the parasitics presented to the -IN is the critical node.
The length of wire connected to -IN must be minimised.
All the feedback components should be at the -IN node.
The wires on the other side of the feedback components are generally at lower impedances and less susceptible to interferences/capacitances/inductances.
Or maybe several guards driven at different amplitudes.
If you picture the resistor chain then the amplitude of the voltage at the beginning of the chain is greater than at the summing node end. It appears to me that you can match one part of the resistor chain but not another if only one guard is used.
Or am I completely misunderstanding?
Agreed that the feedback resistors are critical for THD. Should be 10X lower than THD of the amplifier. But note, voltage nonlinearity distortion of resistors is probably soft and mostly third.
How low ttc do you think we need to go to get voltage nonlinearity to less than 0.0001% at 40V rms?
I have not measure voltage nonlinearity, but it would be interesting to measure it on a 100ppm metal film resistor at a high enough frequency that temperature distortions would be much smaller.
Does resistor voltage distortion go up linearly with voltage or as the square or square root?
Cheers,
Bob
One of our Members built a "Blameless" using the Signal Transfer Company PCB.
It had very high distortion.
He reported his quite long story here and eventually found it was due to getting near the voltage rating (I think he was under the resistor's voltage specification limit) of the feedback resistor.
I think he solved it by series connecting two half value resistors so that each saw only half the peak output voltage.
This solution covers two different distortion mechanisms.
It reduced the voltage distortion and it also reduced the temperature induced resistance variation distortion.
It had very high distortion.
He reported his quite long story here and eventually found it was due to getting near the voltage rating (I think he was under the resistor's voltage specification limit) of the feedback resistor.
I think he solved it by series connecting two half value resistors so that each saw only half the peak output voltage.
This solution covers two different distortion mechanisms.
It reduced the voltage distortion and it also reduced the temperature induced resistance variation distortion.
May I ask what is considered "very high distortion" in the context of a Blameless? 🙂
I got something like 0.008% out of mine at 10kHz and 100W (80kHz measurement BW) At 1kHz the distortion due to the amp was below the 0.002% floor of my old Tektronix analyser. As the power output was reduced, the THD reading became dominated by noise.
These figures are quite high compared to the ones I see from Spice simulations using more advanced error correction, but I think they are respectable for a real amplifier measured on the bench.
The feedback resistor was a fairly ordinary 10k, 0.6W, 1% metal film part with a voltage limit of 250V.
I got something like 0.008% out of mine at 10kHz and 100W (80kHz measurement BW) At 1kHz the distortion due to the amp was below the 0.002% floor of my old Tektronix analyser. As the power output was reduced, the THD reading became dominated by noise.
These figures are quite high compared to the ones I see from Spice simulations using more advanced error correction, but I think they are respectable for a real amplifier measured on the bench.
The feedback resistor was a fairly ordinary 10k, 0.6W, 1% metal film part with a voltage limit of 250V.
May I ask what is considered "very high distortion" in the context of a Blameless?
A TMC LIN should be around <.001% THD 20K @100W/8R.
One builder went SMD with my LIN , got 3-5ppm 20K. His power supply and
test cable routing would make the results fluctuate.
Output stage choice is the limiting factor for this particular topology.
Driving a error correcting class A (or headphone output stage) -sub PPM.
OS
How were these results measured? An Audio Precision or a soundcard or what?
FWIW, I used plain dominant pole compensation and a double EF output stage. Higher order compensation might have pushed the figure a bit lower, but I didn't think it was worth the risk.
FWIW, I used plain dominant pole compensation and a double EF output stage. Higher order compensation might have pushed the figure a bit lower, but I didn't think it was worth the risk.
Last edited:
Just a question about the VAS :
I am currently reading an older version of the book, but I wonder what the effect would be exchanging the darlington in the VAS by a triplet? Would that cause instability or just raise the open loop amplification?
I am currently reading an older version of the book, but I wonder what the effect would be exchanging the darlington in the VAS by a triplet? Would that cause instability or just raise the open loop amplification?
A very good commercial soundcard - link ..... http://www.diyaudio.com/forums/solid-state/248105-slewmaster-cfa-vs-vfa-rumble-66.html#post3965982
5th element ported and tested the "wolverine" IPS..
I'm glad I re-read his entries - he used "dirty" mje340/350 for the pre-drivers and
2sc/.a 3503-1381 for the vas. This is what I have to build mine- hope I get his results.
OS
I doubt that making the VAS a Triple would make any improvement, even if it was stable with the local Miller compensation feedback.
Cheers,
Bob
By re-reading Bruce Hofer's article, I believe that the answers are given in it.
Non linear model for voltage distortion is given and it is an absolute value proportionality ! giving :
HD2 = 0 HD3 = |Vcr x Vs| / 5.9 HD5 = HD3/7
According to BH good metal films have Vcr between 0.1 and 1 ppm. This must be ( I believe)for résistors with ttc < 15ppm. In data sheets only ttc are given but according to BH it correlates well with Vcr.
He wrote that a resistor with a respectable Vcr of 10ppm Under 50V (typical of a feedback resistor) gives 0.05% distortion 😱 If we apply his formula for HD3 it is only 0.01% or there is something I miss, but still😱
The dynamic temperature coeff distortion is also given and showed to be significant Under 200hz;
Based on this he gives two rule of thumbs beside using lowest ttc résistors:
keep voltage drop Under 3Vrms and power dissipation Under 20mW peak;
This would require 20 résistors in serie for a feedback resistor Under 50V; With 15ppm ttc it is perhaps economically feasible but is it overkill?
What is the highest ppm ttc you would recommend
Thanks
JPV
Just like if a 3EF is supposed to be "better" than a 2EF then a 4EF must be the holy grail. But in my experience the 4EF gave nothing extra just more problems.
Found the same with the VAS triplet. Other distortions become dominant.
This was all done in sims.
Paul
More drivers only worsens problem of the voltage distortion between the input and output of the OPS. Even if you've eliminated that somehow, the Early effect and Cob of the drivers tends to limit how high the input impedance can be.
I agonized for some time over whether to use a double or triple EF output stage. I was most worried about thermal compensation and stability of the triple.
In the end, I figured that given the high beta of OnSemi's new output transistors, I could just about get away with a double, if I used the beta enhanced VAS and ran it at 10mA.
In the end, I figured that given the high beta of OnSemi's new output transistors, I could just about get away with a double, if I used the beta enhanced VAS and ran it at 10mA.
It's all trade offs... I more worry about thermal issues than stability with 3EFs.
Wondering which is the bigger evil a beta enhanced VAS with 3EF or a triple VAS and a 2EF (something I've never investigated)?
Don't be afraid of ouput triples, quite easy to tame. I use my TT amp daily with no problem what so ever. Here is the schematic http://www.diyaudio.com/forums/solid-state/182554-thermaltrak-tmc-amp-15.html#post3065943 just there is an error in the bias spreader. It was done according Cordell book, but there is an error(in the book) and R44 should be a pot of 2K and P1 should be fixed resistor of 6k8.
Damir
Damir
I would like to know the Voltage coef for these affordable résistors. If Vcr is lower than 1ppm it is ok
Thanks for this info, JPV.
There is something that seems amiss:
"He wrote that a resistor with a respectable Vcr of 10ppm Under 50V (typical of a feedback resistor) gives 0.05% distortion If we apply his formula for HD3 it is only 0.01%".
In my experience there seems no way that even a 100ppm metal film resistor could create 0.05% distortion, even under full output voltage conditions (say 40V rms or more) as long as the dissipation was under about 10% of rating. I don't even think a carbon film would be this bad. Could it be that the 0.05% number is off by a couple of orders of magnitude?
Bruce is extremely smart and seems never to be wrong, so maybe we are somehow mis-reading what he is saying or not getting the context right.
Assuming resistor voltage nonlinearity is not frequency-dependent, how could we then routinely get sub-0.001% THD-1 on many amplifiers where 100ppm metal film feedback resistors are used?
I wonder if BH is referring to very low-wattage SMT resistors, if it makes a difference. Is it possible that very low wattage SMT resistors are more prone to voltage nonlinearity than larger conventional discretes?
Do you have a link to the Hofer article you are referring to?
Unless I can somehow find a difference from my experience, I would say a 100ppm metal film with a very generous dissipation rating would be fine.
One approach would be to take a very good existing amplifier, measure its THD-1 in the most sensitive possible way (maybe using a twin-T notch and a spectrum analyzer, or some combination using the distortion magnifier), and then successively putting in crappier feedback resistors until the measured distortion increases.
Cheers,
Bob