Hello, I will try to answer few questions here.
If you run the kx-Amp off +-35 volt rails in class A you will have a standing dissipation of 84 watts per channel. For two pairs of OP devices this is very high and I strongly recommend that you do NOT do this. However, some of you will ignore this advice and do it anyway
. On the standard kx-amp running on +-28V rails I recommend each channel’s heatsink weighs 1.5 kg minimum. If you run in class A at +-35V, you should go for enormous heatsinks if you want reliability - so try 2.5 to 3 kg minimum. Again, personally I would not do it.
On my unit I use every day, I can switch between class A and class AAB (1.2A standing current vs circa 600 mA per channel). Note that if you run full class A at +-35V distortion will go up on the peak output. On class AAB the kx-Amp still has a very smooth sound.
If you have a transformer you want to use, but the voltage is a bit high, just dial the standing current down a bit to keep the total dissipation within reason. The way to ascertain this is to ensure the heat sink temperature settles out at about 55 Deg C max. You should not run much above this for reasons of reliability since you want this thing around for another 20 or 30 years.
Bridging. I have not tried this but it should work ok. The theoretical power is 4x the rating of a non- bridged amp. Since we only have 2 OPS pairs per amp, you will be asking it in bridged mode to deliver 100 Watts at very high temperatures, and much higher (double) Vce excursions. So, I would suggest you are conservative with the rails and do not go above 28 Volts and preferably keep them in the +-24 V range.
Bridge Wiring. I will try in the next day or so to propose a scheme. For the Zobels, I will have to think about it, but at this stage my view is each amp output should go to its own Zobel which goes back to 0V.
If you run the kx-Amp off +-35 volt rails in class A you will have a standing dissipation of 84 watts per channel. For two pairs of OP devices this is very high and I strongly recommend that you do NOT do this. However, some of you will ignore this advice and do it anyway
. On the standard kx-amp running on +-28V rails I recommend each channel’s heatsink weighs 1.5 kg minimum. If you run in class A at +-35V, you should go for enormous heatsinks if you want reliability - so try 2.5 to 3 kg minimum. Again, personally I would not do it. On my unit I use every day, I can switch between class A and class AAB (1.2A standing current vs circa 600 mA per channel). Note that if you run full class A at +-35V distortion will go up on the peak output. On class AAB the kx-Amp still has a very smooth sound.
If you have a transformer you want to use, but the voltage is a bit high, just dial the standing current down a bit to keep the total dissipation within reason. The way to ascertain this is to ensure the heat sink temperature settles out at about 55 Deg C max. You should not run much above this for reasons of reliability since you want this thing around for another 20 or 30 years.
Bridging. I have not tried this but it should work ok. The theoretical power is 4x the rating of a non- bridged amp. Since we only have 2 OPS pairs per amp, you will be asking it in bridged mode to deliver 100 Watts at very high temperatures, and much higher (double) Vce excursions. So, I would suggest you are conservative with the rails and do not go above 28 Volts and preferably keep them in the +-24 V range.
Bridge Wiring. I will try in the next day or so to propose a scheme. For the Zobels, I will have to think about it, but at this stage my view is each amp output should go to its own Zobel which goes back to 0V.
Last edited:
Many kind thanks.
>>> If you run the kx-Amp off +-35 volt rails in class A you will have a standing dissipation of 84 watts per channel. For two pairs of OP devices this is very high and I strongly recommend that you do NOT do this.
I suppose that is a tad hot ... A simple 0,3 K/W heat sink per board would not cope with that, I presume.
>>> If you run in class A at +-35V, you should go for enormous heat sinks if you want reliability - so try 2.5 to 3 kg minimum. Again, personally I would not do it.
OK. You convinced me. No class A with 35 Volts.
>>> I can switch between class A and class AAB (1.2A standing current vs circa 600 mA per channel). Note that if you run full class A at +-35V distortion will go up on the peak output. On class AAB the kx-Amp still has a very smooth sound.
So I suppose that if I stick to the existing 24VAC toroidals (c.a. 35V DC), I could simply resign from class A and remain hard wired with AAB. That would be one strategy.
>>> If you have a transformer you want to use, but the voltage is a bit high, just dial the standing current down a bit to keep the total dissipation within reason.
By dialing down the standing current, do you mean that as:
Fiddling/increasing the values of R1 and R18?
Or rather/better fiddling/decreasing the value of R8?
Or none of the above? A different way?
That would be the "other" strategy.
Dialing down the standing current: 1.2A ==>> 0,9A at high bias, and 600mA ==> 400 mA .... would that do the trick (at 35V DC rails)?
Obviously, if plan A and plan B are not advisable, then probably there is plan C: "Simply purchase the proper toroidal". Does it make at all sense to push for re-use of the existing 24VAC, or should I simply abandon such an idea?
>>>The way to ascertain this is to ensure the heat sink temperature settles out at about 55 Deg C max.
OK. I suppose that placing a crude, 70 degrees, NC, screw-on thermal sensor, in close proximity to the transistors, would hence not provide any added value? Or maybe a lower temperature rated NC sensor? Makes no sense? As a means of a last resort thermal run away cutoff?
>>>You should not run much above this for reasons of reliability since you want this thing around for another 20 or 30 years.
Indeed, I would want a long life for the device. So the temperature of the transistors should not exceed a certain absolute ceiling value anyway ... I was betting for 70 degrees, but if you say 55 degrees, I will stick to your recommendation.
>>> Bridging. I have not tried this but it should work ok. The theoretical power is 4x the rating of a non- bridged amp. Since we only have 2 OPS pairs per amp, you will be asking it in bridged mode to deliver 100 Watts at very high temperatures, and much higher (double) Vce excursions. So, I would suggest you are conservative with the rails and do not go above 28 Volts and preferably keep them in the +-24 V range. ... And probably stick to the hard wired AAB then ?
>>> Bridge Wiring. I will try in the next day or so to propose a scheme. For the Zobels, I will have to think about it, but at this stage my view is each amp output should go to its own Zobel which goes back to 0V.
Please do. Think about it and provide some thoughts on this. In any case, I would assume that it will not hurt if the two PSU boards are oriented, located, rotated, in such a manner, so that the GND terminals are adjacent and in direct proximity to each other, so as to keep the Inter-PSU-GND-Jumper at a minimal length. If properly oriented, I suppose that the Inter-PSU-GND-Jumper could be 10 AWG or 12 AWG (4..6 mm2) and as short as 2-3cm (?)
Last edited:
Hi Zjjwwa,
Actually, your 24vac dual secondary transformer should work perfect with your KX amp build. After the capMx voltage drop and sag from the amplifier load mine is just about dead on at +/-28vdc
Bonsai is right about running 35vdc rails with conventional aluminum heatsinks. Just too much heat to dissipate safely. You would need a more hi-tech approach with cpu coolers and fans with that higher voltage setup.
Good to hear that the sag is so hugeHi Zjjwwa,
Actually, your 24vac dual secondary transformer should work perfect with your KX amp build. After the capMx voltage drop and sag from the amplifier load mine is just about dead on at +/-28vdc
Bonsai is right about running 35vdc rails with conventional aluminum heatsinks. Just too much heat to dissipate safely. You would need a more hi-tech approach with cpu coolers and fans with that higher voltage setup.
In my case, that could actually save the day.
Are you using a traditional / standard rectifier bridge, or some low forward voltage variety?
Actually, I have two (stereo) bars, each made of electrolysis grade clean copper, dimensions 10mm x 40mm x 380mm (i.e. spanning two 190mm heat sinks per channel, 0,3K/W each). The idea being, that the transistors be mounted to the copper bar, and as copper is twice as good a thermal conductor than ALU, with the cross section of 1cm x 4cm, it would serve as a thermal spreader across the width of two heat sinks, kind of like "optimizing" the thermal spread across all of the fins of the heat sinks in a more uniform manner. Copper is cheap in Poland ... (relatively).
{{ example: Płaskownik miedziany miedź 40 x 10 mm - 50 cm 8267099772 - Allegro.pl 1 USD is ~= 4 PLN }}
As I see it, the PCB has the transistors spread to both sides, so that would imply two copper bars per channel ...
B.t.w. This is the chassis that I have and can use for the project.
DIY AUDIO Chassis Toroidy DT4 RADIATOR NATURAL ALU - Shop Toroidy.pl
Polish stuff. It has 4 x 0,3K/W heat sink sections. Each 5cm x 20cm x 19cm. That would essentially mean 0,15 K/W per each side of the enclosure. It is somewhat very basic in terms of looks, but made of steel, the thickness of which is satisfying, and with the 8mm ALU anodized front, I can live with it. Just wondering, what could I expect with 0,15 K/W per side, in terms of dissipation capability.
Should I go for the stereo unbridged version, with 0,15K/W per channel, or would it cope/manage to cool a bridged mode configuration within the same enclosure, with 4 KX boards, each with only 0,3 K/W under it?
Last edited:
You would have to adjust the value of R1 and/or R8 depending on how high the rails would be.
I’ve heard the Toroidy transformers are very good. Make sure you order it with a flux band (also called a ‘goss’ band) and an interwinding screen between the primary and secondary for the best results.
For temperature, it’s important to note that the output device ratings go down as temperature increases. For the ones the kx-Amp uses, it’s -1.4 watts per deg C above 25 C so if you operate at 70 Celsius you will be reducing the device capability from 200 W to about 140 W.
I’ve heard the Toroidy transformers are very good. Make sure you order it with a flux band (also called a ‘goss’ band) and an interwinding screen between the primary and secondary for the best results.
For temperature, it’s important to note that the output device ratings go down as temperature increases. For the ones the kx-Amp uses, it’s -1.4 watts per deg C above 25 C so if you operate at 70 Celsius you will be reducing the device capability from 200 W to about 140 W.
Attachments
Last edited:
OK. One more, very basic question, but I suppose I need to ask.
NJW3281 ... is this the "same" thing as 2sc3281 or KSC3281, or is it unique and different from all the rest of them?
I mean are they "direct replacements" for the NJW, ... or is this NJW very unique with it's sustained beta feature, a feature which the others do not have, albeit boasting the same numbering scheme in their names?
...
As for beating down the rail voltage, I suppose that I could optionally put in heat-sink equipped custom ripple eaters in front of the board, with adjusted resistor networks, so as to get rid of the one or two extra volts that the KX boards would not be happy about.
That does not get rid of the "heat problem", because the heat will still be there, but at least it will be "exported" onto a different module and a different heat sink. As in: away from the main heat sinks of the KX PCB's.
NJW3281 ... is this the "same" thing as 2sc3281 or KSC3281, or is it unique and different from all the rest of them?
I mean are they "direct replacements" for the NJW, ... or is this NJW very unique with it's sustained beta feature, a feature which the others do not have, albeit boasting the same numbering scheme in their names?
...
As for beating down the rail voltage, I suppose that I could optionally put in heat-sink equipped custom ripple eaters in front of the board, with adjusted resistor networks, so as to get rid of the one or two extra volts that the KX boards would not be happy about.
That does not get rid of the "heat problem", because the heat will still be there, but at least it will be "exported" onto a different module and a different heat sink. As in: away from the main heat sinks of the KX PCB's.
Last edited:
Hello again.Tthat was busy day.
I wanted to help you guys and I've compiled DIGIkey BOM.
here are basket id's:
ripple eater:
Web ID: 303061083
Access ID: 76504
kx-amp for TWO PCB (enter only 1 assembly):
Web ID: 303092174
Access ID: 61572
baskets are valid for a month i think. After opening You can download it in excel compatible with their webpage, and upload whenever want.
I wanted to help you guys and I've compiled DIGIkey BOM.
here are basket id's:
ripple eater:
Web ID: 303061083
Access ID: 76504
kx-amp for TWO PCB (enter only 1 assembly):
Web ID: 303092174
Access ID: 61572
baskets are valid for a month i think. After opening You can download it in excel compatible with their webpage, and upload whenever want.
The 3281/1302 are all electrically equivalent. The ones specified for the kx-Amp (and sx and nx) are all the TO-3P package.
Ripple eater - yes - you will lose about 1.5V across each rail IIRC by using the ripple eater. Indeed you will be 'exporting' it to a different place but it is a good solution. I recommend you use it anyway. Be sure not to test both channels at full power 4 Ohms with the ripple eater in place - it will probably fail. The ripple eater is designed to be used with music signals, not full power sine wave testing.
Hey, by a simple addition of two extra resistors, the ripple eater can be converted into a dual function device:
a). Ripple Eater, and
b). Proportional Voltage Reduction device.
Suffice to add a resistor on the back side of the board, to the pins of C5 and C7. (a parallel connection).
This new resistor, which is a shunt to ground, will constitute a voltage divider, when combined with R3, and in the other branch, with R4.
The Ripple Eater that currently be, is currently equipped with a Shunt_Resistor of infinity ohms (because it does not exist).
Therefore, the voltage divider is equal to:
Fraction = Infinity / (R3 + Infinity) = 100%.
Therefore, Voltage_Output = Fraction * Voltage_Input - Vbe - Vbe,
which basically boils down to a loss of two Vbe's:
Voltage_Output = 100% *Voltage_Input - 0,7V -0,7V.
But for any "other" fraction, other than 100%, for example 96%, and for a tad-too-high input voltage of say 35V, you would get:
Voltage_Output = 96% * 35 - 1,4V = 33,6 - 1,4 = 32,2 Volts.
The drawback: This approach will generate HEAT.
The power transistor of the Ripple Eater would obviously need a bigger heat sink, to accommodate for the bigger losses.
There will be more dropout voltage between its collector and emitter. Times the current, and it can get really hot.
So: be careful with that fraction, so as not to fry the Eater.
Always double check the heat losses and use a bigger heat-sink, accordingly.
OK. So how do I create the "Fraction"?
Fraction = R_shunt / (R_shunt + R3), therefore:
R_shunt = R3 * Fraction / (1 - Fraction)
Specifically, if R3 = 3k9, and Fraction = 0,96:
R_shunt = 3k9 * 0,96 /0,04 = 94k
This is a WONDERFUL way to fry MJE transistors !!!
a). Ripple Eater, and
b). Proportional Voltage Reduction device.
Suffice to add a resistor on the back side of the board, to the pins of C5 and C7. (a parallel connection).
This new resistor, which is a shunt to ground, will constitute a voltage divider, when combined with R3, and in the other branch, with R4.
The Ripple Eater that currently be, is currently equipped with a Shunt_Resistor of infinity ohms (because it does not exist).
Therefore, the voltage divider is equal to:
Fraction = Infinity / (R3 + Infinity) = 100%.
Therefore, Voltage_Output = Fraction * Voltage_Input - Vbe - Vbe,
which basically boils down to a loss of two Vbe's:
Voltage_Output = 100% *Voltage_Input - 0,7V -0,7V.
But for any "other" fraction, other than 100%, for example 96%, and for a tad-too-high input voltage of say 35V, you would get:
Voltage_Output = 96% * 35 - 1,4V = 33,6 - 1,4 = 32,2 Volts.
The drawback: This approach will generate HEAT.
The power transistor of the Ripple Eater would obviously need a bigger heat sink, to accommodate for the bigger losses.
There will be more dropout voltage between its collector and emitter. Times the current, and it can get really hot.
So: be careful with that fraction, so as not to fry the Eater.
Always double check the heat losses and use a bigger heat-sink, accordingly.
OK. So how do I create the "Fraction"?
Fraction = R_shunt / (R_shunt + R3), therefore:
R_shunt = R3 * Fraction / (1 - Fraction)
Specifically, if R3 = 3k9, and Fraction = 0,96:
R_shunt = 3k9 * 0,96 /0,04 = 94k
This is a WONDERFUL way to fry MJE transistors !!!
Last edited:
... Caveat: The time constant.
The current time constant of the Ripple Eater is t = RC seconds, which is 1/(RC) Hertz:
t = 3900 ohms * 0,001 Farads = 3,9 seconds. Or: 0,25 Hertz.
Way outside the audio band. Basically, a concrete slab of inertia.
Now, by adding the shunt resistor, we change the impedance at the top of the capacitor, hence changing the time constant, but only very slightly, so this should not be a concern or degradation of the ripple filtrating capability of the circuit:
3k9 || 94k = 3k74. Or in other words, a change, mark-down, by no more than 4%.
This will not harm the filtration capability of the Ripple Eater in any meaningful way.
Obviously, the "fraction" can not diverge to much from 100%, because otherwise, the transistor with fry, and the diminished filtration capability will be the least of our problems.
The current time constant of the Ripple Eater is t = RC seconds, which is 1/(RC) Hertz:
t = 3900 ohms * 0,001 Farads = 3,9 seconds. Or: 0,25 Hertz.
Way outside the audio band. Basically, a concrete slab of inertia.
Now, by adding the shunt resistor, we change the impedance at the top of the capacitor, hence changing the time constant, but only very slightly, so this should not be a concern or degradation of the ripple filtrating capability of the circuit:
3k9 || 94k = 3k74. Or in other words, a change, mark-down, by no more than 4%.
This will not harm the filtration capability of the Ripple Eater in any meaningful way.
Obviously, the "fraction" can not diverge to much from 100%, because otherwise, the transistor with fry, and the diminished filtration capability will be the least of our problems.
Last edited:
I have not found any issues on either the sx, nx or kx-amps so I would say no the thermal stability wrt offset is very good considering the feedback is DC coupled and there is no servo. The output offset is to < 5mV within 2-3 minutes of powering up and to half that (2-3 mV) within 5 minutes of powering up.
What is the exactly schematic of the ripple eater that you speaking about?
Last edited:
Is it possible / permissible / advisable to exchange the input transistors
Q12 - BC546B
Q14 - BC556B
With lower max voltage, but low noise and high gain grade transistors?
Q12' - BC550C
Q14' - BC560C
The assumption being that the collectors of Q12 and Q14 are protected by the 10V Zener diodes, so they are essentially working under low voltage conditions.
...
The Q13, Q15 transistors would stay as they are, because of the high voltage conditions that they work in. Or maybe not so? Does the Vce on Q13, Q15 ever exceed 50V, as permissible for the BC550C / BC560C to work with?
With the low value of R25, in comparison to the 5x1,8 that are exposed to speaker output swing, the midpoint of R34, R35 and R25 seems not to swing way too far out from ground potential, does it? If indeed this node is "hanging around" ground potential, then maybe it would be possible to use the low noise, high gain BC550/BC560C in the positions of Q13, Q15 too?
...
Looking at Q1 and Q2, these also seem to have a Vce never exceeding Rail to Ground voltage of 35V. Is it possible to use the low noise, high gain BC550C / BC560C here?
Q12 - BC546B
Q14 - BC556B
With lower max voltage, but low noise and high gain grade transistors?
Q12' - BC550C
Q14' - BC560C
The assumption being that the collectors of Q12 and Q14 are protected by the 10V Zener diodes, so they are essentially working under low voltage conditions.
...
The Q13, Q15 transistors would stay as they are, because of the high voltage conditions that they work in. Or maybe not so? Does the Vce on Q13, Q15 ever exceed 50V, as permissible for the BC550C / BC560C to work with?
With the low value of R25, in comparison to the 5x1,8 that are exposed to speaker output swing, the midpoint of R34, R35 and R25 seems not to swing way too far out from ground potential, does it? If indeed this node is "hanging around" ground potential, then maybe it would be possible to use the low noise, high gain BC550/BC560C in the positions of Q13, Q15 too?
...
Looking at Q1 and Q2, these also seem to have a Vce never exceeding Rail to Ground voltage of 35V. Is it possible to use the low noise, high gain BC550C / BC560C here?