[kipman725]

Hello I have designed a follower to operate from a single supply for the driving of headphones. It has a JFET input stage for high input impedance (aprox 1Gohm without the divider) and bipolar output stage for low output impedance (aprox 4.7ohm - 0.01175ohm).
The matched JFET current source helps ensure linearity as both JFETS should have the same Vgs and as they are matched the same Ids. The resistor was selected from the JFETS data sheet to ensure a reasonable voltage drop across it to improve linearity without presenting a too high impedance source for the bipolar follower.
The output stage has a current source aprox 100mA, this is a little high as it would allow peaks of 30V! into a 300ohm headphone such as the HD650 (greater than the supply voltage) but it was primarily set at this level for 30ohm headphones like grados. If you only planned on using this follower with higher impedance phones defiantly reduce this.
The input capacitor has been set to 10x the required value because 1uF non-electrolytic are easy to find and the capacitors with worst distortion are ceramic discs, using a 1uF cap stops people using them. The output capacitor has been sized with grados in mind but it should help prevent bass rolloff and isn't excessive.




Anyway, I hope this is usefull, please tell me what you think and if you have any improvements you could suggest. Should be pretty cheap to build as in the UK the actives come to under £2. Quite simple to run off a split supply (removing the input cap) (raising the input impedance massivly in the proccess) if that is desired, however the output coupling cap is needed!

(PS can anyone recommend a good SPICE tutorial as I would like to simulate THD+N etc)

 

[kvant]

Looks quite similar to the "JFET-MOSFET Headphone Driver" shown in Headwize library. Your schematic here is somewhat prone to power supply ripple leaking into the JFET's gate due to the simple divider biasing. You might want to check the Headwize article for improved biasing strategy that includes additional RC filtering.

As for the SPICE, I find eCircuit Center to be a useful site. They have a lot of examples in their circuit collection.

EDIT: One more thing... 2sc2910 is not the right transistor for the duty it is asked to perform here. You want 100mA through it at some 9V, which is 900mW of heat dissipation. If I read the correct datasheet, the 2sc2910 maximum is 70mA and 900mW (at room temperature!). You will want something in TO-220 package, or at least TO-126. These, however, will have lower current gain and might load the JFET more than you would like to.

 

[kipman725]

the headwise project is interesting it apears to be mitigating the effect of the JFETS input capacitance mainly, however the jfet I plan on suing only has a very small capacitance and so that will not matter at audio frequancies and reducing the effects of supply ripple. However using a mossfet like that is a recipe for distortion and the JFET folower is a worse design.
Improved biasing is definatly a must though I will look into it. Thanks

*the 32ohm resistor on the output is a dummy load I was just using the check it worked... probobly replace with about 10K.

 

[kipman725]

well I have made the improvements you suggest kavant and also raised the input impedance at signal frequancies. Thanks for the heads up about the transistor. Is there any disadvantage to using a darlington in this aplication? (it was the only way I could see of getting enough gain and remaining inside the transistors SOA).

 

[kvant]

A few more thoughts...

Quote:

Originally Posted by kipman725 /img/forum/go_quote.gif the headwise project is interesting it apears to be mitigating the effect of the JFETS input capacitance mainly...

Not exactly. It mitigates the MOSFET's input capacitance by means of the extra JFET buffer.

Quote:

Originally Posted by kipman725 /img/forum/go_quote.gif Is there any disadvantage to using a darlington in this aplication? (it was the only way I could see of getting enough gain and remaining inside the transistors SOA).

I believe that darlington is fine, at least as long as it is not switching on and off, which could lead to incoherence between its constituents. But here we are in class A, thus no problems of this kind. Anyway, I don't see too many simple alternatives here - the other one is MOSFET, but you object to its somewhat higher distortion.

What is the function of C4 in your last schematic?

Does the JFET input stage do any good here? If you drop it, the distortion lowers, I guess, and the input impedance with a darlington alone is not that bad either. Let's say you build it of BC560C (hFE>400) followed by BD135-16 (hFE>100). With 32 ohm load, this should give something like 32x100x400=1.3 Meg, so that biasing is a bottleneck again. BTW, I am using a single-rail class-A diamond buffer as a stand-alone "amp", which is also a double BJT follower, and it drives Grado to my satisfaction (however, I have no tools to measure its technical performance).

 

[kipman725]

C4 is a bootstap capacitor that greatly raises the input impedance. voltage at the output of the JFET follower is aproximatly the input voltage (as it's a follower). At signal frequancies the impedance of C4 is low therefore the voltage accross R8 is aproximatly 0. IF the voltage accross R8 is 0 then the current that flows though it is 0 and the input impeadance is very large. The biasing network actualy has an input impedance of aproximatly 10Mohm.. I could use larger resistors aswell but I hesitate to do this due to increased resistive noise. I shall see if I can use SPICE to simulate this with and without the JFET input.

 

[kipman725]

Well I used the distortion tool of simetrix to get some prelim numbers with pesimistic source impedance (300 ohms from sound card) and supply impedance (1ohm).

JFET+BJTdarlington into 330ohms: 0.000810%
BJTdarlington into 330ohms: 0.000563%
JFET+BJTdarlington into 32ohms: 0.000729%
BJTdarlington into 32ohms: 0.000485%

personaly I think these results are petty meaningless and just show that all situations are pretty close to one another. The JFET does have the advantage of higher input impedance though.

*there has to be something suspect about these results as they are better than the B22 (although those are real life results).

 

[kvant]

Quote:

Originally Posted by kipman725 /img/forum/go_quote.gif *there has to be something suspect about these results as they are better than the B22 (although those are real life results).

One thing I find strange is that your THD figures don't increase with load. I played with it a bit and I am getting (at 1kHz and 1V peak on input):

0.065% into 32 ohm
0.0037% into 330 ohm

This is for your schematic with JFETs included. Output stage is running those cca 100mA.

 

[kipman725]

Well I built it and powered by my 3A 18V supply it goes very loud before any distortion kicks in. I made a few changes to the scematic; Increased 8ohm bias resitors to 10ohm due to lack of 8ohm resistors (bias is still overkill), increased the 1uF caps to 2.2uF as I could get them for almost the same price, also bumped up output capacitors to 940uF and added bleader resitors (10K) to ground from the output.
I am not good at making subjective mesurments of audio quality but it sounds better than the PEcongress power amp I usualy use probobly because of lack of bass rolloff.
Although I did encounter problems connecting the amp to my computer as I got a pretty horiffic ground loop, I have changed my power supply so that the output ground is connected to mains ground now which should fix it. If that dosen't sort it I could build a power supply that runs off my computers power supply, although it should fix it as my other amps have no problems. I am a little concered that there may be something dangerous with my power supply however as I am pretty sure there was a blowing up semiconductor kind of bang from it at one point, however it apears to be operating corectly so it could have just been my fears of mains powered equipment (weird that I always end up tinkering with it). I am currently using a 3A slow blow fuse in it which is way overated but the only way I can avoid the inrush current from blowing the fuse. is there some kind of simple inrush limiter I can add as I would prefer to use a lower fuse value?

 

[kipman725]

The ground loop was fixed by making the power supply gorund = mains ground and it now works fine with my pc. Must have been my over active imagination with the the explosions. My sister says a bird flew though the house so I may have in my concentration interprited that as an explosion.

 

[cotdt]

nice project. sadly, it seems in a few years no one will know what a JFET is anymore.

 

[nikongod]

Quote:

Originally Posted by cotdt /img/forum/go_quote.gif nice project. sadly, it seems in a few years no one will know what a JFET is anymore.

why?

 

[cotdt]

Quote:

Originally Posted by nikongod /img/forum/go_quote.gif why?

Fewer and fewer amps seem to use them JFETs these days. A few of them use them in the input stage but that's about it. I'm interested in seeing JFETs used in the output stage of an amp because of its triode-like region in its curve at low gate voltages. Should make for an interesting tube-like sound.

 

[kipman725]

wow holy thread revival batman!, thanks for the nice comments
I am now a first year EEE student at imperial so I shall defiantly be carrying on designing circuits using JFETS. Anyway what I'm thinking of at the moment is an amplifier with no analogue amplification at all directly converting PCM into a delta sigma bitstream and then using that to drive some power fetts resulting in ultra high efficiency. It should be quite simple to program an FPGA to do this. Only issue I have come across so far is that there is no way I can think of without using an ADC of taking feedback from the final output stage and crunching the maths for higher than 2nd order modulators is hard. I'm tempted to just program the FPGA without doing the maths and just seeing if it's stable.