Hi Dave,
The 15 ohm resistors were used to prevent current hogging. I also ran the amps without them and didn't notice any difference in the sound. I also didn't notice that one 109 pair was heating up more than the other.
Putting a resistors into the circuit at this location reduces the overall gain (not required for the AX) and can also linearize the stage and reduce distortion. Neither of these factors entered into my thinking.
As you can see I didn't get too rigorous about the details. The big deal is the use of JFETs over MOSFETs. That's the magic twanger.
You could use your extra JFETs as cascodes if you want to. If it was me I would save my JFETs for something more worthwhile. Good JFETs are getting scarcer. Good BJTs still seem to be cheap and abundant.
You're absolutely right about not messing with something that ain't broken and sounding great. That's why I haven't done any more messing with my amps.
Have a great holiday Dave.
Graeme
The 15 ohm resistors were used to prevent current hogging. I also ran the amps without them and didn't notice any difference in the sound. I also didn't notice that one 109 pair was heating up more than the other.
Putting a resistors into the circuit at this location reduces the overall gain (not required for the AX) and can also linearize the stage and reduce distortion. Neither of these factors entered into my thinking.
As you can see I didn't get too rigorous about the details. The big deal is the use of JFETs over MOSFETs. That's the magic twanger.
You could use your extra JFETs as cascodes if you want to. If it was me I would save my JFETs for something more worthwhile. Good JFETs are getting scarcer. Good BJTs still seem to be cheap and abundant.
You're absolutely right about not messing with something that ain't broken and sounding great. That's why I haven't done any more messing with my amps.
Have a great holiday Dave.
Graeme
Hi Tad,
The only JFET I know of that you could directly substitute for the 2SJ109 is a matched pair of 2SJ74's. I used 2SJ109BL's so it would be the 2SJ74BL's that you would want for this circuit. I just bought 100 of them from bdent for my stash.
The 2SJ109 pair is already matched but the match isn't super close. You can get close enough by just and matching the 2SJ74's for Idss. This is covered in a lot of places. Macka built an AX100 amp by using matched 2SJ74BL's and things worked great. Passlabs does this now too I understand. There is a thread here somewhere covering macka's (Ian's) journey.
There are other P-channel JFETs around but nothing currently available comes to mind as a quick and easy sub here.
And a great holiday back at you!
Regards,
Graeme
The only JFET I know of that you could directly substitute for the 2SJ109 is a matched pair of 2SJ74's. I used 2SJ109BL's so it would be the 2SJ74BL's that you would want for this circuit. I just bought 100 of them from bdent for my stash.
The 2SJ109 pair is already matched but the match isn't super close. You can get close enough by just and matching the 2SJ74's for Idss. This is covered in a lot of places. Macka built an AX100 amp by using matched 2SJ74BL's and things worked great. Passlabs does this now too I understand. There is a thread here somewhere covering macka's (Ian's) journey.
There are other P-channel JFETs around but nothing currently available comes to mind as a quick and easy sub here.
And a great holiday back at you!
Regards,
Graeme
aleph1.2 to AX100 jfet mod and initial check out
GL, I had some time over the holidays to exierment with replacing my 9610 mosfet front end with the J74 fets that I got matched from Erno Borbely. I used your schematic 01 as a start point.
The differences in my cirucuit from yours are:
1. I used 10uf input caps vs 4.7uf as this is what I had on hand.
2. I did not use the jfet source resistors as I didn't have the
15 ohm values on hand but I do have some 27ohm that I can
stick in later.
3. I used all of your resistor value changes for the macmillian and
feedback resistors.
4. I repalced your 845ohm R12 with a high quality 2kohm pot so I
I could adjust the gain between the plus/minus amps. Using a
1k square wave I adjusted the resistor until I nulled the two
output signals (add function on the dual trace scope), I had
amp connected to a 5ohm 100 watt resistor.
My first check out revealed the following after about 4 hours of running on the test bench:
1. Relative offset adjusted to 70-80mv but I believe I can get it
lower but it was New Years eve and time to party.
2. The absolute offset increased to 1 volt after warm up which is
about 300-500mv more than my mosfet front end set up.
3. The relative DC at the balanced inputs after the cap
and into the board is the same 1 volt. This is not
surprising as the feedback network is connected to the
100 ohm resistors to ground at the outputs.
Everything else looks the same when I checked the dc voltages against the mosfet version. This amp runs hard all the time, I have to check the source resistors on the output fets yet but I am using the 100k for R14/R33 so my out put is still around 120 to 140 watts.
I am not sure how to get the absolute dc offset lower so as to eliminate the input caps. You are correct that this is a mystery as the mosfet version did not need input coupling caps but was connected in the same way. The only difference is that the feedback voltage divider on the input was changed (10k to 22k and the macmillian resistor was increased to 22k from 10k. Perhaps this is the reason for the higher absolute values.
I want to run it for another day and keep adjusting it to see what I can do. Perhaps lowering the 100ohm resistors may help.
What happens to a speaker if one leg of the amp gets disconnected? Does that put a volt across the drivers or just nothing since there is no return path?
happy new year to all and I am sure that Nelson is drinking some doobie brothers wine issue from BR Cohn. somewhere in the valley. dave
GL, I had some time over the holidays to exierment with replacing my 9610 mosfet front end with the J74 fets that I got matched from Erno Borbely. I used your schematic 01 as a start point.
The differences in my cirucuit from yours are:
1. I used 10uf input caps vs 4.7uf as this is what I had on hand.
2. I did not use the jfet source resistors as I didn't have the
15 ohm values on hand but I do have some 27ohm that I can
stick in later.
3. I used all of your resistor value changes for the macmillian and
feedback resistors.
4. I repalced your 845ohm R12 with a high quality 2kohm pot so I
I could adjust the gain between the plus/minus amps. Using a
1k square wave I adjusted the resistor until I nulled the two
output signals (add function on the dual trace scope), I had
amp connected to a 5ohm 100 watt resistor.
My first check out revealed the following after about 4 hours of running on the test bench:
1. Relative offset adjusted to 70-80mv but I believe I can get it
lower but it was New Years eve and time to party.
2. The absolute offset increased to 1 volt after warm up which is
about 300-500mv more than my mosfet front end set up.
3. The relative DC at the balanced inputs after the cap
and into the board is the same 1 volt. This is not
surprising as the feedback network is connected to the
100 ohm resistors to ground at the outputs.
Everything else looks the same when I checked the dc voltages against the mosfet version. This amp runs hard all the time, I have to check the source resistors on the output fets yet but I am using the 100k for R14/R33 so my out put is still around 120 to 140 watts.
I am not sure how to get the absolute dc offset lower so as to eliminate the input caps. You are correct that this is a mystery as the mosfet version did not need input coupling caps but was connected in the same way. The only difference is that the feedback voltage divider on the input was changed (10k to 22k and the macmillian resistor was increased to 22k from 10k. Perhaps this is the reason for the higher absolute values.
I want to run it for another day and keep adjusting it to see what I can do. Perhaps lowering the 100ohm resistors may help.
What happens to a speaker if one leg of the amp gets disconnected? Does that put a volt across the drivers or just nothing since there is no return path?
happy new year to all and I am sure that Nelson is drinking some doobie brothers wine issue from BR Cohn. somewhere in the valley. dave
Hi Nelson & Graeme,
Sorry I have been away from this great thread for so long.
I wanted to ask a few question about current share with Jfet front end of an X Aleph
Have you found that proportions other than 50/50 are preferable?
Its been a while since I read up on this but does 60/40 mean the current source is sharing 60% of the a/c current share?
Ian
Sorry I have been away from this great thread for so long.
I wanted to ask a few question about current share with Jfet front end of an X Aleph
Have you found that proportions other than 50/50 are preferable?
Its been a while since I read up on this but does 60/40 mean the current source is sharing 60% of the a/c current share?
Ian
Hi Nelson,
I was referring to the current of the output stage (with JFet front end like the Aleph J).
I sense having heard a diy Aleph J that less is more (better) when adjusting the A/c current share..as I recall the resister is like 1200R.
Despite the low transconductance of single Jfet pair it works beautifully in that design!!
I am taking an educated guess here here but with the improved linearity of the Jfet over the 9610 the 50/50 share and cancellation of distortion of the output current share may not work to best advantage.
Then again perhaps it just arbitrary.
Anyways I have bought some 10 k linear dual gang pots and propose to strap them across 1300R with 2K in series with the pots.
This will give me an adjustment range of about 800- 1200 ohms for the current share resister on the fly. The linear pots track quite closely.
You can tell I like to tinker.
Ian
I was referring to the current of the output stage (with JFet front end like the Aleph J).
I sense having heard a diy Aleph J that less is more (better) when adjusting the A/c current share..as I recall the resister is like 1200R.
Despite the low transconductance of single Jfet pair it works beautifully in that design!!
I am taking an educated guess here here but with the improved linearity of the Jfet over the 9610 the 50/50 share and cancellation of distortion of the output current share may not work to best advantage.
Then again perhaps it just arbitrary.
Anyways I have bought some 10 k linear dual gang pots and propose to strap them across 1300R with 2K in series with the pots.
This will give me an adjustment range of about 800- 1200 ohms for the current share resister on the fly. The linear pots track quite closely.
You can tell I like to tinker.
Ian
aleph1.2 to AX100 jfet mod and initial check out
GL and others, my second full day of checkout after converting my ax100 to jfets went as follows:
One full day of running the amp hooked up to a 5ohm 125watt resistor on the outputs with the inputs shorted.
Input:
330mv dc on the gates of the input fets and this remained steady so I will use the input coupling caps as GL recommends. I am attributing this to the different resistor values used as the Mosfet version did not have any DC.
80mv dc relative offset voltage on the outputs. I am leaving this adjustment alone. The amp starts up at 40mv and after 8hours is ramps up to 80mv.
The absolute dc on the output starts out at 13 volts and slowly drops to about 3.5 volts dc after several hours. This poses no safety hazard so for now I am not going to try playing with resistor values to get it lower. I would be interested in what others are getting? I think this number is temp dependent. My mosfet verstion the absolue dc was about 800mv after a long warm up. The rail voltage is a little higher than GL's at 32 volts.
In checking my output devices (they are going on 9 years old as they were part of my original Aleph 1.2 I built in 2000); I am getting 600mv across the 1 ohm source resistors. This is an average as some outputs are 700mv and some are 450mv. I guess you can say I didn't match them but I am getting lazy. I am using 24 total per side or 6 per quad.
I did run some square waves and they look ok but I have to use the generator as a single ended input and I don't have a shielded cable to connect it to the amp (I am using an ordinary wire). Therefore I think the 10k square wave on the + input looks really clean but the - input which is connected to ground looks a bit rounded. I will listen to some music this week and see if I hear any distortion but my test equipment is ancient.
I also cleaned up some wires and drilled some additional holesin the amps top cover. More to come in later. dave
GL and others, my second full day of checkout after converting my ax100 to jfets went as follows:
One full day of running the amp hooked up to a 5ohm 125watt resistor on the outputs with the inputs shorted.
Input:
330mv dc on the gates of the input fets and this remained steady so I will use the input coupling caps as GL recommends. I am attributing this to the different resistor values used as the Mosfet version did not have any DC.
80mv dc relative offset voltage on the outputs. I am leaving this adjustment alone. The amp starts up at 40mv and after 8hours is ramps up to 80mv.
The absolute dc on the output starts out at 13 volts and slowly drops to about 3.5 volts dc after several hours. This poses no safety hazard so for now I am not going to try playing with resistor values to get it lower. I would be interested in what others are getting? I think this number is temp dependent. My mosfet verstion the absolue dc was about 800mv after a long warm up. The rail voltage is a little higher than GL's at 32 volts.
In checking my output devices (they are going on 9 years old as they were part of my original Aleph 1.2 I built in 2000); I am getting 600mv across the 1 ohm source resistors. This is an average as some outputs are 700mv and some are 450mv. I guess you can say I didn't match them but I am getting lazy. I am using 24 total per side or 6 per quad.
I did run some square waves and they look ok but I have to use the generator as a single ended input and I don't have a shielded cable to connect it to the amp (I am using an ordinary wire). Therefore I think the 10k square wave on the + input looks really clean but the - input which is connected to ground looks a bit rounded. I will listen to some music this week and see if I hear any distortion but my test equipment is ancient.
I also cleaned up some wires and drilled some additional holesin the amps top cover. More to come in later. dave
aleph1.2 to AX100 jfet mod and initial check out
GL after running the amp on the test bench all day today and bolting it back up I noticed the output absolute offset went down below a volt but the relative offset climbed up to 300mv. I have modified the amps cover to allow more cooling (I drilled 1 inch holes along the top sides (like boulder) and covered them from the inside with black screen and I think this has affected the bias as the amps run cooler.
I think I will install the source resistors but I may use a 10 ohm pot and attach the wiper to the fet current source and each leg to the +/- signal input jfets. I can then adjust this to see if I can zero the relative offset. I read up on the threads and it seems patience is the best solution. dave
GL after running the amp on the test bench all day today and bolting it back up I noticed the output absolute offset went down below a volt but the relative offset climbed up to 300mv. I have modified the amps cover to allow more cooling (I drilled 1 inch holes along the top sides (like boulder) and covered them from the inside with black screen and I think this has affected the bias as the amps run cooler.
I think I will install the source resistors but I may use a 10 ohm pot and attach the wiper to the fet current source and each leg to the +/- signal input jfets. I can then adjust this to see if I can zero the relative offset. I read up on the threads and it seems patience is the best solution. dave
Hi Dave,
Minimum relative offset IMO is governed by close matching. It also plays a role in the absolute offset although other factors play a large potential role there too.
Like you say, your output transistors don't appear to be matched as well as they could be. I matched mine to within .01 - .02 volts on the Vgs. The voltage drop across the Source resistors differed by only a few millivolts as I recall. The 10 gain (output) transistors in each channel were matched to each other and the 10 CCS transistors were matched to each other.
Definitely use the coupling caps.
Graeme
Minimum relative offset IMO is governed by close matching. It also plays a role in the absolute offset although other factors play a large potential role there too.
Like you say, your output transistors don't appear to be matched as well as they could be. I matched mine to within .01 - .02 volts on the Vgs. The voltage drop across the Source resistors differed by only a few millivolts as I recall. The 10 gain (output) transistors in each channel were matched to each other and the 10 CCS transistors were matched to each other.
Definitely use the coupling caps.
Graeme
aleph1.2 to AX100 jfet mod and initial check out
GL, I have been fighting some heat gremlins. After successfully converting to the all JFet front end and checking things out with the cover off --perfection--70mv dc offset and all electrical dc voltages spot on. The absolute offset took a bit longer to reach about 1 volt than the previous mosfet version.
However after I put the cover on the relative dc on the outputs shot up to 300mv after about an hour. So what to do.
My plan now is to reduce the bias on the 24 outputs (6 per quad) to .450. from about .600. This will give me 5.4 amps per leg or from the PS perspective a total of 10.8 amps at idle. This should reduce the heat load considerably.
I have drilled new cooling holes in the top but this did not help.
Adjusting U2 did not get it lower.
I measured the voltage across the R23 and R25 392 drain resistors and got the following:
+side=4.91v
-side=4.68 v This is a difference of .23v with a current of 12.6ma (4.91/392=12.6). So I thought I should put in a 14ohm source resistor on the +side (230mv/12.6ma=14ohms) but that caused the relative dc to jump to 1.68volts on the outputs with no other changes.
So I am trying the reduced bias first as in the cold condition everything is ok (the square waves looked good). I have 10uf ppcaps on the inputs in balanced mode.
I am not sure what exactly is causing this as the voltage across R23 and R25 did not change much with the addition of the source resistor. I may try to put in a 30ohm pot with the wiper connected to the current source and each leg to one of the pot connections and adjust this after the amp gets hot.
The way I had this bias set I would be getting 14.4 amps at idle and estimated 200watts of class A which I do not need.
I am replacing the 100k R11/23 with a 100k pot set for 75k today and adjust from there. I will keep in touch-dave
I will keep in touch-dave
GL, I have been fighting some heat gremlins. After successfully converting to the all JFet front end and checking things out with the cover off --perfection--70mv dc offset and all electrical dc voltages spot on. The absolute offset took a bit longer to reach about 1 volt than the previous mosfet version.
However after I put the cover on the relative dc on the outputs shot up to 300mv after about an hour. So what to do.
My plan now is to reduce the bias on the 24 outputs (6 per quad) to .450. from about .600. This will give me 5.4 amps per leg or from the PS perspective a total of 10.8 amps at idle. This should reduce the heat load considerably.
I have drilled new cooling holes in the top but this did not help.
Adjusting U2 did not get it lower.
I measured the voltage across the R23 and R25 392 drain resistors and got the following:
+side=4.91v
-side=4.68 v This is a difference of .23v with a current of 12.6ma (4.91/392=12.6). So I thought I should put in a 14ohm source resistor on the +side (230mv/12.6ma=14ohms) but that caused the relative dc to jump to 1.68volts on the outputs with no other changes.
So I am trying the reduced bias first as in the cold condition everything is ok (the square waves looked good). I have 10uf ppcaps on the inputs in balanced mode.
I am not sure what exactly is causing this as the voltage across R23 and R25 did not change much with the addition of the source resistor. I may try to put in a 30ohm pot with the wiper connected to the current source and each leg to one of the pot connections and adjust this after the amp gets hot.
The way I had this bias set I would be getting 14.4 amps at idle and estimated 200watts of class A which I do not need.
I am replacing the 100k R11/23 with a 100k pot set for 75k today and adjust from there. I will keep in touch-dave
I will keep in touch-dave
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.