Low-Noise 7805/7812 Drop-In Replacements

[@t1d100]

As a related topic... The traditional LM78xx style linear regulators do a good job of taming ripple. However, the EE hobbyist community is configuring them slightly incorrectly. This mistake is limiting the 78xx's ripple rejection abilities... Here is the problem... Hobbyist are only supplying the 78xx with its bare minimum drop-out of headroom voltage... Say +1.5V. However, the 78xx requires a minimum of 3 volts of headroom to attain its best ripple rejection. The manufacturers' marketing groups have obscured this detail in the data sheets, to de-emphasize this higher power requirement. But, this detail is in the data sheets. Go to the graphs and look at the test circuits... You will find that they do, indeed, test at +3V of headroom, to achieve the lowest ripple. An additional mistake is that hobbyist build their supplies according to the data sheet application circuit. These circuits only include the small capacitors that the 78xx needs for its operations. The hobbyist does not know that the manufacture engineers expect the user to know what other components are needed... Specifically, an adequate smoothing capacitor. So, this gets left out. Now, the circuit draws down the power and things do not work well, including excessive ripple. Of course, this same condition can occur with the reviewed switch mode supplies, if adequate smoothing caps are not provided by the builder. I hope that helps someone. Cheers.

[@theovannieuwenhuizen5756]

Great remarks! Ditto this.

[@boriscat1999]

I was always taught to add a BFC to my power supplied, Big "Filter" Capacitor (or another word that starts with F). And if the is ripple or noise under load, to start measuring these points in the power supply. Either bad caps or bad regulator. It's really hard to find and fix stuff that was built out of spec, like too low of voltage for a regulator.

[@root42]

I think the 1541 uses relatively high voltage for the input. Even more so as my drives are still wired for 220V but we use 240V nowadays, so the transformer runs a bit high anyway.

[@theovannieuwenhuizen5756]

The comments on this topic are getting more and more in-depth. Correct me if I'm wrong but using a switching power supply vs a linear (old school) also contributes to a larger noise level. But please measure these with a typical load attached since that will be a better representation of the working condition. Love to see the follow up on this video.

[@DrJunge]

While picking at the hobbists: Don’t forget the power diode from output to input to protect the 78xx from a higher input than output voltage when powering down. Also a discharge resistor across the ‘big filtering capacitor’ is a niche touch.

[@jwstolk]

It looks like a very neat design, and the datasheet is just icing on the cake.

[@root42]

The author designs power supplies for a living, so I trust that he knows what he’s doing. And the datasheet is adorable!

[@MrFloppyPCB]

Thank you. The datasheet was done by my girlfriend H.EXE, I provided her the raw data and she created a datasheet out of it. We thought it's a nice touch.

[@phoneaccount6907]

One day I've replaced soviet 7805 called КРЕН 5 in my old ZX Spectrum to cheap ali dc-dc module. Got moving lines on screen, because of its noise. Interesting , but with linear power regulator image is very clean, and very good quality on lcd tv. Despite ZX is a digital device, connected with analog rgb lines.

[@smotryash]

Кренки отечественные кстати мошьнее чем импортные

[@lohikarhu734]

As we often say, "there's no free lunch", and 'dropping in' an SMPS replacement for a 7805 is another "it depends" problem. A *very good* SMPS can give low noise, but elimination of switch noise is non-trivial in PCB layout, the pinout of the IC, the quality of the inductor and capacitors, the choice of capacitor size(s)... And, BTW, measurement of the noise is something that most hobbyists don't do well. As a former AE for test & measurement companies...

[@root42]

Yup, I don't even have remotely the experience or the equipment to do the proper measurements. However Mr Floppy -- the author of this replacement -- did use proper equipment and also has the proper technological background. I can merely show off his amazing work...

[@MrFloppyPCB]

Thank you for your feedback. The Mr. Floppy regulator design provides a Z0 probing technique, with a attenuation rate of 20:1. The oscilloscope input needs to be terminated with 50Ohm. The MMCX connector provides the possibility to probe Vout very accurate by picking up as small as possible radiated EMI from outside. And you get a properly defined 50Ohm impedance over nearly the full signal path from TO-3 regulator output to the oscilloscope input. Of course I would have chosen a bigger RF connector, but space was limited as you can see. Talking about switching noise, I want to add not to forget reverse recovery noise, which is always the noise with the highest frequency components. At the expense of some efficiency, I added a tiny snubber element within the SW-node area. However, I would be curious to see proper EMI measurements.

[@movax20h]

Nice product and concept. I would say measuring ripple without a load is mostly pointless. You need to load the circuit, maybe 1-2A, but also max load would be good. Only then you will see how it behaves. Also the output of the regulator will go to some caps, and for switching regulators you probably want more modern low ESR caps, which has huge effects. And then you showed that with no load, the ripple is 300mV. That is pretty horrible, but again, that is because you have very little output capacitance. The best would be to do sanity check, install in the drive, then actually measure ripple. Which you did. Nice. Also, as others said, the linear regulator will most likely have worse ripple if your input is too close to the drop out voltage. You need to give them probably 2.5V over the target voltage (this is with modern versions, some might be less, but older ones will require even more, like 3V). The "plastic thingy for isolation" is not plastic. It is usually a mica sheet. Has good thermal conductivity, but electrically insulating. This is also why there is a thermal paste there. They are absolutly needed, otherwise you might short casing of the package to tracks on the board. (Usually these regulators are mounted to heatsinks, and will also have extra ceramic insulation sleeves for the pins. As of the video, It is a bit sparse. First 23 minutes could be condensed in 3 minutes easily. Just saying.

[@root42]

Yeah, my videos are unscripted, so I sometimes just ramble my train of thought. The mica sheet is interesting, I didn't know those were that transparent. I always thought they had a more brownish tint. But these here are totally glass-like and a bit flexible. I installed them under the replacements, but didn't put thermal paste on them, as the switchign regulators run super cool. I think convection is good enough to dissipate the heat.

[@MrFloppyPCB]

Thats absolutely right. I can assure you, the ripple measurement which is plotted on the datasheet was done under full load of 3A.

[@retronexusio]

Just found your channel! excellent content

[@tsbrownie]

Very interesting. Thanks for making this video.

[@theovannieuwenhuizen5756]

Interesting video. I could understand why you would want to run things cooler. What I don't quite understand is why you leave the old and aged capacitors on the circuit board. At least test these before reuse. My guess is most will be deteriorated and disfunctional. 10x probe means decreasing the measurement by 10. 20x on the scope means it will adjust the measurement by a factor 20. So 10x probe on 20x input means you'll measure 2x the voltage. But it would seem to me you would be able to set the channel probe on the scope to any factor. Also you should limit the bandwidth to 20MHz for measuring the industry standard noise. The manual for your scope is online, and changing the bandwidth limit and probe factor are described in 2-3

[@root42]

Nice! That’s helpful. I might do a short follow up to show exactly that. Thanks.

[@paulstubbs7678]

Nice, although having the components exposed is a problem for regulators mounted on back panels, hopefully a regular TO3 cover will fit.

[@MrFloppyPCB]

You are absolutely correct, a cover would be the best scenario to shield radiated EMI. Before starting this design, I did a quick research, where to get empty TO-3 packages (or at least the covers), but was unable to find one. We then decided it would be too expensive to manufacture those covers in a small batch, so we did the design coverless. If there is a hobbist out there who wants to do proper radiated EMI tests, I'm happy to get in touch.

[@paulstubbs7678]

@@MrFloppyPCB No, you have me wrong, I was referring to the plastic covers used to protect TO3's on back panels of gear where stray wires can short them to ground and cause all sorts of trouble. I have a HiFi amp with TO3's on the back, so you've got to be real careful when groping around the back trying to push a lead into a socket, 30V straight from the power supply filter caps can easily take out a speaker

[@PrabhakarSharma-qg4ov]

Very nice information Great work for you Sir ji

[@alabamacajun7791]

Day-em is that real. Nice

[@magicsmokeescapist4069]

I might misinterpreted what you said, but your probe and the scope is basically a voltage divider between the probe and scope. Whit this in mind if the scope has 1M(which most are) and your probe to 10x and measure let's say 5Vdc(on board) you'll read 0,5Vdc, if the scope is set to 1X and vice verse. Your DSO input are 1M and you probe are 9M or 1M depending on your setting. So the DSO probe setting has to match to give you a "valid" reading, but for the scope itself it doesn't matter at all, it's your agony when nothing matches. ;). This is why Signal generators has a setting for 50ohm and Hi Z, so when you want 2vpp @ 1kHz you should measure that, because most SIGS are 50ohm output impedance no matter the settings, they just calculate the voltage output so it matches reasonable within what you want. I came by a Rodhe Swcharz RMS URE2 which gave me some headache before I realized that the probe was set to 10X... It has 1M input... So every reading was of by a factor 10. 😂

[@root42]

Thanks for the info! Makes sense.

[@chainq68k]

Another great video. I've heard of these switching regulators being noisy, but never seen it quantified, or visualized. On the other hand, would have been interesting to see the difference in-circuit. The caps and the general impedance of the powered circuit might smooth out some of those spikes, no? (I'm not an expert.)

[@root42]

Yeah someone with better equipment and especially knowledge than me can perhaps do some in circuit measurements. Fact is: I have been using this drive now for several hours and it seems just fine. Even the other drive with the more noisy ezSBC regulators has been running fine for over a year… and the other unmodified drives run fine too, albeit a bit warmer. With the 8050 it actually makes a significant difference as it has multiple regulators and gets quite toasty.

[@InssiAjaton]

Two thoughts about the noise measurements… First, why not turn the meter to AC mV setting and read numbers? Well, the noise appears to be spiky, so the meter may not capture them, but still interesting to try. Second one is about the scope measurements on the provided terminal. I expect the terminal to be made for direct use of the scope probe, i.e. remove the hook sleeve and there is the ground around the pin. Just plug in! Or am I assuming too much?

[@root42]

You mean measure in AC mV mode on my multimeter? Might be interesting approach, yes. It probably smoothes things out which might actually be desirable...? Not sure. The scope probe has the GND ring yes, but I don't think it fits inside the MMXC plug, as there is a plastic insulator in between. I might have to buy that MMXC -> BNC adapter. But I don't really want that, just for this one measurement, as it costs ~20 EUR shipped. The 50 Ohm adapter is similar, but I can use that for other measurements, too.

[@MrFloppyPCB]

Actually I never did such a ripple measurement with a normal meter in AC mode. I'm pretty sure the bandwidth is so low you won't see much. Especially because you don't just want to see switching frequency, but also reverse recovery noise of the power stage within the regulator.

[@Jenny_Digital]

Whilst I'm no genius or qualified expert in regulator design, I certainly learned a thing or two over the years. One thing I did notice about switching regulators is that the noise that comes out of them generally increases with load. Another thing is that some switching regulators aren't unconditionally stable with differing amounts of capacitance at the output. Finally, what's the _minimum_ load?

[@root42]

Yeah, I think there could be many more and better measurements made. I think I will finally have to invest the 20 EUR for the 50 Ohm adapter... For the 1541 the loads will be pretty stable for the 5V rail, as the old ICs are probably continuously drawing a certain amount of current, and don't really "sleep" -- at least the CPU etc. The 12V rail is obviously only loaded when the stepper and drive motor engage, and is otherwise unloaded, as I don't think there is anything else drawing 12V.

[@MrFloppyPCB]

There is a minimum load on the PCB, so the regulator does not fall into DCM. Talking about maximum capacitive load capability, we tested up to around 2mF. I need to check my records and probably update the datasheet.

[@uki352]

The glimmer shields (or the silicon sheets) do isolate the voltage regulator from the cooling block. They are needed if the ground from the regulator casing shall not be shorted to the ground that the cooling block connects via its screws. There are systems that have a "Shield" that connects all casing related mechanics but also all shields on cables connecting to other units in a system. Another reason may be if a negative and a positive regulator are mounted on the same block. Negative regulators do not have ground on their casing!

[@root42]

I didn’t know negative regulators don’t have GND on their case! Very interesting! How do they connect to ground? I have a 7912 in my PET…

[@uki352]

@@root42 Internally the negative regulator is sort of reverse constructed. Input is GND and GND is Input. The case of the TO3 is therefore connected to the unregulated -12V input and what is the Input Pin on the positive regulator is now the GND pin for the negative one. That also counts for the other case variants like TO220.

[@filips7158]

A switching regulator like this one will pollute well into the 100s of MHz region. The case will be packed with HF currents you'd surely want to decouple from your system. Then there's the common mode noise generation, the input needs a choke.

[@MrFloppyPCB]

There is no general rule to always use a common mode choke. In fact, which would split up the grounding, which usually leads to other issues. While there is no common mode filter in this design, there is still a differential mode filter at the input. It should pass at least the old CISPR22 Class B by design, but this was not tested.

[@KF-bj3ce]

Very neat, have never seen it.

[@YogSothoth1969]

Super interessant! Gibt es für die 5V und12V Versionen von Mr.Floppy eine Bestellmöglichkeit, z.B. einen Link, wo man an so etwas herankommen kann? Vielen lieben Dank im Voraus! Gruss, Michael

[@MrFloppyPCB]

Hallo Michael. Ja wird werden uns da irgendwas überlegen müssen. Das Design habe ich als Open-Source-Hardware deklariert, die Publikation folgt noch. Ich wollte erst ein paar reale Feedbacks abwarten. Aber wir werden auch käufliche, fertig bestückte PCBs anbieten. Infos wird es garantiert auf DOS Reloaded geben. Ganz ehrlich? Das grosse Interesse hatte ich nicht erwartet ;)

[@YogSothoth1969]

@@MrFloppyPCB Ja find ich echt super hatte nämlich letztens noch gedacht, was machst Du, wenn Du diese Dinger mal brauchst und es gibt keine mehr. Daher bin ich auf jeden Fall interessiert 😊

[@tenminutetokyo2643]

How do you set the output level?

[@root42]

I think there are resistors you can change to generate different voltages.

[@rickpontificates3406]

Clearly, those drop in replacements are reliant on external capacitors for better filtering. I dunno why they didn't put a couple of larger SMD tantalum caps on there. The higher output voltage is often caused by lack of sufficient capacitance on the output

[@MrFloppyPCB]

Tantalum capacitors are well known for not doing well with higher ripple currents. However, higher capacity does not mean better filtering, especially high frequency content. The design is optimized for ceramic capacitors and already provides a very low output ripple. However, you can still ad some more bulk capacitance on the output, thats up to the user. Why did I not add a big and chunky cap? Because there is simply no space.

[@punpck]

how can a switching regulator have lower noise than a linear regulator

[@enginerdy]

40 years of electronics advancement and good design!

[@MrCarlsonsLab]

The answer is: "They just don't." When you start with little to no noise, it's pretty easy to end up with no noise, (or close to.)

[@root42]

@mrcalsonslab exactly. I guess the OP misunderstood. The linear regulators are also low ripple but these switching regulators are for situations where you don’t have linear regulators as a replacement or maybe they put out too much heat.

[@GermanMythbuster]

The scope probe gives a x20 reduced voltage to the oscilloscope. So lets say you have 2V so your scope sees 0.1V But the scope only compensates for 10x so 0.1V x 10 = 1V therefore your scope only shows you *1/2 the actual voltage.*

[@speedsterh]

"So we will repair one of my broken 1541s" I wonder where that YT suggestion came from. Now I know.

[@paulperry7091]

Three pin regulators are not as simple as they look. Some have a minimum load in order to operate. And, while they regulate going forward, and noise generated on the power rails can (in some devices) flow backwards quite happily, leading to interference elsewhere.

[@MrFloppyPCB]

There is a small load added on this board, to keep the regulator running and not falling into DCM.

[@phoneaccount6907]

Turn. Turn it. Just turn. Turn replacement variant. Turn it . Want to see how pins are made. Turn.

[@phoneaccount6907]

4:20, he heard.

[@jwstolk]

and 5:11 for the new design, which also seems to use a different style pin.

[@viperwizard491]

with x1 oscilloscope will get better signal to noise ratio

[@root42]

Then it seems I really need to get that 50 Ohms adapter, try to limit the bandwidth and measure again!

[@viperwizard491]

@@root42cable length and signal bandwidth . otherwise must terminate properly to stop reflections at high freq. for 50ohm need DC blocking capacitor too (flat for desired frequency range)

[@TheCrakkle]

That high frequency noise will also effect the life of the capacitors in the device.

[@boriscat1999]

old power supplies often run better (lower ripple in the upstream/central power supply) with a bit of a static load on them from all these power hungry regulators. I worry I might be playing a game of wack-a-mole if I replace old components with lower power ones and then I have to replace more and more to modernize my entire power system. I've actually put large 5W or 10W load resistors on my linear power supply to quiet the ripple, very wasteful of course but old TTL components do not tolerate noise in the power supply very well.

[@root42]

As I said: usually the linear regulators are just fine. This is more for people who really want everything to run cooler. Nowadays there’s also so much interference from switchmode power supplies, WiFi, mobile, etc… completely different world compared to 40 years ago.

[@boriscat1999]

@@root42100% agree. I mostly run into the issues when fixing old equipment that were designed around very different assumptions than we have today. I made a 4-track multitracker recorder a little worse trying to do the minimal fixes to it, in the end I had to go in and fix a lot more than I initially planned on.

[@swrzesinski]

I would actually start by replacing that frako capacitor. They are dodgy ones at this age.

[@root42]

I could test them. Might be a good idea. Haven’t got a real LCR meter, but my multimeter does have a function for that. I definitely have no axial ones though. Will have to order them if I replace them

[@alexmendez6738]

Para que youtube pone videos en yo quieto saber ingles

[@drstrangelove09]

"problematic" -> a word with an evil history

[@firasgh871]

شكرا لك لو استخدمت راسم إشارة تناظري وليس دجتل سيكون أفضل

[@wiwingmargahayu6831]

periodic video and thoisoi 2 and some other youtube channel

[@charlesg5085]

You are not am electrical engineer... that is obvious. I am suprused you are able to function on a daily basis.

[@MrFloppyPCB]

I'm sorry about your depression.

[@stevenbliss989]

If the ground line is 1) not ground, or 2) the ground is feeble, because we are dealing with a SYNC BUCK, which needs a HARD ground otherwise the output noise will be HORRID and V regulation crap in case of really crap ground.