John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
Disabled Account
Joined 2012
Right. So now do it. Lets give that number. we all have an idea of what THD and noise level is too high (audible) -- use them. Typical output voltage of 1-2v for line levels. And a 3-term reg for everything except PA. Then typical ripple on rails of PA using 5K or 10K cap and bridge rect. (1-10W/8). High FB VFA. We have all this info... A short PSR range is Ok also. So if we throw in some typical values, what do you end up with?



THx-RNMarsh
 
Last edited:
www.hifisonix.com
Joined 2003
Paid Member
My take on it Richard is if your PSRR is below other sources of distortion or noise, you are ok.

So, on a preamp, ref 1 volt output, you should design for > 90 dB PSRR (should be easy - lots of techniques out there). This puts it at or below all other extraneous sources. If you then add them up root mean square style, you can readily see this approach ensures optimum PSRR performance for least effort and $

Power amps are more problematic, but 90 dB ref 1 W out is also achievable. Expect this number to degrade when the amp is delivering lots of power as rail ripple goes up and IR potentials are developed in the ground wiring. You will have to spend more effort and money if you want it better.

These IMV are practical objective numbers. You can go overboard but that's plain wasteful.
 
Last edited:
AX tech editor
Joined 2002
Paid Member
My take on it Richard is if your PSRR is below other sources of distortion or noise, you are ok.

So, on a preamp, ref 1 volt output, you should design for > 90 dB PSRR (should be easy - lots of techniques out there). This puts it at or below all other extraneous sources. If you then add them up root mean square style, you can readily see this approach ensures optimum PSRR performance for least effort and $

Power amps are more problematic, but 90 dB ref 1 W out is also achievable. Expect this number to degrade when the amp is delivering lots of power as rail ripple goes up and IR potentials are developed in the ground wiring. You will have to spend more effort and money if you want it better.

These IMV are practical objective numbers. You can go overboard but that's plain wasteful.

But you can trade off power supply purity against PSRR. As an extreme, an amplifier with zero PSRR would work fine with an ideal psu with zero ripple and noise.

Of course there is no such psu, so you need to make a guesstimate for the psu and then you have the required PSRR for your target output contamination like -90dB below 2V.
And you can get that actual psu performance from a couple of tests in real world with some measurements. But someone has to do it ;-)

Jan
 
One of the roles of a real engineer is to take a vaguely defined goal and solidify it with specifications. Once the goals are defined then you can examine what or if trade-offs are needed.

The PSSR issue is really at what level is power supply noise an issue. As the range of human hearing without damage is around 140 dB a design goal could be 150 dB to allow some margin for other things such as noise and distortion that also contaminate the signal.

Now I have built power supplies that are that quiet.

So for any given circuit that has some intrinsic PSSR the power supply needs less.

For the perfect digital CD standard my arithmetic says 94 dB to the output from a nominal 5% ripple. Some from the regulators and some from the circuit under power.

The other point is that circuits with low PSSR can actually have other advantages. In some the low power supply impedance can lower the effective total circuit impedance and noise level.

Then there is the big fallacy about PSSR and how it is measured and specified vs how it really behaves in real circuits.
 
Member
Joined 2016
Paid Member
I had quality circles, TQM, SPC and then six sigma. Given the individually tuned RF circuits in some critical locations we chortled at trying to apply something that works for camshaft deliveries to Ford. But, the one that annoys me and makes me glad I'm not working for a major telco with a red logo any more is the six sigma black belt. Six sigma is fine, even if I suspect Motorola unleashed it to both allow themselves sloppier margins on some part and to slow down the competition as they raced to adopt it. But it only works on large sample sizes and not on people, gantt charts etc.

I came across a QA guy at a PES once who calculated the expected fail rate of a product, neat little bell curve. Where it became bizarre was that if the fail rate was more OR less than his curve there was obviously a problem with the product!!!

Love Dilbert. I used to blow up pointy head manager cartoons and pin em to the walls by the CEO's office at a place I was at. I was leaving, so had nothing to lose... :)
 
www.hifisonix.com
Joined 2003
Paid Member
Here is a simple cheap way to generate a clean PSU rail.

I am reminded of Denis Colin's ([sp?] LP797 with the single ended front end where he used a 10 000uF cap in a RC network to smooth the supply rail because the PSRR is terrible in that configuration. Using an opamp is cheaper, smaller and the performance is manifestly better.

Re how far to go with this thing. You have to be pragmatic here. If you have an amp that has 1% distortion and an S/N of 60 dB, why would to build a super low noise, high reg PSU - the benefit is completely swamped by amplifier issues. Ditto the reverse situation. Indeed you do have to look at this from the systems level so, that's why my guidance is as described in my earlier post.
 

Attachments

  • Opamp PSU033.pdf
    80.2 KB · Views: 103
AX tech editor
Joined 2002
Paid Member
Here is a simple cheap way to generate a clean PSU rail.

I am reminded of Denis Colin's ([sp?] LP797 with the single ended front end where he used a 10 000uF cap in a RC network to smooth the supply rail because the PSRR is terrible in that configuration. Using an opamp is cheaper, smaller and the performance is manifestly better.

Re how far to go with this thing. You have to be pragmatic here. If you have an amp that has 1% distortion and an S/N of 60 dB, why would to build a super low noise, high reg PSU - the benefit is completely swamped by amplifier issues. Ditto the reverse situation. Indeed you do have to look at this from the systems level so, that's why my guidance is as described in my earlier post.

I can improve on that:

- make the 1st opamp output 18VDC to feed the 2nd opamp;
- divide down that 18VDC as input to the 2nd opamp.

Of course it works only with single opamps, but hey, you want performance or not! And even with 2 x 5534 it is still cheaper than a single 4562!
Have your cake and eat it too!

One other thing Andrew. Don't try to confuse us with logic. SY is already a victim; that's enough for now ;-)

jan
 
Talking about string length may be better than talking about string colour, as people do in other threads. Of course, we all know that green string holds things in place in a restful way while red string makes things bounce around and get agitated. White string has the widest bandwidth.

The best string, of course, is homemade from unbleached natural materials; it doesn't last as long as commercial stuff made in factories by conventional production engineers and is more variable in characteristics so you need to use more of it (and it is expensive) but quality is always worth paying for.
 
Member
Joined 2016
Paid Member
Talking about string length may be better than talking about string colour, as people do in other threads. Of course, we all know that green string holds things in place in a restful way while red string makes things bounce around and get agitated. White string has the widest bandwidth.

The best string, of course, is homemade from unbleached natural materials; it doesn't last as long as commercial stuff made in factories by conventional production engineers and is more variable in characteristics so you need to use more of it (and it is expensive) but quality is always worth paying for.

That rough hairy string is the best.... You can do anything with that... It has quantum properties...
 
we chortled at trying to apply something that works for camshaft deliveries to Ford. .

For me, it was a military hybrid manufacturer being led about by some "experts" who's big accomplishment was the Keebler Cookie Factory.

That hybrid manufacturer went belly up. 600 people to zero. You know you're in trouble when production ships empty hermetic packages to the customers to make the dollar target. They would steal my incoming samples from storage and mark them for a customer.

One of the roles of a real engineer is to take a vaguely defined goal and solidify it with specifications.

Hmm..what is a vaguely defined goal? I guess some engineers actually get at least that much...sigh, that would be a luxury...


I came across a QA guy at a PES once who calculated the expected fail rate of a product, neat little bell curve. Where it became bizarre was that if the fail rate was more OR less than his curve there was obviously a problem with the product!!!

My current employ required MTBF stuff long before the designs were even out to prototype. It satisfied the pencil pushers, but had nothing to do with the production of actual hardware. Most of the autopsies I've performed so far fall into four distinct categories. Counterfeit components, bad component lots due to production errors at the component factory, damaged components due to manufacturer or screening handling, or damaged components due to process errors or bad process design.

There will not be diffusion type failures in my lifetime..so the arrrrhhh (enious) equation is moot.


Saw some shots from a documentary yesterday abuot Amsterdam - seeing the Gay Pride boats sail by gives a whole new meaning to the word 'string' ;-)

Jan
Was in Manhattan last weekend. The gay pride parade there totally wiped traffic from about 65th to Battery Park. Add in the street fairs on 6th(7th?), man what a mess.

John
 
Member
Joined 2011
Paid Member
A three part cascade consisting of 2X these little jewels, plus Walt Jung's / John Curl's discrete shunt mode Super Regulator, is probably more than good enough for me. Even when powering a hypothetical moving coil headamp whose amplification circuitry hypothetically has zero PSRR of its own. These are low dropout regulator ICs so they don't squander a significant portion of the output from your expensive Rcore transformer.

Use the typical stunt of completely independent xfrmr secondaries, completely independent rectifiers, completely independent filter caps, and build two identical regulator cards using "positive regulator" chips in each. And don't forget RFI filtering! These 6 ampere SMD feedthrough capacitors may help. Three terminals: input, common, output. Woo!

_
 

Attachments

  • figure1.png
    figure1.png
    30.3 KB · Views: 198
  • figure2.png
    figure2.png
    39.9 KB · Views: 197
  • figure3.png
    figure3.png
    28.1 KB · Views: 196
  • figure4.png
    figure4.png
    61.9 KB · Views: 225
Last edited:
A three part cascade consisting of 2X these little jewels, plus Walt Jung's / John Curl's discrete shunt mode Super Regulator, is probably more than good enough for me. Even when powering a hypothetical moving coil headamp whose amplification circuitry hypothetically has zero PSRR of its own. These are low dropout regulator ICs so they don't squander a significant portion of the output from your expensive Rcore transformer.

Use the typical stunt of completely independent xfrmr secondaries, completely independent rectifiers, completely independent filter caps, and build two identical regulator cards using "positive regulator" chips in each. And don't forget RFI filtering! These 6 ampere SMD feedthrough capacitors may help. Three terminals: input, common, output. Woo!

_

The only trick you missed is four transformer windings into four bridges combined for phase/anti-phase combining.

Thanks for the regulator cite. 75 apparently real PSSR! Even more the surface mount feedthrough capacitors. I also like the shunt bit.

I don't know which will end up having greater legs, my almost forgotten June 2007 AX article "Shunt Regulator: The Almost Forgotten Circuit" which seems to have started the revival in the circuit, or the bit on resistor distortion in Linear Audio V1.

The only part I think most folks missed on shunt regulators is using a resistor from a regulated voltage source has lower noise than using a current source.

So much thanks.
 
Here is a simple cheap way to generate a clean PSU rail.

I am reminded of Denis Colin's ([sp?] LP797 with the single ended front end where he used a 10 000uF cap in a RC network to smooth the supply rail because the PSRR is terrible in that configuration. Using an opamp is cheaper, smaller and the performance is manifestly better.

Re how far to go with this thing. You have to be pragmatic here. If you have an amp that has 1% distortion and an S/N of 60 dB, why would to build a super low noise, high reg PSU - the benefit is completely swamped by amplifier issues. Ditto the reverse situation. Indeed you do have to look at this from the systems level so, that's why my guidance is as described in my earlier post.
1. It will sound very "tight"
2. The negative supply is better made as dual tracking.
3. etc.
 
Status
Not open for further replies.