If it were a major problem, i.e. above other limitations of the the opamp, would not show up in any of the curves given by manufacturers or any enterprising 3rd party? (looking squarely at Samuel Groner's excellent work with testing opamps)
I mean if one's ensemble sins are sitting at < -120 dB DC to 20kHz, and being used in precision instrumentation, then how is it limiting audio? Even *if* standard sweeps don't test the right thing, there's just not much of a budget for errors regardless. (with the presumption that "signal-out = constant*signal_in" is the goal; if we're aiming for a characteristic effect, then all bets are off.)
Ideologically, fine, if one doesn't like using monolithic circuits that's okay (or needs something they can't do easily), but the premise of your inquiry seems absurd.
Last edited:
Brian it's certainly fair to ask, but I can see nothing that would be a limitation. The major cap would be the comp cap and this is rarely more than 50pF or so and this would be in parallel with the impedance at the high gain node which will in most cases also have junction parasitics also associated with it. BTW if you find the LT1028/1115 ISSCC paper you will see at least one very large horrible junction cap in the compensation network, and it still has many friends in the DIY community.
The 797 was designed to use cancellation by symmetry of junction parasitic capacitance and a single gain stage to facilitate a simple one capacitor compensation. Jim Williams once joked to me that the LT1028 breadboard was compensated with all the decade boxes they had in the lab.
Last edited:
Innocent question, I suppose, then -- what would be the best way to torture test this out? I mean the dominant pole starts early with as much DC gain as many modern devices have.
And that's why my current project uses an AD744 + HA-5002: to make sure the opamp sees a negligible, high impedance "load", and to make sure the 600 ohm "true output load" is driven by a special high current beast with enormous safety margin. Plus, I once interviewed for a job at Harris and found their D.I. process to be wildly intriguing.
I hope you will share your results, the 744 had enough internal nodes brought out that you can challenge some pretty good discrete designs with minimal parts. You can bypass the output stage entirely or simply add an external PNP with no added bias components.
> the 744 had enough internal nodes brought out that you can challenge some pretty good discrete designs with minimal parts.
But for the highish input voltage noise ?
Patrick
But for the highish input voltage noise ?
Patrick
K, that was my presumption, but wasn't sure if I was missing something. Thanks!
And, Mark, I'm curious as well the improvement.
Not so bad for line stages. 14nV is -114dB SNR re: 1Vrms in 20k BW. Walt and I used 744's as second stages in split RIAA circuits and there was no audible noise at all.
If the simplified schematics is to go by then it is not difficult to make a DIP version from Toshiba SMD transistors.
🙂
Patrick
🙂
Patrick
Certainly one of the pleasures of the HA-5002 is the fact that you can only buy it in surface mount. Keeps out the "proud to be a lame old fart" riff raff. I suppose I am willing to sell one of my (thru hole) HYBRID parts from whom the HA-5002 descended, namely the (in)famous LH-0002 "diamond buffer", but the bidding begins at USD 300.00 plus shipping and insurance. Got a couple 033s and 063s too but it'll take a boatload of cash to induce me to let go of them.
You can get LH0033 for reasonable money.
And I actually made a discrete SMD versiopn (size and pin compatible).
Even lower noise than the original.
😉
Patrick
And I actually made a discrete SMD versiopn (size and pin compatible).
Even lower noise than the original.
😉
Patrick
Almost so. It was a fast inverting hybrid amplifier based on uA709, buffer output stage + fast parallel branch. 250V/us, fT 10MHz, settling time 200ns/1%. Late seventies of 20th century.
Attachments
Actually, the Aleph P did both. The preferred customer setting that Wayne
mentioned was for the input diff pair resistance.
I was just reading the AD797 datasheet. In the theory of operation section it states:
Did the Marketing Department have a hand in this? 😉
No, why do you ask? The statement is true up to the point of device BW limitations at high frequencies all secondary effects. jcx posted some analysis years ago and we discussed it, does it behave like neutralization, positive feedback, a gyrator, does it matter?
Anyone can make a 20dB gain stage and connect/disconnect the cap and observe that the frequency compensation does not need any adjustment nor does the BW change appreciably.
EDIT - The AES preprint has pictures and actual measurements, they speak for themselves.
Last edited:
The statement can only be true for infinite NFB or bandwidth. It seems to me to use the same mathemagical licence as HEC. I don’t have that AES paper but I’ll chew on it some more.
I’m not dissing the IC, I’m sure it is very good, but that claim seems a tad hyperbolic. 🙂
I’m not dissing the IC, I’m sure it is very good, but that claim seems a tad hyperbolic. 🙂
Last edited:
PMA, thanks for the 'dated' spec sheet on a 709 hybrid. This schematic shows true design engineering, converting a usually very slow IC with other problems to something that is really elegant. When I first worked with 709's in 1966-7, we were designing military equipment too, but we did not enhance them. Later I switched to the uA741, which was perfect for the tape transport servos that I was designing, but of course, these parts were lousy for relatively high speed instrumentation or even audio. Back in late 1967 when I moved to the Ampex Instrumentation tape recorders division, I recall another engineer trying to use the uA709 for instrumentation tape recorders and he gave up in frustration. My regards to the (Russian?) engineer who made the uA709 into something more than it originally was.
Traderbam, Scott made this breakthrough in circuit linearization and it is why his AD797 is so extra special as an IC in test equipment. I used them myself in my ST1700 Thd analyzer.
Thanks. I have great respect for Scott’s achievement. Not at question at all.
If my challenging of the data sheet wording is crass and inappropriate in this thread I apologise and I’ll leave it at that.
- Status
- Not open for further replies.