Dual J176 balanced against a single J113 for each cell.
This PCB is intended for idle current of 100mA total. Class A Peak current ~200mA.
It should be happy up to + / - 30V rails.
It should deliver 12.8 v p-p class A into 32 ohms.
I am sharing some progress. I have figured out how to include the actual JFETs in-hand into the basic model for LTSPICE. LTSPICE requires Vpinch (Vto) to be measured at 1nA. You really have to measure at 1nA. Measuring the actual JFET at 3nA gives a very different Vpinch. I have the instrumentation to force/sink exactly 1nA at the source pin while applying 12v to the drain pin. Beta = Idss / (Vpinch ^2).
Attached is the basic cell and the individual JFETs demonstrating Idss and Vpinch. Just run the file as is and you can poke around and see what is going on.
These are actual JFETs that I am using on a breadboard to probe for DC and AC behavior of the basic cell.
On the breadbaord, if R17 and R18 are a short, the output has a high freq oscillation riding on top of the signal but only on the lower half of the sine wave. Adding R17 and R18 kills the oscillation.
I am going to order some 200 ohm trimmer pots for the next build.
The JFETs are USD $0.10 ~0.15 each in large quantity.
One thought is to gang up 10 to 20 JFETs into one virtual large JFET and just perform the balancing once instead of balancing each individual cell. Any thoughts/comments on this strategy are welcomed.
Attached is the basic cell and the individual JFETs demonstrating Idss and Vpinch. Just run the file as is and you can poke around and see what is going on.
These are actual JFETs that I am using on a breadboard to probe for DC and AC behavior of the basic cell.
On the breadbaord, if R17 and R18 are a short, the output has a high freq oscillation riding on top of the signal but only on the lower half of the sine wave. Adding R17 and R18 kills the oscillation.
I am going to order some 200 ohm trimmer pots for the next build.
The JFETs are USD $0.10 ~0.15 each in large quantity.
One thought is to gang up 10 to 20 JFETs into one virtual large JFET and just perform the balancing once instead of balancing each individual cell. Any thoughts/comments on this strategy are welcomed.
Attachments
You can save yourself a lot of trouble by determining the actual ratios you desire, and then think about the ratios you might want in fully-comp, quasi-comp or CCS single ended. All will deliver the goods, and there is a large aesthetic factor to the decision.
🤔
🤔
"You can save yourself a lot of trouble by determining the actual ratios you desire, and then think about the ratios you might want in fully-comp, quasi-comp or CCS single ended. All will deliver the goods, and there is a large aesthetic factor to the decision."
Thank you for the reply. Very kind of you.
The overall goal will be to produce 1 amp RMS in 8 ohms and, if possible, 1 amp RMS into 16 ohms.
At 8 ohms, it would be 8 watts and at 16 ohms it would be 16 watts.
1 amp RMS at 16 ohms is desirable since I have a collection of Audax HD13B mid/woofs that I can configure into a truncated array of 4 mid/woofs per side. They can be arranged into pairs of 16 ohms. One PCB per pair and two pairs per speaker/channel would be 32 watts per channel. These are the mid/woofs that were used in one version of the Rogers LS3/5A. I built an open baffle version years ago with 4 mid woofs per side and they sounded wonderful. They were paired with DIY 4-foot tall ribbon tweeters. The system was tri-amped with stereo subwoofers.
In this version, I will use a Scanspeak tweeter or a SEAS tweeter for each speaker and connect it to one PCB amplifer. 8 watts should be plenty for the tweeter.
1 amp RMS would be ~1.4 amp peak.
At ~10mA idle per cell the number of cells needed are
70 cells for fully-comp
140 cells for CCS single-ended. I suppose that with CCS-single ended, I could use one power MOSFET to replace all of the individual JFET CCS devices. The attraction of individual JFETs for all positions is that no heat sink is required.
I am not sure about quasi-comp for idle vs. peak. Can quasi comp deliver 2x idle current for both halves of driver pair?
Quasi-comp requires a bias spreader to drive it. Is that correct?
I abandoned the J175 to use with the J113 since the J175 has similar Idss but much less gm and the J175 runs out of current long before the J113 runs out of current. Qty 2 of J175 is just slightly more gm than qty 1 of J113 so the the current through the J113 runs out just slightly before the qty 2 of J176.
These are my design goals and observation. If I missed something or have something incorrect, I welcome feedback.
Thank you for the reply. Very kind of you.
The overall goal will be to produce 1 amp RMS in 8 ohms and, if possible, 1 amp RMS into 16 ohms.
At 8 ohms, it would be 8 watts and at 16 ohms it would be 16 watts.
1 amp RMS at 16 ohms is desirable since I have a collection of Audax HD13B mid/woofs that I can configure into a truncated array of 4 mid/woofs per side. They can be arranged into pairs of 16 ohms. One PCB per pair and two pairs per speaker/channel would be 32 watts per channel. These are the mid/woofs that were used in one version of the Rogers LS3/5A. I built an open baffle version years ago with 4 mid woofs per side and they sounded wonderful. They were paired with DIY 4-foot tall ribbon tweeters. The system was tri-amped with stereo subwoofers.
In this version, I will use a Scanspeak tweeter or a SEAS tweeter for each speaker and connect it to one PCB amplifer. 8 watts should be plenty for the tweeter.
1 amp RMS would be ~1.4 amp peak.
At ~10mA idle per cell the number of cells needed are
70 cells for fully-comp
140 cells for CCS single-ended. I suppose that with CCS-single ended, I could use one power MOSFET to replace all of the individual JFET CCS devices. The attraction of individual JFETs for all positions is that no heat sink is required.
I am not sure about quasi-comp for idle vs. peak. Can quasi comp deliver 2x idle current for both halves of driver pair?
Quasi-comp requires a bias spreader to drive it. Is that correct?
I abandoned the J175 to use with the J113 since the J175 has similar Idss but much less gm and the J175 runs out of current long before the J113 runs out of current. Qty 2 of J175 is just slightly more gm than qty 1 of J113 so the the current through the J113 runs out just slightly before the qty 2 of J176.
These are my design goals and observation. If I missed something or have something incorrect, I welcome feedback.
Hi Mark
For me, the J175 is problematic.
I built a 1A idle Beast using the J113 and J175. The J175 half runs out of gas at 1.5A peak wheres as the J113 half happily puts out 2.3A before clipping.
Also the J175 has too little power dissipation capability to keep up with the J113 for some designs.
For me, the J175 is problematic.
I built a 1A idle Beast using the J113 and J175. The J175 half runs out of gas at 1.5A peak wheres as the J113 half happily puts out 2.3A before clipping.
Also the J175 has too little power dissipation capability to keep up with the J113 for some designs.
Just for fun, I updated Patrick's JFET Circlotron with actual Vpinch and Beta for a J113 that I have in-hand. Idss of this J113 is 21mA. I set idle current for 11mA. 21mA Idss is somewhat in the middle of the pack for the population J113s that I have.
If I want 1.4A pk, I only need 50 cells.
That would yield 8 watts into 8 ohms which would be plenty for an efficient speaker.
I adjusted rail voltages down to be less than maximum per the datasheet.
If I want 1.4A pk, I only need 50 cells.
That would yield 8 watts into 8 ohms which would be plenty for an efficient speaker.
I adjusted rail voltages down to be less than maximum per the datasheet.
Attachments
"You might expect to use significantly more 175's than 113's in a complementary circuit."
Agreed. A solitary J175 can match a solitary J113 for Idss but the J175 has around 1/2 of the gm of the J113.
That is why I switched my focus to the J176. You can find lots of J176 that have exactly 1/2 of the Idss of a J113 The J176 has close to 1/2 of the gm of the J113.
When paralleling qty 2 of J176, the combined gm is fairly close that of the J113. I am matching pairs of J176 and recording the Idss of the ganged pair. I will be matching the pairs of J176 to solitary J113 of the same Idss. It is very time consuming. Not a problem for DIY. I only plan to make enough for one stereo setup.
I have one such cell on a breadboard and it works. The paired J176 gm is slightly larger than the J113. The J113 reaches current limiting a little before the paired J176 reaches current limit.
Agreed. A solitary J175 can match a solitary J113 for Idss but the J175 has around 1/2 of the gm of the J113.
That is why I switched my focus to the J176. You can find lots of J176 that have exactly 1/2 of the Idss of a J113 The J176 has close to 1/2 of the gm of the J113.
When paralleling qty 2 of J176, the combined gm is fairly close that of the J113. I am matching pairs of J176 and recording the Idss of the ganged pair. I will be matching the pairs of J176 to solitary J113 of the same Idss. It is very time consuming. Not a problem for DIY. I only plan to make enough for one stereo setup.
I have one such cell on a breadboard and it works. The paired J176 gm is slightly larger than the J113. The J113 reaches current limiting a little before the paired J176 reaches current limit.
I took some measurements.
The conditions of the measurement are.
12v Vds
Operate the JFET at 40% to 60% of Idss. This is how each Beast cell will be operated.
J113 gm ~ 13.2 Idss 17.5 mA , Idle 10mA
J176 gm ~7.3 Idss 11.37mA, Idle 5mA
gm was measured at idle current by changing Vgs + / - 0.1v
My conclusion is that 2x J176 are a good match for 1 x J113. This agrees with LTSPICE sim using measured Vpinch and measured Idss.
The conditions of the measurement are.
12v Vds
Operate the JFET at 40% to 60% of Idss. This is how each Beast cell will be operated.
J113 gm ~ 13.2 Idss 17.5 mA , Idle 10mA
J176 gm ~7.3 Idss 11.37mA, Idle 5mA
gm was measured at idle current by changing Vgs + / - 0.1v
My conclusion is that 2x J176 are a good match for 1 x J113. This agrees with LTSPICE sim using measured Vpinch and measured Idss.
I breadboarded Patrick's J113 circlotron. Thank you, Patrick, for sharing.
The triple power supply is not isolated enough to make this circuit work. I had to switch to independent power supplies.
The current measure function of the independent power supplies are off by a large % at 11 mA.
Bias voltage of ~0.5 required to throttle 21mA Idss J113 down to 11mA.
I think that the beauty of this design is that the JFET pairs match for Idss and gm.
The downside will be more power transformers compared to an ordinary amplifier.
I think that the benefits outweigh the downside.
The triple power supply is not isolated enough to make this circuit work. I had to switch to independent power supplies.
The current measure function of the independent power supplies are off by a large % at 11 mA.
Bias voltage of ~0.5 required to throttle 21mA Idss J113 down to 11mA.
I think that the beauty of this design is that the JFET pairs match for Idss and gm.
The downside will be more power transformers compared to an ordinary amplifier.
I think that the benefits outweigh the downside.
Single cell J113 circlotron operating. 50+ volts p-p into 1k ohm. 11mA idle. 25mA pk. Nice. Input driven with an EDCOR 150/10k transformer.
Thanks to Mr Pass's suggestions, I have considered how to make a fully-comp Beast that is simple to build and simple to balance. My builds and measurements so far lead me to think that 1xJ113 for every 2xJ176 will yield a well balanced comp Beast. The attached schematic is the basic idea for a Beast PCB that would be simple to build and simple to balance after populating all of the JFETs. No need for meticulous matching. You can characterize the JFETs into bins of like Idss and then populate the board(s) using various bins as needed for coarse overall matching. The power resistors can be selected for fine balancing.
The below schematic may be the 20xJ113 + 40xJ176 block that is driven by one cell as in the original Pass comp Beast.
Each block would idle at ~200mA and have class A peak current of ~500mA.
No potentiometers. Fairly quick build.
The below schematic may be the 20xJ113 + 40xJ176 block that is driven by one cell as in the original Pass comp Beast.
Each block would idle at ~200mA and have class A peak current of ~500mA.
No potentiometers. Fairly quick build.
And the headphones on the left are 90's technology, nowdays everything is full of bga and tiny resistorsWe've come a long way
The attached picture speaks to the progressing genius of the human race for advancing this technology and downsizing its components/systems.
The green PCB is double sided; for an [RF] SONY stereo headphones which connect to a TV for private listening. The orange-dot parts are LEDs from a spent 800 Lumens Philips household lamp.
I keep these amazing creations because I feel bad trashing them.
Best
Anton
Following along in your journey has been very fun, @woofertester ! I decided to buy a small amount of J113, J175 and J176 to see how i like sorting them, and it will be a good reason to start doing some REW measurements along with listening tests.
@Kris0603 Thanks for the kind words. I am currently measuring and sorting into bins 1000s of J113 and J176. The two designs that I think will be useful are
1. Complementary J113 and J176. Around twice as many J176 as J113.
2. XEN J113 Circlotron.
I have pivoted multiple times on final architecture after fabricating some circuit boards and building some protoboard cells.
Lots of thanks to Mr. Pass's for his original idea and for his suggestion of multiple JFETs in parallel per buffer cell.
1. Complementary J113 and J176. Around twice as many J176 as J113.
2. XEN J113 Circlotron.
I have pivoted multiple times on final architecture after fabricating some circuit boards and building some protoboard cells.
Lots of thanks to Mr. Pass's for his original idea and for his suggestion of multiple JFETs in parallel per buffer cell.
- Home
- Amplifiers
- Pass Labs
- Beast with 1000 JFETs redux?