Bob , I have a "apples and oranges" question. In my final choice of devices for your triple output stage ("DBT") , I have 2 choices. #1 is to keep the same (1381/3503 - njw0302/0281 - njw0302/0281) as my pre-driver-output choices. All these are 30 mhz + Ft devices and I have run them successfully in the triple.My other choice is the NJW21193/4 ,they are only 4 mhz w/ much higher SOA and 60 Hfe .... just as cheap , (2$) at mouser.
My real question is 2 part , would it be detrimental to have devices with "mismatched" Ft in a BJT triple ... for example (mje340/15mhz - njw0281/30mhz - njw21194/4mhz) pre-drv-OP ? And if you did use these would it increase the margin for stability beyond the triple with the faster Ft devices. I pondered this after researching the Leach amp with it's slow TIP's (10mhz Ft) and super slow to-3 MJ's (2mhz). My concern is with how I will implement it.
My voltage stage , predriver and final njw0281/0302 drivers will be on a small PCB and the final 3 pair OP devices , zoble , and OP inductor on a separate board bolted to the main heatsinks. These will be connected by 2 wires (+drive and - drive) to the driver/ VS board. Being of such low impedance and much shorter than a typical leach build (I have seen real long wires) , will I have problems with parasitics ?
Just bought a vintage altec lansing 2204 for US$20 (rare-below) to put my "bobamp" in , huge 800VA trafo and large heatsinks ... just have to choose my outputs and final layout.
OS
Hi OS,
In general, I strongly prefer the faster RET-like output transistors to the older 21194/21193 devices, and I don't think this choice will result in stability problems with the Triple with good layout and wise use of power supply filtering and base stoppers. I always suggest using transistors from the same complementary family for the PNP and NPN pairs in an output stage.
However, I don't see preceding the 21193/4 output transistors with faster drivers as a problem. I just prefer the faster output devices in general.
Of course, I am also a very big fan of the TermalTrak devices. Once you have mastered the art of biasing an output stage with ThermalTraks, you'll never look back.
Cheers,
Bob
Hi effebi,
That would be asking for trouble, big-time. It will destroy power supply rejection in the output stage, for one thing.
As you know, there is very little I like about CFP output stages. The only time I ever used a CFP in an output stage was when I had to in the bad old days when we had to go quasi-complementary - the dark ages in my opinion.
Cheers,
Bob
I suspected so. But would you mind to elaborate your concept for the less blessed people?
effebi
Last edited:
Thanks for the explanation, it makes a lot of sense.
However I do not consider the Quasi a "real" CFP ( of what I am, without a technical reason, a fan) but, just as you said ,a byproduct of the dark ages.
effebi
Last edited:
PET vs. RET ??
I know they are faster (RET - ring emitter transistor) , but are not as durable as a PET (perforated emitter transistor). The To-3 MJ's are PET's and sound great , actually .. I don't perceive any difference between PET/RET in the same circuit...with a light load (8R).
(rod elliot -Semiconductor Safe Operating Area ) with a light load. Driving a short or a sub , a PET in a triple OP might show some superiority. Rod says at relativity slow linear use it does not matter but the PET "puts a lot more emitter in a given piece of silicon" than a RET.
For a EF2 , RET's are the only answer .. PET's don't have the beta. (njw21193 = 40-50/ avg. and njw0302 = 90-110/avg. I sure the T-trak's are the same BJT die as the NJW's ( with an added mur120 ).
OS
I know they are faster (RET - ring emitter transistor) , but are not as durable as a PET (perforated emitter transistor). The To-3 MJ's are PET's and sound great , actually .. I don't perceive any difference between PET/RET in the same circuit...with a light load (8R).
(rod elliot -Semiconductor Safe Operating Area ) with a light load. Driving a short or a sub , a PET in a triple OP might show some superiority. Rod says at relativity slow linear use it does not matter but the PET "puts a lot more emitter in a given piece of silicon" than a RET.
For a EF2 , RET's are the only answer .. PET's don't have the beta. (njw21193 = 40-50/ avg. and njw0302 = 90-110/avg. I sure the T-trak's are the same BJT die as the NJW's ( with an added mur120
).OS
Diamond Buffer Triple
Gasp! thanks to OS I had a look on the DBT on the book. It looks a wonderful OP scheme. I just did not understood what DBT was meaning (we are not all acronymous-dependent as our US friends!).
I really believe that it solves elegantly the thermal compensation issues of the Triple. So with a Vbe double transistor stage fitted on one simple H/S together with the pre-drivers and drivers and a couple of diodes (either integrated or discrete) on the main H/S all should be done. Simple and robust.
effebi
Gasp! thanks to OS I had a look on the DBT on the book. It looks a wonderful OP scheme. I just did not understood what DBT was meaning (we are not all acronymous-dependent as our US friends!).
I really believe that it solves elegantly the thermal compensation issues of the Triple. So with a Vbe double transistor stage fitted on one simple H/S together with the pre-drivers and drivers and a couple of diodes (either integrated or discrete) on the main H/S all should be done. Simple and robust.
effebi
Like this ? :
NJW21194/3 - 4mhz - OP
NJW0281/0302 - 30mhz - driver
Ksc-a3503/1381- 150mhz - predriver (and VAS)
I am not trying to overcomplicate things , but with bob's "DBT", the driver/predriver Vbe vs. temp. should be the same (or close). With the ksa/c's and the NJW's on the same heatsink there is a slight positive coefficient (driver current increases slightly with temperature ). Mje350-340's and NJW's just about cancel out (0 coefficient) ,Vbe wise ... leaving just the OP's Vbe to compensate for. This is one of the main attractive features of this circuit that caught my attention in Bob's book.
Also , in it's prototype form , it (the dbt) works so well ... I actually might go all the way and use thermal track outputs to have the perfect BJT current stage (thermally and electrically).
I just wondered what effect the MJE340-50 would have on the whole "package" used as the predriver.
OS
You make it sound as if Bob invented the diamond output stage, here is an app note from 1979 about Nationals high speed buffers. Note the LH0002 schematic, and the LH0063 with 6000V/uS slew rate:
http://www.national.com/an/AN/AN-227.pdf
It is nice to see something different promoted around here but it is also good to know at least a bit about the history. The idea was used at least 20 years ago in audio power amps.
No , I said that reading bob's book brought it to my attention. I knew of the nakamichi's DB output stages and even discussed it to death on the forum. Bob's example and detailed explaination finally made me build one - this is good , as I will share the tested and perfected result back to the community. BTW , thanks for the App sheet - it will be assimilated
.Just about everything here at the DIYA SS forum is 20 years old , that is why I am reverse engineering every vintage amp I get my hands on and collecting every class AB power amp schema I can DL.
Adjust these old circuits to the newer, faster devices , a little DIY over-engineering and really have some great amps. One can even achieve predriver/driver "null" (0 Vbe) with the right combinations of devices and close physical thermal layout ("face to face").
OS
Last edited:
Yes , the new $30K luxman is just the 20 YO M-120a with more OP's and regulation ... "over-engineered" - better be , for $30K. New Sherwood "full discrete" receiver is basically a "blameless" with sankens. Most linear subwoofer plates use a symasym (hitachi app note) voltage stage run at high bias tied to a simple EF2. All the same , just with new parts.
OS
Hi Pete,
I applaud OS for his interest in the diamond buffer triple output stage and am humbled by his kind words for my discussion of it in the book. It really doesn't matter who invented what, and I certainly did not invent the diamond buffer. In fact, we first used it in op amps in the 1970's at Bell Labs when we developed the junction-isolated complementary BJT process for ICs.
What's important is that a good circuit gets explained and popularized. I also cringe a bit when a properly executed version of the old Linn circuit is referred to as the Blameless.
It is not a great technical leap to put a diamond buffer in front of a pair of output transistors and arrive at a nice Triple. I am, however, unaware of specific commercial power amplifiers that use this circuit. I'd like to reference these in the next edition of my book. Can you provide me those examples?
Thanks!
Bob
DBT amps...
The nakamitchi 620 is close (below 1). I simulated , but did not build it. It was not as strait forward in it's T- comp (D 003/004 and Q 009/010 are coupled to the main HS). It also is a voltage follower style current amp (buffer), but still a triple... with 100K+ hfe. As I was not yet skilled enough 2 years ago , I passed on it and did an easy EF2.
Bryston also has a similar idea , I found a thread on it here .. Discrete Buffer Design
Some sort of crazy CFP - diamond hybrid
Neither is as "slick" as the "DBT"
OS
The nakamitchi 620 is close (below 1). I simulated , but did not build it. It was not as strait forward in it's T- comp (D 003/004 and Q 009/010 are coupled to the main HS). It also is a voltage follower style current amp (buffer), but still a triple... with 100K+ hfe. As I was not yet skilled enough 2 years ago , I passed on it and did an easy EF2.
Bryston also has a similar idea , I found a thread on it here .. Discrete Buffer Design
Some sort of crazy CFP - diamond hybrid
Neither is as "slick" as the "DBT"
No , it isn't ... but to "popularize it" and create a stable , durable alternative to CFP's and EF2's for driving demanding loads is very much needed here at DIYA. I hate to see some paralleling 6 or more pairs of devices with an emitter follower (ef2) and calling it a high current , high power amp .
OS
Attachments
Last edited:
Hi Kevinh,
No, I haven't looked at them yet. Here I have to plead ignorance
.Cheers,
Bob
I know they are faster (RET - ring emitter transistor) , but are not as durable as a PET (perforated emitter transistor). The To-3 MJ's are PET's and sound great , actually .. I don't perceive any difference between PET/RET in the same circuit...with a light load (8R).
(rod elliot -Semiconductor Safe Operating Area ) with a light load. Driving a short or a sub , a PET in a triple OP might show some superiority. Rod says at relativity slow linear use it does not matter but the PET "puts a lot more emitter in a given piece of silicon" than a RET.
For a EF2 , RET's are the only answer .. PET's don't have the beta. (njw21193 = 40-50/ avg. and njw0302 = 90-110/avg. I sure the T-trak's are the same BJT die as the NJW's ( with an added mur120
OS
Hi OS,
I'm not an expert on different geometric device layouts, and I've generally just thought that OnSemi PET was their answer to the RET. I think in both cases the objective is to arrive at an arrangement that effectively looks like lots of smaller transistors in parallel (analogous in some ways to the HEXFET appproach for MOSFETs).
The 21193 and ThermalTrak die are almost certainly quite different, as the performance is quite different (huge speed advantage for the TT's). Take a look at the spec sheets.
Cheers,
Bob
Sorry , I meant the njw 0281/0302 (ret) not the njw21193/4 (pet). I have broken both with a hammer (burnt "victims" ) . The BJT die of a njw0281 = mjl3281 = njl3281 +diode (mur120 ?) , all 6.5mm sq. njw's have just a little less SOA with a smaller slug which makes them less thermally proficient.
You are right , the NJW21193 is a bigger die and looks different. also , has about 15% less SOA (smaller slug) than a MJL or MJ (most powerful).
OS
) . The BJT die of a njw0281 = mjl3281 = njl3281 +diode (mur120 ?) , all 6.5mm sq. njw's have just a little less SOA with a smaller slug which makes them less thermally proficient.You are right , the NJW21193 is a bigger die and looks different. also , has about 15% less SOA (smaller slug) than a MJL or MJ (most powerful).
OS
Last edited:
Bob I think you are only going to find 2, maybe 3 and all by the same manufacturer, namely Pioneer. The circuit was used by them in around 1983-1988 in their exclusive series of amps. One of the models is the M6.