Hi Juma,
I still have 4 pairs of the original 50/135 laying around and am intregued by your Cubie2 design.
Also I do like the Malcolm Hawksford enhanced cascode that you use.
Only I am wondering why you place the C2 capacitor parrallel to the (constant-)current resistor R14. I think this will degrade the PSRR as it could modulate the basis of Q6/7.
Would it not be better to place a cap over the LED's as originally done in the M-H design?
I still have 4 pairs of the original 50/135 laying around and am intregued by your Cubie2 design.
Also I do like the Malcolm Hawksford enhanced cascode that you use.
Only I am wondering why you place the C2 capacitor parrallel to the (constant-)current resistor R14. I think this will degrade the PSRR as it could modulate the basis of Q6/7.
Would it not be better to place a cap over the LED's as originally done in the M-H design?
Hi Rick,
C2 equalizes eventual small HF fluctuations over R14 and since the amp is symmetrical it should improve PSRR.
Anyway, it's easy to tests and measure your ideas and see what you get...
So, with the moving restrictions and obligatory stay at home due to the Covid-19 pandemic, it's a great opportunity to get my hands on projects left aside for a long time now. Hence, I'm bringing back to life this thread to ask juma a couple of questions regarding Cubie2.
Since two years now, I had gathered all the components, but never had the chance to go further. First of all, what I did was an extensive testing and matching of the input jfets: I have a lot of K170 and enough J74 jfets, so I managed to get two decent pairs. Next, I had to test a bunch of BCs (around 30 from each type), in order to pick-up some nice triplets. Following this, I tested 6 pairs of lateral MOSFETs (namely the EXC10N20 and EXC10P20 Exicons) in order to check the Vgs voltage at an Id of 0.7 A. The EXC10N20s read around 1.4 V and the EXC10P20s read around 1.9 V.
So my questions to juma are the following:
1. The jfets I chose have an Idss of 7.8 mA. How can I calculate the value of R11 and R13 resistors in order to have a current o 10 mA flowing through Rbias? Can I leave the original 68 Ohm in place and adjust Rbias so that I have a current of 0.7 A through the output MOSFETs?
2. Have you any experience regarding the increase in output current? I mean, if I raise the output current to 0.8 or 1.0 A, what is the acoustic difference? I understand (from what Nelson Pass has told) that, for each output MOSFET, there is a "sweet spot", where it sounds best.
3. I would like to try a 0FB version. For that, I intend to build a FB and a non-FB board and try them listening just one channel. In your description, your instructions are clear regarding the changes needed. What is not mentioned, though, is where the wiper of the offset trimmer (P1) should be connected.
That's all for the moment.
Stay well, healthy and safe,
Vagelis
Since two years now, I had gathered all the components, but never had the chance to go further. First of all, what I did was an extensive testing and matching of the input jfets: I have a lot of K170 and enough J74 jfets, so I managed to get two decent pairs. Next, I had to test a bunch of BCs (around 30 from each type), in order to pick-up some nice triplets. Following this, I tested 6 pairs of lateral MOSFETs (namely the EXC10N20 and EXC10P20 Exicons) in order to check the Vgs voltage at an Id of 0.7 A. The EXC10N20s read around 1.4 V and the EXC10P20s read around 1.9 V.
So my questions to juma are the following:
1. The jfets I chose have an Idss of 7.8 mA. How can I calculate the value of R11 and R13 resistors in order to have a current o 10 mA flowing through Rbias? Can I leave the original 68 Ohm in place and adjust Rbias so that I have a current of 0.7 A through the output MOSFETs?
2. Have you any experience regarding the increase in output current? I mean, if I raise the output current to 0.8 or 1.0 A, what is the acoustic difference? I understand (from what Nelson Pass has told) that, for each output MOSFET, there is a "sweet spot", where it sounds best.
3. I would like to try a 0FB version. For that, I intend to build a FB and a non-FB board and try them listening just one channel. In your description, your instructions are clear regarding the changes needed. What is not mentioned, though, is where the wiper of the offset trimmer (P1) should be connected.
That's all for the moment.
Stay well, healthy and safe,
Vagelis
Giasou Vageli,
1. Yes, keep original values and yes, you can adjust R bias to get the output stage bias you want.
2. Exicons have total Vgs of 1.4V + 1.9V=3.3V - the same value I got with 2sk1058/2sj162.
Acoustic difference is subjective so you'll have to try different values, listen and decide what you like the best.
3.See schm. below and be patient when setting DC offset, it is more thermally sensitive without the feedback loop.
P1 still sets the DC offset. Relation Ra/Rb roughly sets the gain so for Ra=120R and Rb=1k you'll have about 20db gain.
Output bias current is set with Rx,Ry i.e. Rx=Ry=150R-220R (higher resistor value means higher current so start with lower values). Rx and Ry don't have to be equal - you can change the DC offset that way too, but it will change the bias current at the same time so you can experiment with that freely (but carefully).
Best regards and sorry for my brevity - I'm extremely busy these days
1. Yes, keep original values and yes, you can adjust R bias to get the output stage bias you want.
2. Exicons have total Vgs of 1.4V + 1.9V=3.3V - the same value I got with 2sk1058/2sj162.
Acoustic difference is subjective so you'll have to try different values, listen and decide what you like the best.
3.See schm. below and be patient when setting DC offset, it is more thermally sensitive without the feedback loop.
P1 still sets the DC offset. Relation Ra/Rb roughly sets the gain so for Ra=120R and Rb=1k you'll have about 20db gain.
Output bias current is set with Rx,Ry i.e. Rx=Ry=150R-220R (higher resistor value means higher current so start with lower values). Rx and Ry don't have to be equal - you can change the DC offset that way too, but it will change the bias current at the same time so you can experiment with that freely (but carefully).
Best regards and sorry for my brevity - I'm extremely busy these days
Attachments
Geia sou Juma,
Thank you very much for your response.
While waiting for your reply, I tried to calculate the current through the Rbias string the hard way: I breadboarded the whole input/current mirror stage using the actual parts and found out that the current was a little lower (9.2 mA), so I had to put a Rbias resistor of 360 Ohm in order to get a voltage of 3.3 V across Rbias. A little trivial task, but it worked just fine.
Regarding the zero feedback version, I thought that the wiper of the offset trimmer should be referenced to ground, but my knowledge wasn't that specialized. So, your help was great in this matter.
Thank you again for you time and I will keep you informed as soon as I draw my impressions (unless something else comes up).
It's very good that you are busy at these hard times. On the contrary, I'm out of work for two weeks now and I'm trying to keep up with all those things that I had left behind.
Best regards,
Vagelis
Thank you very much for your response.
While waiting for your reply, I tried to calculate the current through the Rbias string the hard way: I breadboarded the whole input/current mirror stage using the actual parts and found out that the current was a little lower (9.2 mA), so I had to put a Rbias resistor of 360 Ohm in order to get a voltage of 3.3 V across Rbias. A little trivial task, but it worked just fine.
Regarding the zero feedback version, I thought that the wiper of the offset trimmer should be referenced to ground, but my knowledge wasn't that specialized. So, your help was great in this matter.
Thank you again for you time and I will keep you informed as soon as I draw my impressions (unless something else comes up).
It's very good that you are busy at these hard times. On the contrary, I'm out of work for two weeks now and I'm trying to keep up with all those things that I had left behind.
Best regards,
Vagelis
Geia sou Juma,
Just a quick question regarding the power supply.
In the Cubie2 thread, in post #62, you published a capacitance multiplier circuit, that is actually two positive rail circuits stacked to act as a symmetrical PS.
Later, in the Cubie3 thread, you published a cap multiplier circuit that is pure symmetrical, using complementary MOSFETs. Is there any differnce between the two? Is the pure symmetrical better than the stacked one?
Also, in the Cubie2 PS, you use a series RC (Zobel?) circuit on the output, while in the Cubie3 PS, there is a parallel RC network. What's the difference? Which one should I use?
Regards,
Vagelis
Just a quick question regarding the power supply.
In the Cubie2 thread, in post #62, you published a capacitance multiplier circuit, that is actually two positive rail circuits stacked to act as a symmetrical PS.
Later, in the Cubie3 thread, you published a cap multiplier circuit that is pure symmetrical, using complementary MOSFETs. Is there any differnce between the two? Is the pure symmetrical better than the stacked one?
Also, in the Cubie2 PS, you use a series RC (Zobel?) circuit on the output, while in the Cubie3 PS, there is a parallel RC network. What's the difference? Which one should I use?
Regards,
Vagelis
Well, I just finished building one channel and, as this is my standard practice, I put it on my workbench to check everything before connecting it to my speaker. For the initial testing, I always use a bench power supply, which is symmetrical and has also adjustable current limit, so that I can be on the safe side in case something goes wrong.
So, my initial findings are as following:
1. To my surprise, though I had checked the MOSFTETs regarding their Vgs values at 0.7A bias, the values I got were way different than those anticipated. In particular, initially, I noted a large current draw, that exceeded 1,5 A in the negative rail. So, my first task was to connect a trimpot in parallel with Rbias and adjust it in order to lower the current through the MOSFETs. What I noted is that there is a large discrepancy in current draw between the two rails. In particular, there is a difference of 0.25 A between the positive and negative rail. When I had checked the MOSFETs, there was also a discrepancy in Vgs values between the N and P channel ones. In the current configuration, when I adjust Rbias so that I have a current of 0.7 A through the N channel, the current from P channel is 0.95 A. So my question is the following: is there a problem with this? Also, a strange finding is that, when checked on the bench with a current feed of 0.7 A, Vgs was 1.4 V for the N channel and 1.9 V for the P channel. On the actual circuit, those values are 1.14 and 1.59 V, respectively.
2. I made a full power output measurement and found out that max power output (at the beginning of clipping on the oscilloscope) is just 8 W. Is that normal?
Looking forward for further enlightenment,
Best regards,
Evangelos
So, my initial findings are as following:
1. To my surprise, though I had checked the MOSFTETs regarding their Vgs values at 0.7A bias, the values I got were way different than those anticipated. In particular, initially, I noted a large current draw, that exceeded 1,5 A in the negative rail. So, my first task was to connect a trimpot in parallel with Rbias and adjust it in order to lower the current through the MOSFETs. What I noted is that there is a large discrepancy in current draw between the two rails. In particular, there is a difference of 0.25 A between the positive and negative rail. When I had checked the MOSFETs, there was also a discrepancy in Vgs values between the N and P channel ones. In the current configuration, when I adjust Rbias so that I have a current of 0.7 A through the N channel, the current from P channel is 0.95 A. So my question is the following: is there a problem with this? Also, a strange finding is that, when checked on the bench with a current feed of 0.7 A, Vgs was 1.4 V for the N channel and 1.9 V for the P channel. On the actual circuit, those values are 1.14 and 1.59 V, respectively.
2. I made a full power output measurement and found out that max power output (at the beginning of clipping on the oscilloscope) is just 8 W. Is that normal?
Looking forward for further enlightenment,
Best regards,
Evangelos
Set Iq to some low value (100-200mA) then use P1 to set the DC offset at output to 0V and after that increase Rbias to Iq you want. DC offset at output must be set to 0V.
When you measure power you have to state the load and precise voltage - for example : rail voltage is +-20V, the load is 8R and output voltage is 15Vpeak or 11Vrms or 30Vpeak-to-peak...
When you measure power you have to state the load and precise voltage - for example : rail voltage is +-20V, the load is 8R and output voltage is 15Vpeak or 11Vrms or 30Vpeak-to-peak...
Rail voltage is +-17 V, Rload is 8 Ohms, Vrms is 8.2 V.
I will use the technique you mention and report back tomorrow. In my previous configuration, I had already set output offset to 0 volts, but after setting Iq to 700 mA (+rail, -rail draws 950 mA). But I will try starting from a lower Iq first.
if there is 0V at the output the current draw from both rails must be the same. If it's not the case there's an error somewhere.
As for output power, with +/-16 V rail voltage I got: "Full power before clipping is 34W_peak at 4R load with 1.3V_peak input signal and about 19W_peak at 8R load with 1.4V_peak at input" - it is close to what you measured, differences are probably due to rail voltage sag under max load and/or lower gm i.e. higher Rs of Exicons vs. Renesas. Anyway, a watt or two more or less is unimportant..
As for output power, with +/-16 V rail voltage I got: "Full power before clipping is 34W_peak at 4R load with 1.3V_peak input signal and about 19W_peak at 8R load with 1.4V_peak at input" - it is close to what you measured, differences are probably due to rail voltage sag under max load and/or lower gm i.e. higher Rs of Exicons vs. Renesas. Anyway, a watt or two more or less is unimportant..
Hi Juma,
I applied the proposed method, lowering down Iq to 100 mA, adjusting output voltage offset to exactly 0.0 mV and, then, increasing the Iq to 700 mA. Still the same finding here: when positive rail current is set to 700 mA, negative rail draws 930 mA.
So, where should I search to spot a possible error? Could this discrepancy be due to the large difference of Vgs between the two laterals?
I applied the proposed method, lowering down Iq to 100 mA, adjusting output voltage offset to exactly 0.0 mV and, then, increasing the Iq to 700 mA. Still the same finding here: when positive rail current is set to 700 mA, negative rail draws 930 mA.
So, where should I search to spot a possible error? Could this discrepancy be due to the large difference of Vgs between the two laterals?
Sorry Vageli,
but that is not possible.
When there is 0V at the output the current draw has to be equal through both MOSFETs because they are connected in series.
It's the law, there's no two ways about it.
You are obviously doing/measuring something in a wrong way - keep looking until you find it...
but that is not possible.
When there is 0V at the output the current draw has to be equal through both MOSFETs because they are connected in series.
It's the law, there's no two ways about it.
You are obviously doing/measuring something in a wrong way - keep looking until you find it...
Sorry Vageli,
but that is not possible.
When there is 0V at the output the current draw has to be equal through both MOSFETs because they are connected in series.
It's the law, there's no two ways about it.
You are obviously doing/measuring something in a wrong way - keep looking until you find it...
Juma,
Your words lit up a light-bulb in my head: Of course! How could I miss that? When voltage at the output node is 0 volts, the two currents flowing to that node should be equal and in reverse direction.
So, I immediately figured out and found out what was going wrong. The bench PS I have uses just one panel meter to give readings both for voltage and current both for the two rails, via a 4 position rotary switch. So, I never used any other means to read the current draw except this on-board meter. It seems that the negative rail current adjustment has gone out and I was getting a wrong reading.
Things cleared out immediately, soon as I put a digital multi-meter in series on each rail and, as a miracle, current draw through each rail was absolutely the same!
It seems that, not so rarely, our mind sticks on a certain spot and cannot think any further...
So, my next move is to connect it to a loudspeaker and do my initial acoustic testing (in mono). As a next step, I intend to build the other channel with 0 feedback and do a comparison with the feedback version. My loudspeakers are 2-way horns and I have found that they always benefit from a no-feedback amp and low damping factor. But, I will report later regarding this.
Just another thought in relation to the power output. What puzzles me is that in the first Cubie presantation, you write that:
"The oscillograms show 24 Vpeak-to-peak output (the clipping starts at about 26Vpeak-to-peak)",
which is ca. 8.5 Vrms, very close to my findings.
But, in the Cubie2 presentation, you site the following:
"Full power before clipping is 34W_peak at 4R load with 1.3V_peak input signal and about 19W_peak at 8R load with 1.4V_peak at input."
I have the suspicion that you misspelled W for V. I don't think it would be ever possible that with a power supply of +/- 16 V and a biasing current of 0.7 A this amp could give an output of 19 Watts at a 8 Ohms load. Am I right? In that case, do you mean 19 V peak or peak-to peak?
OK, I think I said a lot so far, so here I stop.
Thank you very much, once again.
Best regards,
Vagelis
OK, current draw panic is solved, now let's see about the power..
If you read carefully, I wrote "19W_peak at 8R load" and 19W_peak we get from 12,3V_peak (that's about 8.69Vrms or 24.6Vpeak-to-peak) at 8R load :
P=V*V/R = 12.3V * 12.3V / 8R = 151.29 / 8 = 18.9 W_peak - from V_peak you get W_peak, ok ?
I use peak value rather than avg or RMS because it shows real voltage/power excursion limit (where it hits the bottom/ceiling), not a derivative.
So everything is ok, you just need to read more carefully what I wrote...
If you read carefully, I wrote "19W_peak at 8R load" and 19W_peak we get from 12,3V_peak (that's about 8.69Vrms or 24.6Vpeak-to-peak) at 8R load :
P=V*V/R = 12.3V * 12.3V / 8R = 151.29 / 8 = 18.9 W_peak - from V_peak you get W_peak, ok ?
I use peak value rather than avg or RMS because it shows real voltage/power excursion limit (where it hits the bottom/ceiling), not a derivative.
So everything is ok, you just need to read more carefully what I wrote...
Well, after finishing and checking meticulously on the bench one channel, I dedicated this morning to a long listening test. My initial finding is that this little amp sound fabulous! Voices are just magical, without any sign of sibilance, clear and warm. Bass is extensive, yet firm. Mids are excellent. Highs are crystal clear. The speed of the amp is super high.
In comparison, I was listening back and forth to my "reference" amplifier, a 300B SE (built in monoblocks). Well, honestly, I could not tell a distinctive difference between the two. Maybe, the tube amp had a slightly better separation when multiple instruments were playing, but I cannot say for sure regarding this. If there is a minor difference, it's so minuscule that is not distinctive.
Also, I must note that the amp was still supplied from my bench power supply. I believe that its sound will be much better with its own PS.
One thing that I also tried was to raise Iq to ca. 900mA, to check if I hear any difference. To be honest, I didn't hear any difference. So, since I have plenty of heatsinking (after 1,5 hours, the temperature of the heatsink had raised 16 degrees Celsius about ambient just next to the MOSFETs), I will leave it to the higher bias current.
One thing that I have noticed and which worries me is something I had seen since I was checking the Cubie2 on the bench: when I switched on the signal generator (which was already connected to the amp), I saw the DC voltage in the output going up to the full B+ and, in a couple of seconds, going back to the set offset voltage. I thought that this was a problem of the generator, whose output is DC coupled. Having this in mind, I connected it to my system without connecting the loudspeaker to the output of Cubie2, I switched on the amp and, then I turned on the preamp. Here I have to note that my tube preamp has an output capacitor and has a turn-on delay of ca. 30 seconds, with its output shorted to ground until the heaters are OK. What I saw was exactly the same: DC voltage in the output going up to the full B+ and going back to the set offset voltage within a few secs. What can I do regarding this? Of course, I could install an output protection circuit, but, since I had tried this previously on another SS amp, I noticed that it degraded the sound of the amp, since this circuit is in parallel with the output.
So, my next move is to build the other channel with zero feedback and do a comparison listening test between the feedback and non-feedback version to decide which one suits me best.
In comparison, I was listening back and forth to my "reference" amplifier, a 300B SE (built in monoblocks). Well, honestly, I could not tell a distinctive difference between the two. Maybe, the tube amp had a slightly better separation when multiple instruments were playing, but I cannot say for sure regarding this. If there is a minor difference, it's so minuscule that is not distinctive.
Also, I must note that the amp was still supplied from my bench power supply. I believe that its sound will be much better with its own PS.
One thing that I also tried was to raise Iq to ca. 900mA, to check if I hear any difference. To be honest, I didn't hear any difference. So, since I have plenty of heatsinking (after 1,5 hours, the temperature of the heatsink had raised 16 degrees Celsius about ambient just next to the MOSFETs), I will leave it to the higher bias current.
One thing that I have noticed and which worries me is something I had seen since I was checking the Cubie2 on the bench: when I switched on the signal generator (which was already connected to the amp), I saw the DC voltage in the output going up to the full B+ and, in a couple of seconds, going back to the set offset voltage. I thought that this was a problem of the generator, whose output is DC coupled. Having this in mind, I connected it to my system without connecting the loudspeaker to the output of Cubie2, I switched on the amp and, then I turned on the preamp. Here I have to note that my tube preamp has an output capacitor and has a turn-on delay of ca. 30 seconds, with its output shorted to ground until the heaters are OK. What I saw was exactly the same: DC voltage in the output going up to the full B+ and going back to the set offset voltage within a few secs. What can I do regarding this? Of course, I could install an output protection circuit, but, since I had tried this previously on another SS amp, I noticed that it degraded the sound of the amp, since this circuit is in parallel with the output.
So, my next move is to build the other channel with zero feedback and do a comparison listening test between the feedback and non-feedback version to decide which one suits me best.