Calculating capacitors after bridge rectifier and before regulator.

[ruffrecords]

The circuit characterization or specification is called PSRR or Power Supply Rejection Ratio. This quantifies how much power supply ripple (or whatever) shows up in the circuit's output. For op amp circuits the PSRR is typically referred to the input so PSRR reduces the PS ripple noise, and the circuit forward gain amplifies it again.

JR

JR is right, the PSRR of the circuit being powered is the important factor. Most tube circuits have a very poor PSRR compared to op amps so their supplies need a lot more smoothing. For example, a typical single ended tube output stage might have a PSRR of only 20dB. This means any noise on the HT line will appear at the output just 20dB lower. So. if you want HT noise at the output of the amp to be better than -80dBu for example then the HT ripple must be less that -60dBu or about 1 millivolt. This is just an example. In practice you can achieve lower levels of ripple than this with a three stage RCRCRC circuit.

Cheers

Ian

 

[warpie]

What about discrete transistor circuits like neve, api, etc...?

 

[JohnRoberts]

What about discrete transistor circuits like neve, api, etc...?

better than tubes, not likely better than modern op amps. But they will be more like op amps than tubes.

PSRR is a serious design consideration for modern op amp design engineers.

JR

 

[ruffrecords]

What about discrete transistor circuits like neve, api, etc...?

A lot depends on topology and negative feedback. Assuming a discrete transistor design takes care to decouple the input stages, then it is the output stage that is the most vulnerable to PSRR. However, add a chunk of NFB and your PSRR improves by exactly that amount (it is one of the basic properties of NFB).

The Neve 2N3055 single ended output stage will not have a very good PSRR but at typical closed loop gain settings there could easily be 40dB of NFB. Even so it is not brilliant. Now you can see why the stage is usually decoupled with 1000uF and a few ohms.

Op amps, even discrete ones, can have open loop gains exceeding 100dB so add some NFB to make a 40dB amplifier and you still have improved the PSRR by 60dB. In addition, push pull emitter follower transistor output stages have a much better intrinsic PSRR than say the Neve single ended type (the same way as a cathode follower has a much better PSRR than a common cathode stage)

I find it interesting that op amp manufacturers only quote their incredible distortion figures at unity gain. I am not sure under what conditions they quote PSRR but with 100dB of NFB anyone can achieve a 100dB PSRR.

Cheers

Ian

 

[merlin]

The Neve 2N3055 single ended output stage will not have a very good PSRR but at typical closed loop gain settings there could easily be 40dB of NFB. Even so it is not brilliant. Now you can see why the stage is usually decoupled with 1000uF and a few ohms.

In fact it's worse than that, it has negative PSRR -it amplifies power supply noise! Neve power supply needs to be scrupulously quiet.

 

[ruffrecords]

In fact it's worse than that, it has negative PSRR -it amplifies power supply noise! Neve power supply needs to be scrupulously quiet.

I know that David Rees was quite scathing of Rupert's single ended output stage and even published an improved version where the output stage is an emitter follower driving the transformer. Despite its limitations, that output stage regularly achieved output noise levels in the -90 to -100dBu range.

Cheers

Ian

 

[rh001]

Hi!

As others have mentioned 1000uf per needed amp is correct,. Your first calculation gave you a vale of 2200uf which is probably right for 1.5A cuz a little extra is always better. Best way to go for high quality audio is to allow only 0.2 vac p-p ripple. The formula given by KA is good.

But you must also have this consideration: (Voltage on output of rectfier with rectfier capacitor) - 4 volts - (voltage regulator voltage drop + 0.5 volts margin) = needed voltage. What I am saying here is that if you want finished product to run at 113 vac rms you have to have an extra approx 4 volts of pre-regulator voltage. This is a consideration that even if normal mains supply is 120 vac rms, during summer peaks that can dip. Dips of more than 3 volts are uncommon in much of the USA but primarily in mountain areas or at end of line dips can be more than 3 volts. In second and third world locations who knows what the dip will be?

In the end you have to calculate then build and test with variac: with full output level into expected load turn down variac until you see ripple coming out of regulator and then turn up the variac until ripple just goes away and measure output of variac.....that's the voltage at which your circuit will work. I recommend 113 vac rms cuz one may need to keep working during peak demand due to deadlines or the fact that all the musicians are there and so one does not want to inconvenience them.

Note: the more pre-regulator voltage margin you provide the hotter will be the regulator when mains are running at 120vac rms. So you have to take that into consideration. TO-220 regulators can run at 6 watts....some would say more but then it is hot to the touch. Not that anyone is touching it but hot running power supplies are less reliable after a decade or so of near daily use. I use 6 watts as a ceiling and heatsink to chassis and then it's just comfortably warm to the touch and does not create hot spot on chassis.

 

[Analog_Fan]

2 days ago, i finished my Class AB pcb and send it out.
https://wiki.analog.com/university/courses/engineering_discovery/lab_14

Now i gonna make a "function generator" pcb to see if the amp caves in at some frequencies or not.
I just watched some German transformer producers websites, haven't asked yet.

I also watched some video's on EMI, EMC filtering, that will be subject by the end of years or so.

Although it might be only 19 Watts @ 19.6 volt (rectified), i would like to execute it like this or something like:

This amp has 66.000uF and 24 carat gold bus bars across the capacitors.

 

[Analog_Fan]

In the end you have to calculate then build and test with variac: with full output level into expected load turn down variac until you see ripple coming out of regulator and then turn up the variac until ripple just goes away and measure output of variac.....that's the voltage at which your circuit will work. I recommend 113 vac rms cuz one may need to keep working during peak demand due to deadlines or the fact that all the musicians are there and so one does not want to inconvenience them.

I have 2 equal transformers connected back to back, so i work on 220V safely, not triggering the differential or automatic fuse.

 

[ruffrecords]

It is one thing to go for an exotc PCB with 24 carat gold plating but it seems more of a fashion statement to me than good engineering practice. If currents are that high why not simply have some 1mm thick pure copper bus bars laser cut to the size and shape.

Cheers

Ian

 

[Analog_Fan]

It is one thing to go for an exotc PCB with 24 carat gold plating but it seems more of a fashion statement to me than good engineering practice. If currents are that high why not simply have some 1mm thick pure copper bus bars laser cut to the size and shape.

Cheers

Ian

It's probably gold plated, but to buy this amp you need to trade in 3 x Fiat 500 base models.
It's made in Swiss, so it could be real solid gold, it also has 8 mm thick hardened glass to see the inside and even this was carefully thought out, so loud music doesn't disturb the inside electronics.

It seams to output +/-50 Milli Volt @ 150 Watt RMS


Never seen such circuit, until i saw the video below.

The circuit is world wide patented.
See patent, pdf.
It took the guy 16 years to figure that out.

But Chinese are selling clones, people on ebay sell PCB's.

A German properly tested a Chinese clone and it's perfect.
It doesn't cave in with frequencies beyond 200kHz.

 

[merlin]

The Swiss love diamond buffers, I don't know why. Every Swiss amp uses one these days, they're like SRPPs in tube amps. Maybe the word 'diamond' reminds them of bank vaults and watches with jewel bearings.

 

[C H]

Can I be the first to rofl?

 

[Cqwet Dbdfte]

The capacitor bank metal rails are most likely anodized aluminum and dyed yellow.
Keeping (rectifier) noise out is good, those Zeners won't do much, the other schematic in the video uses LM317. Crossing over at say 200 Hz leaves very little power required above, most is consumed below 200.
This would make speaker powering much easier, using a separate amp for the sub.
I still agree with the concept of massive energy storage if no regulation is used.
Interesting circuit, no global feedback.

 

[Disco Volante]

200kHz... They mainly target the canine market...

 

[Analog_Fan]

The capacitor bank metal rails are most likely anodized aluminum and dyed yellow.

the website states 24 carat gold.

Keeping (rectifier) noise out is good, those Zeners won't do much, the other schematic in the video uses LM317. Crossing over at say 200 Hz leaves very little power required above, most is consumed below 200.
This would make speaker powering much easier, using a separate amp for the sub.
I still agree with the concept of massive energy storage if no regulation is used.
Interesting circuit, no global feedback.

There is no rectifier in the original, it's unregulated.
The transformer is 500 watt, @ full power 900 watt is used (according to website).

 

[Analog_Fan]

The Swiss love diamond buffers, I don't know why. Every Swiss amp uses one these days, they're like SRPPs in tube amps. Maybe the word 'diamond' reminds them of bank vaults and watches with jewel bearings.

Well, Swiss is the country with the highest amount of millionaires in the world.
Pretty much every home owner.

 

[Cqwet Dbdfte]

the website states 24 carat gold.



There is no rectifier in the original, it's unregulated.
The transformer is 500 watt, @ full power 900 watt is used (according to website).

No rectifier. Mkay.
Gold is really opaque and a 0.03 - 0.23 micron film is enough for PCBs. Cost is minimal, but audiophillies may like it. Ooooo... shiney!

 

[Analog_Fan]

No rectifier. Mkay.
Gold is really opaque and a 0.03 - 0.23 micron film is enough for PCBs. Cost is minimal, but audiophillies may like it. Ooooo... shiney!

All pcb i have made where ENIG, OSH park, default and only option, accept the current class ab amp, witch are done at Aisler, it would have cost 8€+ more for ENIG.
It doesn't corrode like normal pcb's, if you have bad luck.

Recently, i found CNC shop who is willing to take small projects.

Not adverting for other forums, but i was learning more about Dartzeel
https://www.diyaudio.com/community/threads/dartzeel-amp-schematic-build-this.134362/page-112

Here is a guy (pic) with 2 pcb's each having 11 x 6 x 1 uF for each rail!
Cool idea, since today i got like 90mF (divide by 4700uF) at my disposal.

I only wanna power a like 12 cm diameter speaker and don't cost more power ( annual electric bill) to replace pc speakers or marginal.

Even if you double or more the output transistors doesn't automatically mean you double the power.
It actually stays the same, the current gets shared equal on both bases, you gain nothing.
The design i made behaves like a voltage follower, it tracks perfect in the sim, i used the current indicated on the wires and multiplied that with HFE and that was indicated by the sim on the next signal in the circuit.
i use tl074 with small resistor (1k, 4k99) values to increase the signal.

Some people have build this on breadboard and it sounds gut with real size speakers.
So, lets' see.

I did include 2 pole Sallen key filter @ 48kHz.