@woofertester
So your methodology is to sort devices for Idss.
Use devices for the CCS position with an Idss that is lower than your chosen cell current and trim with source resistance and then pick a device with a higher Idss for the cascode.
(Ref. Mark's suggestion at top of page)
Do I have that approximately right?
So your methodology is to sort devices for Idss.
Use devices for the CCS position with an Idss that is lower than your chosen cell current and trim with source resistance and then pick a device with a higher Idss for the cascode.
(Ref. Mark's suggestion at top of page)
Do I have that approximately right?
Have to say, my favourite part of that post #3377 was:I thought so, when I published the circuit schematic and Gerbers for the "IPS6" input stage on the M2x power amp. It's at M2x post # 3,408 and the test method+jig is at # 3,377
However it may be true, that reasonable people might agree: it's reasonably likely, that others will probably get reasonably similar results.
You weren't a lawyer in a past life were you? 😆
Here is a link to the thread page: M2X
@woofertester
So your methodology is to sort devices for Idss.
Use devices for the CCS position with an Idss that is lower than your chosen cell current and trim with source resistance and then pick a device with a higher Idss for the cascode.
(Ref. Mark's suggestion at top of page)
Do I have that approximately right?
Yes, under certain circumstances.
The compelling reason that I am going down the cascode B1 road is that I have some JFETs that I think will drive low impedance loads better than the run of the mill J111/J112/J113. If it turns out that the compelling reason does not exist, then the cascode is a waste of an additional JFET.
For my newest PCB version with a cascode, the PCB can be populated to ignore the cascode by shorting that JFET D-S pins and not populating the resistor for the cascode JFET gate.
Using a J113 in the cascode positon creates less than 1V across the gain JFET. That means that Id will be less then Idss.
Using a J111 in the cascode position creates between 5 and 6 volts across the gain JFET and Id is every bit all of Idss.
The tradeoff is that a J111 in the cascode position requires 5 to 6 more volts in the + power supply to make clipping symmetric.
The J113 in the cascode position allows symmetric power rails but does not squeeze out all of the current from the gain JFET.
If your as-yet-to-be-named jfet does not work as planned, doesn't the cascode still give some benefit in terms of reduced distortion due to operating at constant voltage?
Good question. I do not know the answer. It will merit building some boards with and without the cascode and taking some measurements.
Time to build to evaluate performance. Prelim micro Beast PCBs are in hand.
I populated a vero board with sockets so that I can adjust the POT for zero DC output offset and then move the group to a PCB for soldering.
The vero board is a layout I did a few years ago for just this sort of endeavor. This is my first build on one of these boards.
The POT is next to a J113 acting as the current source. The JFET in the middle is the current gain device. The JFET farthest from the POT is a J111 servimg as a cascode.
The IDSS of the gain JFET is around 20mA. The source resistor of the gain JFET is 1 ohm.
DC output offset is less than 1 mv.
The cascode voltage is around 5v. I added 5v to the + supply to attempt to have symmetric voltage clipping.
No capacitors in the signal chain. Fully DC coupled.
Next is connecting an AC signal source and evaluating output impedance and clipping behavior.
I populated a vero board with sockets so that I can adjust the POT for zero DC output offset and then move the group to a PCB for soldering.
The vero board is a layout I did a few years ago for just this sort of endeavor. This is my first build on one of these boards.
The POT is next to a J113 acting as the current source. The JFET in the middle is the current gain device. The JFET farthest from the POT is a J111 servimg as a cascode.
The IDSS of the gain JFET is around 20mA. The source resistor of the gain JFET is 1 ohm.
DC output offset is less than 1 mv.
The cascode voltage is around 5v. I added 5v to the + supply to attempt to have symmetric voltage clipping.
No capacitors in the signal chain. Fully DC coupled.
Next is connecting an AC signal source and evaluating output impedance and clipping behavior.
The buffer idles at ~ 20 mA.
My sig gen puts out 20v p-p max.
Sig gen freq 100Hz.
Unloaded, buffer output is 20.011 v p-p
With a 500 ohm load, the buffer output is 19.044 v p-p with no visible clipping
With a 400 ohm load, the negative going half has visible clipping.
Delta V is 0.967 p-p
Delta I is 19.044 / 500 ohms = 38 mA p-p
Output impedance is Delta V / Delta I = 0.967 / 0.038 = 25 ohms
RMS output voltage = 19.044 / 2.82 = 6.753 v RMS
6.753 ^2 / 500 ohms = 91 mW power delivered
A 10-cell board could deliver 910 mW into 50 ohms.
A 72-cell power stage could deliver 6 1/2 watts. 72 cells is what my latest PCBs have for the power stage.
My sig gen puts out 20v p-p max.
Sig gen freq 100Hz.
Unloaded, buffer output is 20.011 v p-p
With a 500 ohm load, the buffer output is 19.044 v p-p with no visible clipping
With a 400 ohm load, the negative going half has visible clipping.
Delta V is 0.967 p-p
Delta I is 19.044 / 500 ohms = 38 mA p-p
Output impedance is Delta V / Delta I = 0.967 / 0.038 = 25 ohms
RMS output voltage = 19.044 / 2.82 = 6.753 v RMS
6.753 ^2 / 500 ohms = 91 mW power delivered
A 10-cell board could deliver 910 mW into 50 ohms.
A 72-cell power stage could deliver 6 1/2 watts. 72 cells is what my latest PCBs have for the power stage.
"A 10-cell board could deliver 910 mW into 50 ohms."
Sounds like a headphone amp...
Do you have the capability to measure frequency response and distortion Woofertester?
(Optimally into both 120 ohm and 8 ohm loads)
Sounds like a headphone amp...
Do you have the capability to measure frequency response and distortion Woofertester?
(Optimally into both 120 ohm and 8 ohm loads)
@BeardyWan Maybe. I have a piece of gear that may work but have not used it for this yet. I am still experimenting and building.
In no way is this a panic request! 🙂
It would be interesting to see data that is comparable with that available for other circuits as this would allow one to muse in a contemplative and strokey-beard manner on the relative merits of different approaches. Just curious. cheers.
It would be interesting to see data that is comparable with that available for other circuits as this would allow one to muse in a contemplative and strokey-beard manner on the relative merits of different approaches. Just curious. cheers.
Here is the first 10-cell board layed out by JPS64. It is the Cascode series B1.
I have populated the first cell in order to shake down the PCB.
I plan to populate two versions.
1. 20 mA per cell with + / - 15v rails
2. 10 mA per cell with + / - 30v rails
The idea is to operate the J111 and J113 at 300mW each of which is about 1/2 of max power each.
The cascode voltage of the gain JFET is around 5V.
I am populating this PCB with ~20mA Idss gain devices.
I have populated the first cell in order to shake down the PCB.
I plan to populate two versions.
1. 20 mA per cell with + / - 15v rails
2. 10 mA per cell with + / - 30v rails
The idea is to operate the J111 and J113 at 300mW each of which is about 1/2 of max power each.
The cascode voltage of the gain JFET is around 5V.
I am populating this PCB with ~20mA Idss gain devices.
If I cannot do that, I will find someone who can and send a PCB to them.In no way is this a panic request! 🙂
It would be interesting to see data that is comparable with that available for other circuits as this would allow one to muse in a contemplative and strokey-beard manner on the relative merits of different approaches. Just curious. cheers.
It should be the same or very close to Mr Pass's Beast.
I think that the version with 10 mA per cell with + / - 30v rails will drive an F4.
I plan to use a populated board to try to drive the complementary Beast that I built a while back.
I have put an EDCOR 150/10k transformer in front of this buffer. The transformer has a voltage gain of ~7.
The EDCOR will drive this buffer to the + / - 30v rails.
I plan to use a populated board to try to drive the complementary Beast that I built a while back.
I have put an EDCOR 150/10k transformer in front of this buffer. The transformer has a voltage gain of ~7.
The EDCOR will drive this buffer to the + / - 30v rails.
I am waiting for some potentiometers to show up for the series micro beast build to make progress.
In the meantime, I dusted off the original complementary beast build.
Power supply rails are + / - 12v. Idle current is 1A per rail. 24 watts total idle power.
At 10 kHz, unloaded, the maximum output voltage before clipping is 6.93 VRMS.
Placing an 8 ohm power resistor across the output loads the output down to 6.06 VRMS.
If I calculate correctly, the output impedance of the complementary Beast is 1.15 ohms.
Power into 8 ohms is 4.6 watts.
In the meantime, I dusted off the original complementary beast build.
Power supply rails are + / - 12v. Idle current is 1A per rail. 24 watts total idle power.
At 10 kHz, unloaded, the maximum output voltage before clipping is 6.93 VRMS.
Placing an 8 ohm power resistor across the output loads the output down to 6.06 VRMS.
If I calculate correctly, the output impedance of the complementary Beast is 1.15 ohms.
Power into 8 ohms is 4.6 watts.
Last edited:
Scope shot with output loaded
6 VRMS before clipping from 20 Hz to 10 kHZ
The input capacitance of the Beast plays havoc with the sig gen output requiring adjustment of the sig gen output when changing from high freq to low freq.
6 VRMS before clipping from 20 Hz to 10 kHZ
The input capacitance of the Beast plays havoc with the sig gen output requiring adjustment of the sig gen output when changing from high freq to low freq.
I sorted out the sig gen. Got a good square wave capture. 1kHz, 18v p-p square wave output into 8 ohms.
This is just about 10 watts real power for the square wave. It is essentially 9 volts continuous into 8 ohms. The 8 ohm resistor is nicely warm on a cold day.
This is just about 10 watts real power for the square wave. It is essentially 9 volts continuous into 8 ohms. The 8 ohm resistor is nicely warm on a cold day.
- Home
- Amplifiers
- Pass Labs
- Beast with 1000 JFETs redux?