John Curl's Blowtorch preamplifier

[janneman]

Originally posted by john curl
Jan, why do you persist in questioning anything that we do or have knowledge of?

John,

I hope I don't! What I do is ask 'why' if someone comes with the latest trick to significantly improve his/her system with an unlikely change. If someone tells me the color of the wire is audible, and when asked why, answers: "because I clearly heard it", I shrug.

I learned some interesting things from you in that discussion on input long tail pairs some time ago. That was a factual discussion with engineering arguments from you that made a lot of sense.

But then when you say that output coils are audible but cannot provide any arguments beyond "I clearly heard it, so you better accept it as The Truth (tm)" I'm disappointed. You're not just any old DIY-er John, you're John Curl! People look up to you for explanations and arguments, not "because I say so, and if you don't accept that, hey, tough luck". And if you don't know, I don't think anybody has a problem with "I don't really know, I'm not sure, possibly I'm wrong". Saying you don't know if you don't once in a while only makes one more convincing when you say that you're certain. But people very quickly see when one is tapdancing.

You're John Curl. Noblesse Oblige.

Jan Didden

[cliffforrest]

Originally posted by john curl
Jan, why do you persist in questioning anything that we do or have knowledge of?
__________________
"Condemnation without Examination is Prejudice"

Mr Curl, do you not see a contradiction in your two sentences?

a gentle

[jcx]

Originally posted by john curl
Which was better JCX? Noisewise.

my choice wasn't relevant to audio apps-

our strain gage transducers use 350 and 700 Ohm bridges and we are looking at “DC” (most users collect <100 s data sets) -1KHz so both of these "low noise" op amps are marginal in that i_n is dominant over v_n and the mix of bridge impedances requires a compromise solution

my reading of the LT1028 data sheet specs edged out the AD797 - but I'd guess process variations for both companies could make it go either way

since we're not directly competing on the “DC” noise spec the cost to characterize parts on a lot by lot basis and wouldn’t be justified

I'm interested in the new OPA211 in the same application

if pressed for better DC performance the next step would be to go to AC bridge excitation - demodulation becoming "trivial" in software as DSP hardware is already used for filtering/decimation - but mechanical effects like creep and thermal drift limit transducer performance so pushing further for low “DC” noise isn’t too important in our applications

[john curl]

Thanks, JCX, I did notice that the current noise was about 1/2 as much in the 1028 type product, yet the voltage noise was about the same. Offset is not too important to me, and the only other factor that really concerns me is the Iq of the output stage. I believe that both are seriously compromised, so I am putting a current source to -V on the output. Now, to optimize that.

[john curl]

JCX, the 211 looks pretty good, but I need a minidip package. The layout is already done.

[jcx]

Unusually, both LT1028 and AD797 data sheets show the output stage bias current in odd corners of the data sheets - they both indicate 500uA is used in the output Qs

with 500uA bias 1mA Class A push-pull output is available

the LT1028 over-comp pin is used with a Cherry-style C around the output stage

the AD797 also has a unusual output stage error canceling scheme

[john curl]

Kind of low, as I use 50 ma in my JC-2 preamps. Oh well, 100/1, I guess I can improve the situation. Now for some 'inside knowledge' what is the peak beta current of the output devices used in the IC's?

[KSTR]

Well, that's really inside knowledge... I would use, as an educated guess, 1/10th or so of the short circuit current as external class-A bias, which is hopefully close to the beta and fT max but if it misses it's probably on the more benign left side of these curves.

- Klaus

[john curl]

But Scott DESIGNED the AD797 IC. He can tell me, and it would help me optimize his part further. This can't be a secret, I should think. Better to ask the designer, than to speculate.
Just the NPN should be OK, unless the PNP is even a more linear and better device.

[john curl]

Does the 2N2222 ring a bell?