Not bad, That's a very engaging analogy there, Kt. I sense you are about to start exploring the very silicon structures for fuller answers to this. Be sure to return with full sanity, though. You may need it later. You seem to hit the nail on the head with the delicacy of the LTP current balance... As a transconductance amp. with a gain around 4, it is the balance we want to preserve. I suspect this also reinforces the view that we need a a large bag full of unobtanium 100V matched duals to do it.
Still, the pith of the matter has not yet been touched on; having scrubbed the signal of its even harmonic richness, how can we now avoid the clinical coldness of low THD and still maintain these subtle soundstage details?
Ian Finch
You don't need expensive duals to accomplish this. The LS352 and LS313 certainly do make life easier, but BC550C , MPSA18 etc. are dirt cheap, and you can buy a hundred or so and get plenty of matched pairs to bond together. You just need patience. Many of those from the same batch are pretty close together in HFE and VBE anyway. I use both the LS devices and bonded pairs in my Class A preamp,Class A Headphone amp, and 15W /Ch.Class A.
Other DIYAudio members in Sydney match 2SA970 etc for their 100W/Ch SC ULDs etc.
Clinical coldness of low THD ? That is absolute garbage !
At least in Class A amplifiers and preamplifiers with these mods. They have a beautiful natural warmth, with a great 3D soundstage on good material.
Hell, even the American soapie "Days of our Lives "on DTV usually has a very good soundstage and great LF effects .You also need a far better than average DAC as well though.
SandyK
You don't need expensive duals to accomplish this. The LS352 and LS313 certainly do make life easier, but BC550C , MPSA18 etc. are dirt cheap, and you can buy a hundred or so and get plenty of matched pairs to bond together. You just need patience. Many of those from the same batch are pretty close together in HFE and VBE anyway. I use both the LS devices and bonded pairs in my Class A preamp,Class A Headphone amp, and 15W /Ch.Class A.
Other DIYAudio members in Sydney match 2SA970 etc for their 100W/Ch SC ULDs etc.
Clinical coldness of low THD ? That is absolute garbage !
At least in Class A amplifiers and preamplifiers with these mods. They have a beautiful natural warmth, with a great 3D soundstage on good material.
Hell, even the American soapie "Days of our Lives "on DTV usually has a very good soundstage and great LF effects .You also need a far better than average DAC as well though.
SandyK
it seems that there is a body of opinion that agrees that true balance of the LTP is audibly important.
Can your summary above be applied to a jFET input LTP?
I would state that the much wider tolerance of jFETs makes it more important to remove unbalance in the LTP parameters to achieve that craved for sweet spot.
For a jFET ltp the summary becomes:
matching Vgs and transconductance of the LTP and balancing them could give a real boost of the sonic and soundstage quality.
I will state that selecting perfectly matched Idss devices does not get anywhere near creating a balanced pair.
Comments????????
this reminds me of Leach's explanation of linearising his front end.
Are you both in the same ballpark?
Andrew
Perhaps it would need to be left to the semiconductor manufacturers ?
Linear Systems seems to make some fairly closely matched duals.
Alex
http://www.linearsystems.com/datasheets/LSK389.pdf
Last edited:
I do not know what parameters are matched, nor do I know the tolerance allowed in the matching of those parameters.
What I do know is that both Toshiba and Linear Systems claim a matching tolerance for Idss that is better than 0.9:1
That is upto 11% Idss imbalance. That is worthless.
However, measuring actual duals shows that Idss matching is much better than 0.9:1 possibly around and better than 0.96:1
I found 1 good 389 in my batch of 20 and it was nowhere near good enough to get into my super-matched pairs that I have and sell.
The big advantage of the 170 pairing in the 389 is temperature tracking. But is that close a temperature track of such importance if transconductance and Idss and Vgs at operating current are well out of decent matching?
What I do know is that both Toshiba and Linear Systems claim a matching tolerance for Idss that is better than 0.9:1
That is upto 11% Idss imbalance. That is worthless.
However, measuring actual duals shows that Idss matching is much better than 0.9:1 possibly around and better than 0.96:1
I found 1 good 389 in my batch of 20 and it was nowhere near good enough to get into my super-matched pairs that I have and sell.
The big advantage of the 170 pairing in the 389 is temperature tracking. But is that close a temperature track of such importance if transconductance and Idss and Vgs at operating current are well out of decent matching?
Hello Andrew
I found much easyer to find very well match bjt, I buy them in stippe so they come from the same batch.
I've did buy some 170 jfet and I did not found any good match, I did not want to buy a load to find some match pairs, so I end-up puting them in my parts bin.
Some preamp circuits like those from Bob Cordell and Boberly need well match jfet. And, for me, it was to much trouble to have good batch of well match jfet to do them. So I've go back to bjt circuits.
I use the M-Cubed Electronix Semiconductor Analyser to match my transistors, bjt and jfet.
Bye
Gaetan
I found much easyer to find very well match bjt, I buy them in stippe so they come from the same batch.
I've did buy some 170 jfet and I did not found any good match, I did not want to buy a load to find some match pairs, so I end-up puting them in my parts bin.
Some preamp circuits like those from Bob Cordell and Boberly need well match jfet. And, for me, it was to much trouble to have good batch of well match jfet to do them. So I've go back to bjt circuits.
I use the M-Cubed Electronix Semiconductor Analyser to match my transistors, bjt and jfet.
Bye
Gaetan
Last edited:
....................................................
The MF A5 trivista job suits me fine for CDs
Ian
Ian
Assuming that you have a well matched input pair, you could try cutting the connection between the CM collector and the LTP collector and inserting a selected 1% MF resistor.You can work out the value by knowing the actual emitter current of the RHS device, and reading the difference voltage between the LTP collectors.. PCB layout in the 15W Class A makes this easier to do with the SC15W Class A than the SC 20W Class A. The SC ULD2 is VERY easy, as it just involves replacing a link in the front middle of the PCB with a MF resistor of typically 130 ohms, or a trimpot to get even closer.A couple of Sydney DIYAudio members can verify that this works, and Alfred confirmed this in a ULD2 post.
"There is a large and growing body of Forum members who also openly express dislike for low THD product and philosophy. I am sure that you are aware of this too."
Especially in Hugh's little area here !!! I agree with Hugh that the ear seems capable of being much more sensitive to artifacts at vanishingly low levels than they are supposedly capable of. Perhaps even finding extremely low crossover distortion residuals annoying too ?
Alex
Assuming that you have a well matched input pair, you could try cutting the connection between the CM collector and the LTP collector and inserting a selected 1% MF resistor.You can work out the value by knowing the actual emitter current of the RHS device, and reading the difference voltage between the LTP collectors.. PCB layout in the 15W Class A makes this easier to do with the SC15W Class A than the SC 20W Class A. The SC ULD2 is VERY easy, as it just involves replacing a link in the front middle of the PCB with a MF resistor of typically 130 ohms, or a trimpot to get even closer.A couple of Sydney DIYAudio members can verify that this works, and Alfred confirmed this in a ULD2 post.
"There is a large and growing body of Forum members who also openly express dislike for low THD product and philosophy. I am sure that you are aware of this too."
Especially in Hugh's little area here !!! I agree with Hugh that the ear seems capable of being much more sensitive to artifacts at vanishingly low levels than they are supposedly capable of. Perhaps even finding extremely low crossover distortion residuals annoying too ?
Alex
Last edited:
Hi Alex,
Good point. THD embraces all the harmonic artefacts, from H2 to hells gate. But we know that H2, H3 and H4 are all harmonious from simple music theory, at least on the diatonic western scale (the principal, seven note scale used in western music). If THD measurements omitted the second, third and fourth harmonics, leaving only H5 and beyond, we would have a more sympatico measurement system which should give greater correlation between the objective measurement system and the subjective listening experience.
This can of course be done with a simple spreadsheet, after inputing the measured distortion analysis. The measured levels of H2 and H3 can then define subjectively whether the sound is 'warm', or 'sharp'.
I have in fact found this useful in my analysis. If, for example, 9% of the thd is H5 and beyond (as in one of my amps), and thd as measured is 0.0404% for 25W 1KHz into 8R, the clear indication is that 91%, or 0.0368%, is H2, H3 and H4, and only 0.00364% is H5 and beyond. Such an amp sounds better than one where thd is identical, but only 70% is H2, H3 and H4.
Cheers,
Hugh
Good point. THD embraces all the harmonic artefacts, from H2 to hells gate. But we know that H2, H3 and H4 are all harmonious from simple music theory, at least on the diatonic western scale (the principal, seven note scale used in western music). If THD measurements omitted the second, third and fourth harmonics, leaving only H5 and beyond, we would have a more sympatico measurement system which should give greater correlation between the objective measurement system and the subjective listening experience.
This can of course be done with a simple spreadsheet, after inputing the measured distortion analysis. The measured levels of H2 and H3 can then define subjectively whether the sound is 'warm', or 'sharp'.
I have in fact found this useful in my analysis. If, for example, 9% of the thd is H5 and beyond (as in one of my amps), and thd as measured is 0.0404% for 25W 1KHz into 8R, the clear indication is that 91%, or 0.0368%, is H2, H3 and H4, and only 0.00364% is H5 and beyond. Such an amp sounds better than one where thd is identical, but only 70% is H2, H3 and H4.
Cheers,
Hugh
But what if there are some people out there who enjoy listening to H5 and beyond? In which case the statement that it sounds better is a prejudice against a particular subset of the population.
What about all the other factors which affect whether it sounds better or not, have these gone out of the window?
Thanks for the tips, Alex. I will look at that but I did get a very fortuitous balance here when I sweated over the CM last year. The amp was used as a ref. in a Hi fi audition lounge and the comments were not as the owner and I anticipated. We both thought it a really great amp. It's at another lounge now as there is technical interest in class A but I think they are hoping to hear JLH SE quality sound!
There is no discrimination here, just building a product that fits a criteria that Hugh believes in. He doesn't expect to please everyone; if you're not satisfied you buy something that does satisfy you. If he tried to please everyone, he wouldn't impress anyone.
There's no reason to assume other factors have gone out the window. We just aren't discussing them right now.
- keantoken
Oh, and why don't we try this?
If we want to exclude H2, 3, and 4 from the THD analysis we can run it on the 4th harmonic. So if our test signal was 1KHz we would set the distortion meter at 4KHz, then it would just measure the 5th harmonics and beyond. In LTSpice you would do this like:
.four {freq*4} V(Vout)
Thoughts?
- keantoken
If we want to exclude H2, 3, and 4 from the THD analysis we can run it on the 4th harmonic. So if our test signal was 1KHz we would set the distortion meter at 4KHz, then it would just measure the 5th harmonics and beyond. In LTSpice you would do this like:
.four {freq*4} V(Vout)
Thoughts?
- keantoken
Ah... Um... Ahem.
I just realized why that won't work. It would only look at harmonics of the 4th harmonic, so it would look at 8th harmonic, 12th harmonic...
Oh well. It was thought.
Sorry!
- keantoken
I just realized why that won't work. It would only look at harmonics of the 4th harmonic, so it would look at 8th harmonic, 12th harmonic...
Oh well. It was thought.
Sorry!
- keantoken
Nice try though...... I think it needs to be subtractive of H2, H3, and H4, and since it does compute each and every harmonic, this should be possible. Not sure how to do it however.
Another attempt perhaps, KT?
Another attempt perhaps, KT?
Well, we could do it by hand, but getting the process automated is the trick...
I don't think there is any way to change the way the .four command calculates. I suppose what you could do is run two separate .four commands, one computing all harmonics, the other computing only H2, 3 and 4, and then subtract the latter from the former?
But then again, 50% distortion can still sound good. So it seems we need to calculate the harmonics in relation to one another, maybe even to the exclusion of the fundamental?... I've heard that odd order harmonics aren't so bad as long as they have even order counterparts.
It would be nice to rig up a microcontroller-based harmonic distorter, and program in different transfer curves for different harmonic profiles, have it switch randomly and gather subjective results...
Hugh, I read you used to do stuff with organs, and have a deep background in music theory? I hadn't heard about this. Do you play the moonlight sonata on a massive pipe organ, deep into the night? Do you know of any good links on the musicality of harmonics?
I have a very annoying sounding cheap electronic keyboard, it is very unmusical. Passages full of sour notes can sometimes sound good in a strange way, and I now suspect this is because the unmusical harmonics actually harmonize when you play something sour... But of course the effect falls apart when you play something right. I guess it is in favor of bad players, and this is a selling point I think...
- keantoken
I don't think there is any way to change the way the .four command calculates. I suppose what you could do is run two separate .four commands, one computing all harmonics, the other computing only H2, 3 and 4, and then subtract the latter from the former?
But then again, 50% distortion can still sound good. So it seems we need to calculate the harmonics in relation to one another, maybe even to the exclusion of the fundamental?... I've heard that odd order harmonics aren't so bad as long as they have even order counterparts.
It would be nice to rig up a microcontroller-based harmonic distorter, and program in different transfer curves for different harmonic profiles, have it switch randomly and gather subjective results...
Hugh, I read you used to do stuff with organs, and have a deep background in music theory? I hadn't heard about this. Do you play the moonlight sonata on a massive pipe organ, deep into the night? Do you know of any good links on the musicality of harmonics?
I have a very annoying sounding cheap electronic keyboard, it is very unmusical. Passages full of sour notes can sometimes sound good in a strange way, and I now suspect this is because the unmusical harmonics actually harmonize when you play something sour... But of course the effect falls apart when you play something right. I guess it is in favor of bad players, and this is a selling point I think...
- keantoken
A few people (me included) who bothered to try this close balancing all heard a difference for the better.
Most small transistors are cheap so its not expensive to buy a bunch and match up, certainly worth doing imo