If I put my notes here, I might be able to find them again later!
Szekeres 2015
Posted 20th May 2015 at 06:00 AM by rjm
Updated 18th June 2015 at 11:26 PM by rjm (added schematic of original version)
Updated 18th June 2015 at 11:26 PM by rjm (added schematic of original version)
The circuit was originally hosted on Headwize, but the site seems to have gone offline.
It was a single stage resistively-loaded MOSFET follower, a unity gain current buffer used to drive headphones.
Some updated versions provided below. As noted in the comments the "Reverso" version with the CCS on the V+ and a p-channel mosfet has better PSRR performance, especially with voltage divider network R6,R7,C4 on the collector of Q2.
So good in fact that I switched around the n-channel version to use a negative voltage rail to obtain the same result!
It was a single stage resistively-loaded MOSFET follower, a unity gain current buffer used to drive headphones.
Some updated versions provided below. As noted in the comments the "Reverso" version with the CCS on the V+ and a p-channel mosfet has better PSRR performance, especially with voltage divider network R6,R7,C4 on the collector of Q2.
So good in fact that I switched around the n-channel version to use a negative voltage rail to obtain the same result!
Total Comments 14
Comments
-
It would be interesting to compare the sound between returning the headphone 'ground' side as shown and then to the +ve rail so the rail draw becomes constant and the FET is acting as a 'shunt' amp.Posted 25th May 2015 at 07:09 AM by abraxalito
-
With the circuit as presented above, returning the signal to the positive rail degrades the PSRR.
I imagine with a bit of tweaking of the current source you might get it to work though equivalently though.Posted 26th May 2015 at 12:43 AM by rjm
-
PSRR is only part of the equation - the other half of it is how much signal-correlated noise gets induced on the rails by the load current variations in the first place. PSRR is only needed to compensate for noisy rails. With the signal returned to the positive rail, no load-induced variation will occur on the rail other than imperfections in the current source, which can presumably be compensated for with additional devices.Posted 26th May 2015 at 02:02 AM by abraxalito
-
The variation in supply voltage with load current depends on the power supply impedance. Properly bypassed, there is little impediment to the return of load currents from circuit common to the supply, nor are the load currents (headphones!) so large as to be a concern. Not enough that I'd toss away 10-20 dB of PSRR for at any rate.
In the spirit of experiment however it's an interesting notion.Posted 26th May 2015 at 03:01 AM by rjm
-
I recall there was one such follower in wireless world - very popular... The usp was its reference to the positive rail and the shunt action of the amp... Can't remember what it was called thoPosted 27th May 2015 at 12:59 PM by kasey197
-
The 'reverso's PSRR could probably be improved still further with Elvee's mod to the current source - adding a resistor to cancel out the supply voltage variation of the biassing current. I seem to recall Elvee said 20dB improvement was possible.
I suspect you're getting much better PSRR with the P-channel FET because the source follower doesn't do too well on PSRR, a CCS has the potential to do much better.
A listen to both side by side would be very interesting.Posted 27th May 2015 at 01:57 PM by abraxalito
-
The improvement was so dramatic I've replaced the schematic above with the modified CCS.Posted 28th May 2015 at 12:28 AM by rjm
-
ah here it is :
the Power FolllowerPosted 28th May 2015 at 01:42 AM by kasey197
-
In pursuit of even more PSRR you could also try changing Q1 to 2SA1381 - Bob Cordell has the model on his website. It has better VAF than the BC327-25 I seem to recall. Downside is its only good for 100mA so two would need paralleling for higher biasses which will negate some of the VAF advantage.Posted 28th May 2015 at 01:45 AM by abraxalito
-
@kasey197 I uploaded my version a split second before you posted that! It matters not though: we arrive at the same point. Good.
@abraxalito In a practical build it must be something bigger than the BC337/327 at any rate.Posted 28th May 2015 at 01:50 AM by rjm
-
Sorry, yes I just realized I hadn't read the small print on your schematic. The BD136 data from Siemens looks quite decent for VAF - probably no advantage going to the 1381. There does though look to be a 'quasi saturation' region below 10V for VCE where the output impedance reduces. Above 10V all looks dandy for IC up to 200mA. I wonder how much variation there is between manufacturers in this though - NXP doesn't show any data like this in their DS.Posted 28th May 2015 at 01:59 AM by abraxalito
Updated 28th May 2015 at 02:03 AM by abraxalito -
Great and nice stuff on this blog - many thanks for posting
Posted 28th May 2015 at 07:12 AM by kasey197
-
Here's what happens to PSRR if you re-reverse the Reverso:
You could improve things again (plus general performance) by reducing the value of R1, the input series resistor, but it would still remain a fair bit worse, and maximum PSRR obviously remains limited by the MOSFET's gain.Posted 15th January 2016 at 09:58 PM by sgrossklass
-
That looks promising! I was thinking that split supplies could be sued to good effect, if care was taken to ensure the output coupling cap couldn't be reverse biased...
Unfortunately the text on your posted image is too small to read clearly, but I get the general idea.Posted 4th February 2016 at 12:55 AM by rjm
