No ... and yes, Andrew!
No-one suggested I build a linear PSU but I did this.
I'm certain I posted the result of some comparisons we did between this and the SMPS which came as standard with the AC speed controller, within the first 16 posts. But to summarise:
* 48v linear PS sounded better than 48v SMPS
* 48v linear PS with isolating transfer "in front of it" sounded even better
* but SMPS with isolating transfer "in front of it" sounded better still
* 48v SLA supply sounded even better.
So, given the hash-filter improves the sound of the SMPS+Isotran ... the next test is to see how this compares to the battery supply. Why I'm going to all this trouble is that - while I suspect the battery supply will 'win' ... it will cost over A$500 to build one (case + 8x12v SLAs + input & output sockets + some smoothing caps + charger) so I'm hoping the cheaper SMPS + addons will win the day!
Regards,
Andy
Simple - bcoz I have absolutely no idea what it is ... and what it is supposed to do.
Perhaps you might like to explain?
Andy
Spot on Andrew with the proviso you don't do what I did and put everything to ground on a large piece of copper and call it zero volts. My recent experiments used a LM317 as a benchmark. What a great benchmark. The way the noise can be shunted to ground makes a big difference both in hum and hiss. All of Johnathan's parts can be added.
I did a project with the Hypex SMPS. I was shocked how dirty it was. Knowing Hypex not to be stupid I guessed they knew something I didn't. None the less I made some special filters. The aux output to drive a phono pre amp in a box not unlike Naim Nait 2. I got this project right. After all my hard work the owner thought a conventional PSU better as he wanted to power a Raspberry Pi and DAC also. It took two attempts and for one idea I used I suspect I own much to Bob Carver. Energy storage PSU. I won't say more becuase I might get paid for that one ( one day ). The Hypex was excellent but always like looking through very clean glass with a slight tint. I would rather have no glass. The Hypex SMPS makes measuring the Hypex UCD 180 difficult. That must mean something. I have started a project to synchronize that same SMPS with the output at 1/4 the 400 kHz. Maybe ? How that works is the SMPS starts up then takes a new reference from the output stage via a Schmidt trigger and flip flop divider. I see no reason to think it won't work. The change over can be on a simple NE 555 and zero cross detector. I doubt the switching frequency needs to be better than whatever +/- 20% ? The synchronising is to allow only beat frequencies or hetrodynes common to the higher frequency.
My LM 317 mistake was cured by a solid wire bar between 2 x 10000 uF resulting in a letter T shape of folded metal. 1.5 mm dia is about right. The T upright about 3 cm and 1 cm at it's base soldered to the big sheet of 1 oz copper. How I got into trouble is my early experiments worked fine ( as below ). I said to myself computer designers just use loads of copper and don't bother with star grounding. We are talking of a 22 dB of error.
One thing I feel I should say. The most primitive way of all looks wonderful on my tests. That is zeners diodes and a TIP3055 transistor or better ( 2SC/TTC5200 ? ). I choose this for how well it looses heat. 2N3055 also. I added 1 uF at the output. When talking about shunt regulators recently I realised in the modern world none of our dream solutions do anything like the job we suspect important. Also universally it is these last capacitors that do most. Sometimes for stablity and often to do the job of a high speed battery ( almost ). Inside your equipement the designer has assumed the external PSU to be low grade. The machine will function OK. OK = circa 90% and the 10 % is worth having. In Audiophile speak the noise floor. Noise floor is very weird. A diode signing it's head off in the MHz will make audio band hiss. My advice is soft recovery diodes and no suppression. This begs a simple question. If a simple diode is so bad what of SMPS ?
What is very wonderful about the PSU I was told to forget in 1974 is it has no obvious stability problems. That is the zener shunt with a series current amplifier. If you think about it a voltage regulator is a power amplifer. It will work as a go-between. Although unlikely to be very defective it will have a sonic signature. The amplifed zener should be the most innocent. One can use it's close cousin the capitance multiplier. It can be a pre regulator or can be a hybrid with a zener. The single TIP3055 and be made into a complimentary feedback pair of BD139, 100R and TIP2955. It should not oscillate as it is not in an external feedback loop. I have not done this exact test. I have done a more complex version that were fine with TL431 ( below ). I suspect for a novice TL431 is likely to fail. Don't take my example as good. It was something I did out of my head and I think it conflicts with graph in the data sheets. I did take things off and got mostly the same. The TL431 can be replace with a conventional zener with no 9K1 and 2K4. The readings relate to 1 V and my needs 12 V. Thus I say different to the readings in my added text. The TL431 soulution is within 5 dB of where I finished.
This says more than I can the details.
http://wwhttp://www.tnt-audio.com/clinica/regulators_noise2_e.html
I did a project with the Hypex SMPS. I was shocked how dirty it was. Knowing Hypex not to be stupid I guessed they knew something I didn't. None the less I made some special filters. The aux output to drive a phono pre amp in a box not unlike Naim Nait 2. I got this project right. After all my hard work the owner thought a conventional PSU better as he wanted to power a Raspberry Pi and DAC also. It took two attempts and for one idea I used I suspect I own much to Bob Carver. Energy storage PSU. I won't say more becuase I might get paid for that one ( one day ). The Hypex was excellent but always like looking through very clean glass with a slight tint. I would rather have no glass. The Hypex SMPS makes measuring the Hypex UCD 180 difficult. That must mean something. I have started a project to synchronize that same SMPS with the output at 1/4 the 400 kHz. Maybe ? How that works is the SMPS starts up then takes a new reference from the output stage via a Schmidt trigger and flip flop divider. I see no reason to think it won't work. The change over can be on a simple NE 555 and zero cross detector. I doubt the switching frequency needs to be better than whatever +/- 20% ? The synchronising is to allow only beat frequencies or hetrodynes common to the higher frequency.
My LM 317 mistake was cured by a solid wire bar between 2 x 10000 uF resulting in a letter T shape of folded metal. 1.5 mm dia is about right. The T upright about 3 cm and 1 cm at it's base soldered to the big sheet of 1 oz copper. How I got into trouble is my early experiments worked fine ( as below ). I said to myself computer designers just use loads of copper and don't bother with star grounding. We are talking of a 22 dB of error.
One thing I feel I should say. The most primitive way of all looks wonderful on my tests. That is zeners diodes and a TIP3055 transistor or better ( 2SC/TTC5200 ? ). I choose this for how well it looses heat. 2N3055 also. I added 1 uF at the output. When talking about shunt regulators recently I realised in the modern world none of our dream solutions do anything like the job we suspect important. Also universally it is these last capacitors that do most. Sometimes for stablity and often to do the job of a high speed battery ( almost ). Inside your equipement the designer has assumed the external PSU to be low grade. The machine will function OK. OK = circa 90% and the 10 % is worth having. In Audiophile speak the noise floor. Noise floor is very weird. A diode signing it's head off in the MHz will make audio band hiss. My advice is soft recovery diodes and no suppression. This begs a simple question. If a simple diode is so bad what of SMPS ?
What is very wonderful about the PSU I was told to forget in 1974 is it has no obvious stability problems. That is the zener shunt with a series current amplifier. If you think about it a voltage regulator is a power amplifer. It will work as a go-between. Although unlikely to be very defective it will have a sonic signature. The amplifed zener should be the most innocent. One can use it's close cousin the capitance multiplier. It can be a pre regulator or can be a hybrid with a zener. The single TIP3055 and be made into a complimentary feedback pair of BD139, 100R and TIP2955. It should not oscillate as it is not in an external feedback loop. I have not done this exact test. I have done a more complex version that were fine with TL431 ( below ). I suspect for a novice TL431 is likely to fail. Don't take my example as good. It was something I did out of my head and I think it conflicts with graph in the data sheets. I did take things off and got mostly the same. The TL431 can be replace with a conventional zener with no 9K1 and 2K4. The readings relate to 1 V and my needs 12 V. Thus I say different to the readings in my added text. The TL431 soulution is within 5 dB of where I finished.
This says more than I can the details.
http://wwhttp://www.tnt-audio.com/clinica/regulators_noise2_e.html
Looking again at my example I can see where the error started. The result has a 50 Hz peak. If looking at the ripple spectrum of about 400 mv peak to peak the 50 Hz component is low. The TL431 has not been able to eliminate this. Thus it's reference point is not great. Looking for the 100 Hz it is almost under noise ( odd noise at that ) . I have access to a better analyser. It will be interesting to see where I really got. My target was - 145 dB which is a bit unrealistic. - 135 dB seems likely with - 127 db 100Hz at 1.5 a load.
Whilst I remember. In my quest to learn I built a version of the most primitive transistor amplifiers in the style of Mr H C Lin circa 1957. An input transistor, VAS with Cdom and single output device. The best reason to keep away from this deign is the need to set DC offset. For a voltage regulator it is a virtue. The amplifer can have gain without any complexity added. The dominant pole capacitor can be our choice. Loop gain can be reasonable. This allows without much extra complexity the sophistication of standard regulator and the simplicity of the amplified zener. My tests were very positive, sadly I keept no detailed drawings. BC327/337/ MJ3001 were what my useful box provided. BC327/337 are low noise devices. MJ3001 is far better than it's specs suggest. I can not say for cetain if it outperformed a LM317. I suspect subject to Cdom it did. My main inspitation came from the JLH 10 watt amplifier for this. I was surprised just how well a simple op amp like this works. Removing the need for 0 volts output helps greatly. It seemed much more stable than most ready made regulators. I would argue like a TL431 in it used of PCB space, but not as diffiicult to stabilise. It is a delight to need only manipulate one capacitor. I started with 220 pF. I suspect 10 pF would have worked. From memory TL 431 drops to 40 dB correction at 10kHz. That is not unusual. My little device can beat that without any problems. The VAS should have a CCS. It is a moot point if a bootstrapped one as I used was doing much. It would if called upon I guess ? Better to uses a conventional one as the PCB space is less.
Yes, I agree ... on the surface, it does seem incredible.
But if you were a 'Linnie', you would know there are audible differences between the various AC motor controllers which Linn brought out over the decades, for the LP12.
But there were 5 pairs of ears involved in these tests - not just mine. And when all agree - and can describe the differences - then AFAIAC they are real ... and not imaginary. And one of the pairs of ears belonged to a guy who is nearly blind - so, while it wasn't a formal "DBT", you could say it was a "HBT" (half blind test
)! Regards,
Andy
The silly thing about the LP 12 is is is mostly the voltage that matters. I did a project taking months over this and concluded the Lingo was more about 66V than class A output. The Valhalla 90 V. I don't much like the Lingo. Naim had a big transformer and capacitor for the LP12. The PSU I built was for my friend who was getting over cancer. I took more trouble and had doubts I would have time. He was OK and loved it. 80 V is about right with moderately low distortion.
I liked the suggestions in this thread to improve SQ when using SMPS so far.
For a simple amp-project I still have to finish, I also decided on a SMPS. The high frequencies of the thing worried me, for they will be still present in the DC.
Thought about several ways to manage this problem, but I think I'll opt for the K multiplier in a version that allows for more current.
For a simple amp-project I still have to finish, I also decided on a SMPS. The high frequencies of the thing worried me, for they will be still present in the DC.
Thought about several ways to manage this problem, but I think I'll opt for the K multiplier in a version that allows for more current.
I think you should read this in the link. If anyone knows of the best ferrite beads to use it would be interesting. As transistor and regulators are often TO220 the same will do for either.
I wanted to post my Hypex graphs. Alas they won't be very helpful as I seem to have lost the clean up ones. All you need know is audio band noise output was - 86dB below 1V. A Conventional PSU with LM7812 20 dB better. LM7812 is usually one people avoid. Hypex is likely to be the benchmark. I greatly doubt a capacitance multiplier will do much if the noise is not reduced by 1920's compomnents in modern form. I used a LCL double T Pi Filter. 10 mH 1 uF 10 mH. Before that a 30 mH common mode choke and a zener version of the multiplier to finishe. The end of the T seeing 1000 uF before the zener regulator. There were then many 10 uF when needed. This gave results as good as conventional although residual noise was untidy. The cost of these parts about $10 which made it worth doing as the Hypex gives you this output virtually for nothing. It also was compact. Double Pi was +/-23 v down to +/- 14V.
http://cds.linear.com/docs/en/application-note/an101f.pdf
I wanted to post my Hypex graphs. Alas they won't be very helpful as I seem to have lost the clean up ones. All you need know is audio band noise output was - 86dB below 1V. A Conventional PSU with LM7812 20 dB better. LM7812 is usually one people avoid. Hypex is likely to be the benchmark. I greatly doubt a capacitance multiplier will do much if the noise is not reduced by 1920's compomnents in modern form. I used a LCL double T Pi Filter. 10 mH 1 uF 10 mH. Before that a 30 mH common mode choke and a zener version of the multiplier to finishe. The end of the T seeing 1000 uF before the zener regulator. There were then many 10 uF when needed. This gave results as good as conventional although residual noise was untidy. The cost of these parts about $10 which made it worth doing as the Hypex gives you this output virtually for nothing. It also was compact. Double Pi was +/-23 v down to +/- 14V.
http://cds.linear.com/docs/en/application-note/an101f.pdf
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.