Calcs for op amp loading vs THD?

[JohnRoberts]

25mA is WAY too low. You need around 26mV across the emitter resistor of each transistor to get optimum "Class B" according to Mr Self (most of us call that Class AB). If you use 0.22R emitter resistors then:

26mV/220mOhm = 118mA

your math seems accurate, the random call to authority (no offense to Douglas Self) seems arbitrary.

I recall class AB amps using lateral mosfets being biased up above 100mA quiescent to enjoy the positive temperature coefficient region of those particular power devices for stable thermal behavior. The high class A current coincidentally improved crossover distortion while the high gate capacitance was not easy to slew.

Kind of.

It is actually a circuit that prevents the "non-active" transistor from cutting off

the old demon slew limiting/transient intermodulation distortion ... Marshal Leach published an interesting AES paper showing how to avoid saturating the input LTP in amps

and provides a stable, thermally independent bias.

yup

As a result essentially all of the common "Class AB" distortion mechanisms EXCEPT GM doubling/halving (doubling if you from a Class B view, halving if you view from a Class A view) which is generally amenable to using NFB to correct.

It is a clever little circuit.

I have something simpler that does the same job, but that's beside the point.

Thor

I stand by my observation that typical bipolar amp class A bias is closer to 25 mA*** than 120 mA. That said I have been out of the amplifier product management/design trenches for decades.

JR

**** this is per device current, so large power amps with multiple devices in parallel, and high voltage power rails can result in significant no load idle power dissipation. Preventing thermal runaway from this no load heat involves much design effort. I've seen amps with marginal thermal stability literally self destruct (not a pun) with no load.

 

[Admiral-dk]

I will agree with JR on the 25mA in just about ALL PA Amps I have repaired so far - and I haven't heard any Cross-Over Distortion on those + I even repaired two with Zero mA (0-1mA) BIAS, that sounded Great (though some might claim that you hear the Drivers below 0.5W on those).

I believe that the 120mA is a figure floating around among the Hi-Fi 'Freaks' (though a number of those, want Amps as Idle) ....

Per

 

[thor.zmt]

your math seems accurate, the random call to authority (no offense to Douglas Self) seems arbitrary.

Well, I do not consider Mr. Self as "Authority". He does show a lot of detailed data however why this "optimum bias point" is what it is. Note, it is NOT any specific current, that depends on the emitter resistors used.

I recall class AB amps using lateral mosfets being biased up above 100mA quiescent to enjoy the positive temperature coefficient region of those particular power devices for stable thermal behavior.

Yes, it worked quite well too.

the old demon slew limiting/transient intermodulation distortion ... Marshal Leach published an interesting AES paper showing how to avoid saturating the input LTP in amps

Nothing to do with that. Turning off the non-active transistor(s) etc. causes extra distortion at HF. Again, no "Authority", all the detailed data is Self's articles and books.

I stand by my observation that typical bipolar amp class A bias is closer to 25 mA*** than 120 mA.

For Sound reinforcement, to guard against thermal runaway - maybe.

Many PA Amplifiers use compound feedback pair outputs, these have different optimum bias points compared to Emitter follower outputs. Data in the usual articles and books.

I will agree with JR on the 25mA in just about ALL PA Amps I have repaired so far ... even repaired two with Zero mA (0-1mA) BIAS, that sounded Great (though some might claim that you hear the Drivers below 0.5W on those).

The optimum Bias current depends on the output topology and emitter resistor value. As we were discussing a specific circuit with a specific topology and thus a specific optimum bias current.

Many PA Amplifiers these days all copy the same kind of circuit, all with grounded collector (compound feedback pair) outputs where 25mA sounds about right.

and I haven't heard any Cross-Over Distortion on those

Are you sure? Do you how crossover distortion at HF sounds like?

even repaired two with Zero mA (0-1mA) BIAS, that sounded Great

Could this be another topology which is in effect "current dumping" similar to Quad, with a Class A biased drivers linked to the output via resistors with the actual output transistors cutting in only above a certain current level?

In this case there is usually a lot of quiescent current in the driver transistors. I remember the Zeck Bimos Amplifiers from the 80's with classic 100mA biased Hitachi lateral FET's (and a Hitachi App Note FET Amp) that drove banks of bipolar current dumpers running zero bias.

They were among the better sounding ones, after I got my first unit I quickly dumped my previously used Peavey CS800 & CS1200's (sorry) and got Zeck Bimos instead.

But my experience is that using a driver circuit that is not linked to the output and correctly biasing the outputs correctly provides better performance (in the sense of lower THD with much lower high order/odd order HD).

I believe that the 120mA is a figure floating around among the Hi-Fi 'Freaks' (though a number of those, want Amps as Idle) ....

I believe that you believe that.

However this number (26mV across the emitter resistor for bipolar based emitter follower outputs, NOT 120mA) is down to the work of one Barney Oliver at Hewlett Packard in the late 60's to make transistor laboratory amplifiers with the lowest possible distortion and has been again and again confirmed.

Optimum bias is a clearly defined condition, but NOT a specific current. It is just so that for the SPECIFIC CIRCUIT being discussed 118mA (0.22R Re) or 130mA in the article for 0.2R Re is the correct optimum bias.

Thor

 

[JohnRoberts]

Well, I do not consider Mr. Self as "Authority". He does show a lot of detailed data however why this "optimum bias point" is what it is. Note, it is NOT any specific current, that depends on the emitter resistors used.

Many do consider Douglas Self an authority, I am not in his circle.

Optimum bias current wrt emitter resistors sounds like some of the old thermal stability equations, but it also needs to know the thermal resistance junction to ambient. I have another old Peavey anecdote on the subject but I won't bore you.

Yes, it worked quite well too.



Nothing to do with that. Turning off the non-active transistor(s) etc. causes extra distortion at HF. Again, no "Authority", all the detailed data is Self's articles and books.

this is well known in the art... if we allow bipolar transistors to turn off, it takes a finite time to turn on again. This is not just for the input LTP but anywhere in the signal path. Common practice in amp design was to use anti-sat diodes to prevent bipolar devices from going into saturation. This can be seen in how cleanly power amplifiers recover from clipping... poorly designed amps can stick to the rail, and misbehave for the brief time that negative feedback is not in charge.

For Sound reinforcement, to guard against thermal runaway - maybe.

maybe I'll share that story later... One of my patents at Peavey was for a power amp heat sink design. (6,515,859 Roberts , et al. February 4, 2003)

Many PA Amplifiers use compound feedback pair outputs, these have different optimum bias points compared to Emitter follower outputs. Data in the usual articles and books.

Not sure I follow. One problem with some early amps to deal with the early bipolar PNP power devices being wimpy, was to cobble together a PNP driver with a robust NPN output device to behave like a PNP power device. This topology was called quasi-complementary as used for making amplifiers that didn't blow up. (My personal DIY power amp that I built back in the early 70s used quasi-comp compound power devices for the PNP side. 2n3773s were plenty robust for the NPN... I haven't powered up that amp recently but it was working last time I turned it on. Now for my home system I use class D amps designed by Bruno Putzey. his amps don't suck.

The optimum Bias current depends on the output topology and emitter resistor value. As we were discussing a specific circuit with a specific topology and thus a specific optimum bias current.

we?

Many PA Amplifiers these days all copy the same kind of circuit, all with grounded collector (compound feedback pair) outputs where 25mA sounds about right.

true... Crown was big on using qusai-comp topology long after robust PNP devices became possible.

Are you sure? Do you how crossover distortion at HF sounds like?

Low level HF content sounds fuzzy and distorted ..... duh.

Sorry I have better stuff to do.

JR

Could this be another topology which is in effect "current dumping" similar to Quad, with a Class A biased drivers linked to the output via resistors with the actual output transistors cutting in only above a certain current level?

In this case there is usually a lot of quiescent current in the driver transistors. I remember the Zeck Bimos Amplifiers from the 80's with classic 100mA biased Hitachi lateral FET's (and a Hitachi App Note FET Amp) that drove banks of bipolar current dumpers running zero bias.

They were among the better sounding ones, after I got my first unit I quickly dumped my previously used Peavey CS800 & CS1200's (sorry) and got Zeck Bimos instead.

But my experience is that using a driver circuit that is not linked to the output and correctly biasing the outputs correctly provides better performance (in the sense of lower THD with much lower high order/odd order HD).



I believe that you believe that.

However this number (26mV across the emitter resistor for bipolar based emitter follower outputs, NOT 120mA) is down to the work of one Barney Oliver at Hewlett Packard in the late 60's to make transistor laboratory amplifiers with the lowest possible distortion and has been again and again confirmed.

Optimum bias is a clearly defined condition, but NOT a specific current. It is just so that for the SPECIFIC CIRCUIT being discussed 118mA (0.22R Re) or 130mA in the article for 0.2R Re is the correct optimum bias.

Thor

 

[thor.zmt]

Optimum bias current wrt emitter resistors sounds like some of the old thermal stability equations,

Not related. It is strictly the "static" crossover distortion (e.g. the gm / output impedance change.. As said, it has

this is well known in the art... if we allow bipolar transistors to turn off, it takes a finite time to turn on again.

Stored charge in the reverse biased base junction.

Not sure I follow.

Most modern dap PA Amplifiers do not use Emitter follower outputs.

MC2 is one of the few that do (and I like their amplifiers better than many others).

Most modern PA Amplifiers use a circuit originated by IIRC Crest.

An Op-Amp drives a compound feedback pair with gain. Here a version from a Behringer Amp:


This circuit has a few really interesting benefits, but like this can use some improvements.

Example - D15/16 would best be replaced with a Diamond style driver.

One problem with some early amps to deal with the early bipolar PNP power devices being wimpy, was to cobble together a PNP driver with a robust NPN output device to behave like a PNP power device. This topology was called quasi-complementary as used for making amplifiers that didn't blow up. (My personal DIY power amp that I built back in the early 70s used quasi-comp compound power devices for the PNP side. 2n3773s were plenty robust for the NPN...

Yes, CS-800 & CS-1200 used these.

Now for my home system I use class D amps designed by Bruno Putzey. his amps don't suck.

We may have to agree to disagree here. Well, not exactly that S-word and I guess next to an underbiased quasi-complementary ca mid-70's Amp yes, it would be an improvement.

we?

You and me, talking about that Class I example.

Later Thor

 

[JohnRoberts]

Not related. It is strictly the "static" crossover distortion (e.g. the gm / output impedance change.. As said, it has

?

Stored charge in the reverse biased base junction.

correct... this was even worse for early mosfet amps trying to drive all that gate capacitance.

Most modern dap PA Amplifiers do not use Emitter follower outputs.

"dap"? I just did a search.... an EU Brand?

MC2 is one of the few that do (and I like their amplifiers better than many others).

Most modern PA Amplifiers use a circuit originated by IIRC Crest.

nah... but widely used by most most major amp manufacturers. The beauty of that driven rail topology is that the front end is all low voltage, so possible to be implemented inside a 36V IC. I was tasked with investigating making a full custom Peavey IC, but the project stumbled when the junior engineer assigned to the project didn't even understand how OTAs work. We needed an OTA integrated on chip to execute a clip limiter.

FWIW I was on the team that kicked the tires at Crest when Peavey bought them. JD Bennet a well known Crest amp designer had already quit and moved the Peavey before that, but I got to meet the guys who were left in NJ.

An Op-Amp drives a compound feedback pair with gain. Here a version from a Behringer Amp:

View attachment 107142
This circuit has a few really interesting benefits, but like this can use some improvements.

Example - D15/16 would best be replaced with a Diamond style driver.

Uli copied from many companies... I think Crown was early on the driven rail bandwagon, but pretty much everybody used that topology AFAIK.

Yes, CS-800 & CS-1200 used these.

HUH? Now you are going to school me about Peavey? My office was right next to Jack Sondermeyer RIP (he designed the CS800 et al).

The CS800 and CS1200 were traditional class AB using full complementary bipolar output stages, no driven rail, and no quasi complementary. FWIW the CS1200 was the poster boy for heavy iron. You don't want to drop one on your foot.

We may have to agree to disagree here. Well, not exactly that S-word and I guess next to an underbiased quasi-complementary ca mid-70's Amp yes, it would be an improvement.


You and me, talking about that Class I example.

Later Thor

No just you...

JR

 

[thor.zmt]

"dap"? I just did a search.... an EU Brand?

Auto incorrect strikes again.

nah... but widely used by most most major amp manufacturers. The beauty of that driven rail topology is that the front end is all low voltage, so possible to be implemented inside a 36V IC.

Not just that. Grounded collectors mean no isolation to the heatsink.

In non-PA amplifiers I believe driven rail topology was pioneered by Jim Strickland in the late 70's using Hitachi MOSFETs. I have not seen any earlier references, but I have not looked very hard.

I was tasked with investigating making a full custom Peavey IC, but the project stumbled when the junior engineer assigned to the project didn't even understand how OTAs work. We needed an OTA integrated on chip to execute a clip limiter.

OTA = standard op-amp without buffer between VAS and load... He would have understood instantly.

I think Crown was early on the driven rail bandwagon, but pretty much everybody used that topology AFAIK.

Possible, Crown was early on a lot of stuff.

HUH? Now you are going to school me about Peavey? My office was right next to Jack Sondermeyer RIP (he designed the CS800 et al).

The CS800 and CS1200 were traditional class AB using full complementary bipolar output stages, no driven rail, and no quasi complementary. FWIW the CS1200 was the poster boy for heavy iron. You don't want to drop one on your foot.

I never dropped one on my foot. Just used them. And replaced them with what basically a Hitachi App-note 100W Amp with bipolar current dumpers.



This sounded much, much better than the CS-1200/800 we used up to then.

No just you...

Good to know you did not comment on the Class I output stage and did not dump all over it without actually fully analysing the circuit, because the circuit is biased at 130mA which gives lowest distortion and not at 25mA like every other Amp.

Thor

 

[abbey road d enfer]

Most modern dap PA Amplifiers do not use Emitter follower outputs.

Most modern PA Amplifiers use a circuit originated by IIRC Crest.

Actually most modern PA amplifiers are switch-mode.
Crown, Lab Gruppen, Powersoft...

 

[abbey road d enfer]

The beauty of that driven rail topology is that the front end is all low voltage, so possible to be implemented inside a 36V IC.

The uglyness of driven rail topology is that the audio output is injected in the mains via the xfmr leakage capacitance.
The original Microtech 1000 was particularly guilty of it.
It became particularly evident at one gig at the largest sport arena in Paris. Driving the MT's would interfere with all the other amps in the system.
The only cure was to rearrange the PE circulation.

 

[thor.zmt]

The uglyness of driven rail topology is that the audio output is injected in the mains via the xfmr leakage capacitance.

Screen Winding? Should be mandatory in any audio power transformer.

Thor

 

[abbey road d enfer]

Screen Winding? Should be mandatory in any audio power transformer.

Thor

Yes, but the screen is connected to PE, so the signal is capacitively coupled to PE. If the actual impedance of the PE wiring is inadequate, it propagates all over through L & N.

 

[JohnRoberts]

The uglyness of driven rail topology is that the audio output is injected in the mains via the xfmr leakage capacitance.
The original Microtech 1000 was particularly guilty of it.
It became particularly evident at one gig at the largest sport arena in Paris. Driving the MT's would interfere with all the other amps in the system.
The only cure was to rearrange the PE circulation.

Sorry that was the product manager in me speaking. Driven rail amps are attractive to manufacturers and customers for being cheap to build. That makes them popular for fixed install (background music) amplifiers. They are not hifi amps but when well executed work adequately.

JR

 

[thor.zmt]

Yes, but the screen is connected to PE, so the signal is capacitively coupled to PE. If the actual impedance of the PE wiring is inadequate, it propagates all over through L & N.

Screens can be connected wherever, depending on voltage rating.

Who says it cannot be connected to GND? Of course, that is still Earth, but ideally with a soft connection.

If the impedance of the PEN is high enough to cause issues from this audio leakage, it may just be EN, without the "P for "protective".

Thor

 

[JohnRoberts]

Auto incorrect strikes again.

?

Not just that. Grounded collectors mean no isolation to the heatsink.

What the collectors get connected to circuit-wise has nothing to do with heat sinks. Some manufacturers took advantage of the grounded collectors to eliminate the mica washers reducing the thermal resistance junction to sink. Of course ground is just a node and in a high current power amp voltage drops from the heavy current need to be accounted for.

In non-PA amplifiers I believe driven rail topology was pioneered by Jim Strickland in the late 70's using Hitachi MOSFETs. I have not seen any earlier references, but I have not looked very hard.

I am not familiar with his work. I recall the white papers written by Hitachi promoting their class H (multi voltage rail) power amps to reduce dissipation.

OTA = standard op-amp without buffer between VAS and load... He would have understood instantly.

Apparently you don't understand either. "OTA" = operational transconductance amp. Peavey used an OTA for the gain element in their DDT "clip limiter". I wanted to incorporate the DDT into the single IC amp front end, but the junior engineer didn't appreciate the difficulty to make one at IC level that wouldn't suck (the input LTP for an OTA is not trivial).

Possible, Crown was early on a lot of stuff.



I never dropped one on my foot. Just used them. And replaced them with what basically a Hitachi App-note 100W Amp with bipolar current dumpers.

View attachment 107159

This sounded much, much better than the CS-1200/800 we used up to then.

The CS800 and CS1200 were well regarded. The valid criticism of the CS1200 was that it used class AB topology so was heavy and not as efficient as later class G/H amps. The later Peavey CS1800 made the same power as the former CS1200 in less rack spaces, less weight, and less cost.

The CS800 was pretty popular... back when I was product manager for them I calculated that we had sold 500,000 units worldwide.

Good to know you did not comment on the Class I output stage and did not dump all over it without actually fully analysing the circuit, because the circuit is biased at 130mA which gives lowest distortion and not at 25mA like every other Amp.

Thor

My interest was checking out the 3 transistor differential stage, and that wasn't it.

I don't argue with people on the WWW about what they say they hear.

JR

 

[abbey road d enfer]

Sorry that was the product manager in me speaking. Driven rail amps are attractive to manufacturers and customers for being cheap to build.

You don't need to be sorry. I sold hundreds (thousands?) of the MT/MA range when they were current. Not so long ago the 5002VZ was hard to beat, in particular with highly compressed subwoofers, where switching amplifiers, rated at 20kW and plugged on 3.5kVA outlet, see their limits.
The problem I mentioned was due to multiple distant racks. In most cases it was not an issue.

 

[abbey road d enfer]

Screens can be connected wherever, depending on voltage rating.

Who says it cannot be connected to GND? Of course, that is still Earth, but ideally with a soft connection.

Even with a soft connection, currents are dumped in the PE conductor, and radiate in the L & N. I don't think it would change much.

If the impedance of the PEN is high enough to cause issues from this audio leakage, it may just be EN, without the "P for "protective".

In this case, it's the inductance that floats the boat. At 50 Hz, the protective role is assumed.

 

[thor.zmt]

Even with a soft connection, currents are dumped in the PE conductor, and radiate in the L & N. I don't think it would change much.

In this case, it's the inductance that floats the boat. At 50 Hz, the protective role is assumed.

If you have a grounded shield circulating currents stay inside the chassis. Kirchoff and all.

Thor

 

[abbey road d enfer]

If you have a grounded shield circulating currents stay inside the chassis. Kirchoff and all.

What do you do with the currents that are induced on the L & N wires? Do they stay inside?
This is like saying that an smps in a shielded chassis has no conducted emissions.

 

[thor.zmt]

?

I typed one thing, the "auto (in) correct" function corrected it without me noticing to DAP = Digital Audio Player (read generic for iPod).

What the collectors get connected to circuit-wise has nothing to do with heat sinks.

Except that metal that transmits heat from the transistor die to the heatsink is connected to collector (BJT), drain (Vertical FET) or Source (Lateral/Hitachi FET).

Some manufacturers took advantage of the grounded collectors to eliminate the mica washers reducing the thermal resistance junction to sink.

Yup. And the difference is quite material.

I am not familiar with his work.

Hitachi Lateral FET, Grounded Source, Driven Rail topology. Very clever really.


This is one of his late designs for Hafler.

He originally designed and marketed Electrostatic Speakers under the Acoustat Brand and found all sorts of Amp's, even supermassive black hole level Pro Audio Amp, would blow up driving his speakers.

So he set out to design that counts drive his speakers.

Apparently you don't understand either. "OTA" = operational transconductance amp.

Really? Operational Transconductance Amplifier is literally a Voltage to current converter.

Or, if we look at the circuit topology, a conventional op-amp with the output directly from the VAS Stage, no miller compensation, no output buffer.

Output is thus a current from a high impedance node that is proportional to the input voltage.

So it is really very basic and nothing special, except for a fancy name it is just an uncompensated Op-Amp minus output buffer, which is what I wrote.

Those who understand something can explain to those that don't in a way that allows to proceed with their work based on what the supervisor needs and they can understand.

Peavey used an OTA for the gain element in their DDT "clip limiter". I wanted to incorporate the DDT into the single IC amp front end, but the junior engineer didn't appreciate the difficulty to make one at IC level that wouldn't suck (the input LTP for an OTA is not trivial).

I disagree, an OTA is entirely trivial, as is any non OTA op-amp. Just a minor variations. You can even easily turn any 5532 into an OTA using a slightly different Circuit (known as Howland current pump).

Later edit:
I looked at the DDT circuit, what is used there is not stritctly a classic "OTA". It is an OTA with adjustable current in the differential input and thus adjustable transconductance:



Technically speaking, as implemented in DDT it is a crude VCA, not an OTA.

The CS800 and CS1200 were well regarded.

As said, I/we used for a long time. Heavy, but reliable and decent sound at high power.

But they were biased in Class B (not AB as you state) and sounded quite gritty. Worse if hot, as this shifted the output stage deeper into class B. CS800 also was quasi-complementary.

When I got the first Zeck Bimos Amp's they went on the HF horns and super tweeters first, after that we rapidly converted the rest.

My interest was checking out the 3 transistor differential stage, and that wasn't it.

It's a 3- Transistor differential alright, just not the same as yours.

I think we dragged this sufficiently OT (if interesting) sufficiently lomg, let's return to boring generic op-amp's.

Thor

 

[Lampie519]

Jim Strickland, Hafler's chief engineer, developed the first Transnova amplifier in 1981 for Acoustat,

https://www.stereophile.com/content/hafler-transnova-9500-power-amplifier
A schematic of the Acoustat TNT-120 :