Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › Dirt Cheap Stax Amp DIY - new schematic updated!
New Posts  All Forums:Forum Nav:

Dirt Cheap Stax Amp DIY - new schematic updated! - Page 4

post #46 of 99

the Tube Cad articles on SRPP do mention padding the cathode with additional R, there is a continuum from ccs to "optimum push-pull" to "mu follower" and the articles describe SS versions as "SRPP derived"  - clearly there's a strong similarity in operating principles - but it does seem like "SRPP" label is pushing the wrong buttons for some of you

 

http://www.tubecad.com/2009/09/blog0171.htm

 

with depletion MOSFET having high gm R padding is needed to get the "optimum" "push-pull" relation at a specified (low) current - looking at the circuit as a modulating current source is useful - "impedance multiplier in the blog

 

as I pointed out above a possible use of the depletion mode MOSFET is to cascode another current regulator circuit - Walt Jung audioXpress ccs articles shows some options - should give circuit designers some ideas

 

tapping the current sense resistor to get the push-pull current modulation "impedance multiplication" is applicable to several of the ccs circuits, the stability and TC of the set current, accuracy of push-pull matching can be improved, made nearly independent of the depletion cascode Vth

 

 

while I haven't measured my test circuit that carefully - sims with a square law gm model extracted from real parts measurements show the modulating current source simply looks like the load Z has been reduced by 1/2 to beyond 1 MHz

 

 

and I don't see the justification for such blunt dismissals of a output stage circuit on grounds of its "sound" - seems too simplistic

 

if a circuit can drive the load reasonably linearly over the range of expected signal then the rest of the amplifier circuit can control it to varying degrees of accuracy via global and/or local feedback, "distortion compensation" potentially give a range of "sound" - or preferably just be "transparent"


Edited by jcx - 3/28/13 at 9:37am
post #47 of 99
Thread Starter 

special thanks to Frank. He sent me an STAX transformer amp ( SRD-7), still in excellent condition and I really want to thank him for his generosity. I will do some tests with it( loaded and unloaded and see how it performs, obviously I only have one TA2020 amp to drive it, not great power amp, but at least I can use the oscilloscope to compare the input vs. output.) I will post pictures before I tear it apart to build the dirt cheap amp in it. 

post #48 of 99
Thread Starter 
Quote:
Originally Posted by kevin gilmore View Post

the schematic posted above is that of the sennheisser hev60.

Most people who have listened to that amp don't like it.

 

Its only SRPP if the load impedance matches the cathode resistor.

In this case the source of the top fet. Therefore its not SRPP but

rather constant current.

 

In fact, due to C20 its more like a white cathode follower done

with solid state.

with R43 and R44, C20 and C20, if the output point is at the negative end of C20, then we have a CCS loaded circuit, not SRPP. But since the output is taken between R44 and R43, I still think this is a SRPP, but with R43 and R44, I guess the amplification is more linear? I dont really quite understand how the two resistors work together. 

post #49 of 99
Thread Starter 
Quote:
Originally Posted by jcx View Post

the Tube Cad articles on SRPP do mention padding the cathode with additional R, there is a continuum from ccs to "optimum push-pull" to "mu follower" and the articles describe SS versions as "SRPP derived"  - clearly there's a strong similarity in operating principles - but it does seem like "SRPP" label is pushing the wrong buttons for some of you

 

http://www.tubecad.com/2009/09/blog0171.htm

 

with depletion MOSFET having high gm R padding is needed to get the "optimum" "push-pull" relation at a specified (low) current - looking at the circuit as a modulating current source is useful - "impedance multiplier in the blog

 

as I pointed out above a possible use of the depletion mode MOSFET is to cascode another current regulator circuit - Walt Jung audioXpress ccs articles shows some options - should give circuit designers some ideas

 

tapping the current sense resistor to get the push-pull current modulation "impedance multiplication" is applicable to several of the ccs circuits, the stability and TC of the set current, accuracy of push-pull matching can be improved, made nearly independent of the depletion cascode Vth

 

 

while I haven't measured my test circuit that carefully - sims with a square law gm model extracted from real parts measurements show the modulating current source simply looks like the load Z has been reduced by 1/2 to beyond 1 MHz

 

 

and I don't see the justification for such blunt dismissals of a output stage circuit on grounds of its "sound" - seems too simplistic

 

if a circuit can drive the load reasonably linearly over the range of expected signal then the rest of the amplifier circuit can control it to varying degrees of accuracy via global and/or local feedback, "distortion compensation" potentially give a range of "sound" - or preferably just be "transparent"

I was thinking about this yesterday. and please help me out ( correct me if I am wrong)

 

- Resistor load: most stable and sounds nice and linear, but with ES phones, high freq roll off is really some issue if no feedback is in place. 

 

- CCS load: I feel this one is very tricky to get right without DC feedback. a very small change in Ic the lower transistor( or tube in the schematic below) will cause a significant change in the DC output. I dont know if tubes are very stable or not since I never played with CCS for one of my tube amp I built years ago, but many people are using CCS load for tubes and sees to get very linear frequency response, however if the lower tube is changed to a transistor, the output looks very different, in stead of being linear, the output can get clipped very easily( with voltage swing from 0v to V supply), because of the difference in the V-I graph between tube and transistor. So its very tricky to use CCS load without global feedback. and similar things may also apply to SRPP. I feel its very hard to get a flat response with SRPP I build with NPN transistors, even though it has very low output impedance, but the gain factor changes so much when input frequency changes ( say under the say 0.1V input, I get +-50V output at 100hz to 1000hz, but the output rolls off to +-10V at 15KHZ, under a proto NPN SRPP circuit I built) 

 

 

post #50 of 99

What are the drive requirements here? I'm unlikely to ever own a pair of Stax phones, but so what...

post #51 of 99

at least 1600 volts peak to peak stator to stator.

 

2000 volts is better.

post #52 of 99

stators are pretty much always driven balanced with 2 driver stages per channel

 

the load seems to range from 70-120 pF by published ES headphone specs - as seen by each individual stator driver the load is effectively double if you calculate based on just that driver's output V since the other side is being driven in opposition

 

cable is significant part of the Cload, and extensions are sold - could add to the max load design target

 

the relation for current vs load is different than usual audio amps/transducers - at low frequency next to no current is needed so "big bass" only depends on Vswing

 

relatively high current is needed to charge/discharge the Cload at higher audio frequency - while we never listen to 20 kHz sine at full amplitude that could be one goal for setting max current output

 

piling together all worst case assumptions you may want 20 mA or so peak output current, practically anything over 5 mA should be quite listenable - most music has <5 kHz power bandwidth - even exotic moving coil phono cartridges won't track that high

 

for digital source, close miced, it is possible to record faster signals but rare slew rate limiting distortion is even less audible than rare clipping - which really isn't as easy to detect as most would think - it really is crossover/low level distortion that is clearly audible - and all direct drive ES headphone amps are at least "deep A" Class AB biased and avoid many low level problems

 

most project amps are Class A - simpler if power hungry - sliding bias or AB would be seen by most as unescessary compromise and circuit complication - but again the E9 energizer sold with the Koss ESP-950 is listenable, the electrostat quality shows thru despite the things the E9 amp "does wrong" by audiophile standards


Edited by jcx - 3/29/13 at 8:37am
post #53 of 99

Thanks jcx, my next question was going to be what kind of a load do the phones present...

post #54 of 99

driving the feedback R can be an annoyingly large fraction of the output current budget - but again doesn't impact low level linearity which is most obviously audible

 

with ~ 100 dB SPL @ 100 Vrms sensitivity spec few hundred kOhm feedback R are drawing less than 1 mA @100 dB - but the feedback R could be soaking up 1/2 of the current at higher 500 Vpk (one side) and low bias 5 mA SE Class A

post #55 of 99
Thread Starter 
Quote:
Originally Posted by jcx View Post

driving the feedback R can be an annoyingly large fraction of the output current budget - but again doesn't impact low level linearity which is most obviously audible

with ~ 100 dB SPL @ 100 Vrms sensitivity spec few hundred kOhm feedback R are drawing less than 1 mA @100 dB - but the feedback R could be soaking up 1/2 of the current at higher 500 Vpk (one side) and low bias 5 mA SE Class A

So what about this: my schematic in the first stage and a ccs loaded emitter follower in the second stage, I can do 1ma at first stage and 5ma in second stage. That comes to 24ma in total, with + - 300v supply we are looking at 12w in total that will give 1200v pp output with a very low output impedance and no feedback resistor. Damn now I need a heatsink.
post #56 of 99

post #57 of 99
Thread Starter 
Quote:
Originally Posted by tyre View Post


This one does not have feedback. With load resistor as high as 60k, a significant roll off will happen beyond 15khz
post #58 of 99
Quote:
Originally Posted by wdiabc View Post


This one does not have feedback. With load resistor as high as 60k, a significant roll off will happen beyond 15khz

 

It's just a derivation of an amp from tubecad. The load resistor value could be adjusted.

 

http://www.tubecad.com/november99/page9.html

post #59 of 99

The SRM-T1/T1S/006T all use 66k for plate resistors with the 6CG7.  SRM-600 uses 60K and the 007t uses 47K.  The latter has been the "golden standard" by Stax since the 60's though. 

post #60 of 99
Thread Starter 

Today I come up with this circuit, not really SRPP, not really a cathode follower. its a hybrid, dont know what to call it, but it is now my newer version of the dirt cheap amp. 

please see the attached circuit. 

 

New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › Dirt Cheap Stax Amp DIY - new schematic updated!