John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
Jump to Latest
#40,681
john curl said:
I actually try things, before I condemn them, and I sometimes even open one up (destroying it, naturally) to INDEPENDENTLY verify if I am being lied to, or taken for a chump.
Click to expand...

Ears-only testing will tell you much more quickly and more reliably. I know it's difficult to believe, but there may be some things you'd take apart that you don't actually understand- or "understanding" that you have which isn't borne out by using real sensory testing. Crazy talk, I know, but some of us actually trust our ears instead of just reciting that as a mantra.
 
#40,683
haha now its me silly again; angry math fail. I think I only did the less than half in my head, or just a minor brain explosion.

revised. 16A@ 250V ~4000W (not 1500W 😕 ) for each mono amp channel more than meets my needs OK ...

it seems to have met these specs and all of the relevent regulations including IECEE, UL, DVE under IEC 60 529 and for breaking ability, which the old one didnt have, so couldnt be used legally here for standard AC cords.
 
#40,684
Alexandre said:
Relevant to this discussion, from the AD1862 datasheet: Output Impedance (±30%) 2.1 kΩ

I think this figure of ±30% has to do with the code dependence.

-Alex
Click to expand...

I would be surprised if it was the code dependency that causes the variation.
If that were true, a current-input I/V (with virtual gnd input) would register huge distortions due to the output current varying with code (separately from the code value itself of course).
I would expect this to be process variations part-to-part.
No prove but seems logical to me.

jan
 
#40,686
Quote:
Originally Posted by Alexandre
Relevant to this discussion, from the AD1862 datasheet: Output Impedance (±30%) 2.1 kΩ

I think this figure of ±30% has to do with the code dependence.

-Alex

That's fascinating, thanks! If indeed that is code dependence, it shows how important a low input impedance for the I-V is. I wonder if Scott W. knows anyone at ADI who could shed some light? [end previous post]

Further: the suggested non-drop-in replacement for the obsolete AD1862, the AD1955, does not give output impedance information, other than a capacitance maximum of 100pF, but a remark is made that passive terminations will seriously degrade performance. The opamps shown in the applications section are AD797.
 
#40,687
jan.didden said:
I would be surprised if it was the code dependency that causes the variation.
If that were true, a current-input I/V (with virtual gnd input) would register huge distortions due to the output current varying with code (separately from the code value itself of course).
I would expect this to be process variations part-to-part.
No prove but seems logical to me.

jan
Click to expand...
I don't see that following Jan.

My first thought was the nominal variation in on-chip resistors, but then I noticed that they are said to be thin-film. It is mentioned that some things are trimmed for a small variation in net gains, to make it easier to do the I-V conversion to a given output voltage.

With most beasts like this, things are made ratiometric as far as possible, and on-chip tracking is exploited. It's a bit akin to buying potentiometers: you will have a well-specified ratio of input to wiper voltage as a function of position, especially for linear taper and of course an unloaded wiper. But try finding a tight spec on end-to-end resistance. So I think here they trim some and count on tracking of others.

Well, it's time to measure some DACs I think, and maybe call for the industry to start specifying these things.

As you know I've gone round with EUVL on this issue, and criticized the use of a 20k source (your AP generator with a 20k resistor) to characterize some of his I-V discrete converters. This high an impedance greatly alleviates the distortion arising from a current-dependent variable input impedance, and would be valid if the DAC were a 20k output R --- but the part used was not that high. When I mentioned the two issues of code dependence and distortion arising from a lower DAC output resistance, and showed how to reduce such effects, it was not particularly well-received, with one person commenting that he didn't see any advantage. This was despite a 19 dB reduction in distortion at 1kHz, IF both circuits were fed from a 1kohm DAC output --- and this neglected any likely but unknown dependence on code.
 
#40,689
Alexandre said:
On a second thought...

Laser trimming was done to make it monotonic, I think, and that variation in Zout could be the result of the trimming.

Would be interesting to know how more about this "golden-age" dac.

-Alex
Click to expand...
That's plausible, although as I mentioned above ADI warns at anything much less than a virtual ground for the suggested replacement part, while not saying anything about how performance might degrade.

My impression was also that thin-film resistors can be much more tightly controlled even before trimming, but I am not an IC designer.

EDIT: If one creates a worst-case strawman DAC using an R-2R ladder with no additional fancy circuitry, the effects of code on output R can be readily seen. I believe there is some material online that discusses that specific case.
 
Last edited:
#40,691
Do the resistors need trimming to account for the imperfect associated switches?

What would be a reasonable criterium for loading impedance? Small compared to a value that caused 1 LSB error with any switch or combination of switches turned on?

My naive thought is that the trimming could be done to match some (practical) finite non-zero load resistance.

Thanks,
Chris
 
#40,692
those same recommendations are made for the ESS and are born out in use. passive resistive IV will not do better than -103dB THD+N and can be much worse. low impedance high GM fets or virtual ground will net you up to -115 (fet) -120dB THD+N
 
Last edited:
#40,697
The reason I ask is because a small enough resistor, say, a gazillionth of an Ohm, is the theoretical "perfect" load. But is a tenth Ohm small enough? How about a half? How about 3 Ohms? At some point the electronics for decent S/N becomes trivial. At some point before that it becomes heroic but possible for the dedicated DIY'er.

Also, of course, a brute force solution can operate in parallel with a transconductance input stage and/or a transformer.

As bcarso has said, we *really* need some numbers.

Much thanks,
Chris
 
#40,698
Not that I am one for flogging a dead 'orse (as we say up in't north of England) but🙂
Unwrapping the cover, I found some lead filled damping material covering what appeared to be some metallic coatings on a ceramic substrate. In any case, once exposing the 'interesting' part of the device to the air, I took it outside, with a fire extinguisher, put it in a metal pan and added water.
Guess what? Nothing happened! The guy on the internet had just guessed at what it was, and speculated further about his guess. It happens around here, sometimes too!
I ultimately had to tell Jack what I did, and I think he came to understand that I could not BLINDLY trust him, I had to test for myself.
Click to expand...
There have been these devices dismantled, and it seems the magic ingredients change regularly, for a military based material is does not seem to be very constant, and now we have lead to add to the equation, so they cant be sold due to RoHS and other worldwide restrictions on hazardus substances, never mind breaking ITAR restrictions...
 
#40,700
Having been threatened with a lawsuit over these devices, due to my comments, I have rather a bee in my bonnet over them, plus I think they are a shining example of the total lack of Integrity that a lot of high end audio runs on.
Oh the Bybee site seems to be down today, so I couldn't cheer myself up reading the distortions of truth that is the marketing blurb.
 
Status
Not open for further replies.
Top Bottom