance and most surly affects the sound quality as this is just one reason The Cascode Current source made sutch an Improvement. I am evean starting to Like The of an OPA-637. With an AD-825 and AD-823 OH Ya.

ppl, is it necessary to use dual JFET? I was thinking of using two separate ones. I do have some duals but it's really a sacriledge (not to mention a major $$$ issue) using low noise dual JFETs for biasing, instead of differential input stage...

It is Not required to use Dual jfets for the cascode Current source. I Did because i was replacing a CRD and this is the size of a 1/8 watt resistor. My portable amp board is quite tight so space was an Issue. It had to fit in the same space as the CRD and Conect into the same Hole's. Also the LS-840 has a low Pinchoff Voltage so this also was a factor since the lower this voltage the Lower the Voltage will be that the Current source Drops out of regulation. Capacitence is also a issue since i wanted a low capacitence current source. It is true that cascoding reduces the Capacitence vs just the Normal non cascoded Current source but having lees to start with will result in less in the End. I temperly Tried an MPF-102 and these also improved the sound. but i think a lower capacitence type like on semiconductor's type 2N5484 & 2N5486 would be evean better.

ppl,

Did the guy at Linear Systems ever give you a
price on the LS-480? Or anything else?

I asked him twice months before I gave you there
address, and I still haven't received a response.

He said he would on the phone, but it hasn't
happened.

Hmm... So you decided to go with the spendy Dual's. Heck thay are one hell of a great fet. I am going to use some of these as the Diff Amp In my Winter Project An AC powered All Cascoded Discreet component Opti-Fet Dual mono ultra reference Headphone amp. So far i got the Power Supply Done and I done some Tricks to Totaly Eliminate DC Drift with tempature on the Regulators. Evean Kevins and Jungs supper regulator drifts alot with Temp. Heck i Blow air across the Circuitry and get a change of a volt or two. On my latest design No change rock solid +/- 15.0 Volts. I am glad the dramatic amount of work i put into that paid off. Any way sorry i got side tracked. LSI has a sales rep that is sometimes hard to deal with But i Think he sed somthing like $5.50 Ea for the LS-840. in Small Qunt. Ill will check and let you Know. I tell ya this company is one of the Better Nitch Analog fab Plants around. there 60 Volt Dual BJT's actualy spec out on a curve Tracer at 95-115 Volts. try that with a fairchild or Siliconex Device.

I'd also be interested in getting those LS-480's. I want to try them in my portable DAC (which is progressing nicely btw, just about ready to start laying out the PCB). If you guys figure out a way to get them, could you let me know?

I want to play with their BJT's as much as the
ls-840's. You don't see other 600MHz pnp's with
good voltage, let alone good current gain.

I sure wish the heterostructure parts would go
commercial...

Anyway, a price list would sure be nice.

I Requested a price list so when i get it ill email you a copy.I also like some of there BJT's but i use the dual PNP/NPN Ls-312 & LS-352. These are only 200 MHz. But real low noise and High gain like over 300@ 1Ma Tested. When I am done with my Opti-Fet headphone amp that uses these as well as the LS-840 Ill email you a copy. So far the PS is done and sounds Killer on the Gilmore amp i am using to tweek the Regulator circuit. Boy after listening to this I may never go back to IC reg';s again Night and day difference. IC regulators are still in the dark ages. I think evean a simple Series pass transistor and zener reference sounds better than any IC reg evean the Linear tech types. and i doo tweeks to designs using these most people dont Bother with, these tweeks improve the sound of the IC's alot But still way short of a discreet component design, but i gess there is only so mutch you can do with a sow's ear.