Scott, no I had and still have no idea what your post was intended to say. The language was convoluted and unclear. If it was an opamp... never mind. 🙄
Scott, we are not YET at the point where one might conduct a "properly controlled test".
Scott, what does this mean:
"there is no I know which is which but I'm going to be honest and say I can't hear a difference. This opportunity does not present itself. " ??
Do you mean to say that this does not happen IF a "properly controlled test" is conducted?
You seem frustrated or vexed that apparently someone else is hearing something that you are not? Or something else that is touching you in a reasonably strong way is going on. I don't know what it is, or why anything in Ed's proposed test would do that.
All that I said is that I would report what I heard without any attempt to hear something that I did not hear.
The things I will be listening for will likely not be tremendously effected by the noise floor, unless the noise floor sounds like a white noise generator - in other words I can hear the noise floor as being loud enough so that it manifests as audible noise. That's a rather redundant way to say that.
And, should any be grouped, or indistinguishable to me, I'll report that.
So, what's wrong with this?
It's entirely preliminary as far as I am concerned.
A first, baby step.
Right, but that eliminates anything practical i.e. will ever happen. We have been there a few times parties have withdrawn.
How many times do I have to say I could care less what you think you hear, let's let Ed referee this if he wants to I have no interest in wasting my time. If you don't understand that listening to two amplifiers with audibly different noise floors removes the B from DBT I am disappointed.
Well, now that you go through all this, it would be interesting to look at the preference and compare it with 'the usual' measurements - noise, THD, frequently response should catch most measurable differences. But you or someone else can do the measurements afterwards; just don't throw them away!
Edit Ed actually you should not have disclosed what you are going to put in the tin. Now we know the opamps, now Bear already wants to limit it to specific types. The test is no longer totally double blind now. So the potential for discussions afterwards already opens up. Too bad.
Edit2 Scott''s comment brings memories of the Tiefenbrunn test where one listener nailed the differences perfect after he realised the noise floor was different. So the purpose of the test to hear difference with or without an ADC DAC chain in the system was lost.
Jan
Last edited:
Got a full BOM, I would be interested in doing a little PCB for a bit of mental exercise, as I have been away from my beloved work computer now for over 4 weeks due to the bloody op.
Single 741 running at 40dB vs 6 buffers in cascade would make totally different distortion profiles (like with 19+20k IMD). No loop gain available for feedback above a few kHz, so it running open loop at 20kHz! Even with the gain drop accounted for in the pre-scaler this is not a reasonable test. With a typical 10MHz GBW opamp it's not fair either, with only ~20dB of gain reserve at 10kHz. Noise issue was mentioned already.
I think there is no way around testing the DUTs at realistic gains (<20dB).
I think there is no way around testing the DUTs at realistic gains (<20dB).
Exactly. That's why 2-3 opamps in series sound better than a single one for the same gain. And that's why RIAA correction with opamps has to be in the feedback loop.
I don't know. I would think I could isolate the 741 at 40dB.
OTOH I know I can't reliably tell a chain of about 10 MC33078's from much better spec'd OPA1612's, in a complex active crossover + signal conditioning, all running at low gains, non-inverting, and with more load current a 33078 would be easy with, spec-wise (wrt to Samuel Groner's work).
OTOH I know I can't reliably tell a chain of about 10 MC33078's from much better spec'd OPA1612's, in a complex active crossover + signal conditioning, all running at low gains, non-inverting, and with more load current a 33078 would be easy with, spec-wise (wrt to Samuel Groner's work).
Last edited:
Well here is the problem, I want to test a single opamp. If the input resistance is more than 60 ohms for some of the devices the input resistance will dominate the noise sources. With 20 dB of gain that would put the input resistance at 600 ohms which is lower than many consumer pieces can drive. The IEC standard is 2K ohms.
By the metric that 40 dB of gain is too much, 600 ohms is too low then the conclusion would be opamps require additional circuitry to perform.
By the metric that 40 dB of gain is too much, 600 ohms is too low then the conclusion would be opamps require additional circuitry to perform.
Ed maybe you can maintain the gain structure but adapt resistor absolute values so that all sets have similar noise performance?
Jan
Jan
Please specify what levels of noise floor, for this proposed test, and in your expert opinion will be "audible"?
This would be the first step to avoiding such a problem.
_-_-
Jan, you misunderstand.
I suggested that we start with a limited number of test subjects.
Trying to group a dozen +/- "black boxes" (some of which may or may not be duplicates) is not simple, EVEN IF a majority of them do present audible differences.
So, I suggested to start with TWO that I have already heard as being audibly different in specific situations, plus a third.
ALL of the "black boxes" are unknown to me as far as what is inside.
So, assuming this works as expected based on my prior experience with the 5534 and the 797, plus the third "wild card" being added in too, THEN it makes sense to consider more opamps.
Consider this as a preliminary effort. A test of the test.
I still have various concerns about the proposed method/circuit.
These can wait a bit to be voiced.
This is why, again, I would have preferred to work out these details (at least in the main) behind the scenes, before bringing it forward in public.
Please excuse this remark, but I thought that a lot of people were saying that ALL of this distortion is below (can we say "most") audible thresholds??
Last edited:
^ Well, yes, but only when used properly. For a 741 this would mean very low gain and low output amplitude. I don't know, never tried a 741 but I supect an inverting buffer with something like 2...3k source impedance at both pins, driving 10k to 1Vrms could well be 'transparent enough' for most people (except for the sound change from polarity flip which must be accounted for elsewhere in the listening chain).
If I may construct a very simplified explanatory story in response to bear's comment: It looks like opamp data sheets typically specify distortion for 1 kHz input signal and with the amplifier configured for unity gain. Presumably, data sheets may bear some similarity to resumes, where the objective is to paint the candidate in the most favorable light in order to obtain an interview. Next, people, without doing any particular circuit design, may look at numbers given on a data sheet and compare that directly with existing numbers for human distortion perception, and then conclude nobody could possibly hear that. Once forming that conclusion, it's not too much more of a leap to conclude that anyone who does claim to hear distortion from opamp circuits must be imagining things.
This could become quite complex. We could end up with a clear preference in the use cases you propose, but someone could rightfully make the case that the use case is not realistic.
Sort of like trying to establish whether two car types have a different drive experience by looking at how much they drift when hard cornering at 120 miles/hr.
Jan
this seems to be a common misapprehension about DBT - the "Blinding" is only applied for the actual listening trials
you can know everything about the devices, test setup
training with all test, device, signals knowledge to guide is perfectly acceptable if the test is trying to get information on human hearing thresholds
that's also what the A/B switching is about too - you can spend as long/as many trials unblinded to establish any audible difference in A vs B
only A/B identity need be randomized and Blind during scoring listening trials
- Status
- Not open for further replies.