John Curl's Blowtorch preamplifier part II

[jcx]

are the discrete op amp design crowd going to at least read Groner's op amp distortion paper, try to master the design details he has published for discrete op amps

SG-Acoustics · Samuel Groner · Discrete OpAmps

Today it is not needed. I would take one low noise dual opamp and combine it with one more single fast opamp that has better output stage, adding few resistors. The result will be one opamp that you can't beat by discrete designed opamp.

I certainly agree that many op amp parameters can be improved by using composite/multiloop monolithic designs; extreme low noise, hi V swing are the main limitations - limits that are only relevant for a very few home diy audio applications

 

[scott wurcer]

And as remarked, he thought he was the first to conceive of the use of inductors in the emitters to limit gain at high frequencies, but it was, like so many things, prior art, which (like they do more times than not) the USPTO missed.

To be fair the schematics of the modules were not published until much later. In 1966 they would have been a trade secret, and if Dick invented it I have no idea why it was not patented at the time.

 

[EUVL]

> you guys will have to help on the discrete NPN/PNP choices

2SC3324 / 2SA1312

> I was going to use LSK170/74's on the input.

Can the general public get LSJ74 ?
Since when, and where ?

If not, I suggest we only use N-devices (BF862), or MMBF5458/5462.

I shall open a new thread now.

http://www.diyaudio.com/forums/analog-line-level/218373-discrete-opamp-open-design.html#post3136148


Patrick

 

[Wavebourn]

I certainly agree that many op amp parameters can be improved by using composite/multiloop monolithic designs; extreme low noise, hi V swing are the main limitations - limits that are only relevant for a very few home diy audio applications

High voltage opamps for an output stage are available. It is not a problem. And an opamp made of several opamps will be still discrete. Or, I can make a hybrid IC containing couple of case-less chips, one with low noise dual opamp, another with high voltage opamp. It can even fit into existing DIP sockets.

 

[31697B]

Scott -- so whats the end goal?

discrete circuit to beat the 990 using 990 topology?

 

[scott wurcer]

Scott -- so whats the end goal?

discrete circuit to beat the 990 using 990 topology?

No, FET input and as simple as possible. I'd say it needs to outperform an OP627 on all audio specs.

EVUL's right it needs to use cheap readily available parts. The 2mV max offset is unrealistic with hand matched devices even most discrete duals are worse.

 

[31697B]

Here is My challenge for a descrete opamp --> using any topology and transistor types (except expensive and/rare/obsolete/exotic types): no more than 8 transistors; drive 30 ohms or lower with no more than -105db harmonics, 20-20Khz; S/N ref 1v -unweighted to be at least -110db (input shorted). ; no use of dc servo, allowed - dc offset at output less than 2mV over time and normal use temp range. all when used at gain = 10 or less.

How much better than these requirments/specs can be made??

RNM

 

[31697B]

No, FET input and as simple as possible. I'd say it needs to outperform an OP627 on all audio specs.

EVUL's right it needs to use cheap readily available parts. The 2mV max offset is unrealistic with hand matched devices even most discrete duals are worse.

Trimmable to 2mv and stays there without servo.

Ok - 627 is goal. without dc servo in low gain app?? Cant require an IC added for dc servo in low gain apps. Should be required to Not use coupling caps anywhere -- too expensive and bulky.

 

[morinix]

HUGE thread! 27,000+ posts and this is Part2!

So, can I be bold enough to ask what exactly is this thread about? I assume it was to work out what was in John Curl's pre-amp?
Has this been figured out yet?

No doubt some smart-ar$e will suggest reading of the entire thread...

Andy

For parts one and two together it's been 6 years, 7 months, 15 days As John says, go and read the first few 100 posts in part one.

 

[Wavebourn]

Scott -- so whats the end goal?

discrete circuit to beat the 990 using 990 topology?

I think, the question should start from what the market would prefer: an analog of a famous opamp, or an opamp that would beat the famous opamp, or a preamp that will beat preamps that used that famous opamp?

 

[john curl]

Note, everyone, that what is useful is something both buildable and BETTER, even in theoretical approach. For example all balanced is better than single sided driver stage.
The design should be fast, low noise, and open loop LINEAR. Open loop bandwidth is debatable. Look at the topology of the AD797, for a good example.

 

[Steve Eddy]

Before I go further let's see if we can get agreement on a really simple circuit.

Pavel, Wavey, Scott, SY and anyone else interested, would you each please post an answer to what is the power dissipated in the resistor in the attached circuit?

10 volts what? Peak? RMS? Average? Peak to peak?

se

 

[john curl]

Well back to business: I wish that I had a better schematic maker. It would be easier to show the evolution of an electronic gain block, with pictures.

 

[miklos]

For parts one and two together it's been 6 years, 7 months, 15 days As John says, go and read the first few 100 posts in part one.

Last he said the first 80 pages, you save 20

 

[miklos]

I think, the question should start from what the market would prefer: an analog of a famous opamp, or an opamp that would beat the famous opamp, or a preamp that will beat preamps that used that famous opamp?

...or just one without a ton of gain and a ton of NFB.

 

[morinix]

Last he said the first 80 pages, you save 20

I don't know just start reading from the top 'till the nuggets show

 

[Wavebourn]

...or just one without a ton of gain and a ton of NFB.

...or the one with praised name, in a praised box, with praised components, without things that believed to "compromise" the sound.

 

[simon7000]

10 volts what? Peak? RMS? Average? Peak to peak?

se

RMS. Average would be zero. Peak would be noted. There really are standard conventions.

 

[Steve Eddy]

RMS. Average would be zero. Peak would be noted. There really are standard conventions.

Average would be zero? Far as I'm aware, the average of 10 VRMS is 8.99 V.

se

 

[bcarso]

Average would be zero? Far as I'm aware, the average of 10 VRMS is 8.99 V.

se

Yes; the a.c. average is 2/pi times the peak value, the a.c. rms value 2^(-1/2) times the peak value.

And, actually, most simulators have generators that require specifying the peak value.

Of course, in the absence of an offset, the average d.c. value over an integer number of cycles is zero.