Actually, the error voltage is inversely proportional to the loop gain. The error magnitude can be calculated as:
em = |1/(1+1/T(s))|-1, where T(s) is the loop gain as function of (complex) frequency. The error magnitude approaches em = |1/T(s)| as |T(s)| approaches infinity. In other words, errors are attenuated by |T(s)| for large values of |T(s)|. At low values of |T(s)|, such as the 5-10 dB you mention, you need to use the entire equation and the amount of error correction will be slightly smaller than |T(s)|.
This app note may provide some insight: http://www.ti.com/lit/an/slyt374/slyt374.pdf
Tom
Last edited:
Ah, that makes sense (I'd forgotten that, and thanks for the ap-note!).
I was more referring to the fact that adding heaps more feedback to a situation will knock out a bunch of non-linearities, but not all. At a certain point, those other non-linearities (that were buried) will dominate the performance characteristic. E.g. noise. In short, I'd be very surprised if the performance of your circuit would change (much) if somehow you had another 20 dB of loop gain to apply. You're so far up the diminishing returns curve. (That's not to slight your offering in the least! It's just slamming up against what you can achieve here, and so far past audibility).
I was more referring to the fact that adding heaps more feedback to a situation will knock out a bunch of non-linearities, but not all. At a certain point, those other non-linearities (that were buried) will dominate the performance characteristic. E.g. noise. In short, I'd be very surprised if the performance of your circuit would change (much) if somehow you had another 20 dB of loop gain to apply. You're so far up the diminishing returns curve. (That's not to slight your offering in the least! It's just slamming up against what you can achieve here, and so far past audibility).
I truly appreciate Tom's physics based approach 🙂 It only helps that the Mod-86 sounds great too!
Thanks Tomchr for that simple plain english answer. The difference between 225 & 250 is not important in the context of the loopgain and it's error correcting capability. The mistype is excusable. You are completely forgiven in my book. I do much worse.
It suspect this cost you quite a bit of work. Thanks again.
Can anybody see that the comment "the total loop gain applied to error-correct the LM3886 is approaching 250 dB" applies only at frequencies at, or below, 30milliHertz?
The comment with respect to error correcting of audio signals could more usefully be stated as 167dB @ 20Hz, or 49dB @ 20kHz or some other value within the audio band.
Last edited:
Yes, the system will (to a large degree) assume the noise of the error input stage. As has been recognized, Tom is probably getting down towards the intrinsic specs of the THAT1200 and LME49710. Liquid nitrogen is probably his next step. 🙂
That's true. I'm not sure anyone understood those were the numbers you were after, though.
I would go even further and argue that 49 dB @ 20 kHz is the important number, under the unstated assumption that the loop gain is higher at lower frequencies. Maximizing the loop gain at 20 kHz was part of the mission of the R1.0 -> R2.0 update.
I'm still working on sourcing the materials for the cooling chute. For the ultimate sonic purity, it must be machined from a solid block of unobtanium by virgins. Sourcing the unobtanium is the easy part... 🙂
On a more serious note: As you mention, the distortion performance is likely limited by the THAT1200. That said, 0.000067 % is pretty darn low THD. One could toy with the idea of implementing the differential receiver with LME497x0s, but this is likely to cause a significant reduction in CMRR while not providing any meaningful improvement of the already vanishingly low THD, so I don't see any reason to go there.
Tom
BTW: Parallel-86 boards are now back in stock. The first batch sold out rather suddenly, so I had to get more made. That's alright by me, though... 🙂
Tom
Tom
Wh-Wh-whoa! 😱
This is DIYaudio not "buy a module and never touch it audio!" (BAMANTIA) 🙂
I, for one, want to lower the gain (past your max/min) so I can basically feed it a classic 2Vrms source through ... NO vol control! And enjoy low to medium listening levels (of course relative.)
Do I need to alter compensations with these lower gains?, which module at which freq? ...
I think you know the best balance of gains and appropriate compensations ... but will save it for PM! 😉
Cheers,
Jeff
Last edited:
erm. It's not a module. It's a PCB that you then solder up yourself. Totally DIY, and certainly what a lot of members want to do.
How much less gain do you want? You can reduce the gain up to 6dB by putting another THAT chip in. There are also some mode for very low power output setups.
How much less gain do you want? You can reduce the gain up to 6dB by putting another THAT chip in. There are also some mode for very low power output setups.
Jeff--I'd be lowering the gain via using a THAT1206 instead. I realize that this means you don't really have to tweak things, but it's probably for the best to leave things otherwise alone. 🙂
I actually wanted to lower the gain of the MOD86. A gain of 1 V/V (straight wire, no gain 🙂) would be nice. I was not able to make the compensation work for that without sacrificing loop gain at 20 kHz, hence, performance. I've decided to live with a gain of 10 V/V (20 dB).
As others have pointed out, you can lower the total gain of the MOD86 or PAR86 by swapping the THAT1200 for a THAT1206.
Tom
As others have pointed out, you can lower the total gain of the MOD86 or PAR86 by swapping the THAT1200 for a THAT1206.
Tom
semiconductors don't all work well in liquid nitrogen.
I believe many start to give up their useful properties at around -60°C
I believe many start to give up their useful properties at around -60°C
Odd. I clearly remember an experiment dunking a CMOS ring oscillator into LN2 at uni. As expected frequency changed with temperature. Which particular properties are you thinking of?
What changes with temp...White noise, 1/f noise ?.
What happens when approaching and then going lower than -60C ?.
Dan.
I don't know the details.
But a cursory look at datasheets will show a minimum operating temperature in the range -40°C to -50°C
I did find this
"Temperature Dependence of Semiconductor Conductivity
(Originally contributed by Professor E.D.H. Green)"
But a cursory look at datasheets will show a minimum operating temperature in the range -40°C to -50°C
I did find this
"Temperature Dependence of Semiconductor Conductivity
(Originally contributed by Professor E.D.H. Green)"
Last edited:
Member
Joined 2006