Are you still taking orders through your website? I put one through a few weeks ago but have had no response.
got my LDRs 
thanks Uriah!
another power supply question to clarify. After a brief glance on similar topics:
-7080 reg + bc550
-Salas mini shunt idea
-wall wart + 7805 comparison
I've decided to go for REG 1 + REG 2 (7805), hence the questions:
1. it is specified here that 470uf cap should be placed before the reg2 and vice-versa here. What would be the best option? Also may I use 0.1uf mkt cap for bypassing?
2. Will original (not mini) Salas shunt reg be kind of an overkill as a reg1? Would it be enough to use mini shunt or some alternative shunt reg on LM317 +BD139?
3. If Salas shunt will be used, what is the best sequence of caps here?
ex.: PS 2200uf (4700 ?!) -> Salas 100uf (output) -> 7805 470uf + 0.1 mkt (output)
4. This one is the most important that doesn't allow me to sleep at nights
Will I reduce PS noise by adding another voltage regulation circuit over existing one? The answer is yes as George's solution (wall wart + 7805) confirms, but if I add 10 more regs after the last one, will that bring more noise or there is some general rule here?
In many DACs and preamps the following scheme is used:
LM317 (alone or with some transistors/ zeners) -> 7805 (or any other fixed voltage reg),
so if I want to tune such or similar circuit should I replace it completely with Salas reg, for instance, or add Salas before the last 7805 reg or add Salas before LM?
thanks Uriah!another power supply question to clarify. After a brief glance on similar topics:
-7080 reg + bc550
-Salas mini shunt idea
-wall wart + 7805 comparison
I've decided to go for REG 1 + REG 2 (7805), hence the questions:
1. it is specified here that 470uf cap should be placed before the reg2 and vice-versa here. What would be the best option? Also may I use 0.1uf mkt cap for bypassing?
2. Will original (not mini) Salas shunt reg be kind of an overkill as a reg1? Would it be enough to use mini shunt or some alternative shunt reg on LM317 +BD139?
3. If Salas shunt will be used, what is the best sequence of caps here?
ex.: PS 2200uf (4700 ?!) -> Salas 100uf (output) -> 7805 470uf + 0.1 mkt (output)
4. This one is the most important that doesn't allow me to sleep at nights
Will I reduce PS noise by adding another voltage regulation circuit over existing one? The answer is yes as George's solution (wall wart + 7805) confirms, but if I add 10 more regs after the last one, will that bring more noise or there is some general rule here?
In many DACs and preamps the following scheme is used:
LM317 (alone or with some transistors/ zeners) -> 7805 (or any other fixed voltage reg),
so if I want to tune such or similar circuit should I replace it completely with Salas reg, for instance, or add Salas before the last 7805 reg or add Salas before LM?
A 7805 or 317 is perfectly good enough for the LDR supply.
You may find that making the supply slightly adjustable, say from 3Vdc to 6Vdc will make the volume adjustment "feel" more to your liking. This would suit the 317 reg. This voltage also changes the input impedance seen by the source.
The lower adjustment leg of a 317 can have a LED inserted instead of a resistor. The output voltage is then 1.25V+Vf of LED. i.e. about 3V1 depending on Vf.
MKT=MKS=PES=polyester has a slightly higher esr than the "high quality" caps normally used for audio. This higher esr makes them ideal for decoupling. It reduces the risk of ringing on the supply rail due to the damping of the resistance element in the esr.
You may find that making the supply slightly adjustable, say from 3Vdc to 6Vdc will make the volume adjustment "feel" more to your liking. This would suit the 317 reg. This voltage also changes the input impedance seen by the source.
The lower adjustment leg of a 317 can have a LED inserted instead of a resistor. The output voltage is then 1.25V+Vf of LED. i.e. about 3V1 depending on Vf.
MKT=MKS=PES=polyester has a slightly higher esr than the "high quality" caps normally used for audio. This higher esr makes them ideal for decoupling. It reduces the risk of ringing on the supply rail due to the damping of the resistance element in the esr.
Last edited:
You're overthinking it. The supply needs to be dead quiet is all. Improvement would be using a current source but after that put caps on the led side and be done with
it.
Read Walt Jung's "Sources 101" papers. Especially part 2. Note that a great solution would be lm317 voltage reg followed by two series lm317 ccs. That and caps. 100uf and .1-.47uf in parallel with it on each led. Oh, use 1-10uf tantalums on each lm317 leg.
I have been using 10-13v. I like it. Will be trying 3.3v in a new design because of cheap low noise surface mount parts designed to supply CPUs. Plus better caps in lower volts.
it.
Read Walt Jung's "Sources 101" papers. Especially part 2. Note that a great solution would be lm317 voltage reg followed by two series lm317 ccs. That and caps. 100uf and .1-.47uf in parallel with it on each led. Oh, use 1-10uf tantalums on each lm317 leg.
I have been using 10-13v. I like it. Will be trying 3.3v in a new design because of cheap low noise surface mount parts designed to supply CPUs. Plus better caps in lower volts.
voltage regulator
For what it is worth I have played around with some simple alternatives for powering the light speed.
After using a 7805 regulator, I found that a simple and inexpensive diy regulator described in TNT audio resulted in an improvement. The Zener + emitter follower can be built very easily with a parts cost well below $10.
Simple Voltage Regulators Part 1: Noise - [English]
A well filtered linear power supply before this also makes a measured improvement.
For what it is worth I have played around with some simple alternatives for powering the light speed.
After using a 7805 regulator, I found that a simple and inexpensive diy regulator described in TNT audio resulted in an improvement. The Zener + emitter follower can be built very easily with a parts cost well below $10.
Simple Voltage Regulators Part 1: Noise - [English]
A well filtered linear power supply before this also makes a measured improvement.
I also back AndrewT on this as well.
What I've always said after conducting a 20+ group of "Golden Ears" blind A/B test, is that no one could pick the difference between a linear wall wart + regulated 7805 based supply against battery power. It comes down to a psychological difference to the individual.
One needs to look at the big picture. The LDR's are for want of a better word "self regulating" because the reaction time is so slow to the LED variations, that any noise and spikes on the power supply to the LED's are mostly invisible to LDR's themselves, in other words they cannot react fast enough to these abnormalities "if any" to have an effect on their resistance once set by the led "volume" control.
Cheers George
Last edited:
This is, for instance, how digital isolators work. They use either transformers, radiofrequency, light or capacitance to isolate either DC or unwanted voltages from the signal. Common mode noise is especially a problem with the capacitive isolators and its effect is heightened with more current. In our situation we use a lot of current compared to what digital isolators are using. Capacitive coupling is a great way to transfer digital AND analog signals. This is one of the reasons that capacitors across the LED side are an advantage. Use a big cap for any decent sized fluctuations, which would be noise, and put a tiny, really fast cap, in parallel with it. The two together will deaden a lot of the problem. The reason to use current sources is that they reject noise so well.
Just read the first page of this article
http://waltjung.org/PDFs/Sources_101_P1.pdf
If we can reject noise at all audio frequencies, great. If we can reject above audio frequencies, also great. Even if its out of our hearing it interacts with the frequencies we can hear, causing distortions.
If you guys don't believe me, then just get the crappiest, oldest, lowest frequency switching supply you've got and replace the LDRs power supply with this thing and then tell me you don't hear the noise from the supply. The fact that it will be painfully obvious just means that noise does couple from the LED to the resistive side of the LDR. A 7805 has decent noise rejection but its rather poor compared to what we can get and for a cost only a few times more than the 7805 itself. Look at Jungs supplies. Complicated? No. You can all do it. There isn't even any tuning involved. Besides a couple of them that might oscillate this is childs play now that Walt has done the work for us.
Just read the first page of this article
http://waltjung.org/PDFs/Sources_101_P1.pdf
If we can reject noise at all audio frequencies, great. If we can reject above audio frequencies, also great. Even if its out of our hearing it interacts with the frequencies we can hear, causing distortions.
If you guys don't believe me, then just get the crappiest, oldest, lowest frequency switching supply you've got and replace the LDRs power supply with this thing and then tell me you don't hear the noise from the supply. The fact that it will be painfully obvious just means that noise does couple from the LED to the resistive side of the LDR. A 7805 has decent noise rejection but its rather poor compared to what we can get and for a cost only a few times more than the 7805 itself. Look at Jungs supplies. Complicated? No. You can all do it. There isn't even any tuning involved. Besides a couple of them that might oscillate this is childs play now that Walt has done the work for us.
Lets look at a 7805.
Cheapest one is 69 cents on Mouser. Its a Fairchild.
Line and load regulation between 4 and 100mV. That's not so hot.
Output noise is 42microVolts per output volt. So 210uV. 210/1000000 of a Volt. .00021V. This is down a little more than 70dB.
Walt Jungs current sources, at 18V, are putting out between -110dB and -140dB, sometimes hitting -150dB and should cost about the same as the LM7805 as long as you don't buy the big mosfets he uses in some of them. Just use the little TO92 ones. So add a minimum of 2 zeroes after the decimal as an improvement over the 7805.
Output resistance is about equal on both.
With the 7805, by the time we get to 10k its -55db (ripple rejection) and keeps right on going up.
In my estimation the 7805 is a good starting point. Make your Lightspeed with a regular TO92 or TO220 regulator. It will sound amazing. Then start putting small circuits to clean up the regulated power after it. They should just plug n play. This will reap dividends in improved sound.
Cheapest one is 69 cents on Mouser. Its a Fairchild.
Line and load regulation between 4 and 100mV. That's not so hot.
Output noise is 42microVolts per output volt. So 210uV. 210/1000000 of a Volt. .00021V. This is down a little more than 70dB.
Walt Jungs current sources, at 18V, are putting out between -110dB and -140dB, sometimes hitting -150dB and should cost about the same as the LM7805 as long as you don't buy the big mosfets he uses in some of them. Just use the little TO92 ones. So add a minimum of 2 zeroes after the decimal as an improvement over the 7805.
Output resistance is about equal on both.
With the 7805, by the time we get to 10k its -55db (ripple rejection) and keeps right on going up.
In my estimation the 7805 is a good starting point. Make your Lightspeed with a regular TO92 or TO220 regulator. It will sound amazing. Then start putting small circuits to clean up the regulated power after it. They should just plug n play. This will reap dividends in improved sound.
I took the approach by udailey
I started with a 7805. Sound was very good. It did not seem to make much difference if there was a walwort before the 7805 or a 9 volt battery.
Just for interest sake, I played around with different caps after the 7805 that happened to be in my current cap supply. Be it psychological or voodoo, there were definite preferences to be seen just based on the cap.
I have not built a Salas shunt. Seems a bit much to build a $50 regulator to power $10 worth of LDRs. Reading the TNT article cited above, it was clear from their measurements that the Zener + Emitter Follower measures very well vs 7805, 317 and various optimizations of the 417. The parts cost $5 and you can knock one out in a few minutes. I replaced the 7805 with it and the sound is better.
I could not wrap my head around why power supply to LDR would have any impact on LDR resistor. If Udailey's theory of capacitive connection is accurate, then it could explain what my ears have heard.
Seems very reasonable that Salas shunts and current source supplies could yield a flavour to the sound that some would find pleasing. For me, I am not experimenting further as the sound is so darn good now there are other areas of the system that offer far greater gains.
Another thought.. why to we build +5v regulator and then add 100r resistors? Why not tune the regulator down to +4v and drop the resistors. Would a better quality pot have any advantage in reducing noise in the power supply? I currently use the cheapest off the rack part that can be found. I like the simplicity, but it begs the question if there are optimizations to be had in this dpt?
I started with a 7805. Sound was very good. It did not seem to make much difference if there was a walwort before the 7805 or a 9 volt battery.
Just for interest sake, I played around with different caps after the 7805 that happened to be in my current cap supply. Be it psychological or voodoo, there were definite preferences to be seen just based on the cap.
I have not built a Salas shunt. Seems a bit much to build a $50 regulator to power $10 worth of LDRs. Reading the TNT article cited above, it was clear from their measurements that the Zener + Emitter Follower measures very well vs 7805, 317 and various optimizations of the 417. The parts cost $5 and you can knock one out in a few minutes. I replaced the 7805 with it and the sound is better.
I could not wrap my head around why power supply to LDR would have any impact on LDR resistor. If Udailey's theory of capacitive connection is accurate, then it could explain what my ears have heard.
Seems very reasonable that Salas shunts and current source supplies could yield a flavour to the sound that some would find pleasing. For me, I am not experimenting further as the sound is so darn good now there are other areas of the system that offer far greater gains.
Another thought.. why to we build +5v regulator and then add 100r resistors? Why not tune the regulator down to +4v and drop the resistors. Would a better quality pot have any advantage in reducing noise in the power supply? I currently use the cheapest off the rack part that can be found. I like the simplicity, but it begs the question if there are optimizations to be had in this dpt?
damn, I must build separate Salas then
bad thing is that all five-six positive regs I've built slightly change the output voltage by 5-10% after 10-15mins of operating time.
You will be pleased with the result.
damn, I must build separate Salas then
bad thing is that all five-six positive regs I've built slightly change the output voltage by 5-10% after 10-15mins of operating time.
This might fix your problem, although I have only watched this effect after considerable 'warm up' time. Put a cap on each leg of the reg in addition to your input and output electrolytics. These new caps should be film or better yet tantalum. Great results with .47uf films but even better with a few uf of tantalum on each leg. Do not go overboard as to much capacitance on especially the adjust leg will screw it up. Each of these will go to ground. The one on adjust is parallel to the adjust resistor, so you are making sure that the positive lead of the cap is basically on the regulator leg. This will stabilize the voltage a LOT. However you cant put a tantalum on the adjust pin on a fixed regulator.
Also, put some resistance to ground after the regulator. For a regulator to operate correctly it needs a load on it. Some really good ones only need a mA or two but some of them really need to be shown the right path and need 5-10mA of a load which the LDRs will not give them in most volume positions. So put a correctly sized (watts) resistor from output to ground ( a LED in series with this resistor will give you ON indication) to load this guy. By the way you guys can get away with the TO92 regulators in this circuit as they can handle 100mA as long as you make sure you don't drop to much voltage, which in the case of running only LDRs you wont have a problem.
Oh, another way to help improve regulation is to heatsink the regulator in addition to the caps and loading.
Last edited:
http://www.amb.org/audio/sigma25/
This regulator is simply the 7805 we are talking about but with input and output electrolytics, a load resistor and a heatsink. Its a very good start but you should still put caps on the LED side of the LDR at a minimum and you can follow the regulator with a current source which can simply be a LM317 in CCS mode which is show in Walt Jungs paper and also can be found in the LM317 datasheet.
If you are now looking for other parts of your signal chain to improve, I have a suggestion:
Another LDR on each channel. There is likely a series resistor after the Lightspeed, probably directly after the RCA input on your amp. It is in series with the signal and directly between the RCA and input of your amp circuit. Replace this guy with an LDR. A tiny supply that feeds off your amps supply can run this LDR. A LM334 current source with a trimmer as Rset will be a good way to regulate the value of this LDR. It will give another sonic improvement. You can do this on the output of your source as well since there is likely a 47R resistor or higher on the source's output. It will be the last thing the source sees before it gets to the RCA.