Inverted Aleph J headphone amp with 2SK170 and FQP3P50
Hello,
First a short introduction of myself. I'm the kind of silent member. I've been reading a lot of posts for a lot of years, but never made a new post (till now).
I've done a few DIY audio projects in the past. In the beginning with opamps and other semiconductors, but since 15 years or so only tubes. I really like the airy, lifelike sound of good tube stuff, but I would like to give solid state another try. Because of the philosophy of Nelson Pass' stuff (little feedback, single ended) it has to be something based on his designs.
My current pre and power amps are fine, so no need to replace them. My current headphone amp however is not perfect: a SRPP OTL amp is nice for high impedance headphones, but not for low impedance ones. So a Nelson Pass based headphone amp it has to be.
I like the simple designs of the Zen series, but I don't like the output cap. To make a long story short: it has to be an Aleph J amp.
Of course I don't like any of the existing designs
so I would like to make a few changes.
First the design goals:
* 5V RMS (to drive old Sennheisers and Beyers)
* gain around 20 dB
* 100 mA RMS (to drive Hifimans or Grados)
* J fet FE
* small signal Mosfet for the current
To reach these goals I decided to go for:
* 2 x 12 volt for the power supply (about 20V tt, so 7V RMS)
* 120 mA iddle (with a bit of air because of the Aleph CS)
To find a decent J fet is tricky. All the really nice P channel devices are very difficult to get. The only through hole parts left are the J175 or J176. The transconductance of these parts is not great, so I decided to go for the inverted option: use N channel Jfets and P channel power fets.
As a N channel Jfet I thought of the following devices:
* PF5102 (very cheap, but not a very high transconductance)
* BF862 or BF861 (nice Dutch design, but only as SMD)
* 2SK170 (the original, but expensive)
* 2SK117 (a bit cheaper, but lower transconductance)
* 2SK715 (a little gem I think)
This 2SK715 looks very nice. Very high transconductance (but needs a bit of current), very low noise. Through hole package and cheap. Also low deviation (4 grades available). I think it is a Sanyo design. Mouser has the W version on stock. The biggest issue with this device is the maximum voltage (15V), but that's no problem in this case.
For the power end everybody uses the same IRF510 or 610. Boring
The small Zetex fets get good reviews, so why not use a couple of these. The VP2106 version of Supertex should be capable of 1 Watt. Using 30 mA per fet at 12 volts is 360 mW. Should be fine. Four pairs or even more. Why not. They are cheap.
I lowered the zener values a bit because of the low operating voltage of the amplifier. The CS zener is running at a high current. The BC639 and BC640 were chosen because I'm Dutch (they are cheaper). And they are also very robust (needed because of the high FE CS current).
I'm not an electrical Engineer and I don't have nice simulation tools so I hope somebody likes to assist.
I made a drawing with a nice open source design tool, but this tool is a bit limited in its possibilities (or I am). I can't get it to flip transistors, so the schematic looks a bit funny.
It's basically an Aleph H headphone amp (Metalman), inverted (based on the BF862 mini Aleph) with a Aleph J front end (NP).
The parts that I don't understand fully are marked in red (or orange, or pink).
In the Aleph CS are 2 1k resistors. In some schematics these are equal (Aleph H both 750, in Babelfish J headphone both 820), but in Nelsons original Aleph J these are not equal. Why?
The Aleph CS driving resistor is also 1k. This is taken from the Babelfish JH (and the original Aleph J is nearby at 1.2k). In the Aleph H it is much lower. Why? And is this related to the output resistor (0.12 ohm in this case).
If I understand correctly the feedback part are the 221k and 22k resistor. This would result in a 20 dB gain (correct???).
Please comment.
Update: I changed the design to 2SK170 as input and FQP3P50 as output. The schematics are posted on page 3.
Hello,
First a short introduction of myself. I'm the kind of silent member. I've been reading a lot of posts for a lot of years, but never made a new post (till now).
I've done a few DIY audio projects in the past. In the beginning with opamps and other semiconductors, but since 15 years or so only tubes. I really like the airy, lifelike sound of good tube stuff, but I would like to give solid state another try. Because of the philosophy of Nelson Pass' stuff (little feedback, single ended) it has to be something based on his designs.
My current pre and power amps are fine, so no need to replace them. My current headphone amp however is not perfect: a SRPP OTL amp is nice for high impedance headphones, but not for low impedance ones. So a Nelson Pass based headphone amp it has to be.
I like the simple designs of the Zen series, but I don't like the output cap. To make a long story short: it has to be an Aleph J amp.
Of course I don't like any of the existing designs
so I would like to make a few changes.First the design goals:
* 5V RMS (to drive old Sennheisers and Beyers)
* gain around 20 dB
* 100 mA RMS (to drive Hifimans or Grados)
* J fet FE
* small signal Mosfet for the current
To reach these goals I decided to go for:
* 2 x 12 volt for the power supply (about 20V tt, so 7V RMS)
* 120 mA iddle (with a bit of air because of the Aleph CS)
To find a decent J fet is tricky. All the really nice P channel devices are very difficult to get. The only through hole parts left are the J175 or J176. The transconductance of these parts is not great, so I decided to go for the inverted option: use N channel Jfets and P channel power fets.
As a N channel Jfet I thought of the following devices:
* PF5102 (very cheap, but not a very high transconductance)
* BF862 or BF861 (nice Dutch design, but only as SMD)
* 2SK170 (the original, but expensive)
* 2SK117 (a bit cheaper, but lower transconductance)
* 2SK715 (a little gem I think)
This 2SK715 looks very nice. Very high transconductance (but needs a bit of current), very low noise. Through hole package and cheap. Also low deviation (4 grades available). I think it is a Sanyo design. Mouser has the W version on stock. The biggest issue with this device is the maximum voltage (15V), but that's no problem in this case.
For the power end everybody uses the same IRF510 or 610. Boring
The small Zetex fets get good reviews, so why not use a couple of these. The VP2106 version of Supertex should be capable of 1 Watt. Using 30 mA per fet at 12 volts is 360 mW. Should be fine. Four pairs or even more. Why not. They are cheap.
I lowered the zener values a bit because of the low operating voltage of the amplifier. The CS zener is running at a high current. The BC639 and BC640 were chosen because I'm Dutch (they are cheaper). And they are also very robust (needed because of the high FE CS current).
I'm not an electrical Engineer and I don't have nice simulation tools so I hope somebody likes to assist.
I made a drawing with a nice open source design tool, but this tool is a bit limited in its possibilities (or I am). I can't get it to flip transistors, so the schematic looks a bit funny.
It's basically an Aleph H headphone amp (Metalman), inverted (based on the BF862 mini Aleph) with a Aleph J front end (NP).
The parts that I don't understand fully are marked in red (or orange, or pink).
In the Aleph CS are 2 1k resistors. In some schematics these are equal (Aleph H both 750, in Babelfish J headphone both 820), but in Nelsons original Aleph J these are not equal. Why?
The Aleph CS driving resistor is also 1k. This is taken from the Babelfish JH (and the original Aleph J is nearby at 1.2k). In the Aleph H it is much lower. Why? And is this related to the output resistor (0.12 ohm in this case).
If I understand correctly the feedback part are the 221k and 22k resistor. This would result in a 20 dB gain (correct???).
Please comment.
Update: I changed the design to 2SK170 as input and FQP3P50 as output. The schematics are posted on page 3.
Last edited:
Natuurlijk.
MMBFJ271 Fairchild Semiconductor | Mouser
J271 InterFET | Mouser
But Ralph wasn't asking for help to source P-JFETs.
Patrick
MMBFJ271 Fairchild Semiconductor | Mouser
J271 InterFET | Mouser
But Ralph wasn't asking for help to source P-JFETs.
Patrick
Actually I agree with ZM this time.
If 2SJ74 is the problem then you should try to use J271 and leave the rest as the original.
If your goal is however to have a go at designing circuits, then you at least need to learn to simulate.
In amplifier design, you are not only interested in THD, but also frequency response and stability.
LTSpice is free and easy enough to learn.
And 360mW is about the absolute maximum I would push through a TO92 device without extra heat sinking.
But that is just my own preference.
Patrick
If 2SJ74 is the problem then you should try to use J271 and leave the rest as the original.
If your goal is however to have a go at designing circuits, then you at least need to learn to simulate.
In amplifier design, you are not only interested in THD, but also frequency response and stability.
LTSpice is free and easy enough to learn.
And 360mW is about the absolute maximum I would push through a TO92 device without extra heat sinking.
But that is just my own preference.
Patrick
Thank you guys for the answers.
I know the J271. This Jfet has a bit lower transconductance than the original Toshiba parts and it needs more current (so a lower source resistor). This is not good for the gain. Also this Jfet is expensive or you have to buy a SMD. The J175/J176 are the same parts, but less expensive and with even bigger spread of specs (so you have to buy more).
I think the 2SK715 has far better specs. Very high transconductance, very low noise, very low capacity. And cheap in a SPA package.
You read about people looking for a decent P-channel Jfet. Why not use a N-channel Jfet and a P-channel current amp? The only reason I could find is the midrange transconductance issue of the P-channel IR parts. That is the reason I want to use the VP2106 in stead of the IRF9610.
I know a lot of people like to look at huge heatsinks in a power amp. I like them too. But for this amp I would like to use TO92 powerfets, because of the nice specs, especially as a P-channel part
I've to apologize for not simulating. I didn't realize simulation programs are free nowadays. They weren't last time I checked (some 20 years ago )
I'll try to run a simulation and find out what every part does. The most important of this exercise of this amp is the learning. If the amp sounds very nice in the end:
If not: I still have my tube amp.
I know the J271. This Jfet has a bit lower transconductance than the original Toshiba parts and it needs more current (so a lower source resistor).
This is not good for the gain. Also this Jfet is expensive or you have to buy a SMD. The J175/J176 are the same parts, but less expensive and with even bigger spread of specs (so you have to buy more).I think the 2SK715 has far better specs. Very high transconductance, very low noise, very low capacity. And cheap in a SPA package.
You read about people looking for a decent P-channel Jfet. Why not use a N-channel Jfet and a P-channel current amp? The only reason I could find is the midrange transconductance issue of the P-channel IR parts. That is the reason I want to use the VP2106 in stead of the IRF9610.
I know a lot of people like to look at huge heatsinks in a power amp. I like them too. But for this amp I would like to use TO92 powerfets, because of the nice specs, especially as a P-channel part
I've to apologize for not simulating. I didn't realize simulation programs are free nowadays. They weren't last time I checked (some 20 years ago
) I'll try to run a simulation and find out what every part does. The most important of this exercise of this amp is the learning. If the amp sounds very nice in the end:
If not: I still have my tube amp.
I've installed LTspice and made the schematic. For the input JFET I took the BF862, because I couldn't find a 2SK715 spice model.
At this moment I've got the DC values OK (zero volt at the output, 30 mA idle current per output FET).
Now I have to figure out how to make nice frequency plots. Untill now everything looks OK.
I don't know if DIYAUDIO accepts .wmf files. Lets see.
At this moment I've got the DC values OK (zero volt at the output, 30 mA idle current per output FET).
Now I have to figure out how to make nice frequency plots. Untill now everything looks OK.
I don't know if DIYAUDIO accepts .wmf files. Lets see.
No, the .wmf is not here. Let's try .pdf again.
I've changed a few resistor values so that the amp shows symetrical waveforms at full input and a low impedance output.
The Front End CS delivers a current of 21 mA.
All the output mosfets are running at 30 mA.
The Aleph resistor R8 is 470 ohm. If I increase this value, the amp delivers less current at low impedance headphones (makes sense).
I increased R17 to 1 ohm. This too increases the maximum current.
At a 30 ohm load, everything looks OK uptill 12Vtt (a little over 4Vrms). The little fets are generating 400 mAtt combined.
At a 300 ohm load, the max is 21Vtt (7.5Vrms). This should be enough.
I don't know exactly what R17 does (the 1 ohm output resistor). In most schematics this is much lower, but I don't need a very low output impedance. This "high" value gives less distortion (just looking at the waves).
I'll try to get more info from LTspice. Nice program
If anyone has tips and tricks, please let me know.
I've changed a few resistor values so that the amp shows symetrical waveforms at full input and a low impedance output.
The Front End CS delivers a current of 21 mA.
All the output mosfets are running at 30 mA.
The Aleph resistor R8 is 470 ohm. If I increase this value, the amp delivers less current at low impedance headphones (makes sense).
I increased R17 to 1 ohm. This too increases the maximum current.
At a 30 ohm load, everything looks OK uptill 12Vtt (a little over 4Vrms). The little fets are generating 400 mAtt combined.
At a 300 ohm load, the max is 21Vtt (7.5Vrms). This should be enough.
I don't know exactly what R17 does (the 1 ohm output resistor). In most schematics this is much lower, but I don't need a very low output impedance. This "high" value gives less distortion (just looking at the waves).
I'll try to get more info from LTspice. Nice program
If anyone has tips and tricks, please let me know.
I've done a few more simulations and changed the schematic to look more like the original Aleph J.
This version can give 4.1V RMS at 30ohm with a 2nd harmonic of 1%. Not to bad. With lower voltages or higher impedance loads, this number goes down of course (at a 300 ohm load and 1V RMS output the 2nd harmonic is -90dB, 3th even -110dB). A lot less than my current tube amp.
The output impedance is approx. 3 ohm. Also a lot less than my current amp.
I still can't figure out what every component does. R5 and R6 don't have a big influence in the simulations.
But I'm going to order the components for this version with R14 and R7 as trimmer or potentiometer. And the 2SK715 as Jfet.
This schematic is allready overkill for the average headphone. I can reduce the number of mosfets (even 1 pair should be enough), but I must order for at least 65 euros at Mouser. So I might even use more Mosfets.
This version can give 4.1V RMS at 30ohm with a 2nd harmonic of 1%. Not to bad. With lower voltages or higher impedance loads, this number goes down of course (at a 300 ohm load and 1V RMS output the 2nd harmonic is -90dB, 3th even -110dB). A lot less than my current tube amp.
The output impedance is approx. 3 ohm. Also a lot less than my current amp.
I still can't figure out what every component does. R5 and R6 don't have a big influence in the simulations.
But I'm going to order the components for this version with R14 and R7 as trimmer or potentiometer. And the 2SK715 as Jfet.
This schematic is allready overkill for the average headphone. I can reduce the number of mosfets (even 1 pair should be enough), but I must order for at least 65 euros at Mouser. So I might even use more Mosfets.
.model 2SK715 NJF
+ Beta=41.53m Rs=0 Rd=0 Betatce=-.5 Lambda=24.69m Vto=-.7045
+ Vtotc=-2.5m Cgd=5.451p M=.3333 Pb=1 Fc=.5 Cgs=7p Isr=215.2p
+ Nr=2 Is=21.52p N=1 Xti=3 Alpha=10u Vk=100 Kf=1E-18 Af=1
Hi Zen Mod,
I've changed the BF862 to the 2SK715. The .asc, the .pdf and two very short component files are in the zip.
I didn't need to change a lot for the 2SK715. It was very close to the NXP Jfet.
Please tell me what you think of "my" first design.
Thank you Mr. Pass for the inspiration (and the service manual for the Aleph J).
I've changed the BF862 to the 2SK715. The .asc, the .pdf and two very short component files are in the zip.
I didn't need to change a lot for the 2SK715. It was very close to the NXP Jfet.
Please tell me what you think of "my" first design.
Thank you Mr. Pass for the inspiration (and the service manual for the Aleph J).
Power supply
I finally got a simulation for the power supply. I don't know what the correct way is to simulate a transformer, but I found a sollution that might work.
The things I know from the transfomer of choise:
2 x 12 VAC
25 VA
12% regulation
I made in Spice 2 simulations with a simple double 50 hz AC Voltage: one with no load (that should give 13.4 VAC) and one with a 1 ohm load (is full load). That should give 12 VAC. I got these numbers with an AC voltage of 19.2 and a serial resistance of 1.4 ohm in Spice. That 19.2 is a bit higher than expected (12*1.12*1.414), but close enough.
I know from "my" version of the Aleph amp, that the PSRR is 50 dB. I'm aiming for a maximum hum of less then 0.1 mV. This equals 100 dB. So the power supply should have more then 50 dB hum reduction (that is probably not English).
At Mouser you get discounts with most capacitors at 10 pieces or more. So my per channel CRC sollution became a CRCRC sollution. That worked fine for my tube phono amp. So it should work here.
All capacitors are 2200uf, low esr (12 pieces in total). And all the resistors are 9 ohm (double 18 ohm). This gives an output of 13.1 Volt and a rimple voltage of -60 dB. The total rimple voltage simulated at the output is 113 dB.
I finally got a simulation for the power supply. I don't know what the correct way is to simulate a transformer, but I found a sollution that might work.
The things I know from the transfomer of choise:
2 x 12 VAC
25 VA
12% regulation
I made in Spice 2 simulations with a simple double 50 hz AC Voltage: one with no load (that should give 13.4 VAC) and one with a 1 ohm load (is full load). That should give 12 VAC. I got these numbers with an AC voltage of 19.2 and a serial resistance of 1.4 ohm in Spice. That 19.2 is a bit higher than expected (12*1.12*1.414), but close enough.
I know from "my" version of the Aleph amp, that the PSRR is 50 dB. I'm aiming for a maximum hum of less then 0.1 mV. This equals 100 dB. So the power supply should have more then 50 dB hum reduction (that is probably not English).
At Mouser you get discounts with most capacitors at 10 pieces or more. So my per channel CRC sollution became a CRCRC sollution. That worked fine for my tube phono amp. So it should work here.
All capacitors are 2200uf, low esr (12 pieces in total). And all the resistors are 9 ohm (double 18 ohm). This gives an output of 13.1 Volt and a rimple voltage of -60 dB. The total rimple voltage simulated at the output is 113 dB.
Very low current for Jfets
I need a bit of help.
I've changed the input Jfet pair to J112's. Just to see what happens in LTspice. With the same settings as the 2SK715 (so 21 mA's from the CS), the gain is almost 3 dB less, and the harmonic distortion is 5 dB higher (this is with a bit higher input signal to compensate for the lower gain).
This is fine. Cheap switch vs RF Jfet.
What happens when I lower the bias to these Jfets:
First: the gain gets higher (makes sense: the resistor to the powerfets has to be lowered, so the same current generates a larger swing).
Second: the distortion gets lower.
I expected some problems with higher frequencies (lower driving current with the input capacitance). But in Spice I don't see this. Even with a combined Jfet current of 0.5 mA everything looks fine till 100 khz or so. No real difference with the higher current 2SK715.
Something must be going wrong. I really don't know what. I've tried different models for the output fet (different manufacturers), but no real difference. The last model I took:
.MODEL VP2106 PMOS (LEVEL=3 RS=0.73 NSUB=5.0E14 DELTA=0.1 KAPPA=0.0411 TPG=-1 CGDO=6.1716E-10 RD=1.91 VTO=-2.30 VMAX=3.0E6 ETA=0.12098 NFS=3.384E11 TOX=1.0E-7 LD=1.6980E-9 UO=865.45 XJ=6.5E-7 THETA=4.063E-5 CGSO=9.10E-9 L=2.5E-6 W=0.5E-2 mfg=SUPERTEX)
Am I doing something wrong, or is Spice not that good with these kind of simulations? Or are there other reasons to use higher currents? Please help.
Small update: if I lower the CS current even further (to 0.05 mA) and I look at the gain at 50 khz, I see a drop of 6 dB vs 1 khz (or 50 khz at a higher current). So there is an influence, but it is way smaller than I think is OK?
I need a bit of help.
I've changed the input Jfet pair to J112's. Just to see what happens in LTspice. With the same settings as the 2SK715 (so 21 mA's from the CS), the gain is almost 3 dB less, and the harmonic distortion is 5 dB higher (this is with a bit higher input signal to compensate for the lower gain).
This is fine. Cheap switch vs RF Jfet.
What happens when I lower the bias to these Jfets:
First: the gain gets higher (makes sense: the resistor to the powerfets has to be lowered, so the same current generates a larger swing).
Second: the distortion gets lower.
I expected some problems with higher frequencies (lower driving current with the input capacitance). But in Spice I don't see this. Even with a combined Jfet current of 0.5 mA everything looks fine till 100 khz or so. No real difference with the higher current 2SK715.
Something must be going wrong. I really don't know what. I've tried different models for the output fet (different manufacturers), but no real difference. The last model I took:
.MODEL VP2106 PMOS (LEVEL=3 RS=0.73 NSUB=5.0E14 DELTA=0.1 KAPPA=0.0411 TPG=-1 CGDO=6.1716E-10 RD=1.91 VTO=-2.30 VMAX=3.0E6 ETA=0.12098 NFS=3.384E11 TOX=1.0E-7 LD=1.6980E-9 UO=865.45 XJ=6.5E-7 THETA=4.063E-5 CGSO=9.10E-9 L=2.5E-6 W=0.5E-2 mfg=SUPERTEX)
Am I doing something wrong, or is Spice not that good with these kind of simulations? Or are there other reasons to use higher currents? Please help.
Small update: if I lower the CS current even further (to 0.05 mA) and I look at the gain at 50 khz, I see a drop of 6 dB vs 1 khz (or 50 khz at a higher current). So there is an influence, but it is way smaller than I think is OK?
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.