Graham's Class A/JLH output

[jcx]

Graham Maynard presented a jlh output class A amp in the september Electronics World

I’ve abstracted the input as an ideal gm vccs and used ideal current sources to look at the output stage operation –

http://www.zero-distortion.com/tests...ertrans_05.htm

suggests 2sc3281 is similar to Graham’s 2sc5200 output devices, I was able to find onsemi spice model for mj3281 and already had fairchild bd139 model for a driver



(despite the LTSwCad file header names this is the mj3281 sim)

my question/issue is the poor current division as Vce gets within 5 V of the rail with the 200 V device, the ancient 2n3055 actually looks much better in this regard (i1 has to go up to ~150mA due to the low hfe of the 3055 model in LtSwCad) – so is this real or modeling error?

Given that spice transistors are perfectly matched and only deviate due to differential biasing this is probably optimistically good current division so another question is how can the jlh deserve its reputation when built by diyers lacking power curve tracers to match the output devices?

And finally the sim offers an example of how to step a parameter in LtSwCad Spice for mikes

Added spice directive:

.step param A LIST .17 .35 .7 1.4

V5 input voltage source amplitude defined with {A} as parameter which is stepped from LIST in .step directive

SINE(0 {A} 2K)

see LTSpice asc file:

[Geoff]

jcx

I have found similar anomalies when simulating the JLH. Certain power transistor Spice models give device current results that are not what I would expect. So far, I have put this down to errors in the particular Spice models, but possibly someone more experienced in these matters than myself could advise whether the simulation is in fact reflecting reality.

Fairchild second source the 2SC5200 (KSC5200) and a Spice model is available at their website. In case you wish to run the simulations again with a different model I have added it below.

Geoff


* KSC5200 NPN EPITAXIAL SILICON TRANSISTOR
*---------------------------------------------------
* AUDIO POWER AMPLIFIER
* High Current Capability(Ic=15A)
* High Power Disspation
* Wide S.O.A
* Complement to KSA1943
*---------------------------------------------------
.MODEL KSC5200 NPN (
+ IS =4.3031E-12
+ BF =152.1
+ NF =1.0
+ BR =6.155
+ NR =1.028
+ ISE =1.3924E-11
+ NE =1.5
+ ISC =2.7542E-11
+ NC =1.95
+ VAF =60.0
+ VAR =6.51
+ IKF =10.8637
+ IKR =0.1585
+ RB =2.47
+ RBM =0.02
+ IRB =0.08
+ RE =0.04
+ RC =0.015
+ CJE =5.8111E-9
+ VJE =0.6506
+ MJE =0.3357
+ FC =0.5
+ CJC =6.4394E-10
+ VJC =0.5
+ MJC =0.3966
+ XCJC =0.7624
+ XTB =1.0445
+ EG =1.1663
+ XTI =3.0 )
*-----------------------------------------------------
* FAIRCHILD PUCHUN S.KOREA CASE: TO-3P PID:KSC5200
* 2000-03-17 CREATION

[jcx]

thanx Geoff

the KSC5200 model is only slighty better, the current still flattens out at ~ 2.9A (but without the alarming "foldback" visible in mj3281 sim)

i think this is a characteristic of high Vce transistors, their hfe vs ic looks really flat above ~10 Vce bias but the "saturation region" departure from the Early Voltage controlled region starts at a much higher Vce than transistors doped for low Vce operation

[Graham Maynard]

Hi jcx,

Interesting analysis.

I was castigated in another thread for recommending a minimum 5volt bipolar headroom for quality amplification - this from 'hands on' measuring experience, not simulation.

150mA for 2N3055, something wrong somewhere ?
100mA with triple 2N3055 will give reliable 50W - 8 ohm in a real life JLH.

Circuit enclosed.
Write-up is in Sept 2004 Electronics World.
Rb - start at 270 ohms.
Rz approx 470 kilo-ohms for very accurate output zero; set up after half hour. May be omitted.

In my amplifier I ran 2.2A to 2.4A quiescent current, maybe if this were sim'd that peak current 'suckout' would not appear to be so pronounced.

Each output half is a common emitter gain stage, where gain is related to V-ce; thus drive capability falls as each device conducts whilst the overall push-pull gain remains good; this is not like a push-pull complementary emitter/source follower which has built in voltage feedback.

Cheers ......... Graham.

[Ejam]

Graham

I read your article and like the amp. I am from the valve school having build many amps but must confess ignorance when it comes to transistors. I would like to build a 10 - 12W version of your amp similar to the original JLH version. I have very efficient speakers and a small room and no need for the heat of the 25W version. My 8W 300B amp is more than enough. I would think that the rails of +/- 15V would be in the ball park. Can Tr5 be omitted if the amp rarely gets to maximum output ? Hope you can help with some suggestions.

Regards

[jcx]

while the stepped plot is very busy, you can read the ie(Q3,4) off the right axis, ~ 2.3A for the mj3281 cir, in the 2N3055 sim i used a model that Lt supplied (manufacturer "ST") the 3055 seems to be operating at ~ 30 hfe @ 2.2 A ie

i won't add much to this discussion over the next few days...

[Graham Maynard]

Hi jcx,

I saw the green trace and read 1.5A when I was replying. I became confused by your simulated illustration. I always run current and voltage separately one above the other to avoid such confusion. I had actually noted the current at 2.3A but my reply was time constrained, and I made a mistake.

If you trace across from the voltage curve at the same vertical time point where the output current starts to double peak, then your simulator is reducing MJL3281A gain below V.ce = 10 to 12V.
Which I feel is too high for a *hot* running class-A bipolar output device.

In real life the amplifier makes loudspeakers generate excellent audio with especially clean treble definition and thus image accuracy and depth. There is no indication of the output device current sharing imbalance that does arise, even with the JLH types.

The loudspeaker is driven by the difference between class-A output half device currents, and whether these are equal plus opposite, or forward or reverse curvilinear is of little consequence if they behave similarly with equal but opposite input current drive.
__________________________________________________ _

Hi Ejam,

It has always been my experience that valve amplifiers, including SE, 'sound' twice as powerful as resistor measurement rated transistor designs, so I would not recommend you going less than 20V rails. Of course TR5 can be omitted, and the circuit can be knocked up using whatever is to hand to see if you like it before committing to purchasing new parts.

Cheers .......... Graham.

[Geoff]

Originally posted by jcx
i think this is a characteristic of high Vce transistors, their hfe vs ic looks really flat above ~10 Vce bias but the "saturation region" departure from the Early Voltage controlled region starts at a much higher Vce than transistors doped for low Vce operation

Unfortunately, I don't think this is the answer. ON-Semi have now added models for low Vce transistors (eg 2N3055, TIP35C) and these display the same anomalous current waveforms as shown in your original post.

I have tried to pursue this further during the past couple of days. Every ON-Semi model I have used (MJ21194, MJL21194, MJL3281A, MJL4281A, MJ15001, MJ15003, MJ15022, MJ15024, 2N3055, TIP35C) gives similar results, to a greater of lesser extent, to that shown on your graph.

None of the models I have obtained from other sources (2N3055, TIP3055, 2SC3281, TIP35, KSC5200, MJ15003) display this trait, there is merely a flattening of the curve as the current rises.

So, is this a problem in reality or merely in simulation? Are the ON-Semi pspice models more accurate than the others (they certainly have more parameters) or are they all erroneous in some way?

Unfortunately I do not know enough about the intricacies of transistor operation, or spice modelling, to even start to come up with an answer.

[andy_c]

Originally posted by Geoff
Are the ON-Semi pspice models more accurate than the others (they certainly have more parameters) or are they all erroneous in some way?

I haven't looked at the large-signal parameters of these devices, but at one time I looked into the correlation of ft vs current in simulation vs the data sheet values. The models showed an ft vs current that bore little resemblance to the data sheet values, and at the typical current levels for class AB biasing were low by a factor of six. The thread that has the plots is here Evaluation results: MJL3281A/MJL1302A SPICE Models I tweaked the zero-bias Cbe and the ITF parameters to get a better match with the data sheet, but I still wasn't able to match the curve at high currents. This turned out to be a very time-consuming process, so I didn't pursue it further. I began to believe that On Semiconductor might have some kind of systemic problem with SPICE modeling, but the models of the MJE15030/31 show almost perfect correlation of simulated ft vs current to the data sheet values. I don't know what went wrong with the modeling of the high-power devices. A comparison of a parametric sweep of IB and VCE to the data sheet characteristic curves might be interesting.

[lumanauw]

Hi, Graham,

Nice Design. But some things I dont understand. TR2 and TR5 is forming totem-pole. Is that the bias is set by the 390 R, instead of putting VBE multiplier in the middle?