I thing you are too spoiled 
Use whatever you have at hand. If you have Jfet with too high Idss, then use resistor to limit current. Look at Figure 15c at this LINK
With 47R your current will be 6.x mA. This is only for using 2SK170. For other Jfets you should look at their datasheets
Use whatever you have at hand. If you have Jfet with too high Idss, then use resistor to limit current. Look at Figure 15c at this LINK
With 47R your current will be 6.x mA. This is only for using 2SK170. For other Jfets you should look at their datasheets
Last edited:
I also emailed them about the idss measurements I discovered & they said they would send me 40 SK1058 instead of the 40 SK170s I received but that was back on 29th Sep and maybe they realised their mistake?
Ok slightly off topic.
My plan for this regulator was to run the front end of all my amps with it, while keeping passive filtering on the output.
However for some reason when I simulate the effect of pure dc at the front end of my amps with 100mV pk-pk ripple on the output stage. The end result is worse then just having 100mV pk-pk on all gain stages.
Can anyone explain this to me?
My plan for this regulator was to run the front end of all my amps with it, while keeping passive filtering on the output.
However for some reason when I simulate the effect of pure dc at the front end of my amps with 100mV pk-pk ripple on the output stage. The end result is worse then just having 100mV pk-pk on all gain stages.
Can anyone explain this to me?
Last edited:
Hi, honestly, getting 10 working ones out of 18 is not bad. It is about 20c each and I have no complaint. I do hope that the 240 and 9240 are good.
Put the moral thing aside, I do not think it makes sense for people to make fake products for the really cheap parts and if those cheap parts do not work, because the profit would be so low and nobody would buy from them the second time. So I am not worried about something like 2SK170. May be the products are slightly inferior but I think they must be geniune. Say if I buy something like OPA627, then I would be very careful.
Put the moral thing aside, I do not think it makes sense for people to make fake products for the really cheap parts and if those cheap parts do not work, because the profit would be so low and nobody would buy from them the second time. So I am not worried about something like 2SK170. May be the products are slightly inferior but I think they must be geniune. Say if I buy something like OPA627, then I would be very careful.
NutNut:
Can they be cheap because they got a wrongly mixed and stamped batch from the factory in some special price? Don't know, I never had such oddity with BL.
P.S. If you do the emitter follower bit at a point, remember that the emitter should go directly on the gate stopper of the Mosfet and the stopper directly on the gate pin. It does not tolerate inductance between. I saw it in practice 2 times. Some travel to its base and to the emitter load is no problem.
Can they be cheap because they got a wrongly mixed and stamped batch from the factory in some special price? Don't know, I never had such oddity with BL.
P.S. If you do the emitter follower bit at a point, remember that the emitter should go directly on the gate stopper of the Mosfet and the stopper directly on the gate pin. It does not tolerate inductance between. I saw it in practice 2 times. Some travel to its base and to the emitter load is no problem.
Good advice!
I have redone my PCB layout and placed it as close as possible. See the picture below.
The new layout also includes the buffer mod, with a BC560c on the negative side, the resistor mod, and used a 1.1uF MKP (Vishay blue box) in parallel with a 100uF Rubycon ZL to bypass the Vref resistor. Output capacitor is for the time being 100uF Rubycon ZL.
Note that it is not inteded for a printed circuit board. It is used only for wiring point to point using perfboard, therefore, the track width, etc, are only indications. I use single solid wire stripped off from household electrical wire (rated 10A) for the thicker "track", and 22aug (0.8mm) tinted copper wire for the thin "track". Since the component symbols are not in correct sizes so it looks crowded but the components fit quite well, abeit compact in certain places. I could improve on it by placing the components around the ground wires a bit further away to reduce parasitic capacitance.
An externally hosted image should be here but it was not working when we last tested it.
I went to buy all the remaining parts on Saturday. On Sunday I built it, i.e. Salas v1. I got distracted during mid day so I only finished at 10pm last night. It was too late to turn on the misic.
I used a 2.2k 2W metal film resistor in series with a 2k multiturn trimpot as the Vref resistor. The trimpot was set to 0R. Input voltage was 21.5VDC.
Output voltage: + 12VDC, -12.91 VDC. The heatsinks ran only touch warm and were not hot after 20 minutes.
I turned it on and off for a couple of times without problems.
After 20 minutes, I noticed that the negative rail voltage was dropped to -12.56VDC. The positive was not changed at all. This test was repeated twice.
I used a 2.2k 2W metal film resistor in series with a 2k multiturn trimpot as the Vref resistor. The trimpot was set to 0R. Input voltage was 21.5VDC.
Output voltage: + 12VDC, -12.91 VDC. The heatsinks ran only touch warm and were not hot after 20 minutes.
I turned it on and off for a couple of times without problems.
After 20 minutes, I noticed that the negative rail voltage was dropped to -12.56VDC. The positive was not changed at all. This test was repeated twice.
Looks nice, I only have p2p expereience with the buffer 550c C pin directly on the S pin for V+ and E pin directly on the G pin's gate stopper. No negative side at all as you know. I have seen it oscillating with just a bit of longer distance between emitter and gate stopper. If it will not work in your layout, you may revert to slower transistor like BC182 or BD139 and their PNP equivalents. You understand this is something almost new you do here paractically. Especially the neg side buf experiment. Good luck.
Better use fixed resistors after finding the ballpark with the trimmer. Trimmers aren't always very steady. OK if single sided and settles somewhere after some minutes, but symmetrically they can be unlocking for matching. When set to 0, then it must be the ccs NJFETs Tc.
This morning, I connected a 0.1R resistor in between the RAW supply and the CCS-shunt. I measured around 100mA on the postive side and 110mA on the negative side.
For the 2SK170 pairs, I did exactly as your picks in your previous posts. I did not match the LEDs. Bearing in mind the 0.1R wirewound resistor has 5% tolerence.
The output voltage for the postive rail was dead solid at 12VDC with no changes at all. The one on the negative rail started at -12.91VDC.
My daughter had some virus infection so I spent the 6 hours with her having a check up in a hospital. It was good that the result was that she only had an infection.
In the afternoon, I took out the 0.1R resistors and gave the regulator a listen. It was driving a opa627 buffer with nothing else.
After turned on, the positive rail showed +12VDC again. The negative rail showed -12.68VDC. After playing music for about an hour, before I turned it off, it measured -12.21VDC.
I am wondering if it is the 2SK170 playing up. These 2SK170 were ordered from Futurlec and the values vary greatly.
I did not try to replace it with another one due to various reasons.
I did not try to increase the trimpot to match the pos to the neg, because, as the negative rail voltage is changing, there is really no point.
For the 2SK170 pairs, I did exactly as your picks in your previous posts. I did not match the LEDs. Bearing in mind the 0.1R wirewound resistor has 5% tolerence.
The output voltage for the postive rail was dead solid at 12VDC with no changes at all. The one on the negative rail started at -12.91VDC.
My daughter had some virus infection so I spent the 6 hours with her having a check up in a hospital. It was good that the result was that she only had an infection.
In the afternoon, I took out the 0.1R resistors and gave the regulator a listen. It was driving a opa627 buffer with nothing else.
After turned on, the positive rail showed +12VDC again. The negative rail showed -12.68VDC. After playing music for about an hour, before I turned it off, it measured -12.21VDC.
I am wondering if it is the 2SK170 playing up. These 2SK170 were ordered from Futurlec and the values vary greatly.
I did not try to replace it with another one due to various reasons.
I did not try to increase the trimpot to match the pos to the neg, because, as the negative rail voltage is changing, there is really no point.
Change the one in the neg side with a somewhat lower idss than the one in the pos side. And see. 2 things can drift. NJFET or resistor. Haven' had such, but then again we don't know much about the particular njfet. Is it made with the buffer mod or not? The 10mA main ccs difference is due to led and Mosfet Vgs tolerances. No big deal, no problem.
P.S. The slide does not mean bad regulation, just thermal shift over long.
P.S. The slide does not mean bad regulation, just thermal shift over long.
Now the listening impression:
It sounds very good! It betters everything I tried.
I tried the LM317/337. I also tried the John Linsley Hood's C-multiplier / Ripple eater behind that. I tried the ALW Jung Supereg (my implementation), and I tried another CCS-shunt reg kit, the schematic of which may be copied from one of the earlier Mark Levinson regulators.
The Salas v1 betters them all.
It sounds very good! It betters everything I tried.
I tried the LM317/337. I also tried the John Linsley Hood's C-multiplier / Ripple eater behind that. I tried the ALW Jung Supereg (my implementation), and I tried another CCS-shunt reg kit, the schematic of which may be copied from one of the earlier Mark Levinson regulators.
The Salas v1 betters them all.
Last edited:
I guess I will wait until I build v2, or at least the 2SK170 arrival from Spencer and use those parts instead.
It does not make sense to me that the pos rail voltage is dead solid while the neg reduced half a volt every time after playing for 30 minutes. I guess it is a bad part. I have never played it for more than one hour, what if it continues to drop?
The circuit is as in post 1349.
Enthusiastic? I am actually desperate!!!
I spent the past 5 years "day and night" trying to get a world class system. It is 90% done. The line level regulator has been the last remaining obstacle. I really wish to complete it, then build another pair of OB for the sake of returning something back to the Multi-way forum, then say Bye-Bye and wrap up the hobby and enjoy music for the rest of my life.
I have designed my own pair of 4 way open baffle active loudspeakers with 93dB sensitivity, 120dB SPL capability, with almost the best drivers I could find. It has great potential and I have heard many aspects of it. I found changing a capacitor in the line level regulator changed so much in sound, possibly altering frequency response as well. So I did not finalise the active circuit for the loudspeakers. I have not had good sound in my HiFi room for more than 6 months now. I want to find the ultimate regulator first, and use that regulator when I do driver / loudspeaker measurements. I can only trust the measurement results that way, as the regulator will sit there to drive the active circuit.
Is the negative Vref ccs njfet nearer to some heat source that alters its equilibrium? Those parts change idss even by touching. I have seen single sided regs with resistors settling in a minute or two a bit lower. Also see what happens if you physically jump the negative trimmer with a wire. Maybe its wiper plays a little beyond 0 you set it.
So both buffers worked. I hope it is not some negative side periodic instability that alters the final -Vout. Also utilize the trimmers, set it at say +/-15V after it is warmed up, say 10mins, and see if it stays locked.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.