Thanks so much for this.
Anyone have a source for the semiconductors? Can't seem to find them online anywhere.
Anyone have a source for the semiconductors? Can't seem to find them online anywhere.
The use a diode to nullify the source resistor for high current delivery is an interesting idea.
I am a bit concerned though about the diode's forward voltage being a bit close to the normal bias voltage across the 0R5 source resistor. At 1A bias, this should be 0.5V nominal. Figure 1 of the MUR3020 datasheet (Vishay) indicates that the forward voltage is about 0.5V at 0.1A at 25 degC, dropping down to 0.3V at 150 degC. Thus, even when the heatsink is still at room temperature, the diode is conducting at 0.1A, or at least close to conducting. As a small signal is applied, this conductance will vary, but in the non-linear region of the diode.
http://www.vishay.com/docs/94080/mur3020w.pdf
http://www.onsemi.com/pub_link/Collateral/MUR3020WT-D.PDF
Maybe it is a clever means from Nelson to use the non-linearity of the diode to compensate that for the MOSFET, though I tend to think they add rather than subtract.
Let's look at a few more numbers.
Assuming we do not use the obsolete FQAs from Fairchild, but 2SK1530/2SJ201 or IRFP240/9240. They have about 2S transconductance at 1A bias (behaving like a not-too-linear 0R5 resistor). So when connected in series with R_source of 0R5, they will have 1R effective resistance per FET-R_source set. Using 2 sets in parallel will reduce it to 0R5, 3 sets to 0R33, etc.
By using a diode in parallel with the source resistor, the idea is to half this effective resistance per FET-R_source//diode from 1R at normal bias to 0R5 at high current (when the diode fully conducts). R_source sorts of disappears. And the FET alone determines the "effective resistance". But the problem with the diode conducting also at normal bias I already high lighted above.
Can we avoid the diode (and its non-linearity) but still have similar effective resistance (or its reciprocal -> transconductance) ?
Let's consider the UHC-Mosfet pair from Toshiba, namely 2SK3497 / 2SJ618, as found in many recent power amplifiers from Accuphase. They have about 6S transconductane at the same 1A bias. If we now add a R_source of 0R33, we still get a combined R_eff of 0R5, i.e. same as a 2SK1530+0R5//MUR3020 at high current, but double at low current. And we don't need the diode. If we choose 0R22 as R_source instead, we even has more current driving capability as R_eff = 0R38 then.
No, I am by no means saying that the proposed solution with the diode is not a good one. And the above story is a simplified one, since transconductance of a Mosfet is not constant but varies with Id. I merely want to point to an alternative which can give at least similar performance without the already mentioned complications with the diode.
And yes, one of the versions being tested by the F5X test team has 2SK3497 / 2SJ618 with 0R22 source resistors at the output stage. That particular example also has cascoded frontends, and I guess it would be easy enough to parallel multiple output FETs.
Hope Nelson would care to comment. 🙂
Patrick
I am a bit concerned though about the diode's forward voltage being a bit close to the normal bias voltage across the 0R5 source resistor. At 1A bias, this should be 0.5V nominal. Figure 1 of the MUR3020 datasheet (Vishay) indicates that the forward voltage is about 0.5V at 0.1A at 25 degC, dropping down to 0.3V at 150 degC. Thus, even when the heatsink is still at room temperature, the diode is conducting at 0.1A, or at least close to conducting. As a small signal is applied, this conductance will vary, but in the non-linear region of the diode.
http://www.vishay.com/docs/94080/mur3020w.pdf
http://www.onsemi.com/pub_link/Collateral/MUR3020WT-D.PDF
Maybe it is a clever means from Nelson to use the non-linearity of the diode to compensate that for the MOSFET, though I tend to think they add rather than subtract.
Let's look at a few more numbers.
Assuming we do not use the obsolete FQAs from Fairchild, but 2SK1530/2SJ201 or IRFP240/9240. They have about 2S transconductance at 1A bias (behaving like a not-too-linear 0R5 resistor). So when connected in series with R_source of 0R5, they will have 1R effective resistance per FET-R_source set. Using 2 sets in parallel will reduce it to 0R5, 3 sets to 0R33, etc.
By using a diode in parallel with the source resistor, the idea is to half this effective resistance per FET-R_source//diode from 1R at normal bias to 0R5 at high current (when the diode fully conducts). R_source sorts of disappears. And the FET alone determines the "effective resistance". But the problem with the diode conducting also at normal bias I already high lighted above.
Can we avoid the diode (and its non-linearity) but still have similar effective resistance (or its reciprocal -> transconductance) ?
Let's consider the UHC-Mosfet pair from Toshiba, namely 2SK3497 / 2SJ618, as found in many recent power amplifiers from Accuphase. They have about 6S transconductane at the same 1A bias. If we now add a R_source of 0R33, we still get a combined R_eff of 0R5, i.e. same as a 2SK1530+0R5//MUR3020 at high current, but double at low current. And we don't need the diode. If we choose 0R22 as R_source instead, we even has more current driving capability as R_eff = 0R38 then.
No, I am by no means saying that the proposed solution with the diode is not a good one. And the above story is a simplified one, since transconductance of a Mosfet is not constant but varies with Id. I merely want to point to an alternative which can give at least similar performance without the already mentioned complications with the diode.
And yes, one of the versions being tested by the F5X test team has 2SK3497 / 2SJ618 with 0R22 source resistors at the output stage. That particular example also has cascoded frontends, and I guess it would be easy enough to parallel multiple output FETs.
Hope Nelson would care to comment. 🙂
Patrick
Last edited:
very interesting as usual
Just finished to read the pdf. Thank you Mr. Pass for this new gift.
I'm not going to build it cause I don't need more power than std F5, but there's no doubt that Turbo it's of great interest.
Just finished to read the pdf. Thank you Mr. Pass for this new gift.
I'm not going to build it cause I don't need more power than std F5, but there's no doubt that Turbo it's of great interest.
EUVL, since this is a Turbo version (designed to be pushed hard) did you consider what happens when there is 10A or more flowing through the MOSFET and Rs?
See the benefit of using the diodes ?
I mean, it's all milk and honey at 1A...
See the benefit of using the diodes ?
I mean, it's all milk and honey at 1A...
> did you consider what happens when there is 10A or more flowing through the MOSFET and Rs?
The simple answer is yes.
At 10A, the 2SK1530 has a transcoductance of 4S, or effective resistance of 0R25. The diode dominates over R_source, so we need only to add the effective R of the diode (0R03 according to Nelson), giving a total of 0R28 per set (FET+R_source//diode).
At the same current, the 2SK3497 has a transconductance of 20S, or effective resistance of 0R05. Adding this to R_source of 0R22 gives 0R27. So let's call it even.
I guess for my own use I am more interested in the linearity of my amplifier at normal listening level (say 4V 1A), and am not prepared to compromise that in the interest of an occasional burst at low frequencies. They are much better taken care of by subwoofers with large enough piston area, and driven by ultra high power, low Zout Class D amps.
But all very personal choices, as always.
Patrick
The simple answer is yes.
At 10A, the 2SK1530 has a transcoductance of 4S, or effective resistance of 0R25. The diode dominates over R_source, so we need only to add the effective R of the diode (0R03 according to Nelson), giving a total of 0R28 per set (FET+R_source//diode).
At the same current, the 2SK3497 has a transconductance of 20S, or effective resistance of 0R05. Adding this to R_source of 0R22 gives 0R27. So let's call it even.
I guess for my own use I am more interested in the linearity of my amplifier at normal listening level (say 4V 1A), and am not prepared to compromise that in the interest of an occasional burst at low frequencies. They are much better taken care of by subwoofers with large enough piston area, and driven by ultra high power, low Zout Class D amps.
But all very personal choices, as always.
Patrick
In my opinion there are too many amount of capacitors at power supply.In my F5 amp using a pair of huge audio capacitors after resistor give clearer sound than using paralelled caps.
I was thinking about the +1K of BYW99P i buried in the back yard.
Balanced with different output devices and four parallel diodes instead of two may be a way to get rid of some.
Balanced with different output devices and four parallel diodes instead of two may be a way to get rid of some.
Many thanks Nelson Pass. I just got out of rehab after the first set of amps and now this comes along! 😀
What is the advantage of the rectifier bridge arrangement to eliminate ground loop above that of the CL60 thermistor used in the BA-2 design (placed between ground and the chassis)?
Chris
What is the advantage of the rectifier bridge arrangement to eliminate ground loop above that of the CL60 thermistor used in the BA-2 design (placed between ground and the chassis)?
Chris
Nice concept, V2/V3.
I have BYV32 lying around. 🙂 150 Amp peak, 18 Amp normal duty. Ultrafast. Also have EGP50B fast recovery diodes, 50 Amp recurrent peak, 5 Amp constant and ultrafast too. Should work also in a less beefy version (20/22V DC), what do you think?
I have BYV32 lying around. 🙂 150 Amp peak, 18 Amp normal duty. Ultrafast. Also have EGP50B fast recovery diodes, 50 Amp recurrent peak, 5 Amp constant and ultrafast too. Should work also in a less beefy version (20/22V DC), what do you think?
I also noticed that the original F5 schematic published in the F5 turbo article has now R21 + TH1 connected from the source of the mosfet and the drain of the jfet, while the same components were in parallel with R5/P1.
Papa Pass--the dedication in your article. A very "class act". An acknowledgement well deserved.
As for the article itself--I'm already in the parts collection mode! Still several cold months of winter left in Colorado..... a perfect time to "go DIY turbo".....
I already have fan-cooled chassis plenum and large Conrad heatsinks, and an extra PSU cap bank in my classic F5. I'd hate to "regift" it, but it would be a great candidate for a F5 Turbo build.... (So many Pass amps....so little time!)
As for the article itself--I'm already in the parts collection mode! Still several cold months of winter left in Colorado..... a perfect time to "go DIY turbo".....
I already have fan-cooled chassis plenum and large Conrad heatsinks, and an extra PSU cap bank in my classic F5. I'd hate to "regift" it, but it would be a great candidate for a F5 Turbo build.... (So many Pass amps....so little time!)
- Status
- Not open for further replies.