I am in the process of bringing up my NX amplifier and had a problem with the power supply board switch-on delay circuit.
Here is a list of the component values I started with: C2 = C3 = 1uF, R7 = 5.62K, R6 = 100K, and C10 = 100uF. The transformer is an Antek AS4434, 2x 34 VAC at 400 VA. The line voltage is 120 VAC at 60 Hz with a fair amount of flat topped distortion at the sine wave peaks.
On page 28 of the amplifier PDF file item 5 states that the gate of Q3 should be 3V or more to turn on the speaker protection MOSFETs. My board measured 1.2V at the Q3 gate, the Q3 drain was about 48V, and the LED D15 was off. There was zero voltage drop across R1 so I didn't bother lifting it since that was not causing the voltage loss at the gate of Q3.
I coded the switch-on circuit into LTspice to gain some understanding of circuit function. LTspice indicated the Q3 gate voltage should be 1.4 volts, not far from what was measured. The difference could be from the ideal undistorted sine wave input used in the simulation as contrasted with the clipped sine wave measured on my oscilloscope. Raising the value of R7 increased the gate voltage and spice indicated that using R7 = 12K would give about 2.5V at the gate. On the bench it yielded about 2V but now D15 turned on. The drain was about 22V so the 2N7002 was not turned on fully. The final change was R7 = 20K and that gave a gate voltage of 2.5V and a drain voltage of about 0.3V. All other tests for the power supply card now passed.
Although this circuit has had component values adjusted once before, it appears that reliable operation may still require further tweaking as it did for me. The use of a lower voltage transformer or 50 Hz line will tend to reduce the gate voltage and starve the 2N7002. Adjusting R7 seemed to be the easiest way of raising the gate voltage but maybe some other means would be preferred.
Here is a list of the component values I started with: C2 = C3 = 1uF, R7 = 5.62K, R6 = 100K, and C10 = 100uF. The transformer is an Antek AS4434, 2x 34 VAC at 400 VA. The line voltage is 120 VAC at 60 Hz with a fair amount of flat topped distortion at the sine wave peaks.
On page 28 of the amplifier PDF file item 5 states that the gate of Q3 should be 3V or more to turn on the speaker protection MOSFETs. My board measured 1.2V at the Q3 gate, the Q3 drain was about 48V, and the LED D15 was off. There was zero voltage drop across R1 so I didn't bother lifting it since that was not causing the voltage loss at the gate of Q3.
I coded the switch-on circuit into LTspice to gain some understanding of circuit function. LTspice indicated the Q3 gate voltage should be 1.4 volts, not far from what was measured. The difference could be from the ideal undistorted sine wave input used in the simulation as contrasted with the clipped sine wave measured on my oscilloscope. Raising the value of R7 increased the gate voltage and spice indicated that using R7 = 12K would give about 2.5V at the gate. On the bench it yielded about 2V but now D15 turned on. The drain was about 22V so the 2N7002 was not turned on fully. The final change was R7 = 20K and that gave a gate voltage of 2.5V and a drain voltage of about 0.3V. All other tests for the power supply card now passed.
Although this circuit has had component values adjusted once before, it appears that reliable operation may still require further tweaking as it did for me. The use of a lower voltage transformer or 50 Hz line will tend to reduce the gate voltage and starve the 2N7002. Adjusting R7 seemed to be the easiest way of raising the gate voltage but maybe some other means would be preferred.
I ordered a set of Stanton's nx-Amp boards which arrived last week and assembled them on Friday night and Saturday. Today I tested them and have to report that I had the same problems as Terry (offset cannot be dialed out) and low volt drop across the level shifter resistors (jprco noted this issue)
There are two issues here that are somewhat linked
1. The offset adjustment does not cater for the spread in the front end devices
2. There is a biasing issue with level shifters that exacerbates the offset issue and can result in very low or high standing current in the level shifter outputs
The nature of the problem is such that on some boards you will see no issues, either one of the issues, or both issues on the same board. On the two boards I built this weekend, the first board suffered offset issues, and second board has both problems. The original two boards (and the two prototypes prior to that) did not suffer any of these problems.
1. Offset Problem. The offset current adjustment is fed from a +- 10 V supply via a 10 k pot and a 10 k series resistor. The lower gain setting resistor is 15 Ohms to ground. This limits the amount of current that can be injected or extracted from the summing junction.
2. The bias conditions with the values currently shown ideally set both the input buffer and the level shifter at 1 mA. If there are large enough spreads in Vbe (I.e a few mV), you can end up with a situation where you have large offsets, or if the level shifter Vbe's are sufficiently higher than the input buffer Vbe's, low standing current in the level shifters. This is the problem jprco is seeing on one of his boards (while the other board seems ok).
In summary, the cause of these issues has been identified, and I acknowledge that some of the boards will not bias up correctly and there will be offset issues.
I have figured out a fix that involves resistor value changes only. I will be extraordinarily busy this week, so only next weekend will I be able do some further in depth tests. Once confirmed, I will add a 'fix' sheet on my website, and post it up here.
These issues have nothing to due with Stanton's boards - this is a design issue related to component spreads.
Again, apologies - please wait until next week end for the changes required to effect a cure.
There are two issues here that are somewhat linked
1. The offset adjustment does not cater for the spread in the front end devices
2. There is a biasing issue with level shifters that exacerbates the offset issue and can result in very low or high standing current in the level shifter outputs
The nature of the problem is such that on some boards you will see no issues, either one of the issues, or both issues on the same board. On the two boards I built this weekend, the first board suffered offset issues, and second board has both problems. The original two boards (and the two prototypes prior to that) did not suffer any of these problems.
1. Offset Problem. The offset current adjustment is fed from a +- 10 V supply via a 10 k pot and a 10 k series resistor. The lower gain setting resistor is 15 Ohms to ground. This limits the amount of current that can be injected or extracted from the summing junction.
2. The bias conditions with the values currently shown ideally set both the input buffer and the level shifter at 1 mA. If there are large enough spreads in Vbe (I.e a few mV), you can end up with a situation where you have large offsets, or if the level shifter Vbe's are sufficiently higher than the input buffer Vbe's, low standing current in the level shifters. This is the problem jprco is seeing on one of his boards (while the other board seems ok).
In summary, the cause of these issues has been identified, and I acknowledge that some of the boards will not bias up correctly and there will be offset issues.
I have figured out a fix that involves resistor value changes only. I will be extraordinarily busy this week, so only next weekend will I be able do some further in depth tests. Once confirmed, I will add a 'fix' sheet on my website, and post it up here.
These issues have nothing to due with Stanton's boards - this is a design issue related to component spreads.
Again, apologies - please wait until next week end for the changes required to effect a cure.
Last edited:
Thanks Andrew. At least I know that I didn't do anything wrong. Unfortunately, I wasted two full sets of transistors. I worried that my fix of lowering R9 was maybe not the best way to approach it. It worked for getting the offset fixed but maybe another part of the circuit is being strained because of it. The upper mids sound too hot, (sibilant). I will wait for your fix. Thanks for making this effort on our behalf.
Blessings, Terry
Blessings, Terry
Not being an NX or SX builder but following the thread wouldn't this always be a problem when calculating a circuits design aspects when the total component value spread is wider than anticipated? Using matched component values or a narrow range of assumptions rather than a real worst case analysis almost always leads to these kinds of problems. If we looked at the bell curve of spread on some of the many different implementations as it is obvious that many different types of devices are used by different people it becomes another reason not to substitute different types or brands of devices and to stay with the recommended devices that the original designer used to create the circuits. How can someone like Bonsai design for so many different choices of devices in the original design unless he tried every perturbation originally and has multiple solutions for each change in device? This is one reason that saying that yes one device could be substituted for another may be a Rubik's cube of snakes just waiting to bite you in the behind.
Thanks Andrew. Now I know I am not crazy! After a lot of part switching (like Terry), I think I have 3 boards working - still testing though. I am glad it is not related to Jim's Audio boards. Just a caution though, the boards are not too forgiving if you have to remove/replace parts as I said in earlier post. The pads and or traces lift off quit easily even with low heat applied. I will stay tune and see what you come up with.
Jean
Jean
Bonsai,
In the second phase of bring-up I find that my amplifier boards have very low bias current in all transistors that are on the heatsink. The adjustment pot has little effect on the current. Both boards behave this way and I did a quick check of component values without finding anything wrong. Will the update you plan for next week be likely to affect this issue?
Did you have any comment on the power supply start-up delay issue in my last note?
In the second phase of bring-up I find that my amplifier boards have very low bias current in all transistors that are on the heatsink. The adjustment pot has little effect on the current. Both boards behave this way and I did a quick check of component values without finding anything wrong. Will the update you plan for next week be likely to affect this issue?
Did you have any comment on the power supply start-up delay issue in my last note?
Bill
1. On your first point, yes, the changes I WIL propose addresses these issues. The bias can be too high, too low or lopsided. Terry also had this problem.
2. Sorry, I did not look. I proposed to increase the value of the AC feed cap. This issue nil happens when the secondary voltage is low. I will feed back to you this coming week end
I am on the road all this week - Shenzhen and Beijing.
1. On your first point, yes, the changes I WIL propose addresses these issues. The bias can be too high, too low or lopsided. Terry also had this problem.
2. Sorry, I did not look. I proposed to increase the value of the AC feed cap. This issue nil happens when the secondary voltage is low. I will feed back to you this coming week end
I am on the road all this week - Shenzhen and Beijing.
Bonsai,
Regarding the start-up delay circuit, the latest changes including the increased value of the input capacitor were included in my build and in the simulation. The .asc file is attached to my note if you want to run the simulation and play with the component values. The diodes text file contains models needed for the simulation.
Thanks for supporting the NX amplifier builds for us. I'll probably put the project aside until the next update.
Regarding the start-up delay circuit, the latest changes including the increased value of the input capacitor were included in my build and in the simulation. The .asc file is attached to my note if you want to run the simulation and play with the component values. The diodes text file contains models needed for the simulation.
Thanks for supporting the NX amplifier builds for us. I'll probably put the project aside until the next update.
nx-Amplifier - Important Notice
A few people have emailed me concerning offset adjustment problems, and more seriously, low or zero bias voltage across R32 and R33.
I built 2 prototypes and 2 final boards that I used in the amplifier of which you can see photos in the nx-Amplifier write up. In all 4 cases, I had no issues whatsoever and the amplifier voltages were all correct, as calculated and subsequently predicted by the simulator; offset adjustment was also correct at +- 500mV.
However, out of the 25 PCB sets (so around 50 individual amplifier boards) sold so far on Jim’s Audio, there have been 4 serious complaints (I do not know how many have been actually assembled at this stage). I rechecked my two final boards again, carefully measuring the voltages and offset adjustment range - all was as I had originally intended and measured, and both amplifiers work flawlessly. In order to ascertain whether or not the PCB’s were at fault, I ordered a set from Jim’s Audio and assembled both nx-Amplifier boards. The Jim’s Audio boards use the same Gerber files I used – no changes were made between the final boards I used and the production ones from Jim’s Audio. After carefully checking, I can confirm the PCB’s are 100% correct.
To summarize, here are the issues that were being reported by some builders:-
1. Larger than expected offsets that cannot be adjusted out – typically, at the maximum pot travel builders were still seeing 200mV of offset, even though they had hFE matched the input devices
2. Measuring across R32 and R33, one builder measured 700mV instead of 1V, and on one amplifier build, almost 0V – i.e. in this case both the diamond buffer level transistors Q9 and Q10 were essentially OFF which is completely wrong.
In all cases, the zener voltages (D3 and D4) were measured by the builders at 10V +-200mV which is correct and supply voltages used by the various builders were reported as +-40 VDC through to +-50VDC, again, perfectly acceptable and within range.
Q8 and Q10 are high hFE BC5x7C type devices, while Q9 and Q11 are lower gain range, higher Vce BC5X6B device types. Both the buffer and level shifter stages were designed to run at 1mA on the original design. 15 mV degeneration is applied via the 1mA flowing through the 15 Ohm resistors in the emitters of the 4 diamond buffer transistors – R28, R28 and R36 and R37
After some investigation, the problem has been isolated to normal differences in the spread of the Vbe’s of Q8-Q11.
The problem with the front end design as is that if the Vbe’s of the level shifter devices Q9 and Q11 are greater than the front end buffer devices Q8 and Q10, they are starved of current, and in extreme cases can actually be biased OFF, just as the one builder reported.
The standing offset across the degeneration resistors is 15mV and this is not enough to cater for the worst case spreads in the Vbe’s between the transistors - and matching hFE will not solve the problem either. The modifications detialed below (which consist only of resistor value changes) raise the degeneration voltages to 120-150mV, solving the problem.
If you have not assembled your boards yet, or you are having problems, make the following changes:-
R36 and R37 become 150 Ohms
R28 and R29 become 110 Ohms
R1 becomes 4.7k (but first read note below)
If you leave R1 at 10k, you will have around +-500mV of offset adjustment. I would suggest that you change R36/37 and R28/29 and then check the offset adjustment. If you cannot dial out the offset, then change R1 to 4.7k.
Across each of the degeneration resistors R28 and R29, you will read 120mV which is 8x to 9x the original value, while across R36 and R37, you will read 150mV.
The distortion, bandwidth and slew rate of the amplifier essentially remain unchanged.
I have made the changes to the two boards from the set I purchased from Jim's audio and can confirm that they work.
Changes to the nx-Amplifier document will be made to reflect this update in the next few weeks.
Let me also reconfirm, there are no layout issues or errors on the PCB – you can buy these with confidence – just make the value changes when building as detailed above.
Any questions, feel free to post it up on hifisonix.com or email me via DIYAudio.com.
A few people have emailed me concerning offset adjustment problems, and more seriously, low or zero bias voltage across R32 and R33.
I built 2 prototypes and 2 final boards that I used in the amplifier of which you can see photos in the nx-Amplifier write up. In all 4 cases, I had no issues whatsoever and the amplifier voltages were all correct, as calculated and subsequently predicted by the simulator; offset adjustment was also correct at +- 500mV.
However, out of the 25 PCB sets (so around 50 individual amplifier boards) sold so far on Jim’s Audio, there have been 4 serious complaints (I do not know how many have been actually assembled at this stage). I rechecked my two final boards again, carefully measuring the voltages and offset adjustment range - all was as I had originally intended and measured, and both amplifiers work flawlessly. In order to ascertain whether or not the PCB’s were at fault, I ordered a set from Jim’s Audio and assembled both nx-Amplifier boards. The Jim’s Audio boards use the same Gerber files I used – no changes were made between the final boards I used and the production ones from Jim’s Audio. After carefully checking, I can confirm the PCB’s are 100% correct.
To summarize, here are the issues that were being reported by some builders:-
1. Larger than expected offsets that cannot be adjusted out – typically, at the maximum pot travel builders were still seeing 200mV of offset, even though they had hFE matched the input devices
2. Measuring across R32 and R33, one builder measured 700mV instead of 1V, and on one amplifier build, almost 0V – i.e. in this case both the diamond buffer level transistors Q9 and Q10 were essentially OFF which is completely wrong.
In all cases, the zener voltages (D3 and D4) were measured by the builders at 10V +-200mV which is correct and supply voltages used by the various builders were reported as +-40 VDC through to +-50VDC, again, perfectly acceptable and within range.
Q8 and Q10 are high hFE BC5x7C type devices, while Q9 and Q11 are lower gain range, higher Vce BC5X6B device types. Both the buffer and level shifter stages were designed to run at 1mA on the original design. 15 mV degeneration is applied via the 1mA flowing through the 15 Ohm resistors in the emitters of the 4 diamond buffer transistors – R28, R28 and R36 and R37
After some investigation, the problem has been isolated to normal differences in the spread of the Vbe’s of Q8-Q11.
The problem with the front end design as is that if the Vbe’s of the level shifter devices Q9 and Q11 are greater than the front end buffer devices Q8 and Q10, they are starved of current, and in extreme cases can actually be biased OFF, just as the one builder reported.
The standing offset across the degeneration resistors is 15mV and this is not enough to cater for the worst case spreads in the Vbe’s between the transistors - and matching hFE will not solve the problem either. The modifications detialed below (which consist only of resistor value changes) raise the degeneration voltages to 120-150mV, solving the problem.
If you have not assembled your boards yet, or you are having problems, make the following changes:-
R36 and R37 become 150 Ohms
R28 and R29 become 110 Ohms
R1 becomes 4.7k (but first read note below)
If you leave R1 at 10k, you will have around +-500mV of offset adjustment. I would suggest that you change R36/37 and R28/29 and then check the offset adjustment. If you cannot dial out the offset, then change R1 to 4.7k.
Across each of the degeneration resistors R28 and R29, you will read 120mV which is 8x to 9x the original value, while across R36 and R37, you will read 150mV.
The distortion, bandwidth and slew rate of the amplifier essentially remain unchanged.
I have made the changes to the two boards from the set I purchased from Jim's audio and can confirm that they work.
Changes to the nx-Amplifier document will be made to reflect this update in the next few weeks.
Let me also reconfirm, there are no layout issues or errors on the PCB – you can buy these with confidence – just make the value changes when building as detailed above.
Any questions, feel free to post it up on hifisonix.com or email me via DIYAudio.com.
Last edited:
Translation : Diamond Input = EVIL
Only kidding Andrew.
Kudos for your clear & detailed documentation.Have you fixed the protection circuit's sensitivity to the voltage rails?