"Acceptable" level of power-supply noise for a console?

[Kchula-Rrit]

Just wondering what the power-supply noise immunity for the console is...

I bought a MeanWell RT-65A power supply (from Jameco Electronics) for the BATcomputer until I can fix the battery power supply.  The data sheet says the noise should be 80mV peak-to-peak.  I measured it to be roughly +60mV, -50mV with my storage scope and tried combinations of capacitors and inductors to reduce the noise, with little effect.  Then I just turned the power supply off and checked the noise, which was +30mV, -24mV.

If the "unpowered" noise is a component of the total noise, then the noise contributed by the power supply would be +30mV, -26mV.

K-R.

[JJB]

For what it's worth, I have replaced the power supply board of both my 99/4a consoles with a combination of small switching buck and boost converters. I did spend considerable time trying to reduce the ripple. For +5V I ended up at ~40mV (from ~60mV) by adding some capacitance and inductance at the output, purely by experimenting. I didn't bother with the other rails (+12V and -5V) even though the ripple was a bit higher on both (I think). I haven't seen any negative side effects whatsoever but definitely one positive one: cool to the touch 🙂

 

FYI I am building a Geneve stand-alone system and have replaced all those local linear regulators with the same sort of power supply you're using. It works a treat (haven't measured ripple). 

[RickyDean]

I'm starting to change out regulators on my Foundation z80 card, HFDC's, and other to these type of units from Mouser, VXO7805-1000 https://www.mouser.com/ProductDetail/CUI-Inc/VXO7805-1000., Non-Isolated DC/DC Converters 8-36Vin 5Vout 1A 5W SIP. I'm waiting for the 12 volt version to come in too.

[Kchula-Rrit]

1 hour ago, JJB said:

For what it's worth, I have replaced the power supply board of both my 99/4a consoles with a combination of small switching buck and boost converters. I did spend considerable time trying to reduce the ripple. For +5V I ended up at ~40mV (from ~60mV) by adding some capacitance and inductance at the output, purely by experimenting. I didn't bother with the other rails (+12V and -5V) even though the ripple was a bit higher on both (I think). I haven't seen any negative side effects whatsoever but definitely one positive one: cool to the touch 🙂

 

FYI I am building a Geneve stand-alone system and have replaced all those local linear regulators with the same sort of power supply you're using. It works a treat (haven't measured ripple). 

 

Thanks for your info, and I am glad that your new supply works.  My console has a USB keyboard adapter and a reset switch where the "coffee boiler" went, so any power is external.

I should have mentioned that my tests were with no load on the supply.  I will have to see if adding the console as a load will help with the ripple. 

[Kchula-Rrit]

35 minutes ago, RickyDean said:

I'm starting to change out regulators on my Foundation z80 card, HFDC's, and other to these type of units from Mouser, VXO7805-1000 https://www.mouser.com/ProductDetail/CUI-Inc/VXO7805-1000., Non-Isolated DC/DC Converters 8-36Vin 5Vout 1A 5W SIP. I'm waiting for the 12 volt version to come in too.

I have been thinking of doing the same thing with my PE boards, also.  Shift838 gave me a Mouser part number for a 2-amp part that does the same thing.  He mentioned it here, in the conversation about the new IDE board.

K-R.

[matthew180]

Computers like clean power, and personally I would care more about the power going to a 40 year old computer than a new computer.  Also, the aging components means they are probably not in spec any more, so the filers are not as good, switching thresholds and such not as tight, etc..  An unstable or noisy power supply is probably going to put move stress on the system, so take care and spend a little more on a good supply.

 

The specs for any power supply will depend on what is being powered, and for the 99/4A is would require a good survey of the components and consider special needs for any particular components.  Assuming the 99/4A engineers knew something about what they were doing, we can look at their choices of parts and make some rough specs.

 

The original 99/4A power supply uses 1N4003 1A rectifiers on the input AC, which is 1A @ ~18V, or 20W of power (since 18VAC rectified and filtered is about ~21VDC).  The -5V it has its own single-diode rectifier, so it would have worst regulation specs of the three voltages, but has about 10W available to it.  So the whole power supply is probably rated for about 30W or so.

 

The -5V circuit consists of a single 79M05C, 1.5A max.

The 12V circuit consists of a single 7812C, 1.5A max with 12mv to 240mV load regulation.

The 5V circuit uses a UA723CN precision voltage regulator in a switching mode, with a TIP31A output stage.  The UA723 has a load regulation of ~8mV to ~24mV, and the TIP31A can do 3A max continuous (15W).  There is also a TL331CP voltage comparator used to set the reference voltage for the US723, and may also help get better regulation.

 

I would expect the 5V is probably very clean and stable.  Since the 12V and 5V share the same rectifier, the 20W of power is shared between them.  If 5V is pulling the full 3A through the TIP31A, that would leave about 400mA for the 12V rail.  There is a current limiter on the 5V regulator, but I don't know what the cutoff is.

 

For a minimum basic spec, I would say something like:

 

5V @ 2.5A, no more than 10mV ripple/noise as seen on an O-scope.

12V @500mA, no more than 20mV ripple/noise as seen on an O-scope.

-5V @500mA, no more than 20mV ripple/noise as seen on an O-scope.

 

Noise/ripple is hard to nail down and understand, and datasheets will not specify these values exactly.  You have to derive the numbers based on the other information, and the specs for parts vary from part to part, and manufacturer to manufacturer.

 

[Kchula-Rrit]

Thanks again, for the advice.  Until I can get my battery supply working again, I am using a "Mean Well" RT-65A power supply, with +/- 5V and +12V outputs to run my console.  It seems to work okay, but I need to re-measure the ripple with the console attached.  The battery supply proved to very noisy, and the NanoPEB most definitely did not like it; I am hoping it was not damaged.

For the "real" PEB, I waffle back-and-forth on swapping-out the power supply.  I want the boards to run cooler, so I will probably try the regulators that Shift838 suggested earlier in this conversation, with my next parts order.

I try to treat my TI stuff like Faberge Eggs- not necessarily fragile, but certainly irrreplaceable!

However, there is that fan that sounds like a jet-plane winding-out for takeoff...

K-R.

[matthew180]

For PEB cards, IMO it would be best if everyone started converting all the cards they have to use a DC-to-DC buck-boost converter module like these:

 

https://www.pololu.com/product/4083

 

They will take 2.8V to 22V and output 5V @ 1A (other versions will do more current) with over 90% efficiency.  The nice thing about using the DC-to-DC buck-boost is they make the PEB-card safe to use in a PEB that has been modified to have a modern switcher with 5V and 12V outputs, as well as the original PEB power supply with 8V and 16V.

 

The CUI regulator referenced in the previous post has the same problem as the original 7805, i.e. it needs 8V input minimum, which leaves you in the same situation when trying to upgrade the PEB power supply using a modern PC power supply.

 

[RickyDean]

52 minutes ago, matthew180 said:

For PEB cards, IMO it would be best if everyone started converting all the cards they have to use a DC-to-DC buck-boost converter module like these:

 

https://www.pololu.com/product/4083

 

They will take 2.8V to 22V and output 5V @ 1A (other versions will do more current) with over 90% efficiency.  The nice thing about using the DC-to-DC buck-boost is they make the PEB-card safe to use in a PEB that has been modified to have a modern switcher with 5V and 12V outputs, as well as the original PEB power supply with 8V and 16V.

 

The CUI regulator referenced in the previous post has the same problem as the original 7805, i.e. it needs 8V input minimum, which leaves you in the same situation when trying to upgrade the PEB power supply using a modern PC power supply.

 

I have bought both the CUI type and the this type from Mouser 490-VXO7805-1000. Glad you brought this up here, as I want to be able to move IDE cards, SCSI, and FDC,s between normal machines and modified ones. Too many times have I accidently blown cards, being in a hurry, due to forgetting the voltage differences over the years.

[matthew180]

I'm sure the CUI devices are very good for what they do, but their specs are:

 

Output Power:     5 W     
Input Voltage, Min:     8 V     
Input Voltage, Max:     36 V     
Output Voltage-Channel 1:     5 V     
Output Current-Channel 1:     1 A

 

That is not going to work with a PEB that is modified with a PC power supply.  Also, anyone making new PEB-cards should just design the power subsystem right on the board as a buck-boost DC-to-DC converter.  The cost is not that prohibitive, especially for a hobby.  If the board uses any newer components then it probably also needs 3.3V anyway, so a power section will be required.

[Kchula-Rrit]

The CUI units would probably work in an unmodified PE Box, shouldn't they?  The specs look just like the 7805.  My PE Box is still completely "stock," down to the jet-plane sounding fan.  The only modification I made was to install a 3.5-inch and a half-height 5.25-inch floppy-disk drive.

 

Thanks for the info on the Pololu units.  Sounds like a great idea for a "universal" PE board, to go into a modified or unmodified Box.  The specs say that it can get very hot in normal operation, so it might not be the best thing for an enclosed PE Box.

 

I'm looking at the 32V-input units for the BATcomputer, since the battery I have puts out 21V.  Have to measure the console's current draw first.

 

K-R.

Edited December 28, 2023 by Kchula-Rrit

[RickyDean]

3 minutes ago, Kchula-Rrit said:

The CUI units would probably work in an unmodified PE Box, shouldn't they?  The specs look just like the 7805.

Yes the CUI's work in an unmodified PEB.

[matthew180]

6 hours ago, Kchula-Rrit said:

The specs say that it can get very hot in normal operation,

 

The original 7805 also gets really hot under normal operation, to the point where it should have a heat sink and can easily burn you.  For boards that have a lot of components, it would probably be better to use one of the regulators rated for more than 1A (and use a heat sink).  Double for safety is a pretty basic rule-of-thumb.

 

[+FALCOR4]

18 hours ago, matthew180 said:

For PEB cards, IMO it would be best if everyone started converting all the cards they have to use a DC-to-DC buck-boost converter module like these:

 

https://www.pololu.com/product/4083

 

They will take 2.8V to 22V and output 5V @ 1A (other versions will do more current) with over 90% efficiency.  The nice thing about using the DC-to-DC buck-boost is they make the PEB-card safe to use in a PEB that has been modified to have a modern switcher with 5V and 12V outputs, as well as the original PEB power supply with 8V and 16V.

 

The CUI regulator referenced in the previous post has the same problem as the original 7805, i.e. it needs 8V input minimum, which leaves you in the same situation when trying to upgrade the PEB power supply using a modern PC power supply.

 

@matthew180, other than cost, is there any reason to not use the S13V20F5 or S13V15F5?  Is there a difference, for example, with the output ripple?  Also, I recall a thread that talked about recommended capacitors for these units but can't find it at the moment.

[matthew180]

The main difference between those two are the output current capability, the V20 is 2A, the V15 is 1.5A.  The switching IC used on these is not specified (usually Pololu is good about providing the technical details), and chips that small don't have the actual part numbers on them (only a manufacture specific designation).

 

Also, as I mentioned earlier, the idea of "output ripple" is not something you will directly find in a datasheet, you have to calculate that based on the power supply circuit and the load (which is why no manufacturer will put those kinds of specs in their datasheets).

 

Power supplies are hard and very specific to their situation.  It is usually easier to buy something with the required specs, rather than trying to make you own.  The easiest thing would be to get a few of the Pololu regulators and try them in various cards, measure the output ripple, etc..  These devices also offer: under-voltage lockout, output over-voltage protection, over-current protection, thermal shutdown to prevent damage from overheating, and a soft-start (helps limit in-rush current).

 

As for caps, electrolytics are common for power filtering, just make sure the voltage rating is at least double the max output voltage and you will generally be fine.  Larger capacitance for low frequency, smaller capacitance for higher frequency.

 

[Kchula-Rrit]

17 hours ago, matthew180 said:

 

The original 7805 also gets really hot under normal operation, to the point where it should have a heat sink and can easily burn you.  For boards that have a lot of components, it would probably be better to use one of the regulators rated for more than 1A (and use a heat sink).  Double for safety is a pretty basic rule-of-thumb.

 

The one on my HFDC burned my finger.  The board underneath looked scorched; I always wondered if Myarc sent me a used unit.  Made a crude heat sink from some metal scrap, and it stayed quite a bit cooler, but that HFDC always seemed a bit flakey.  I gave up on the hard disk part and just used it as a rather expensive floppy disk controller.

K-R.

[Kchula-Rrit]

Some CUI P78-2000R units arrived a few days ago, so I plugged one into a breadboard.  Here is the output without any input or output capacitors:

 

And here is the output with 10-mike tantalum and 0.1-mike ceramic caps on the input and output:

 

The noise looks like 170mV without the caps and 60mV with the caps.  The data sheet suggests 22 microfarad caps, so I'll have to order some and repeat the measurements.

The part did not get at all warm to the touch, running with a 12V input and a 1 amp load.

 

K-R.

Edited January 6 by Kchula-Rrit

[Kchula-Rrit]

For grins, I measured the output from a 7805 with the same setup:

About 33mV.  The measurements are pretty "fickle," if I touched the rig everything changed!  If I didn't touch the rig, everything still changed! 8-[

K-R.

[matthew180]

Are those measurements taken with the regulator under load?  That is a lot more noise than I would have expected, especially for the 7805 which is a linear regulator (not a switcher).

[Kchula-Rrit]

2 hours ago, matthew180 said:

Are those measurements taken with the regulator under load?  That is a lot more noise than I would have expected, especially for the 7805 which is a linear regulator (not a switcher).

About 1 to 1.5 Amps, depending if the breadboard and the alligator clips decide to cooperate.

I was hoping the grid lines would have shown up better.  Sometimes I really do feel like Rip van Winkle, trying to re-learn things like this fancy digital oscilloscope!

K-R.

Edited January 6 by Kchula-Rrit

[Kchula-Rrit]

Murphy must be looking the other way for the moment.  I measured the CUI unit output at 48mV peak-to-peak noise:

 

The oscilloscope by itself has about 12mV of noise:

Anyone know if the oscilloscope noise simply adds to the CUI noise, or is it more complicated than that?  I'm assuming it is...  8-(

 

Or am I just too obsessed with power-supply noise?

 

K-R.

[matthew180]

You will have to check the specs of your scope, but I would be surprised it it was introducing 12mV of its own noise.  Are you grounding your scope to the power supply ground for the measurements?

[Kchula-Rrit]

23 minutes ago, matthew180 said:

You will have to check the specs of your scope, but I would be surprised it it was introducing 12mV of its own noise.  Are you grounding your scope to the power supply ground for the measurements?

I'm using the ground pin of the regulator as the scope ground for all of these measurements.  The parts are plugged in to a breadboard, which is probably introducing more noise into the measurements.

K-R.