Jfets are OK. They are NOT as good as tubes, part for part, in linearity. They can be quieter, and they are complementary, so they can be made to be excellent gain stages. Break-in is, to me, a physics problem, that I must acknowledge, to be difficult to 'prove' but easy enough to hear. Break-in for me, is important, but in the total connection and wire sense. I recommend it.
First thing I did when screwing about with amplifiers..is look at the theoretical aspects..then I headed right for the feedback criticals. coupling, PS, and that dang resistor (all concerning the feedback loop).
i started by swapping the resistor out, turning it around, different types, brands, models, solders, each time, a full and proper burn-in in. Skinning the resistors (naked), different mounting, wearing out the leads, via stressing, every thing I could think of. plating the leads, unplating the leads, different field conditions, demagnetizing, etc. It took nearly a year to really get a solid handle on the overall aspects of each and all of those considerations, but since it was the most critical spot in the amp, it ended up explaining much - about all else. This was in the early 90's. I had a couple of basic amp models I used, specifically ones with high feedback, so the effects were multiplied and they (the amps) also had a low number of parallel outputs, so I could actually hear those micro differentials.
Last edited:
It's a trade off between lowered issues of a transient nature vs inductive interference (flubbery inconstant braking) in the field differentials created by the inability of the resistors to be exactly the same under all complex (simultaneously occurring) conditions. My experience, anyway. (I tried parallel and series, physically close and separate, multiple orientations of field coupling interference)
Is the advantage in thermal (overall and related issues) better than the cumulative effect of the rest? Only the specific result in the specific experiment will tell. Too many variables. My ears said you were right, in most ways or the more 'noticeable' considerations.
Last edited:
One main aspect for tubes is their their thermal stability (minor shift in the operating points). Disadvantage is their (more) limited lifetime and that you need filament supply.
In solid state devices, there are big shifts in the device paramters with the signal voltage. There is a discussion on some of this stuff over on Edmond's thread. A classic example is Cob in a VAS transistor, or the input capacitance on mosfets and JFET's. Designers need to be aware of this stuff, because often it frigs with overall stability.
Now, you overlay the thermal impact on device parameters and its clear nothing is really nailed down. Circuits just flop around, hopefully not banging into any serious limitations. Maybe tubes sound good because they run hot (thermally stabilized?) and the paramters don't shift that much with signal - but I am not a tube expert . . . ;-)
Now, you overlay the thermal impact on device parameters and its clear nothing is really nailed down. Circuits just flop around, hopefully not banging into any serious limitations. Maybe tubes sound good because they run hot (thermally stabilized?) and the paramters don't shift that much with signal - but I am not a tube expert . . . ;-)
For linear TC the second 1/10th Watt heats the resistor just as much as the first (the first order model is a thermal resistance). Pre bias over ambient probably just creates more problems with termination thermo-couples, certainly won't help distortion. Heat is actually a diffusion process and has a distributed time constant. A sudden step excitation in a good bridge setup should allow you to extract the thermal curve.
This stuff has been trade secrets for folks making 18bit + modular A/D's and DAC's for decades.
Last edited:
Good thing because I don't think you can get 16 bits out of a carbon comp. resistor!
Ed, you keep pointing that stuff out like nobody ever heard of it before. I once scavanged a discrete DAC made for the missile program in the 60's. It used 2W 1% resistors for an 8 bit R2R ladder.
Thank you, thank you, thank you! I continue to read everywhere that running warmer is somehow entailing lower effects due to self-heating shifts and ambient shifts.
But intuition fails us here. Unless the tempco itself is a strong negative function of absolute temperature, running warmer does nothing for you (and as you point out, invites issues from thermocouple effects if the junctions are not at equal temps).
Brad
Even when we agree, we don't! I pick on 16 bits as that is perfect audio reproduction forever as CD's often mention.
What that means is that audio equipment that uses carbon comp. resistors as an absolute reference (As in a heavy feedback circuit) will be limited to performance less than the digital processes.
Of course the original CD's were only 9 bits linear so even older equipment was often better than that.
So the observation that less feedback sounds better may well be true for those folks who still use carbon comp. resistors. (Yes there are designs where carbon comp. are spec'd to give an amplifier it's "Voicing.")
Now when folks really have to design to meet performance spec.s sometimes many hours are spent figuring out where something went wrong. So that DA module probably got built the way it was because the designers either knew what it would take to maintain 8 bits over the entire environmental range required or it failed at first and they figured out what didn't work.
Now as you know those designs were done with hard well defined performance requirements and real calibrated and traceable measurement results.
To my way of thinking the difference between an Engineer and a Technician is that a Technician given a well defined problem can take the tools at hand and build a product. An Engineer can take a fuzzy specification, refine it until there is a firm definition and then can design or assign the design for the required product.
Now we look at audio reproduction. A fuzzy goal or even a very, very, very fuzzy goal. What is the bandwidth required? 15 -15,000, 20-20,000 and 3.16-192,000 all have been used at some time. What absolute dynamic range? There are places that actually are - 9 dba! Should we be able to reproduce to 120 or even 150 db? Lets not even get into distortion. THD is often quoted as if it is the best method, but there is lots to show it may not be valid past first order estimation.
So duh, they made metal film and wirewound resistors way back because they knew they needed them. All the way back to Wheatstone and Heaviside using strain gauges to design bridges!
Now if some folks here designed bridges:
Wavey would make them big and huge and out of concrete.
SY would use glass!
J.C. would build a beautiful aluminum web and bankrupt us all.
You would use a suspension bridge as that is the classic approach.
T would use stones!
I'd still be testing the foundations.
ES
Running warmer does give you more even order harmonic distortion. Test show that. Now there are certainly folks who like more even order harmonics, but with a low quality resistor you might be able to note the change. Al changes that anyone does always sound best to them.
Here are graphs of tempco of Russian ceramic capacitors (I believe Chinese and American capacitors have similar graphs). No straight answer, is hoter better, or cooler. Some of them have exact temperature of minimal tempco.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Ed: nice simile.
To extend it a tad, may I suggest the work of Robert Maillart and Pier Luigi Nervi: using concrete can be elegant
Exactly! Concrete is very flexible material. I can mold roses from it.
- Status
- Not open for further replies.