F6 Amplifier

My Rs is .25 so the difference is more obvious.I would venture to guess that it is a result of more distortion in the above Rs position. Below Rs, the amp is more focused, but shorter. Now to discover how to keep the THD spectra of the one with the overall distortion of the other, shaping the sound of the amp.
Ilquam, have you measured distortion both ways.
 
I might have actual circuit measurements both ways, but it might take some searching to locate them. Here are Spice sims both ways. My experience is that the simulations do not accurately model the transformer behaviour. I will try to find in-circuit measurements.


My Rs is .25 so the difference is more obvious.I would venture to guess that it is a result of more distortion in the above Rs position. Below Rs, the amp is more focused, but shorter. Now to discover how to keep the THD spectra of the one with the overall distortion of the other, shaping the sound of the amp.
Ilquam, have you measured distortion both ways.
 

Attachments

  • F6-acs4-cap-below.jpg
    F6-acs4-cap-below.jpg
    126.5 KB · Views: 314
  • F6-acs4-cap-above.jpg
    F6-acs4-cap-above.jpg
    121.5 KB · Views: 312
OK, here are the only in-circuit measurements I can find. The left plot is with the cap connection below Rs with Rs=0R47. The right plot is with the cap connection is above Rs, with Rs=0R12. They are very similar in THD, but the the one on the right has much higher open-loop gain, thus more feedback and lower output impedance, thus higher damping factor.

Unfortunately, these measurements are of the amplifier channel with the suspect R100 JFET(s).

My Rs is .25 so the difference is more obvious.I would venture to guess that it is a result of more distortion in the above Rs position. Below Rs, the amp is more focused, but shorter. Now to discover how to keep the THD spectra of the one with the overall distortion of the other, shaping the sound of the amp.
Ilquam, have you measured distortion both ways.
 

Attachments

  • F6-Rs-0R47-Gain-5-fsweep.jpg
    F6-Rs-0R47-Gain-5-fsweep.jpg
    123 KB · Views: 298
  • F6-ACS-2-RS-0R0-Rsen-0R12-right-fsweep.jpg
    F6-ACS-2-RS-0R0-Rsen-0R12-right-fsweep.jpg
    124 KB · Views: 279
OK, here are the only in-circuit measurements I can find. The left plot is with the cap connection below Rs with Rs=0R47. The right plot is with the cap connection is above Rs, with Rs=0R12. They are very similar in THD, but the the one on the right has much higher open-loop gain, thus more feedback and lower output impedance, thus higher damping factor.

Unfortunately, these measurements are of the amplifier channel with the suspect R100 JFET(s).

I need to get my test setup kicking. Want to get a view of the breakdown of the THD, showing 2nd vs third, and so on. There is something going on other than DF, especially in your case, where Rs is minimal.
 
Those plots look to much alike. I am wondering if there was a screw-up on the captioning. I will run appropriate THD sweeps on the "good" channel ASAP.

I need to get my test setup kicking. Want to get a view of the breakdown of the THD, showing 2nd vs third, and so on. There is something going on other than DF, especially in your case, where Rs is minimal.
 
I will try to measure the interstage - phase splitter transformer
this part is from importance in design.
btw
it will be very hard to obtain a good inerstage without the propper driver.
something like already existing on the sch.
But my opp is that it should be on normal laminates not on the ferrite cores.
Normal, not common but preferably thin laminates, 0.27mm or 0.15mm
little bit of higher relative permeability lets say 800 cca (common have of 200-400...)
little bigger window or size to have a space for insulations to reduce capacitances
lets say 5-6 layers to reduce the leakage inductance. proper wire dia even it would be thin...
shield layers and chasses tap. for the proper grounding options
and lets say 1:1 transfer ratio to eliminate transfers of reactive components trough.
voltage gain can be acheived with a stage before the interstage tr.
 
Last edited:
But my opp is that it should be on normal laminates not on the ferrite cores.
Normal, not common but preferably thin laminates, 0.27mm or 0.15mm
little bit of higher relative permeability lets say 800 cca (common have of 200-400...)
little bigger window or size to have a space for insulations to reduce capacitances
lets say 5-6 layers to reduce the leakage inductance. proper wire dia even it would be thin...
shield layers and chasses tap. for the proper grounding options
and lets say 1:1 transfer ratio to eliminate transfers of reactive components trough.

Absolutely agree. An M3 E/I core, either 19mm center leg or 25mm center leg with material thickness of 0.229mm.

I am assuming that this IT is going from SE input to PP output? Pardon the question please, the schematic is not clear to me, but then I only design transformers.

The current nickle core items, seductive as they are, cannot have more than a PSSP winding format. The FR above 35kHz would go wild with any more interleaving than this. We do have dielectric to help here and the numbers are fine for these parts, but without numerous interleaves and the proper ratio of interwinding capacitance to distributed capacitance, information from the E Field will be lost. This is small amplitude signal information that portrays internal gradient structures of both tones and transients.

I do have these, materials on hand and can make IT's to suit the actual load, not just 600 ohms or some other convenient constant please. They are more expensive than the small nickle core items and will not have the amount of inductance under normal operating conditions. The maximum permeability at 27 Hz of M3 is 55 k, M6 40 k and that of both grades of Nickle core is 100 k.

Bud
 
Yes with that very high permeability it will huge decrease of number of turns
which will lead to easier achieving of needed primary inductance
more flexible to primary load, and less much less capacitances.
.
Is there anu AC current trough the secondary flow?
because if it is that will saturate a sensitive core.
And the extending of the gap will not make it much better
even with a same Current trough the primary will not help much...
?
Maybe to find some laminates with less permeability
.
And yes You are right there will be some ringing at the top
but that should be measured and corrected via RC network on the secondary?
 
I suggest you obtain one of the Jensen transformers (JT-123-FLPCH) and measure it. Their specification sheets are are honest. The major parameter that imposes limits on frequency response is the interwinding capacitance due to the quadfilar nature of the windings.

Absolutely agree. An M3 E/I core, either 19mm center leg or 25mm center leg with material thickness of 0.229mm.

I am assuming that this IT is going from SE input to PP output? Pardon the question please, the schematic is not clear to me, but then I only design transformers.

The current nickle core items, seductive as they are, cannot have more than a PSSP winding format. The FR above 35kHz would go wild with any more interleaving than this. We do have dielectric to help here and the numbers are fine for these parts, but without numerous interleaves and the proper ratio of interwinding capacitance to distributed capacitance, information from the E Field will be lost. This is small amplitude signal information that portrays internal gradient structures of both tones and transients.

I do have these, materials on hand and can make IT's to suit the actual load, not just 600 ohms or some other convenient constant please. They are more expensive than the small nickle core items and will not have the amount of inductance under normal operating conditions. The maximum permeability at 27 Hz of M3 is 55 k, M6 40 k and that of both grades of Nickle core is 100 k.

Bud
 
Official Court Jester
Joined 2003
Paid Member

I must admit that thread is already longish , but you must find few things by your self :clown: :

- Pa published just basic schm (that one in first post is having secondary phase mistake , later posted proper one )

- Pa also posted partial amp basic schmtc with accent on some impedances (high - low)

- gate of output mosfet is representing 10K-ish load to xformer secondary

-from all these schmtcs (NP's , several others , extended talk ) it's clear that secondary is DC connected to bias net , but that's not critical ........ considering diminutive gate current

so , focus , please

:clown:
 
Jensen makes excellent transformers. They area a very technically oriented company and design to exacting specs. They typically do not have any headroom above these specs, so it is important to realize that their specified maximums are not to be exceeded, under any conditions.

There are other other concerns beyond numerical value and this is where I spend my resources. In a direct comparison between two units from differing manufacturers, there will be noticeable differences in a number of qualities I have yet to find a suitable objective measure for. Except for one. When passing square waves. You will find that my transformers typically have the characteristic rise in permeability from zero BH, at the onset of each half of a square wave, that is found in the onset of a hysterisis curve. You will also find that my transformers do not exhibit polarized core behavior after B saturation, regardless of the quality or thickness of the core. Make of it what you will.

In all other tests there is not an appreciable difference between what I make, what Jensen makes, what Cinemag makes, what Lundahl makes, what Sowter makes, what Intact Audio makes, what Tribute Audio makes. Yet, each of these manufacturers has a distinct "flavor" to their reproduction of a musical signal of great density, great dynamic range, or great transient attack, across the FR bandwidth. At this level of performance, personal taste really is the determining value, the usual tests are too crude and simple to allow for differentiation based solely upon them.

Bud
 
IMO, THe transformer performs very well. AS BudP said, this is a personal opinion, but i do not think that the Xformer is limiting the design in any way. As for how it colors the sound, I am unable to comment, as I have not heard others. As a matter of practce, I will get others and try, as I like this amp very much. IN playing with it, I have come closer than any other amp, to a tubey triode sound. I know, please don't stone me. I am gong to opoena new thread so that i stop cloggin this one. More useful info will be found here. Mine will be full of conjecture and opinion, with little to no quality info. Fun is allowed, as well as stupidity. I'll bring the stupdity, you bring the fun.
 
buzz/others --- any ideas/thoughts on how we might try to get output impedance down to some reasonable number WITHOUT the use of feedback ? I don't think it can be done (unless I guess maybe you parallel 10 pairs of o/p devices to get 4 ohms output z). Reason i ask is that the amp seems like a very nice candidate to be run open loop - but it appears that unless you have horns/full range drivers it would be practically unusable with most mainstream speakers ... likely wishful thinking on my part but can't hurt to ask....
 
buzz/others --- any ideas/thoughts on how we might try to get output impedance down to some reasonable number WITHOUT the use of feedback ? I don't think it can be done (unless I guess maybe you parallel 10 pairs of o/p devices to get 4 ohms output z). Reason i ask is that the amp seems like a very nice candidate to be run open loop - but it appears that unless you have horns/full range drivers it would be practically unusable with most mainstream speakers ... likely wishful thinking on my part but can't hurt to ask....

Make it part of a Compound Power Amplifier. Please see the thread in the Pass Labs Forum with the same title. The clone of F6 runs open loop and its output impedance is minimized to that of its mate voltage source amp [VSA] of output impedance <0.1 Ohm; plus other performance goodies. The VSA can be an F5 or your choice. If you operate your F6 clone open loop; will you still need its front end?