In the schematic you gave me, you modelled the load to be 350mA and the CCS 650mA. Why does the CCS nearly double the load? If the load is worked out to be absolutely no more than 350mA, could I set CCS 350mA, or 400mA? Is there a technical reason behind that (rather than lowering the output impedance) or just for a safeguard ?
With regards to lowering the output impedance, you mentioned and I understand that the higher the shunt current flows, the lower the output impedance becomes. I increased the current from 100mA to 150mA in v1 and found I liked the sound better. From your modelling, how much the impedance decreases versus current increases? In page 1, does the impedance graph correspond to the schematic there which runs 200mA current?
I asked the questions because I am working out the heatsink requirement for my application. Unless the regulator uses power amplifier like heatsinks, the best individual smaller size (still quite bulky) heatsinks would be rated at around 7 degree W. Adding the insulation pad makes it 9 degree W. To dissipate a moderate 7 W (650mA) per heatsink would add 63 degrees, and plus room temporature in Summer (my home can get 33 degrees many days, not talking about extream weather of 40 degrees happened last year) the temporature can approach 100 degrees. I would prefer to run the MOSFETs at less than 70 degrees. It is for the long life of the device and more importantly, safety.
Edit: I read page 1 where you mentioned that the shunt current may be set to 10-20% of the load current. I guess that is good for class A amplifier where the load is constant current. So the CCS would be set to 110%-130% of the required constant current. However, in my application of running opamps in class B, current demand varies. But I guess it would never be more than 200mA-250mA, as I use fairly high impedance load.
With regards to lowering the output impedance, you mentioned and I understand that the higher the shunt current flows, the lower the output impedance becomes. I increased the current from 100mA to 150mA in v1 and found I liked the sound better. From your modelling, how much the impedance decreases versus current increases? In page 1, does the impedance graph correspond to the schematic there which runs 200mA current?
I asked the questions because I am working out the heatsink requirement for my application. Unless the regulator uses power amplifier like heatsinks, the best individual smaller size (still quite bulky) heatsinks would be rated at around 7 degree W. Adding the insulation pad makes it 9 degree W. To dissipate a moderate 7 W (650mA) per heatsink would add 63 degrees, and plus room temporature in Summer (my home can get 33 degrees many days, not talking about extream weather of 40 degrees happened last year) the temporature can approach 100 degrees. I would prefer to run the MOSFETs at less than 70 degrees. It is for the long life of the device and more importantly, safety.
Edit: I read page 1 where you mentioned that the shunt current may be set to 10-20% of the load current. I guess that is good for class A amplifier where the load is constant current. So the CCS would be set to 110%-130% of the required constant current. However, in my application of running opamps in class B, current demand varies. But I guess it would never be more than 200mA-250mA, as I use fairly high impedance load.
Last edited:
It just happened to be an image of the circuit after I was simulating that load, no particular significance. Often I push the circuit to extremes to see how it behaves.
Hard to quantify because Zout is a function of frequency and it doesn't lower equally everywhere. Suffice to say that increasing the shunt current by 50mA doesn't lower Zout much.
I will refrain from giving exact rules because I don't think they will apply. Best is to make sure the shunt works well first at 10-20% of the load; when you're reasonably sure this is the case, then increase the current and voice the device. Before doing this in the real audio chain, it would be good to have done it with a passive load and prepare several resistors knowing what shunt current they'll give. This is how I would do it. Remember that all the generalities that people talk about with class A and B cases are that, just generalities. Those rules do apply, in general. For your specific case I think it's best that you adjust the shunt in real life for the case at hand. IMHO, if you put so much effort already into your first class system, then it's not too much to do the same with a bit of tweaking on the regulator. You can also bypass all this and set the shunt current much higher than you think you need it, perhaps to 1.5A and call it a day
Just make sure you get adequate heat sinks.
I experimented with the irf3205 before the irfbc40 and managed to fry a few of them when taking them to some extreme currents. They have higher input capacitance, which is not so desirable. Still, I don't think that you'd have problems getting them to work. At the moment my prototype v2 uses irfbc40 and works fine with them.
Having not had much time in the past a few days so it has been a bit slow.
I completed the build a couple of hours ago and have not heard it yet. No Bang, no smoke.
Mine is based on the earlier schematic in which all current sources were 2sk170 with the rest identical. This time the 2sk170 came from Spencer's batch.
For the Vref I had 27k resistors and 2k trimpot between gate and source of the 2sk170. I found this was good for only 17V plus unless I increase the trimpot to 5k. So I reduced the resistors to 20k and kept the 2k trimpot and the voltage available for adjustment started at 12V (2k on trimpot), which was good for me. Lowering the trimpot increases the voltage.
I have turned it on and off a couple of times.
The initial CCS current was set to very low with a 68k Vref resistor and a 2k trimpot set to about 1.2k in the CCS. The current was not measurable and the DMM showed 0mA. However, the voltage was regulated.
When I increased the current to 130mA a few minutes later the shunt heatsinks got very hot. I could not touch it for more than a fraction of a second. I guess it was hot enough to boil an egg. So I turned it off immediately. The heatsinks are rated at 7 degree W. For at most 2W power it got so hot therefore I thought it was definitely not right.
When cooled down I turned it back on and reduced the CCS current to 10mA. Then started to increase it. The CCS current drifted a lot. Eventually settled on above 100mA after 30mins. However, the heatsinks did not get hot this time, only warm.
Turned it on for the third time and had it run for 15mins now. Found the voltage to be very stable since turned on. It did not vary much at all. Maybe only 0.02V at most. The CCS started at 85mA only, and it is now at 93mA. Will it gradually increase to over 100mA? Time will tell.
So other than some shifts in CCS current things seem to be fine. I will turn it off and on one more time then connect it to the circuit and have a listen. It is promising as long as this thing does not oscillate.
Salas mentioned that he needed to solder to gate resistor right at the transistors to prevent oscillation in v1.1. In my PCB design I tried to do so. But the heatsinks are T03 type so the MOSFET legs are quite long. I have not soldered the resistor on the upper part of the leg so there can be a bit of inductance there. I will see if it oscillates. I don't have a scope so I can only judge by ear which is nearly impossible unless it gets the hash down to the audioband. Ikoflexer, I have seen some of your pictures previously and it seems that you soldered them point to point so you may not get oscillation. For me, it is a different story. Fingers crossed.
I suspect that the 2sk170 is not very stable with temporature.
I completed the build a couple of hours ago and have not heard it yet. No Bang, no smoke.
Mine is based on the earlier schematic in which all current sources were 2sk170 with the rest identical. This time the 2sk170 came from Spencer's batch.
For the Vref I had 27k resistors and 2k trimpot between gate and source of the 2sk170. I found this was good for only 17V plus unless I increase the trimpot to 5k. So I reduced the resistors to 20k and kept the 2k trimpot and the voltage available for adjustment started at 12V (2k on trimpot), which was good for me. Lowering the trimpot increases the voltage.
I have turned it on and off a couple of times.
The initial CCS current was set to very low with a 68k Vref resistor and a 2k trimpot set to about 1.2k in the CCS. The current was not measurable and the DMM showed 0mA. However, the voltage was regulated.
When I increased the current to 130mA a few minutes later the shunt heatsinks got very hot. I could not touch it for more than a fraction of a second. I guess it was hot enough to boil an egg. So I turned it off immediately. The heatsinks are rated at 7 degree W. For at most 2W power it got so hot therefore I thought it was definitely not right.
When cooled down I turned it back on and reduced the CCS current to 10mA. Then started to increase it. The CCS current drifted a lot. Eventually settled on above 100mA after 30mins. However, the heatsinks did not get hot this time, only warm.
Turned it on for the third time and had it run for 15mins now. Found the voltage to be very stable since turned on. It did not vary much at all. Maybe only 0.02V at most. The CCS started at 85mA only, and it is now at 93mA. Will it gradually increase to over 100mA? Time will tell.
So other than some shifts in CCS current things seem to be fine. I will turn it off and on one more time then connect it to the circuit and have a listen. It is promising as long as this thing does not oscillate.
Salas mentioned that he needed to solder to gate resistor right at the transistors to prevent oscillation in v1.1. In my PCB design I tried to do so. But the heatsinks are T03 type so the MOSFET legs are quite long. I have not soldered the resistor on the upper part of the leg so there can be a bit of inductance there. I will see if it oscillates. I don't have a scope so I can only judge by ear which is nearly impossible unless it gets the hash down to the audioband. Ikoflexer, I have seen some of your pictures previously and it seems that you soldered them point to point so you may not get oscillation. For me, it is a different story. Fingers crossed.
I suspect that the 2sk170 is not very stable with temporature.
Last edited:
Some minutes passed now. The CCS current did not increase but dropped to 102mA and 100mA.
By the way, in my PCB design the components are placed as far away from the hestsinks as possible. Since the heatsinks are only warm, they should not affect the temporature of the components. I still think the the 2sk170 is not very stable with temporature.
By the way, in my PCB design the components are placed as far away from the hestsinks as possible. Since the heatsinks are only warm, they should not affect the temporature of the components. I still think the the 2sk170 is not very stable with temporature.
I have now given it an hour listening. The voltage is stable. The heatsinks are just warm.
Sounds good. I have built 4 regulators in the past couple of months. They all sounded a bit different. Over all everyone I built was better than its previous one. Is there somebody who is going to publish a new regulator soon?
Sounds good. I have built 4 regulators in the past couple of months. They all sounded a bit different. Over all everyone I built was better than its previous one. Is there somebody who is going to publish a new regulator soon?
Good job! 
Indeed, as you noticed, the jfets are very sensitive to temperature variations; well known fact! Your circuit will probably boxed at some point, so the temperature will be more constant after warm up. Remember that if you have adequate heat sinks (read large) you can increase the current. My and others' experiments show correlation of higher current with better sound (lower output impedance). Of course this should be done with care, to avoid oscillation; I mean, at some point the circuit layout matters.
Again, well done and thanks for the report! When you got a chance, remember, we like pictures It inspires us!
Indeed, as you noticed, the jfets are very sensitive to temperature variations; well known fact! Your circuit will probably boxed at some point, so the temperature will be more constant after warm up. Remember that if you have adequate heat sinks (read large) you can increase the current. My and others' experiments show correlation of higher current with better sound (lower output impedance). Of course this should be done with care, to avoid oscillation; I mean, at some point the circuit layout matters.
Again, well done and thanks for the report! When you got a chance, remember, we like pictures
It inspires us!- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.