I have done much searching and much reading, but I have not been able to find specific answers to my questions. I ask for patience and forbearance, as I am largely ignorant of and inexperienced in power amplifier design and construction.
In the original F5 article, Nelson Pass specifies BL grade (6-12mA Idss) for the input JFET transistors. He also emphasizes the importance of the transconductance figure of 20mS of the original Toshiba JFETs (2SJ74 and 2SK170). How closely should the Idss of the input pair be matched? Is there a recommended tolerance? How many of the Linear Systems replacements (LSJ74 and LSK170) should I obtain to achieve that match? In addition, the Linear Systems datasheets mention "Full Conductance Transconductance" of 22mS for both the LSJ74 and the LSK170. The LSK170 datasheet also mentions "Typical Conduction Transconductance" of 10mS. How do I interpret these numbers as applied to the F5T?
In the F5T article, Mr. Pass mentions the addition to the circuit of P3, which he describes as aiding in tweaking the symmetry of the circuit for the lowest possible distortion. Does P3 largely negate the effects of unmatched input devices?
Finally, who has built the F5T with the Linear Systems devices, and how did you like the results?
Thanks to all who respond. I appreciate your knowledge, experience, and helpfulness.
In the original F5 article, Nelson Pass specifies BL grade (6-12mA Idss) for the input JFET transistors. He also emphasizes the importance of the transconductance figure of 20mS of the original Toshiba JFETs (2SJ74 and 2SK170). How closely should the Idss of the input pair be matched? Is there a recommended tolerance? How many of the Linear Systems replacements (LSJ74 and LSK170) should I obtain to achieve that match? In addition, the Linear Systems datasheets mention "Full Conductance Transconductance" of 22mS for both the LSJ74 and the LSK170. The LSK170 datasheet also mentions "Typical Conduction Transconductance" of 10mS. How do I interpret these numbers as applied to the F5T?
In the F5T article, Mr. Pass mentions the addition to the circuit of P3, which he describes as aiding in tweaking the symmetry of the circuit for the lowest possible distortion. Does P3 largely negate the effects of unmatched input devices?
Finally, who has built the F5T with the Linear Systems devices, and how did you like the results?
Thanks to all who respond. I appreciate your knowledge, experience, and helpfulness.
Monitor off-set and bias when adjusting P3, in addition to the jfet inbalance. Don't let the bias setting get out of control.
Last edited:
This post might be relevant:
http://www.diyaudio.com/forums/pass-labs/183362-f5x-euvl-approach-79.html#post3258802
It deals with matched vs mismatched output fets but might translate to the input jFets. It shows 2nd harmonic can be adjusted (not the odd).
http://www.diyaudio.com/forums/pass-labs/183362-f5x-euvl-approach-79.html#post3258802
It deals with matched vs mismatched output fets but might translate to the input jFets. It shows 2nd harmonic can be adjusted (not the odd).
Can I accurately assume that as long as an F5T is built with BL grade input JFETs, with one JFET Idss at 6mA and the other at 12mA, that the amplifier will perform to specification without any additional circuitry?
Can someone define the difference between "full conductance transconductance" and "typical condunctance transconductance" and tell if that difference has any significance for an F5T build?
Can someone define the difference between "full conductance transconductance" and "typical condunctance transconductance" and tell if that difference has any significance for an F5T build?
amp will work , but certainly better/more predictable with JFets matched by Idss
you can easily find EUVL's paper/writing about his prefered way of achieving null THD with selected JFets and ther source resistors
you can easily find EUVL's paper/writing about his prefered way of achieving null THD with selected JFets and ther source resistors
Spencer should be able to get you your jfets...
FET Audio | Hi-End Audio Projects
He is reliable and trustworthy and can sell you matched pairs.
FET Audio | Hi-End Audio Projects
He is reliable and trustworthy and can sell you matched pairs.
Thanks, Hikari1. On Spencer's web site, he lists matched pairs of the Toshiba JFETs as having Idss in a range of 6.4mA to 9mA. I want to be sure I understand what the tolerance should be. If the accepted range for a match is as above for BL grade JFETs, does that mean that ranges of 6mA to 8.6mA and 9.4mA to 12mA would also be acceptable? Is the important Idss tolerance for a matched pair 2.6mA or less?
buy selected pairs.
If you agree Euvl's philosophy then the K170 should have a slightly higher/lower (it's a long time since I read Euvl) Idss than the j74 and you use the Pchannel source resistor to bring the transconductance down to match the Nchannel.
i.e. best to NOT match and ask for a selected pair instead.
This also applies to the P3 adjustment. That P3 is inserting a different source resistor to match the "transconductance" for best (lowest distortion) performance.
The in circuit "transconductance" depends on the actual Id used in operation.
A high Idss device has a much lower transconductance when Id is only 30% of Idss.
A lower Idss device run with the same Id will have a very slightly higher transconductance, but it VERY slight.
Transconductance is the slope of the Id vs Vds curve at the Id in use.
If you agree Euvl's philosophy then the K170 should have a slightly higher/lower (it's a long time since I read Euvl) Idss than the j74 and you use the Pchannel source resistor to bring the transconductance down to match the Nchannel.
i.e. best to NOT match and ask for a selected pair instead.
This also applies to the P3 adjustment. That P3 is inserting a different source resistor to match the "transconductance" for best (lowest distortion) performance.
The in circuit "transconductance" depends on the actual Id used in operation.
A high Idss device has a much lower transconductance when Id is only 30% of Idss.
A lower Idss device run with the same Id will have a very slightly higher transconductance, but it VERY slight.
Transconductance is the slope of the Id vs Vds curve at the Id in use.
Last edited:
There is no "Idss tolerance".
The devices from the mfr. are grouped by ranges of Idss, which is a natural result of what comes from the manufacturing process, devices tend to group in ranges. If you actually curve trace a large batch you'll find that most devices will bunch up in terms of Idss in each range. One finds some "out lie-ers", a majority usually in a tighter range.
What you would like to find is a dead on match. Of course without a batch of devices and a curve tracer, matching only Idss matches only the operating point(s). Which is fine, but does not tell you anything about comparative linearity and gain.
One of the issues with this is that a lot of totally fake and bogus (rebadged) devices are out on the market being sold as the real thing. They do have the requisite Idss, but that is all. You probably can not get them to match unless you have actual Toshiba devices (or LSK) and they are both Idss and curve traced matched
At the present time the safest way to go is with the Linear Systems LSK products (sold in the DiyAudio store... or have been). Spencer has been reliable in the past.
Unless you have real Toshiba (or Linear Systems) JFets, matching doesn't buy you much.
In practice this amp has likely been built and is running with bogus parts in the front end many times. Most of the ebay Toshiba parts are apparently bogus.
The other thing is that some have found that the amp actually is more pleasing to listen to when there is less of a good match in the input and/or output devices.
That's another point, the output devices can be any of several different types. The Toshiba (iirc) Mosfets happen to be nearly perfect matches between P and Nch, but some have reported that they do not sound as good as devices that match less well.
The general effect is that you get more harmonics, 2nd and 3rd in particular.
Anyhow, if you socket the input jfets you can always swap them out later or as required.
For the record the LSK devices have slightly different appearance on a curve tracer, but the specs are similar and the P and Nch are good compliments.
_-_-
The devices from the mfr. are grouped by ranges of Idss, which is a natural result of what comes from the manufacturing process, devices tend to group in ranges. If you actually curve trace a large batch you'll find that most devices will bunch up in terms of Idss in each range. One finds some "out lie-ers", a majority usually in a tighter range.
What you would like to find is a dead on match. Of course without a batch of devices and a curve tracer, matching only Idss matches only the operating point(s). Which is fine, but does not tell you anything about comparative linearity and gain.
One of the issues with this is that a lot of totally fake and bogus (rebadged) devices are out on the market being sold as the real thing. They do have the requisite Idss, but that is all. You probably can not get them to match unless you have actual Toshiba devices (or LSK) and they are both Idss and curve traced matched
At the present time the safest way to go is with the Linear Systems LSK products (sold in the DiyAudio store... or have been). Spencer has been reliable in the past.
Unless you have real Toshiba (or Linear Systems) JFets, matching doesn't buy you much.
In practice this amp has likely been built and is running with bogus parts in the front end many times. Most of the ebay Toshiba parts are apparently bogus.
The other thing is that some have found that the amp actually is more pleasing to listen to when there is less of a good match in the input and/or output devices.
That's another point, the output devices can be any of several different types. The Toshiba (iirc) Mosfets happen to be nearly perfect matches between P and Nch, but some have reported that they do not sound as good as devices that match less well.
The general effect is that you get more harmonics, 2nd and 3rd in particular.
Anyhow, if you socket the input jfets you can always swap them out later or as required.
For the record the LSK devices have slightly different appearance on a curve tracer, but the specs are similar and the P and Nch are good compliments.
_-_-
Last edited:
Spencer claims .2mA match at 10V. You choose what value you want (from 6.4-9mA) when you order from him. The "match" range is not 6.4-9mA...he is allowing you to choose your value from this range.
From what little I have seen, the Idss of the fakes are way off, and so it is
usually the quickest way to spot them.
😎
Hi Nelson;
I tried to contact you, but lost your e-mail address, and your private messages on the forum are disabled.
Please advise
Anatoliy
I tried to contact you, but lost your e-mail address, and your private messages on the forum are disabled.
Please advise
Anatoliy
Compare Vp and Idss to the sk170/j74 values shown in the datasheet.
Most fakes will use a cheap low gm device. Even when the fake has a similar Idss, the low gm will show as a very high Vp value for that Idss.
Check both Vp and Idss to confirm you have high gm devices.
Most fakes will use a cheap low gm device. Even when the fake has a similar Idss, the low gm will show as a very high Vp value for that Idss.
Check both Vp and Idss to confirm you have high gm devices.
I have in my possession far too many bogus fakes that have perfect Idss.
think, iirc, they were alleged k170s.
Check the Yfs and all bets are off, and we have discovered which devices were sanded and rebadged - in another thread there are pix that show this (last years posts, iirc)
think, iirc, they were alleged k170s.
Check the Yfs and all bets are off, and we have discovered which devices were sanded and rebadged - in another thread there are pix that show this (last years posts, iirc)
I think the info is in this thread:
http://www.diyaudio.com/forums/parts/82638-my-transistors-original-copy.html
http://www.diyaudio.com/forums/parts/82638-my-transistors-original-copy.html
- Status
- Not open for further replies.