As promised a few weeks ago, here is the new kx2-Amplifier (hereinafter simply the 'kx2').
This design replaces the kx-amplifier that has been withdrawn and offers significant upgrades to performance. It uses the same PCB footprint as the kx-amplifier (and the nx and sx amplifiers before that). The bugs reported by some of the builders have been resolved, making this a trouble-free project.
The full pdf build doc and Excel BOM are here hifisonix kx2-Amplifier (C. 5.7 MB download). There is a link to a short introductory YouTube video on the kx2 as well.
You can buy a set of silk screened through hole plated boards from Jim's Audio here: - kx2-Amplifier PCB Set
Main features of the design:-
Any questions or assistance needed, please feel free to post them up here or to message me at hifisonix.com
Disclaimer: I make no financial gain from the sale of the kx2 boards through Jim's Audio. They are offered as a service to interested DIY builders.
The KSC2690 and KSA1220 driver transistors have been discontinued by ON and are no longer available from reputable distributors like Mouser and Digikey. However, this is no problem for the kx2-Amp. Simply replace Q10 with a KSC3503 and Q11 KSA1220 with the KSA1381. These are the same devices used in the amplifier VAS stages but work perfectly as drivers in the kx-Amp without modification or change in performance.
Update November 10th 2021: 146 PCB sets have been sold through Jim's Audio!
Update November 20th 2022: The schematic part assignment for Q7 and the BOM have been corrected. The PCB silk screen shows a BC560 - it should be BC550
Update November 25th 2022: formula on Slide 18 had R1 and R8 transposed - now corrected
This design replaces the kx-amplifier that has been withdrawn and offers significant upgrades to performance. It uses the same PCB footprint as the kx-amplifier (and the nx and sx amplifiers before that). The bugs reported by some of the builders have been resolved, making this a trouble-free project.
The full pdf build doc and Excel BOM are here hifisonix kx2-Amplifier (C. 5.7 MB download). There is a link to a short introductory YouTube video on the kx2 as well.
You can buy a set of silk screened through hole plated boards from Jim's Audio here: - kx2-Amplifier PCB Set
Main features of the design:-
- switchable between class A, class AAB and class AB
- 15W RMS into 8 Ohms, 28 W peak
- 50W into 4 Ohms, 85W into 2 Ohms
- very low distortion
- Zobel network incorporated onto the PCB allowing other PSU's to be used
- improved offset adjustment
- TMC or Miller Compensation selected via link
- noise floor -115dBr 15W RMS in class A, -120dBr 15W RMS class AB
Any questions or assistance needed, please feel free to post them up here or to message me at hifisonix.com
Disclaimer: I make no financial gain from the sale of the kx2 boards through Jim's Audio. They are offered as a service to interested DIY builders.
The KSC2690 and KSA1220 driver transistors have been discontinued by ON and are no longer available from reputable distributors like Mouser and Digikey. However, this is no problem for the kx2-Amp. Simply replace Q10 with a KSC3503 and Q11 KSA1220 with the KSA1381. These are the same devices used in the amplifier VAS stages but work perfectly as drivers in the kx-Amp without modification or change in performance.
Update November 10th 2021: 146 PCB sets have been sold through Jim's Audio!
Update November 20th 2022: The schematic part assignment for Q7 and the BOM have been corrected. The PCB silk screen shows a BC560 - it should be BC550
Update November 25th 2022: formula on Slide 18 had R1 and R8 transposed - now corrected
Last edited:
Andrew,
Excellent work once again, I’ll need to review the manual thoroughly ;-)
One question. Is there a way to decrease the overall gain? 28dB is very high for an amplifier that is aimed for horns and the like. Especially with modern digital sources that put out at least 3V RMS.
Is there a way to get it close to 20-22 dB?
Best,
Anand.
Excellent work once again, I’ll need to review the manual thoroughly ;-)
One question. Is there a way to decrease the overall gain? 28dB is very high for an amplifier that is aimed for horns and the like. Especially with modern digital sources that put out at least 3V RMS.
Is there a way to get it close to 20-22 dB?
Best,
Anand.
Hello Anand, thank you for your kind words and apologies for the late reply.
I have tested the amp with R25 = 22 Ohms, but not higher. My amp is running with R25 = 15 Ohms. To reduce the gain to c. 22 dB, you would have to increase R25 to 33 or 39 Ohms.
You should not change the feedback resistor (this is the 5 x 1.8 k resistors in parallel called Rf) as this sets the diamond buffer gm which has a direct bearing on the amplifier compensation - this is a CFA topology amp so Rf forms part of the compensation.
Let me stress, I have not reduced gain as much as you need - only a bit by taking R25= to 22 Ohms.
I suggest you try R25 = 33 or 39. Drive a 2-3 kHz signal of c 3V pk-pk into an 8 ohm resistive load. It should be clean with no overshoot. If you do get overshoot, let me know and I will re-comp the amp for you - however, I’m hoping you will not have to do that.
I have tested the amp with R25 = 22 Ohms, but not higher. My amp is running with R25 = 15 Ohms. To reduce the gain to c. 22 dB, you would have to increase R25 to 33 or 39 Ohms.
You should not change the feedback resistor (this is the 5 x 1.8 k resistors in parallel called Rf) as this sets the diamond buffer gm which has a direct bearing on the amplifier compensation - this is a CFA topology amp so Rf forms part of the compensation.
Let me stress, I have not reduced gain as much as you need - only a bit by taking R25= to 22 Ohms.
I suggest you try R25 = 33 or 39. Drive a 2-3 kHz signal of c 3V pk-pk into an 8 ohm resistive load. It should be clean with no overshoot. If you do get overshoot, let me know and I will re-comp the amp for you - however, I’m hoping you will not have to do that.
Andrew!
Excellent!
I have a 5U empty heatsink, along with a bench supply, and a very nice Keysight oscilloscope (with Bode plot capabilities), so we can do all this together prior to committing it to a final chassis.
Can you explain in technical terms why you chose 2-3khz signal? 3Vpp is 1.5Vp or 0.14W rms into 8 ohms. Would you rather have me test at 8Vpp or 4Vp or 1W RMS/8 ohms?
Best,
Anand.
Excellent!
I have a 5U empty heatsink, along with a bench supply, and a very nice Keysight oscilloscope (with Bode plot capabilities), so we can do all this together prior to committing it to a final chassis.
Can you explain in technical terms why you chose 2-3khz signal? 3Vpp is 1.5Vp or 0.14W rms into 8 ohms. Would you rather have me test at 8Vpp or 4Vp or 1W RMS/8 ohms?
Best,
Anand.
Last edited:
Anand
1. When you do a loop stability test, you typically want to do it in the ‘small signal’ regime. At 1-3 V out, the voltage swings on the devices, and particularly the VAS, are low and you are not getting large signal modulation of device parameters like Cob or hFE for example. In other words, you are in effect isololating device parasitics from the loop gain investigation.
2. I choose 2-3 kHz because it’s convenient and again is well away from frequencies where other circuit issues may confound the test. That said, a square wave has a lot of harmonics but if I injected 20 or 30 kKz in, overshoot would not be that apparent because of the action of the input filter, for example.
3. Once you have established that loop is stable and does not suffer from overshoot, you can then exercise the amp with a large signal stimulus at about 80% of the max output. A good test takes a high amplitude sine wave at slow frequency ( say 10-50 Hz) and imposes on it a low amplitude ( few hundred mV) square wave of 2-3 kHz. The LF signal is responsive for shifting the device parameters while the HF part looks at stability over the large signal swing of the amp. Often you will find the amp breaks into HF oscillation at the LF sine wave peaks - a sure sine of parasitic instability and not normally related to the loop stability.
Note, all these tests are into a resistive load. Final testing will include driving capacitive loads, inductive loads etc and then hard/soft overdrive, again places where things like ‘sticky rail’ may manifest, or ringing as the amp comes out of clipping.
1. When you do a loop stability test, you typically want to do it in the ‘small signal’ regime. At 1-3 V out, the voltage swings on the devices, and particularly the VAS, are low and you are not getting large signal modulation of device parameters like Cob or hFE for example. In other words, you are in effect isololating device parasitics from the loop gain investigation.
2. I choose 2-3 kHz because it’s convenient and again is well away from frequencies where other circuit issues may confound the test. That said, a square wave has a lot of harmonics but if I injected 20 or 30 kKz in, overshoot would not be that apparent because of the action of the input filter, for example.
3. Once you have established that loop is stable and does not suffer from overshoot, you can then exercise the amp with a large signal stimulus at about 80% of the max output. A good test takes a high amplitude sine wave at slow frequency ( say 10-50 Hz) and imposes on it a low amplitude ( few hundred mV) square wave of 2-3 kHz. The LF signal is responsive for shifting the device parameters while the HF part looks at stability over the large signal swing of the amp. Often you will find the amp breaks into HF oscillation at the LF sine wave peaks - a sure sine of parasitic instability and not normally related to the loop stability.
Note, all these tests are into a resistive load. Final testing will include driving capacitive loads, inductive loads etc and then hard/soft overdrive, again places where things like ‘sticky rail’ may manifest, or ringing as the amp comes out of clipping.
Last edited:
Andrew,
Your explanation in part 3 is absolutely golden. That is amp testing at the heart! Is this a type of IMD distortion testing?
When you have time (and you have given so much of your time already), I would suggest you write a document on simple amp testing parameters that advanced diy hobbyists that own oscilloscopes can perform, i.e. with both resistive and capacitive loads as well as perhaps simple distortion testing. A QA400 isn't that expensive but even REW can be used to do some simple distortion measurements for peace of mind.
You have a talent for teaching and as long as it doesn't encroach upon your real job (i.e. selling boutique solid state amplifiers) it would help a lot of us hobbyists who don't have an EE but are interested in the science.
You can see in my signature, my NPXP build and I did do some simple testing of the Aleph J, albeit without any capacitive loads only resistive. I have a long ways to go, but I want my NPXP to be an exploration of multiple different amplifier topologies from a subjective standpoint but also a way to sharpen my measurement skills.
Best,
Anand.
Your explanation in part 3 is absolutely golden. That is amp testing at the heart! Is this a type of IMD distortion testing?
When you have time (and you have given so much of your time already), I would suggest you write a document on simple amp testing parameters that advanced diy hobbyists that own oscilloscopes can perform, i.e. with both resistive and capacitive loads as well as perhaps simple distortion testing. A QA400 isn't that expensive but even REW can be used to do some simple distortion measurements for peace of mind.
You have a talent for teaching and as long as it doesn't encroach upon your real job (i.e. selling boutique solid state amplifiers) it would help a lot of us hobbyists who don't have an EE but are interested in the science.
You can see in my signature, my NPXP build and I did do some simple testing of the Aleph J, albeit without any capacitive loads only resistive. I have a long ways to go, but I want my NPXP to be an exploration of multiple different amplifier topologies from a subjective standpoint but also a way to sharpen my measurement skills.
Best,
Anand.
Andrew,
Couple other clarifications if I may since I have a fully built pair of KX boards and Ripple Eater PSU boards.
In order to minimize any loss of investment I was going to use the Ripple Eater PSU boards for the KX2 build but since the Zobel/Boucherot Cell network is included on the amp pcb, I could just avoid that connection for the Ripple Eater PSU (ZNA/ZNB). In addition, I was going to transplant the 0.6uH coil [Thiele network] from my KX pcb to the KX2 pcb. I have not mounted many of the large transistors to the KX pcb so I can use those on the KX2 as long as they are one and the same.
I am planning on making this new KX2 build Class AAB. With 96dB speakers most of my listening is at under 5 watts anyway.
Thanks!
Anand.
Couple other clarifications if I may since I have a fully built pair of KX boards and Ripple Eater PSU boards.
In order to minimize any loss of investment I was going to use the Ripple Eater PSU boards for the KX2 build but since the Zobel/Boucherot Cell network is included on the amp pcb, I could just avoid that connection for the Ripple Eater PSU (ZNA/ZNB). In addition, I was going to transplant the 0.6uH coil [Thiele network] from my KX pcb to the KX2 pcb. I have not mounted many of the large transistors to the KX pcb so I can use those on the KX2 as long as they are one and the same.
I am planning on making this new KX2 build Class AAB. With 96dB speakers most of my listening is at under 5 watts anyway.
Thanks!
Anand.
Last edited:
Anand,
The Bowes test is not an IMD test which is a distortion measurement and particularly useful in measurement systems where you often have a limited bandwidth on the test equipment. A standard 19+20 kHz test will give you an IM tone at 1kHz. In a conventional 20 kHz test, you would be looking for the first distortion harmonic at 20 kHz. The advantage of the IMD test is that the distortion component is 'folded down' into the measurement bandwidth where is can be accurately assessed.
See Bob Cordell's paper here CordellAudio.com - A Fully In-Band Multitone Test for Transient Intermodulation Distortion
The Bowes test simply looks at the loop stability over the full output swing of the amplifier.
Re your kx > kx2 build proposals - that all sounds good and should not cause any issues.
The Bowes test is not an IMD test which is a distortion measurement and particularly useful in measurement systems where you often have a limited bandwidth on the test equipment. A standard 19+20 kHz test will give you an IM tone at 1kHz. In a conventional 20 kHz test, you would be looking for the first distortion harmonic at 20 kHz. The advantage of the IMD test is that the distortion component is 'folded down' into the measurement bandwidth where is can be accurately assessed.
See Bob Cordell's paper here CordellAudio.com - A Fully In-Band Multitone Test for Transient Intermodulation Distortion
The Bowes test simply looks at the loop stability over the full output swing of the amplifier.
Re your kx > kx2 build proposals - that all sounds good and should not cause any issues.