Leach Amp different emitor resistors

[Borc]

I intend to change emitor resistors on output devices from default 0,33 Ohm to 0,22 Ohm (Mills group buy). I removed protection circuit so this is not an issue.
I would like to have some help on the new optimal bios for output devices. Is there any mathematic formula to determine it.
For the same voltage drop on emitor resistors this will be 67,5mA
(from default 45mA), but I think this is too much.

Another Q.
Anybody tried with 1 uF capacitor across common emitor resistor on drivers. Any changes?

Regards

 

[peranders]

0.22 ohms is still quite reasonable.

One method of testing cross-over distortion is to use a 2 channel oscilloscope. Channel 1 at the bases of the output stage (doesn't matter which side, NPN or PNP) and channel 2 at the output. Put the two traces on top of each other. Adjust so the output signal is 1-3 V rms and the frequency 10 kHz. Trim so the channnel 1 looks as much as possible like channel 2. Very easy test with simple means. You also use your ears but then I'll guess the test frequency must be 1-3 kHz, so you can hear any harmonics. Cross-over distortion sounds like a busted speaker (burned voice coil).

 

[Borc]

Thanks Peranders, interesting method!
If i understand correctly:

1. - CH1 HOT to emitor of driver (as I'm without base resistors on output devices) on PNP or NPN side and CH1 COLD to ground
2. - CH2 HOT to the output and CH2 COLD to ground.
3. -Set the CH2 sinal to 1-3V rms and 10 kHz
4. -Trim Iq, so the signal between bases is as much like the output signal.
5. - output load is connected (can be pure R or must be speaker ?)

Regards

 

[peranders]

Both probes (recommend this) grounded.

CH1 at the output of VAS stage, better because you see more distortion but this is a sensitive node(!), a probe is strongly recommended and you may experience disturbance (hopefully not).

The result is better with a real load, at least a pure resistor, but I suspect a speaker is better. This type of distortion is strongly dependent of the load.

When the two curves are alike the distortion is less than a couple of 1% and this is before feedback. If you apply 30 dB (or whatever you have) feedback the distortion (in theory) will be 30 times lower.

A capacitor across the emitter resistor of the drivers can be good because you create low impedance in both directions which is good when is comes to turn-off times of the main power transistors, even more important if you have MOSFET's. I would choose 100 nF, which good HF properties. 100 nF/63 V polyester is quite alright to use.

 

[Borc]

Thanks,
Procedure corrected now.
Regards

 

[john curl]

Most of what is being said is nonsense. The optimum drop across a voltage driven complementary bipolar output stage is 15-25mV across each emitter resistor. Somewhat more or less will still work. Caps? Across .22 ohm?

 

[djk]

"Caps? Across .22 ohm?"

A cap across the emitter resistor for the drivers was suggested, it is more like 150R.

I first saw this done on the ML3.

 

[TonyTecson]

hi djk,
if i want to parallel say, 8 output power transistors, can i increase the emitter resistors from say 0.22, to .5 ohms, reason being a higher emitter resistor promotes better current sharing among the output transistors, please comment....thanks

 

[jam]

Mr. Curl,

Could you please explain your reasoning for setting the bias at 15-25 mls as being optimal.

Another question, would you attach the junction of the driver emmiter resistors to the output or use a single emmiter resistor for the drivers not hooked to the output.

Thank you.

Regards,
Jam

 

[john curl]

This is derived from an article published by HP in about 1971. This is the best area of operation for lowest distortion in the transition between class A and AB.

 

[DrG]

The optimum drop across a voltage driven complementary bipolar output stage is 15-25mV across each emitter resistor.

Why would the voltage drop over the emitter resistor have an optimum? Surely the bias current itself has an optimum, meaning that the voltage dropped across any emitter resistors will vary according to the resistor value?

Could you explain, since the article you refer to is probably inaccessible to most...

S

 

[dimitri]

Another question, would you attach the junction of the driver emmiter resistors to the output or use a single emmiter resistor for the drivers not hooked to the output.

jam, may I answer?
A single emitter resistor is much better for two reasons:
- the output device (base or Cgs) will discharge faster
- the driver transistors will operate in class A, while when the junction of the driver emmiter resistors is attached to the output
the driver transistors will swith off

 

[peranders]

Mr Curl, I have also seen somewhere about the bias current but 1971 is a long time ago. Had this investigation something to do with the transistor types available at that time or was in general?

 

[JensRasmussen]

Optimum Q current

This is derived from an article published by HP in about 1971. This is the best area of operation for lowest distortion in the transition between class A and AB.

But Slone and Self have shown that the optimum class (when not making a class A output stage) is B and not A/B.

I agree with DrG that the voltage measured across an emitter resistor not is the way to determin weather or not the optimum standing current in the OPS has been found.

\Jens

 

[millwood]

Re: Optimum Q current

But Slone and Self have shown that the optimum class (when not making a class A output stage) is B and not A/B.

\Jens

that's pretty much in semantics. a lot of the times, when people say class B, they actually mean conducting slightly less than 180 degrees (or really class C), and class AB means conducting exactly 180 degrees via a bias current (which is really class B).

Class A/B should have meant a amp that was working in class A at small signal (conducting during all phases of the signal).

 

[jam]

Dimitri,

Thanks for your answer. It makes a lot of sense and I suppose this will also reduce crossover distortion.

Regards,
Jam

 

[john curl]

Folks you are just going to have to 'trust me' on this. It comes from a computer simulation that gives a range of possible values. The optimum value is between r(e) and r(e)/2, R(e)=1/G(m) of the transistor.

 

[Cobra2]

interesting....

What about MosFet-amps?

Arne K

 

[peranders]

John, what do you mean by optimum? Does the cross over distortion increase when you turn up the bias?

 

[andy_c]

(...)Does the cross over distortion increase when you turn up the bias?

Yes, the crossover distortion does get worse if you increase the bias too far. This can be seen with simulation using LTSpice.with DC sweep. One can set up a simple complementary output stage with voltage drive and sweep the input voltage from -5V to +5V, plotting Vout vs Vin. Then using Helmut Sennewald's utility from the Yahoo LTSpice users' group, the LTSpice binary output file can be converted to text. Then the text can be brought into Excel. Using Excel, it's easy to do a linear regression on the data of Vout vs Vin. Then the actual Vout data can be subtracted from the best-fit line of Vout vs Vin so the deviation from linearity can be seen at a very high resolution. By doing this at a number of bias currents, it can be seen that there's a "sweet spot" in the bias for which the deviation from linearity is minimized. I did this using the MJL3281A and MJL1302A models with 0.33 Ohm emitter resistors, same as the Leach amp. The optimum bias current for an 8 Ohm load ended up being 40-45 mA. This is the same as the result Leach obtained on the bench using a distortion analyzer. Another interesting result is that if you parallel multiple output devices, the crossover distortion gets less. This has been observed by Self as well.