Bob Cordell's Power amplifier book

[Rafael L]

Hi Rafael,

Your observation is correct, especially if the bias spreader is not much affected by the VAS current.

Yes, I usually like to use TO-126 for predrivers, as they can dissipate some power and they can be mounted to a heat sink or heat spreader.

Note that with a Triple, one can also choose to mount the drivers on the main heat sink and keep the predrivers on the board with their own Vbe multiplier transistor (as part of a two-transistor bias spreader) and have that transistor mounted back-to-back with one of the predrivers (instead of using a heat spreader to mount the two predrivers and the Vbe multiplier). This assumes that the two predriver transistors are at about the same temperature. The extra heat sinking effect of the Vbe multiplier mounted to one of the predrivers will introduce a little bit of error into this assumption.

Cheers,
Bob

Bob, thanks for the tip, will try with drives in even heatsink that output transistors and predrives mounted back-to-back with second VBE multiplier, so I keep the TO-92 as predrives.

I found it strange, you not recommend mount all the transistors (predrives, drives and output) in the same heatsink seems to me the simplest solution, you see any problem in this solution?

 

[michaelkiwanuka]

Hi Mike,

When you say that all current sources are feedback current sources, are you referring to, e.g., the emitter degeneration in a current source formed by a CE transistor whose base is fed a reference voltage, as from a zener?

Yes. A current source using a voltage reference such as a zener diode and a emitter resistor possesses shunt applied-series derived negative feedback, and is, therefore, a negative feedback current source.

In the two-transistor current source that you call a "negative feedback current source", the series derived feedback signal is amplified by the sensing transistor before being shunt applied to the current source transistor's base.

This amplification of the derived feedback signal greatly increases the output impedance of the current source, so much so that there is no need to use a cascode current source in its place.

 

[keantoken]

Same condition as before.
Ordinary CM THD20k=0.019543%
Improved CM THD20k=0.010530%
Wilson CM THD20k=0.021373%
Damir

This confirms what I said about Zcb not being doubled by the buffered CM. I didn't realize your degeneration was 1.3V - pretty huge. Vbe modulation is swamped so there would be no benefit from the Wilson mirror unless you needed lower degeneration. This applies whether cascoded or not.

 

[dadod]

is this an example of the Vsat models not being good, or that the vsat models are correct but that the device is not good as a mirror, or some other sim quirk?

Better current mirror does not mean lower distortion.

 

[AudioLapDance]

Hi Mike,

There is so much good stuff out there, from many different genres!

One of my favorites is a live cut of "Hotel California" by the Eagles. Another one is "Ghetto of my Mind" from Rickie Lee Jone's Flying Cowboys album. Yet another is "New Favorite" on a live Allison Kraus & Union Station album.

Cheers,
Bob

Hotel off Hell Freezes Over is a classic (two tone bass drum at beginning) but I just picked up Allison Kraus Live and love the slide guitar in "New Favorite". And I don't even like Country! Wow, serendipity! PS Did you hear the "it's not a race!" part!

So what are some specific things I should listen for in these two songs?

 

[keantoken]

Better current mirror does not mean lower distortion.

Well if what you mean by better is that it is more linear, then a better current mirror should mean lower distortion. But what kind of mirror is appropriate can vary dramatically depending on the application, so in every application the "better" mirror is different. It is easy to be overwhelmed by all the variables.

 

[Bob Cordell]

I understand that. My questions is : do you think that using diodes (same as transistor) as explaine will give the same performance as your circuit;

The reason why I am asking this is : I use the lme49810 as driver. It has two diodes in series in the vas and has a predriver on chip. This is very handy to make a triple because the predriver is on chip and the 2 diodes must be in very good thermal contact on chip and I hope mached to the predriver.
In this way, the vbe compensation to be implemented will be for the driver and output transistor only.

JPV

Hi JPV,

Thanks for clarifying what you had in mind. I think I also misunderstood your original question.

You are indeed correct about the diodes inside the LME49810. They are very convenient and will track the thermals of the on-chip predrivers very well. As you point out, the standard form of bias spreader can then be used to take care of the external driver and output transistor thermals.

Cheers,
Bob

 

[Zero D]

LTP

In some Amps i see the LTP like this, with a PreSet of around 500R to balance out discrepancies. Obviously a PNP version also applies.

Is this not a good idea, & if so why ?

 

[Bob Cordell]

Bob, thanks for the tip, will try with drives in even heatsink that output transistors and predrives mounted back-to-back with second VBE multiplier, so I keep the TO-92 as predrives.

I found it strange, you not recommend mount all the transistors (predrives, drives and output) in the same heatsink seems to me the simplest solution, you see any problem in this solution?

Hi Rafael,

There is no real problem with mounting the predrivers, drivers and output transistors all on the main heat sink, but it may get a bit crowded. I think that a key objective is that the predrivers and drivers be isothermal with the portion of the bias spreader that is for them.

Cheers,
Bob

 

[Bob Cordell]

Yes. A current source using a voltage reference such as a zener diode and a emitter resistor possesses shunt applied-series derived negative feedback, and is, therefore, a negative feedback current source.

In the two-transistor current source that you call a "negative feedback current source", the series derived feedback signal is amplified by the sensing transistor before being shunt applied to the current source transistor's base.

This amplification of the derived feedback signal greatly increases the output impedance of the current source, so much so that there is no need to use a cascode current source in its place.

Hi Mike,

Yes, the benefits of the amplified feedback current source are well-understood.

Guess those guys that design amplifiers with "no negative feedback" are not considering all those emitter followers they use .

Aaaahhhh, semantics.

Cheers,
Bob

 

[Bob Cordell]

Hotel off Hell Freezes Over is a classic (two tone bass drum at beginning) but I just picked up Allison Kraus Live and love the slide guitar in "New Favorite". And I don't even like Country! Wow, serendipity! PS Did you hear the "it's not a race!" part!

So what are some specific things I should listen for in these two songs?

One of the things in Hotel is how the low end of the system handles those low drums. You can really tell if the bass goes really low and is tight.

I think New Fav brings out the beauty of a really clean high end. In addition, there is a lot of great imaging and sound-stage to listen for.

However, as an amplifier designer, I must in fairness say that the attributes I mentioned above are also strongly influenced by the loudspeakers.

The Sheffield Lab drum and track disc is also exceptional, both in the low end and in imaging. And don't forget the cymbals.

Cheers,
Bob

 

[Bob Cordell]

In some Amps i see the LTP like this, with a PreSet of around 500R to balance out discrepancies. Obviously a PNP version also applies.

Is this not a good idea, & if so why ?

I would not control amplifier DC offset this way. As shown, the pot would have way too much offset adjustment range if it is also serving the function of emitter degeneration. Also, I think passing the LTP tail current through the pot wiper is asking for trouble.

Cheers,
Bob

 

[Zero D]

@ Bob Cordell

OK, Thanks

 

[michaelkiwanuka]

Correction!

Yes. A current source using a voltage reference such as a zener diode and a emitter resistor possesses shunt applied-series derived negative feedback, and is, therefore, a negative feedback current source.

In the two-transistor current source that you call a "negative feedback current source", the series derived feedback signal is amplified by the sensing transistor before being shunt applied to the current source transistor's base.

This amplification of the derived feedback signal greatly increases the output impedance of the current source, so much so that there is no need to use a cascode current source in its place.

On the contrary, a current source using a voltage reference such as a zener diode and a emitter resistor possesses series applied-series derived negative feedback, and not shunt applied-series derived negative feedback.

 

[keantoken]

The trimmer could be paralleled by a much smaller resistor to reduce sensitivity. That said, the app notes I read about trimmers say that having a minimum current through the wiper is the most reliable way to operate a trimmer. I would do it in a prototype where I wanted to reduce soldering time, but not in anything critical.

 

[Bob Cordell]

The trimmer could be paralleled by a much smaller resistor to reduce sensitivity. That said, the app notes I read about trimmers say that having a minimum current through the wiper is the most reliable way to operate a trimmer. I would do it in a prototype where I wanted to reduce soldering time, but not in anything critical.

I just think that there are a lot better ways to control amplifier offset than fooling around in the LTP emitter circuit. I don't like to have to trim it anyway. Using a DC servo is best, and it really does not take up much space - just a dual op amp, an integrating capacitor and a few resistors. This, however, assumes you already have a convenient supply voltage to run the op amp.

Cheers,
Bob

 

[kgrlee]

I just think that there are a lot better ways to control amplifier offset than fooling around in the LTP emitter circuit.

If you MUST trim, the LTP emitter is unduly sensitive.

A much better place to trim is the CM resistors ... shock! HORROR!

That's where the OPA designers do it including Guru Scott Wurcer. Once you relinquish Dave Zan's perfectly 'balanced' CM, you might as well drop the diode connected BJT in the CM and use a cheapo 1n4148.

This ISN'T an obstacle to 1ppm THD20k in many applications as can be seen from my later posts in discrete-opamp-open-design

But as Bob says, 'why not use a simple DC servo?'

 

[Rafael L]

Note that with a Triple, one can also choose to mount the drivers on the main heat sink and keep the predrivers on the board with their own Vbe multiplier transistor (as part of a two-transistor bias spreader) and have that transistor mounted back-to-back with one of the predrivers (instead of using a heat spreader to mount the two predrivers and the Vbe multiplier). This assumes that the two predriver transistors are at about the same temperature. The extra heat sinking effect of the Vbe multiplier mounted to one of the predrivers will introduce a little bit of error into this assumption.

I used your tip, output transistors and drives the same heatsink, Predrives are in board with its own VBE multiplier.
Without compensation VBE in Predrives error is 20% for 10°C, with compensation error is of only 5% for 10°C which is reasonable for a simple circuit VBE multiplier.

Circuit VBE multiplier and simulation are attached, Thanks for the help.

The predriver Vbe multiplier transistors are not diode-connected transistors in what I described. They comprise a complete Vbe multiplier that is in series with the Vbe multiplier for the output transistors. I believe I covered this in my book. Works great.

In his book has several circuits VBE multiplier, but from what I've seen, none of them you write for use with 2 VBE sensor.

 

[Dave Zan]

If you MUST trim, the LTP emitter is unduly sensitive.
A much better place to trim is the CM resistors ... shock! HORROR!
That's where the OPA designers do it including Guru Scott Wurcer.

I just retried this on my simplistic (worse case) input and the results of a 1% variation in the emitter resistors of the LTP and the CM were practically identical. This is with typical LTP 52 ohm and a couple of volts on the 470 ohm CM resistors.
Probably Scott knows about some second order effects, any ideas?

Once you relinquish Dave Zan's perfectly 'balanced' CM, you might as well drop the diode connected BJT in the CM and use a cheapo 1n4148

For a cheap production amp this may make sense.
But a diode connected transistor will track better as temperature and current vary. IIRC Self comments on this.
For a DIY amp it seems a poor economy to save 10c, if that.
Why not aim for a properly balanced CM under nominal conditions?

Best wishes
David

 

[Bob Cordell]

If you MUST trim, the LTP emitter is unduly sensitive.

A much better place to trim is the CM resistors ... shock! HORROR!

That's where the OPA designers do it including Guru Scott Wurcer. Once you relinquish Dave Zan's perfectly 'balanced' CM, you might as well drop the diode connected BJT in the CM and use a cheapo 1n4148.

This ISN'T an obstacle to 1ppm THD20k in many applications as can be seen from my later posts in discrete-opamp-open-design

But as Bob says, 'why not use a simple DC servo?'

Trimming in the current mirror resistors is indeed a much better approach; I had actually forgotten about it because I never trim for DC offset. We should keep in mind that a well-designed amplifier will not need much trimming, if at all, so a trim circuit with a very "light touch" and very little range is all that is needed.

Trimming was mostly used back when some made DC coupled amplifiers where the offsets were multiplied by the gain of the amplifier (and no DC servo was used). Amplifiers with the decoupling capacitor in the shunt leg of the feedback network don't amplify the DC offset of the input stage, and an output offset of 10-20mV is OK anyway. We're not talking about precision op amp circuits here.

The use of an electrolytic cap in the feedback network is undesirable, but I showed in my book that if a non-polar electrolytic with a voltage rating of, say 100V is used, as used in loudspeaker crossovers, their distortion contribution is almost unmeasurable.

Cheers,
Bob