Of course everyone is entitled to hear what he perceives.
in SOME cases I didn't perceive a difference when swapping an amp, but, that's the minority of the cases.
Maybe there is some bottleneck in your chain that equalized the various amps?
Member
Joined 2009
Paid Member
Some tubes pre have high output impedance...
And a ratio closer to infinity is definitely better as keeping out interferences and simplifying the load, so i'm all for up to a megaohm input impedance in the power amp (seen done on GOOD highend amps).
I dont need that much but if i can get 300k in place of 30k i would surely do that, if it does not brings other compromises.
Degenerated BJT LTP can reach few tens kOhm and that would be open loop input impedance of the amp. When NFB is applied equivalent Zin becomes much higher. However, since the feedback decreases with frequency, Zin decreases as well, so the Zin is apparently capacitive. Anyway, resistor that connects the base of the input BJT to GND is dominant so the effective imedance seen by the preamp is approximately equal to this resistor.
A very high input inpedence is picking up easily whatever noise you have around. Then you may be happy buying triple - active- shield - twisted- pure gold - kryptonite-plated cables and connectors. 100K (closed loop, in the audio band) seems to me already a top limit. If you have special problems (i.e. very high output Z from the pre) I would try any other cure, if possible, before raising the Zin of the Power Amp to the sky.
effebi
Last edited:
Hi,
if the source has a Hi Zout, then raising the Zin of the receiver is not a great idea.
Low to medium source impedance is always better. Low Zout ~5r0 to 200r
Med Zout ~220r to 2k
The lowish Zout limits the input noise, even when the Zin can be anywhere in the range 600r to 500k.
If you have a Zout >1k, then investigate a buffer on that output to get Zout <200r
if the source has a Hi Zout, then raising the Zin of the receiver is not a great idea.
Low to medium source impedance is always better. Low Zout ~5r0 to 200r
Med Zout ~220r to 2k
The lowish Zout limits the input noise, even when the Zin can be anywhere in the range 600r to 500k.
If you have a Zout >1k, then investigate a buffer on that output to get Zout <200r
Check out measurements of BJT vs JFET input stage in post #39 of this thread:
http://www.diyaudio.com/forums/analog-line-level/167033-preamplifier-dispre-2-jfet-4.html
http://www.diyaudio.com/forums/analog-line-level/167033-preamplifier-dispre-2-jfet-4.html
MATCHED JFETS in the input LTP generate a lower percentage of odd harmonic distortion when fairly compared with bipolar transistors. This becomes more important in complementary symmetrical input designs because this balanced topology removes a large percentage of even harmonic distortion. Bipolar input LTPs in a complementary symmetrical topology can generate greater 3rd harmonic than 2nd harmonic, while JFET input LTPs measure a 3rd harmonic that is lower than the 2nd. John Curl, Borbely, Hafler, and GOD all put their name on a complementary symmetrical topology with JFETS.
From my experience, it is important to have matched dual JFETS on a single die to get all of the benefits. If I did not have a stash of hand matched J2sk389BL and J2sj109BL I would not use JFETs.
When we move from SPICE to the real world, common-die bipolar transistors like the MAT02, MAT03 and THAT transistor arrays can have measurable benefits if considerable hand matching of individual bipolars is not done. Purchasing a "tape" of transistors usually assures they all came from the same process run, and often the same wafer.
I find that the power supply and output stage quality are more important than the input LTP device type.
From my experience, it is important to have matched dual JFETS on a single die to get all of the benefits. If I did not have a stash of hand matched J2sk389BL and J2sj109BL I would not use JFETs.
When we move from SPICE to the real world, common-die bipolar transistors like the MAT02, MAT03 and THAT transistor arrays can have measurable benefits if considerable hand matching of individual bipolars is not done. Purchasing a "tape" of transistors usually assures they all came from the same process run, and often the same wafer.
I find that the power supply and output stage quality are more important than the input LTP device type.
Attachments
not so clear cut there IMHO
I really don't agree on a couple of those points
http://www.diyaudio.com/forums/soli...ning-not-whats-your-belief-64.html#post501789
shows what I consider a "fair" comparison of degen bjt vs jfet - and bjt wins on linearity when gm, GBW are adjusted to the same for both circuits
3rd harmonic (or odd order saturation) only is the best case possible with diff pair inputs - it means you have exactly cancelled even harmonics and has the highly desirable property that reducing diff input error with high loop gain reduces input diff pair nonlinearity % by the 2nd power of the input Vdiff reduction
and there is no mechanism for dual compelmentary stages to give better 2nd order cancellation of input gm nonlinearity - it all happens locally in each pair due to device matching - unless you claim to "match the mismatch" between bias points and nonlinear terms in typically 2x differing gm and Vt n vs p channel devices?
if you mean dual complementary can balance output conductance/nonlinear capacitive effects between differing n/p channel devices - I think you're dreaming - the n vs p channel devices have different areas, doping profiles
dual complementary diff input has a problem with VAS current setting and needs extra common mode control circuity to use current mirror loads - which are pretty indespensible for linear diff pair operation and give large audio frequency loop gain - which linearizes the odd order error of the input gm as mentioned above
perhaps compementary VAS can give some VAS gm even order nonlinearity cancellation but with a "beta enhanced" locally degenerated, quality ccs fed, single ended VAS linearity isn't an issue in this stage - residual even harmonics or not
I do agree that it doesn't take exotic topology or devices to make input diff pairs, either bjt or jfet, sufficently linear that typical output stages are the major limitation in audio power amps
when you get down to single digit ppm distortion however you will need to add bootstrapped cascodes to the fet inputs due to their larger nonlinear C which causes common mode input impedance nonlinearity with typical audio power amp noninverting gain and mismatched input, feedback Z
I really don't agree on a couple of those points
http://www.diyaudio.com/forums/soli...ning-not-whats-your-belief-64.html#post501789
shows what I consider a "fair" comparison of degen bjt vs jfet - and bjt wins on linearity when gm, GBW are adjusted to the same for both circuits
3rd harmonic (or odd order saturation) only is the best case possible with diff pair inputs - it means you have exactly cancelled even harmonics and has the highly desirable property that reducing diff input error with high loop gain reduces input diff pair nonlinearity % by the 2nd power of the input Vdiff reduction
and there is no mechanism for dual compelmentary stages to give better 2nd order cancellation of input gm nonlinearity - it all happens locally in each pair due to device matching - unless you claim to "match the mismatch" between bias points and nonlinear terms in typically 2x differing gm and Vt n vs p channel devices?
if you mean dual complementary can balance output conductance/nonlinear capacitive effects between differing n/p channel devices - I think you're dreaming - the n vs p channel devices have different areas, doping profiles
dual complementary diff input has a problem with VAS current setting and needs extra common mode control circuity to use current mirror loads - which are pretty indespensible for linear diff pair operation and give large audio frequency loop gain - which linearizes the odd order error of the input gm as mentioned above
perhaps compementary VAS can give some VAS gm even order nonlinearity cancellation but with a "beta enhanced" locally degenerated, quality ccs fed, single ended VAS linearity isn't an issue in this stage - residual even harmonics or not
I do agree that it doesn't take exotic topology or devices to make input diff pairs, either bjt or jfet, sufficently linear that typical output stages are the major limitation in audio power amps
when you get down to single digit ppm distortion however you will need to add bootstrapped cascodes to the fet inputs due to their larger nonlinear C which causes common mode input impedance nonlinearity with typical audio power amp noninverting gain and mismatched input, feedback Z
Finally an excellent post in this thread!
QFT.
This is the problem now. The "replacement" jfets J74 and K170 are no way close in performance to the 389/109 monolithics, and even then a little matching between the two would be beneficial.
The LSK389A are very good, in simulation better than the toshiba. Unfortunately there is no complementry part available yet from Linear.
And if not going complementary IMO doesnt make much sense to use these expensive JFets.
I dont have a stash of 389/109, threfore i'm not using them
I believe that very good results can be achieved with bipolars at higher voltage and current than usual. Or, at least, I'm trying that
Also true
QFT.
This is the problem now. The "replacement" jfets J74 and K170 are no way close in performance to the 389/109 monolithics, and even then a little matching between the two would be beneficial.
The LSK389A are very good, in simulation better than the toshiba. Unfortunately there is no complementry part available yet from Linear.
And if not going complementary IMO doesnt make much sense to use these expensive JFets.
I dont have a stash of 389/109, threfore i'm not using them
I believe that very good results can be achieved with bipolars at higher voltage and current than usual. Or, at least, I'm trying that
Also true
the thermal coupling between the two jFETs in a dual package should be very good, even if they are separate chips from a wafer.
However, the specification of Toshiba and Linear show an Idss maximum spread of 0.9:1 for the two jFETs in the dual package.
They do not specify any matching of any other parameters.
That 0.9:1 is a +11% maximum tolerance on the Idss of the higher value jFET.
That is atrocious as a specification if one needs matched jFETs.
Other than ease of assembly (one package instead of two) and Thermal coupling the dual package has no warranty on matching.
A pair of selected singles can perform better than the dual in all respects, except for those two parameters.
Last edited:
which parameter does the dual perform better than selected singles when there is no specification for the dual's performance?
Distortion, according to spice models.
People who have the real transistors (i.e. Nelson Pass and John Curl) could do some real world measurements, but i believe that the monolithics are overall better performers, especially with a little matching betwen the N and P pairs (which would be prohibitively expensive at today's cost).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.