Jump to content

Switching PSUs noise elimination and protection


skunks

Recommended Posts

Hello guys!

While digging dipper into MidiBox SID project I came across some new questions. Powering questions.

 

Switching PSUs with high switching frequency are regulated, so we have to bypass regulator. But they have a switching noise/crackle. Is it a good idea to grab a switching PSU on e.g. 15v and pass it though a 7812 ? Will this remove the noise?

 

Are computer PSUs good? TK mentioned 50Hz hum if core module is powered from USB. Maybe the same with computer PSUs, but why? Where does this hum come from?

 

Question about protection. I guess both linear and switching PSUs need good protection. What is the reason the C64 PSU fries precious SIDs? I heard the reason is it’s regulators, but MB SID/ MB6582 has it’s own regulators. So I’m confused. The only possibility how it can happen that I can see is that 7805 is bypassed because C64 PSU gives 5v DC. Am I right?

Are cheap PSUs also bad in terms of protection and what can one do about it?

Link to comment
Share on other sites

Hola skunks,

 

many questions, a few (personal) findings - collected over the last years :-)

 

* original C64 PSUs have very old (and I think non-78XX) linear voltage regulators, that can fry the SIDs, when they go into berserk mode (outputting more than 5V), which happens occasionally.

* MIDIbox platform voltage regulators cannot be/are not used, when using the original 5V rail of the original PSU, as all regulators need a few volts to drop.

* High quality switching regulators (with high switching frequencies in the > 100KHz range) can be used and no audible difference is expected.

* Purists might still go for linear 5V regulation (with big heatsinks), even if no audible difference is heard (´prolly nostalgic reasons) :-)

* AT/ATX computer PSUs are not so good, as they may require dummy loads to operate and are very inefficient under low loads - also their switching speed may be low (-> audible interference)

* "Filtering" a switched regulator output with a linear regulator works great (use this for FPV projects all the time, e.g. 30V battery -> 9V switched -> 5V linear), with a scope, you can see all switcher peaks being cut off.

* You can protect any psu output from overvoltage by using a protection circuit, like a crowbar-design.

* You might be able to filter "noisy" (low-frequency) switchers with a LC filter (combination of both a high and a low capacitance and an inductor).

 

Here is a link to the MB6582 PSU i built a few years ago, which still works great and is a great footwarmer on cold winter days (classic linear, with crowbar protection):

 

Hope this was helpful! :-)

Many greets,

Peter

Edited by Hawkeye
Link to comment
Share on other sites

Hello Hawkeye! Thank you for the solid reply !

 

use this for FPV projects all the time, e.g. 30V battery -> 9V switched -> 5V linear

Do your 9V stepper have high (>100kHz) "non-audible" switching frequency? And you still find a positive effect of lowering it via 5v linear regulator?

 

I also thought about a crowbar design! Can you give a schematic? (Retro-Donald link is dead)

 

I have a transformer that outputs 

 

6.2VAC 1.15A (it can be split into 2 * 3.1VAC)

19.2VAC 0.8Ð (it can be split into 2 * 9.5VAC)

29.8VAC 0.4Ð (cannot be split)

17.2VAC 0.32A (cannot be split)

 

I thought about doing something like PSU optimized by getting 5V from 6.2VAC and getting ~14VDC from... aaah.. Don't know yet...

I'm wondering if I have to remove bridge rectifiers from SID modules, but leave regulators alone. It is not stated implicitly, however 14VDC has to be converted to 9v/12v somehow.

Link to comment
Share on other sites

By the way, PSU optimized scheme http://www.ucapps.de/mbhp/mbhp_8xsid_c64_psu_optimized.pdf

shows 9VAC going through bridge rectifier and 7809 regulator in order to get 9VDC (for adding to 5VDC)

But I've tried to pass my 9.5VAC through bridge rectifier and 7809 and got 8.32V DC after bridge and 6.82VDC after 7809.

What I am missing? May be Commodore PSU's 9VAC was not exactly 9VAC, but a little more?

Link to comment
Share on other sites

Hi Skunks,

 

regarding the crowbar schem, i´ve attached a PDF schem (made by Retro-Donald), hope it is helpful - it is also heavily fused, should anything go wrong - it has worked well for me...

 

In the schematic, the section around the TIC 106 is the crowbar. It works quite simple: in case of overvoltage, the zener diode triggers the "crowbar" thrysistor (TIC 106) to short the output rail, blowing the fuse.

 

The transformer output should be 2x 9V AC @ 2Amps each for a fully populated mb6582 (8 SIDs).

 

 

But I've tried to pass my 9.5VAC through bridge rectifier and 7809 and got 8.32V DC after bridge and 6.82VDC after 7809.

 

Regarding the low voltages you reported, I am not sure, but I suspect, that your transformer does not deliver enough amps (thus the voltage drops) under load.

Did you measure under load or under no-load conditions? All capacitors (as shown in the PDF) in place?

 

 

Do your 9V stepper have high (>100kHz) "non-audible" switching frequency? And you still find a positive effect of lowering it via 5v linear regulator?

 

Yes, I used a high-frequency switcher, both components received clean outputs (9V video cam, 5V video transmitter), no problems with noise here (both components are very sensitive to voltage spikes).

 

Many greets,

Peter

MB6582-Netzteil.pdf

Edited by Hawkeye
Link to comment
Share on other sites

..

Switching PSUs with high switching frequency are regulated, so we have to bypass regulator. But they have a switching noise/crackle. Is it a good idea to grab a switching PSU on e.g. 15v and pass it though a 7812 ? Will this remove the noise?

 

...

 

This is how I do it.  I use a Triad 15V 1.5A switching wall wart which feeds a 7812, a R78C9-1.0, and a R78B5.0-1.5 (for the 9V and 5V respectively, those are switching Recom Vregs).  This set up allows me to power the whole thing from a single DC supply (a very small one at that).  I retrofitted my box coming from the big C64 brick and was careful to check for any noise level changes and there were none.  Same thing for the 15V DC brick, initially I used a linear one but it was almost the size of the C64 one and again, no noise going to the switcher

Link to comment
Share on other sites

Thank you, Hawkeye and Altitude! I really appreciate your help, guys!

 

I measured my transformer under load: 9.5VAC turnes into 9.1VAC if current is 1.5A.

But...

 

All capacitors (as shown in the PDF) in place?

 

Arrhh... I missed the big one. :pirate:

Now it is 8.83VDC after 7809. It's output voltage doesn't change even if I feed it with 12VDC instead of 9.5VAC.

So I make a conclusion 9.5VAC is enough to feed it.

Only 7805 needs minimum 6.2V on input...

In order to feed 7809 and 7812 ,  9VAC and 12VAC (respectively) are enough, am I right? 

 

I still wondering why TK designed PSU optimized scheme that way - using 7809, adding 5v to get 14v, then using 7812/7809 on SID modules. I don't think it is to distribute heat between 78xx, but for extra smoothing / denoising. Do I think in right direction?

 

What about Retro-Donald's PDF schem, why only 5V line needs protection?

 

 

Altitude, it's encouraging news. I also wanted to make some tests to compare this approach with linear. But now I'm not sure if it is necessary if you compared it with 2 linear PSUs. 

Link to comment
Share on other sites

Hi Skunks,

 

please forgive the incoherent stuff below, still in partymode! ;-)

 

9VAC = 9x~1.4 = 12.6V peaks of the AC waveform.

Subtract ~ 2V for the rectifier diodes ~ ca 10.6V input to the 7809 - which is a bit close for a non LDO (low-drop) 9V linear regulator.  You might go shopping for a LDO 9V vreg.

Under load, the 7809 may drop below its 9V target, as the input voltage will be lower...

edit: you have 9.5V ac, which might be just enough! :-)

 

Retro-Donald does not protect the 9V line, as it is AC (directly from the transformer, no vreg can go berserk) - the original C64 PSU is 5V DC regulated and 9V AC unregulated.

 

The optimzed PSU scheme makes sense, I see no other easy way to obtain 12V DC for the 6581 from a 9V AC and a 5V DC source, but as always, please correct me if I am wrong!

 

Altitude is right - (high frequency) switchers are the modern way to go - if you can´t reuse existing parts, i´d recommend to go that way. It is also way more power efficient.

 

Many greets,

Peter

Edited by Hawkeye
Link to comment
Share on other sites

Partymode :-) He-he, I took it into consideration while following your calculations. So I decided to experiment a little with plain 78xx, if they doesn't work - will go and buy LDO :-)

 

Retro-Donald does not protect the 9V line, as it is AC (directly from the transformer, no vreg can go berserk) - the original C64 PSU is 5V DC regulated and 9V AC unregulated.

 

But then 9v AC line is converted to DC (in PSU optimized scheme) and regulated. Can that vreg go berserk?

 

Today I made a couple of tests.

12v DC from cheap switcher was regulated to 9v (to power 6580 SID) and then to 5v (to power core & SID). Of course, all bridge rectifiers were bypassed.

The noise was less audible then when getting 5v directly from switcher.

But! It was much-much stronger then noise I got when using transformer (6.2v AC, 9.5v AC).

 

I can attach sound examples if anyone is interested.

 

Maybe it would be better if my cheap switcher would output 15v instead of 12v. Or if it would be expensive switcher :-) may be I've got to get one

Link to comment
Share on other sites

But then 9v AC line is converted to DC (in PSU optimized scheme) and regulated. Can that vreg go berserk?

 

Yes, that could happen, but I think (please correct me, if I am wrong), that

a) we have additional vregs that would create the SID target voltage of 9-12V from the 14V (of course, these could fail, as well, but see c))

b) the 5V line would be raised  to a higher relative voltage than the 9v/12v lines, if the vregs went berserk

c) the 5V digital supply needs to be within close limits, and damage to SIDs would occur with even slight overvoltage, whereas the 9V or 12V supplies the analog part of the SIDs (opamps, etc), and those might be able to stand more abuse (speculative!).

 

I think, the best you can do is follow Altitudes recommendation and use high quality switchers, maybe add a few filtering caps after the switchers (one big, one small cap each, e.g. 470uf and 100nf) and you´re good to go! :-)

 

Many greets,

Peter

Edited by Hawkeye
Link to comment
Share on other sites

Thank you Hawkeye for sharing your thoughts. It sound quite reasonable!

 

While experimenting with different PSU's and different way to connect grounds I found an interesting bug (or feature) during ALL experiments.

If I touch a big electrolytic condenser (or sometimes even regulator) with some metallic tool, core module hangs (no sound anymore, no notes sent to midi are playing, but MIOS Studio says: "Application is up & running!") and needs to be reset.

Is it OK or I have to do smth about it? (I would certainly if I could)

 

Can you give an exact advice on LC filter you are talking about? I'd like to experiment a little :-)

Edited by skunks
Link to comment
Share on other sites

Hi,

 

regarding the Cap-touching and core hang afterwards - it could be possible that that either causes a voltage peak (electrostatics?) or a voltage-drop (brownout) - no need to change that, if the core runs fine otherwise - just leave your fingers and metallic tools off it :-).

The MIOS studio "Application is up&running" message won´t get updated in case of a core hang - so that observation is consistent!

 

Here is a LC-Filter build howto - quite simple, but very effective (for me at least :-)):

http://www.fpvuk.org/how-tos/lc-filters/

 

Many greets and have fun!

Peter

Link to comment
Share on other sites

That LC filter manual is straight and simple, thank you Hawkeye!

 

I read carefully about fighting with noise in your report. You compared new and old PSUs.

 

What powering option did you use for your MB-6582? Was it the same with old and new PSUs? There is no option if you have "PSU optimized" scheme already done. Here I found an invention of Option E for this case: http://www.subatomicglue.com/sidl0g/ (see the post dated "5.12.2008")

 

 

The MIOS studio "Application is up&running" message won´t get updated in case of a core hang - so that observation is consistent!

I pressed "Query" button, so the message got updated. But OK, I won't touch it :-)

Link to comment
Share on other sites

Hi,

 

 

What powering option did you use for your MB-6582?

 

i went for power option A in my MB6582 (only need 9V for the newer-model SIDs) - i also did not change the power option when switching from original C64 PSU to the new DIY PSU - the retro-donald PSU just about perfectly emulates an original C64 PSU (9V AC, 5V DC, but 5V rail protected), as these voltages are what the MB6582 baseboard wants.

 

 

I read carefully about fighting with noise in your report. You compared new and old PSUs.

 

Yes, after moving the 78S05 Voltage regulator caps more closely to the voltage-regulator itself and adding a second large input filter capacitor (2200µF), the DIY PSU was very good, and produces a bit more silent output than the original C64 PSU. I don´t know exactly why this is the case, as both PSUs are linear regulated. I can only guess, that the modern 78S05 and the good filtering caps somehow improve the audio output quality, as probably a more stable 5V output is generated.

 

Many greets,

Peter

Edited by Hawkeye
Link to comment
Share on other sites

In TK's PSU optimized scheme there are also 2 extra 2200uF caps (but on 14V line).

I think 14V is intended not only for older SIDs, but also for good S/N ratio for newer SIDs - 5v drop on 7809 might give much less noise.

Here http://www.ucapps.de/mbhp_sid.html

TK says: 14-15V AC (!!!before!!! the regulation unit) are required for a good and noiseless sound.

 

 14 * 1.4 - 2 ~= 17.6 DC - that's quite a lot...
Link to comment
Share on other sites

The 78xx regulator needs an additional 3V to properly regulate.  If 12V DC output is required you need to input atleast 15V DC.

 

So for the old 6581 SID, 12V AC is just barely enough. But only if in your house the AC is properly 230V. So it's safer to choose 15V AC.

 

For the 8580 or 6582 (new) 9V AC is again barely enough to get the additional 3V. So it's better to opt for the 12V AC. I would advise against using the 15V AC as all the excessive voltage will be dispensated as heat.

 

 

If i can make a suggestion, if you don't want to go for a big fancy SMPS. Then maybe you can look for an LM2596 solution. A lot of companies are selling these as ready made modules. Usually for a couple of euros. These small boards are adjustable, high efficient and have a switching frequency of 150 kHz.

Edited by Shuriken
Link to comment
Share on other sites

Hello Shuriken! Thank you for advice! I've already got LM2596 and found it's rater noisy. I see Altitude and other people here advice more expensive switchers: PT78ST105 - 600Hz ($33 is cheapest I could find on ebay)Traco DC/DC converter - 500Hz (~9 euro), R78C9-1.0 and a R78B5.0-1.5 - 300Hz ($10 each). I have not tested neither of these. It could be great if anyone could give a comparison.

Link to comment
Share on other sites

+1! Traco switchers are awesome (and usually can be bought as very compact drop-in replacements for the standard linear vregs) - got them in my fpv stuff for sensitive analog cameras - they are good!

 

How did you measure the LM2596 noise? Theoretically, there should be no audible difference between 150khz, 300khz, 500khz and 600khz switchers - all well above the perceptible audio range.

Especially when properly filtered with a few small capacitors after the switcher...

 

Many greets!

Peter

Edited by Hawkeye
Link to comment
Share on other sites

Thanks for reviews everyone!

Hawkeye, that's how I measured it previously (I can do other tests if you have another ideas). I didn't connect extra capacitors except that was on board of LM2596 I've bought and Core8 board. I connected LM2596 from SID module's 7809's output to Core module 7805's output (bypassing 7805 and therefore 2200uF cap, but 10uF and 100nF were working). SID module was fed with 9.2V AC.

LM2596 was set to output 5v.

This was noisy.

Then 7805 was inserted between LM2596 and Core, I mean I connected LM2596's output to 7805 input.

LM2596 was set to output ~6.3v.

This was a little better, but also noisy.

Then I thew away LM2596 and connected 7809's output to 7805's input directly.

It was MUCH, MUCH better.

 

I can upload mp3's if you wish.

 

BTW, this way of cascading 78xx is a little more noisy then using separate transformer's outputs. I can't guess why.

Link to comment
Share on other sites

Hawkeye, I just looked at your calculations again

 

9VAC = 9x~1.4 = 12.6V peaks of the AC waveform.

Subtract ~ 2V for the rectifier diodes ~ ca 10.6V input to the 7809 - which is a bit close for a non LDO (low-drop) 9V linear regulator.  You might go shopping for a LDO 9V vreg.

Under load, the 7809 may drop below its 9V target, as the input voltage will be lower...

edit: you have 9.5V ac, which might be just enough! :-)

 

and wondered, why 9V AC input for MB-6582 in PSU Option A is considered OK even with plain 7809 (non-LDO) ?

I make an assumption that it worked in your case because your transformed was good enough (25VA) to sustain 9V AC under load.

Link to comment
Share on other sites

OK, I'll try it. Hovever shouldn't it sound good even without 7805?
In theory yes, but it depends alot on the components in front of the circuit. Did you provide power with a transformer?

Hawkeye, I just looked at your calculations again

 

 

and wondered, why 9V AC input for MB-6582 in PSU Option A is considered OK even with plain 7809 (non-LDO) ?

I make an assumption that it worked in your case because your transformed was good enough (25VA) to sustain 9V AC under load.

a 7809 can never pull more load then 1A, so highly unlikely. It's probably more todo with the amount of AC coming out of the wall socket. If the transformer outputs 9VAC at 230V and you only have 225V or less coming out of the wall.... Edited by Shuriken
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...