The simplistic Salas low voltage shunt regulator

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The voltage drop is too gradual for it to drop below usable levels anytime soon. If it continues at this rate, it will take 11 more hours before it drops below 4.8V. Good enough for me to use it as is. :)

Though I'm feeling the itch to tweak both while I'm waiting for the Reflektor MOSFET and PNPs to arrive. Right now, am using 47uF at C107, want to try switching that for a resistor + MKT cap zobel.

Any tips on how to optimize the zobel, given that I have access to a scope to take an inside look at things?

Edit : it seems to have finally stabilized between 5.173V and 5.174V. It's staying at that range even after 10 mins or so
 
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
It will stop at a point. A green led, a red led, even different brands, put different tc in the mix. A diode has to contribute ~0.6V if it is silicone normal junction like 1N4148.
The ref filter & terminal caps can have an influence too. Use low esr low leakage cap for the ref cap, normal esr for the 47u terminal cap. Use the 5.6uF film + 1R values for the Zobel. Look for no sinewave oscillation on the power rail.
 
Changed earlier the Vref cap from a 330uF Silmic II to a 1000uF Panasonic FM cap. Details seemed to be more in focus, bass more taut and with more impact. Subjectively very much preferred to previous config.

Also switched to a 1R + 4.7uF zobel for one supply. Subjectively louder than previous config. Lost a lot of the magic gained from the Vref switch. Sound seems to be more "confused", or at least it's less coherent in acoustic sections. I'm not sure if this is just a case where I need to give time for cap break in, or if it's with the quality of the cap itself ( cheapo Axon True cap ).

On the measuring side, things didn't look that much different from using lytic to switching to zobel. I may just be measuring it wrong though. Had measurements AC coupled, then went for the lowest time / div my scope could manage, with vertical scale just enough that I could see the entire waveform on my screen.

Here's what that looked like before switching to zobel :
An externally hosted image should be here but it was not working when we last tested it.


And here it is after zobel :
An externally hosted image should be here but it was not working when we last tested it.


Based on measurement, should I switch back to the lytic, or is the zobel stable enough?
 
Last edited:
Went back to the lytic, and added a 47uF lytic at the feed point. Also did the same to the feed point of the 5V supply. Subjective memory is a bit fuzzy, since there's a couple of hours of sleep in between tryout, but I think that the tightness of the bass I first reported after changing the Vref cap still isn't back.

On hindsight, shouldn't have added the 47uF at the feed point of the other supply, as it introduces another variation since the previous attempt.

Below is a capture of the AC waveform. Again, another point of variation from previous is that I used the probe bandwidth limiting option (tho am unsure as to what frequency it actually does the limiting).

An externally hosted image should be here but it was not working when we last tested it.

image sharing

Later, I'll get a capture of the waveform at the other supply.
 
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
Try another thing, use the extra 47uF caps across the sense only outputs at the regs connectors instead. If they are not still improving leave them out. The DS1052E has a digital filter menu too. You could filter out 1MHZ up with an LPF since we know we don't have any repeated waveform (oscillation) now, so to avoid capturing wide band RFI noise in the interface better.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Move to sense lines. Don't keep on feed points.

Wideband noise is a nuisance, difficult to know what is what when measuring with scope on a rail, especially when its digital. Look here EEVblog #441 - How To Track Down Common Mode Noise - YouTube

For own regulator noise we are OK with the SSLV1.1 look at the Linear Audio test fig 7 (middle of page links) with just 220uF Vref cap and ~100 mA CCS, a conservative example. We are just looking to avoid extra things.
http://www.linearaudio.net/index.php?option=com_content&view=category&layout=blog&id=39&Itemid=64
 
diyAudio Chief Moderator
Joined 2002
Paid Member
In a pretty typical way like when stacking batteries. You could pre-test on two dummy load resistors in series between their outer pins and middle point. That is when you want a floating symmetric out of two positives. If you just want a floating 18V positive you make a single but don't associate its cold side with any ground in the client or chassis.
 
Latest version

Hi,

I read from the first post to page 280 of this huge thread, taking notes as I read, and it was up to version 1.2 then. I reckon it's probably more efficient if I could start from the post where the latest version of Salas's shunt regulator design was published.

Could someone point me to the post/page where the latest version was introduced.

Thanks in advance.
 
Last edited:
Phew. It's been a while since I last reported back in this thread. Work and stuff got in the way.

Anyway, I tried out Salas' last suggestion, moving the 47uF to the sense lines. That didn't change anything. Funny thing was tho, when I removed the 47uF from the sense lines, it still didn't change anything : the sound I was hearing was still sub par compared to when I first auditioned the system after improving the Vref cap to 1000uF FMs.

I probably messed something up when I started tinkering around with the zobel caps... A couple of suspects
* re-soldering at the DAC feed point (5.2V, 200ma) resulted in the output wires being not as twisted as before, there's a more than 1 inch loop that's not twisted
* I now get a funny reading when I check for a short between the + and GND points for the XMOS feed point (5V, 400ma). Black lead on GND, red lead on the +, DMM shows no continuity. However, black lead on +, red lead on GND there's a reading of 200k-300k ohms or so. The other feed point doesn't have this. Any idea how to resolve this?

After I fix those problems, I'll try out the 47uF on both feed point and sense lines again, as the earlier experiments seem to be invalid.
 
Question : how much of what I've tried out with the Zobel arrangement for the BiBs applies to the Reflektor?

The dual transistors I originally planned for are still a while out since there was a problem with the order and then there's international shipping, but I already have the pass MOSFET, and I have 48 pcs left over of the BC560C, so I can already include that for my weekend plans. :D
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.