I am using Salas shunts to power a TDA1541 (+5, -5, -15V), as well as a Teralink X2 USB-I2S converter.
Looking at the impedance graphs earlier in the thread, I see that impedance of the Salas Shunt increases exponentially from about 200KHz. This is fine for the analog power supply of a dac (with noise in the audio band), but what are the effects on the digital power supply (with a clock speed of 2.8MHz, and possibly inherent noise) or the USB converter (with a USB clock of 12MHz, and possibly inherent noise) ?
May a Salas shunt not be the optimal regulator for a Dac/Digital? Not sure, please enlighten me...
Thanks!
Looking at the impedance graphs earlier in the thread, I see that impedance of the Salas Shunt increases exponentially from about 200KHz. This is fine for the analog power supply of a dac (with noise in the audio band), but what are the effects on the digital power supply (with a clock speed of 2.8MHz, and possibly inherent noise) or the USB converter (with a USB clock of 12MHz, and possibly inherent noise) ?
May a Salas shunt not be the optimal regulator for a Dac/Digital? Not sure, please enlighten me...
Thanks!
Hi Hesener, thanks for the details. As I wrote in the MPP thread, you are lucky to have good PSRR in the main circuit, else the output noise without filtering would intrude. So you are in a position not to slow it at all with any capacitor, and that's good in your application. Keep in mind that when using active elements, like LEDS, or precision Vsources like LM329s, the impedance is far lower than resistors, hence far tighter CCS regulated and faster, slowing down less with a cap. The Zo you measured followed the simulator for value. I don't know if the simulator's noise analysis would also follow for absolute values or shape, but see the following two graphs comparatively. First is output noise when using a 2k2 Vref element (about what you use) bypassed by 100uF. Second is with zero bypass capacitor. Many are happy with a good film cap up to 10uF. In DCB1 Blue we use LEDS and a small 0.22uF MKP, because gain=1 and line level made us able to prioritize Ref regulation. So the concept stays but is about configuring best for each application.
Attachments
Hi Salas,
ok thanks for the detailed explanation, I understand much better now. And, guess what, I will try and put caps across the reference now, that is an easy fix to do.
In fact, what I didnt mention is that the (external) power supply that I use consists of a two-stage EMI filter, toroid, bridge rectifiers and caps, and then cap multipliers with 1.5V drop, so the stuff that arrives at the shunt reg input is (almost) DC already at a rather low impedance. The cable to the phono stage is shielded. So there is not much input noise filtering required from the shunt.
At the same time, with the load being so benign and using pretty much only DC, the shunt regulator is not really asked to do a lot, but can focus on the microdetail-relevant stuff I guess. So, highest gain around the given operating point is good. Not so sure I achieved that goal to a satisfying degree, in fact your V1.2R looks a lot better in that respect.
Did you ever consider using an opamp in the shunt regulator, e.g. like Allen Wright'S SuperReg?
ok thanks for the detailed explanation, I understand much better now. And, guess what, I will try and put caps across the reference now, that is an easy fix to do.
In fact, what I didnt mention is that the (external) power supply that I use consists of a two-stage EMI filter, toroid, bridge rectifiers and caps, and then cap multipliers with 1.5V drop, so the stuff that arrives at the shunt reg input is (almost) DC already at a rather low impedance. The cable to the phono stage is shielded. So there is not much input noise filtering required from the shunt.
At the same time, with the load being so benign and using pretty much only DC, the shunt regulator is not really asked to do a lot, but can focus on the microdetail-relevant stuff I guess. So, highest gain around the given operating point is good. Not so sure I achieved that goal to a satisfying degree, in fact your V1.2R looks a lot better in that respect.
Did you ever consider using an opamp in the shunt regulator, e.g. like Allen Wright'S SuperReg?
Pre filtering is good because those random very high frequency hashes that come in from mains and rectification usually find their way in the circuits loop areas no matter what PSRR, through many ways. Non the less the noise I showed is created in the regulator, its not something that it passes through, so its another concern. The input CCS is very useful for not reflecting current move outside, beyond what it can filter.
Your regulator is a good V1.0 type topologically with added considerations in being fast and having capable modern Mosfets. What it mostly lacks is remote sensing, especially in your case using long cables to the Riaa PCB's output side. That would ensure the nice 4mOhm flat through audio band Zo you measured its preserved at the delivery points. High gain is not good when not needed for no difficult load IMHO.
OpAmp was decided to be avoided as a concept in this family of regulators right from the start. I left it to the members to check different regulators of other persuasions with their systems and decide what makes the better results to their subjective opinion.
Your regulator is a good V1.0 type topologically with added considerations in being fast and having capable modern Mosfets. What it mostly lacks is remote sensing, especially in your case using long cables to the Riaa PCB's output side. That would ensure the nice 4mOhm flat through audio band Zo you measured its preserved at the delivery points. High gain is not good when not needed for no difficult load IMHO.
OpAmp was decided to be avoided as a concept in this family of regulators right from the start. I left it to the members to check different regulators of other persuasions with their systems and decide what makes the better results to their subjective opinion.
Yes, agree. Sorry I did not explicitely state, the noise caused by the regulator is of course determined (among other things) by the noise in the reference voltage, hence me putting capacitors there.
Agree on the remote sensing, thinking about that too. Shouldnt be too complicated.
Was the opamp decision based on potential sound impact concerns? Sorry for asking, it might be stated on one of the 400+ pages....
Agree on the remote sensing, thinking about that too. Shouldnt be too complicated.
Was the opamp decision based on potential sound impact concerns? Sorry for asking, it might be stated on one of the 400+ pages....
Maybe your Riaa does not care about the reg total noise that much, so the capacitors are a step you will judge for bringing any benefit or just side effect. Will be interesting to update us in what was the effect in your system. Try film 4.7 to 10uF first.
OPAMP It can create difficult high range oscillations in a general simple diy friendly concept, hard for average builders to tame and verify in their personal layouts mostly lacking the equipment, it has high gain and high NFB, so it wraps the simple concept in its own demands. OpAmp regulators can be excellent on paper but those I had come across did not persuade me.
OPAMP It can create difficult high range oscillations in a general simple diy friendly concept, hard for average builders to tame and verify in their personal layouts mostly lacking the equipment, it has high gain and high NFB, so it wraps the simple concept in its own demands. OpAmp regulators can be excellent on paper but those I had come across did not persuade me.
Can you please post details about your setup for measuring the output impedance?
Hi Salas,
Do I get it correct that you didn't come across shunt regulators with OpAmp control that impressed you with the sound quality?
Another question:
Your shunt regulators using sense wires, do the sense wires assist the regulator to dump voltage surges that originate in the client amplifier?
Thanks, Andrew.
Only I'm not clear about the way sensing wires reduce the impedance of the connection between the regulator and the load.
off by a factor of 20...
Hi, for this test the current source was set at 500mA and the output voltage at 12V. I then used a electronic load (selfmade) which was drawing 300mA, and overlaying an AC load current of 20mA. This load has a 1ohm resistor to measure the current so I could easily measure the AC current amplitude. I then used a soundcard with interface to change the frequency of that signal from 20Hz to 22kHz. At the same time, I measured the AC voltage at the output, and it was -50dBu or thereabouts, flat across the frequency span (with no signal, hovering around -90dBu).
And when you do the math, just like I redid just now, you'll find out that the output impedance of this regulator is -50dBu (=2.45mV)/20mA = 0.12ohm, quite lousy. Honestly, I do not know where I did that mistake in the first place - obviously the 4mOhm claimed earlier is false. Shame on me.....
Hi, for this test the current source was set at 500mA and the output voltage at 12V. I then used a electronic load (selfmade) which was drawing 300mA, and overlaying an AC load current of 20mA. This load has a 1ohm resistor to measure the current so I could easily measure the AC current amplitude. I then used a soundcard with interface to change the frequency of that signal from 20Hz to 22kHz. At the same time, I measured the AC voltage at the output, and it was -50dBu or thereabouts, flat across the frequency span (with no signal, hovering around -90dBu).
And when you do the math, just like I redid just now, you'll find out that the output impedance of this regulator is -50dBu (=2.45mV)/20mA = 0.12ohm, quite lousy. Honestly, I do not know where I did that mistake in the first place - obviously the 4mOhm claimed earlier is false. Shame on me.....
30 times out of Spice prediction for the V1 topology? R.A.P was right to look for the garbage can when arguing simulators it seems.
Anyway, I will be off to an island for a week or so, see ya guys, and if you confirm what is the best method and results it will be nice to see then. Maybe the cabling needs be sensed also.
Anyway, I will be off to an island for a week or so, see ya guys, and if you confirm what is the best method and results it will be nice to see then. Maybe the cabling needs be sensed also.
I see. The reason I asked is because I've tried such measurements some time ago. Can you post the schematic of your electronic load?
The methodology which Walt Jung used for supply impedance measurement can be found on his website: http://waltjung.org/PDFs/Regs_for_High_Perf_Audio_1.pdf
The AP analyzers allow you to measure the impressed current by comparing the voltage drop across the internal protection resistor. It's a good idea to measure the current explicitly since the signal generator output impedance isn't constant over frequency.
To make everything consistent I like to use the current and voltage he used in his articles, i.e. 15VDC with 150mA load and 50mA impressed current.
That would be Fig. 2c. It is quite surprising that an AP can drive the regulator straight. I had built a mosfet load whose gate was driven by one of my HP sine wave generators.
One question, he says there's a 2.5V RMS over the 50R load, which works out nicely to the reported 50mA RMS current. But you're talking about 15V DC and 150mA DC load, which would suggest a 100R load. Am I getting something wrong here?
One question, he says there's a 2.5V RMS over the 50R load, which works out nicely to the reported 50mA RMS current. But you're talking about 15V DC and 150mA DC load, which would suggest a 100R load. Am I getting something wrong here?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.