Module requests

[fredrik]

may do some breadboard procrastination later today

I tested with a bit more load, and as expected the drop goes up as the load increases (e.g. at around 200 mA you need a bit over 13 V in iirc). The lowest observed drops are simply the collector-emitter drop across a saturated driver transistor, and there’s absolutely no regulation going on.

So my main takeaway is that the midimuso schematics have never worked (at least not for 12 V supplies), but the circuit uses very little current so the designer (inadvertently?) ended up relying on the saturation dropout, and the regulated 12 V being good enough.

I think a TL431 might be exactly what’s needed here.

[lookmumnocomputer]

wow fredrick you have been doing your homework. yes it would seem so, but nevertheless it has worked for him, me and others, playing in tune and happy. ok well I dont want to put it out in its current state. so ill get one of those aswell. which actually works out a little bit less obtrusive as its only a change of a few tracks.

thanks for looking fredrick I have not had a chance to pull a Midimuso out of the rack since we last spoke

[lookmumnocomputer]

@lookmumnocomputer

lm1084

the only thing that concerns me is the TL431 has a 1 percent regulation amount which is far too much. when the other alternatives are much lower.

EDIT. oh I think thats the A one, trying to find a bigger letter after the model number for more accuracy, if they exist

[fredrik]

That’s the reference voltage accuracy, isn’t it? You’d still need to trim it. But not sure if 1.5 V headroom is enough; I meant to do a quick experiment but when I went looking for TL431s I only found a stash of LM317s. Almost so someone’s trying to tell me something.

(TI’s datasheet points to TL431LI as an improved option, btw, but at some point you’d be hitting the point where builders won’t know what exactly to buy since it all sounds roughly the same, so trying out the LMs sounds like a good idea.)

(Also, did you have to smash up This Old Tony’s nice tile or whatever that was that he had worked so hard on? Guess you could send him a PCB to make up for that.)

(EDIT: Ah, just saw you’ve posted it as well. He posted a YouTube story about your shenanigans yesterday, so saw it there first.)

[lookmumnocomputer]

I may be misinterpreting the information but a drift of 1 percent on a reference voltage that is 10 octaves worth of notes is a drift of a semitone or so. I must be honest with voltage references with such a small headroom im personally in unchartered territory and should have solved this issue much earlier than I have got to it based on the practical experiences I have had with the circuit but like you say trusting on the saturation of the lm317 is not a great idea, especially if someone wants to take the module out of 1 rack and put it into another rack with a slightly different output, and need to recalibrate it all again.

I have all of the aforementioned solutions on order and should be turning up early this week to come to some sort of conclusion.

haha oh great! yeah Colin furze arranged that pass the hammer was really nice of him! I threw the hammer without looking where it was going so yes through my own negligence his beautiful thing he was making has bitten the dust! poor dude.

[fredrik]

I may also be misinterpreting, but afaik it’s how close any given component is to the 2.500 2.495 V ideal voltage out of the box, not how it behaves when running (that’s more temperature drift etc). This matters if you’re doing large scale manufacture and are willing to pay a bit extra to avoid having to manually trim your gear, but unless I’m missing something we’re not quite at those volumes here

@antoine.pasde2 talked about it here:

Kosmo 1222 source a reasonable LM4040AIZ 4.1?

In our application (the #1222), as you say the precision doesn’t matter thanks to the REF trim pot. The main difference between the A, B & C parts and the D&E, is the temperature stability (±100ppm/°C vs ±150ppm/°C), but if you intend to keep the temperature mostly constant or are willing to re-tune when temperature changes significantly, it probably doesn’t matter all that much. If my calculations are correct, the difference isn’t huge. If you had a worst case of going from 0°C in a van to 4…

(I mean, it is labelled as a *precision* reference and there’s even an example that shows to use it to make an LM317 more precise in the datasheet, which wouldn’t make that much sense it it was worse than the other regulators )

Still not sure about headroom, though, so needs testing. Looked a bit more and found some AMS1117s (LM1117 variants), but still no TL431s.

[lookmumnocomputer]

oh nice yes I had not seen that.

ok I understand yes it makes sense. after looking at the numbers. the temperature drift is much more workable!!. well I have them on the way too! and to be honest if the standard TL431 functions remotely well. thats gunna be the winner, purely down to the fact its widely available in a To92 enclosure!!

regarding the 1222 yeah not had a problem with the 4040 when I swapped to a polystyrene timing cap that completely solved any drift. but ok thats fair enoug

[antoine.pasde2]

What are the minimum and maximum current draw on the 10.6V supply?

[fredrik]

@fredrik

still no TL431s

Found them (last place I looked, as always ), and some quick tests with low load (~5 mA) looks promising; you can dip quite a bit below 12 V before the shunt calls it a day.

Here’s the test circuit I used:

This choice of component values worked for me down to ~11.2 V.

The numbers next to the trimmer are the measured values after dialing in 10.60 V, normalized to 10k and rounded. The numbers check out: theoretical Vout = 2.495×(1+7650/2350)−2uA×7650 = 10.602 V.

Note that the correct value is close to 25% of the potentiometer range, so a production version could use a ~5k resistor between 10.6 V and REF, and a 5k trimmer to 0 V.

[fredrik]

Bonus round: one drawback with a shunt regulator is that the circuit consumes the same power whether the load uses it or not; to maintain the voltage drop across the supply resistor, the current needs to go either through the load or the shunt. I suspect the 5 mA I tested with is a bit high, and designing for 11.2 V may be overdoing it a bit, so we can probably increase the resistor value (which would reduce the consumption).

Another approach would be to replace the 100 ohm supply resistor with a small NPN transistor and a 1k resistor, turning the circuit into a precision series regulator:

I tested this with a few loads from from 5 to 50 mA, and the total consumption is now close to the actual current through the load; the shunt gets just over 1 mA via the 1k resistor, to make sure it regulates, but the rest goes to the load.

I used a BC337 in my tests, but pretty much any NPN should work (BC547 etc, 2N3904, …). The trimmer setting is the same as before.

[antoine.pasde2]

@fredrik

This worked for me down to ~11.2 V.

If you reduce the 100ohm resistor you can handle larger loads with a lower input voltage.
You basically want that resistor to be just under (Vin_min - Vout)/(I_load_max+1mA)
(The 1mA is for the TL431)
That way a minimum of current is wasted through the shunt but if Vin increases or the load decreases, the the TL431 shunts additional current to increase the drop through the resistor and maintain the desired output voltage. (This explanation is for any onlooker who would not already know, i.e. not for Fredrik.)

@fredrik

one drawback with a shunt regulator is that the circuit consumes the same power whether the load uses it or not; to maintain the voltage drop across the supply resistor, the current needs to go either through the load or the shunt. I suspect the 5 mA I tested with is a bit high

That’s why I asked about the maximum and minimum currents above if the load current doesn’t vary much, it’s all good, but if it varies significantly, then it might be worth implementing a series regulator as you describe.

@lookmumnocomputer

yes haha. to be perfectly honest I would save myself a lot of sadness if I quoted data sheets more often!!!

Well, actually … while cogitating about this circuit, I made the unfortunate mistake of consulting the data sheet for the LM7812 which is commonly used in Kosmo and Eurorack power supplies, including in the Frequency Central power supply, and I found out that the minimum voltage that it might put out is 11.5V so in theory we would have to design to that constraint.
But with a small enough series resistor I believe the TL431 based circuit can handle a dropout of less than a volt.

[fredrik]

Yeah, that was as much “with the somewhat arbitrarily chosen component values I used for this test” as “I’ve verified that you can get close, and there is no unexpected hidden headroom issue” – I mean, the shunt doesn’t even know what the supply voltage is, so what would that be, but we’re in a subthread about a commercial product using an LM317 in a weird way and people that would notice if it didn’t work telling me it works great, and I’ve only used the TL431 with tiny currents before, and the datasheet mentions a 2.5 V minimum headroom, so not going to take anything for granted

(and yes, (11.2−10.6)/100 = 5+1 mA so things do work as they should )

I still think the 10.6 V rail’s power requirements are very low, based on my experiments with a non-regulating LM317-based supply where even a small load resulted in drops well above 1.4 V, so odds are e.g. 470 ohm (0.9 mA at 11.5 V) or even 680 ohm (0.3 mA) could work.

[lookmumnocomputer]

a reply to a previous comment first.

1. The transistor outputs (gate signals?) will be limited to about 4.4V in that emiter follower configuration, is that high enough to trigger most envelope generators reliably?
   Edit: 3.8V, with the 1N4148 drop.

realised last night minimum component solution would just be a voltage follower and protection resistors, no voltage drop, basically the same as you were saying but less an extramc14504, plus yes I guess its less snipping of component legs, 2 opamp packages instead of 50 component legs.

[lookmumnocomputer]

Well, actually … while cogitating about this circuit, I made the unfortunate mistake of consulting the data sheet for the LM7812 which is commonly used in Kosmo and Eurorack power supplies, including in the Frequency Central power supply, and I found out that the minimum voltage that it might put out is 11.5V so in theory we would have to design to that constraint.

you guys together seem to be tearing apart the fabric of the eurorack community at the same time as solving this voltage reference issue!

First of all thanks to you two, ive been catching up on some well needed, “fill in the blanks on rather patchy know how”, searching various aspects of the chat and getting to the bottom of this.

The current draw from the Vdd reference pins of the ms14504’s, I just went to test this, needless to say the only midimuso I have sitting here has been pilfered for parts much to my dismay. in the words of terminator (who most definitely had headroom for his voltage reference) i’ll be back.

But with a small enough series resistor I believe the TL431 based circuit can handle a dropout of less than a volt.

Antoine this is indeed what I have been working to this whole time. To be honest I assumed because such a large bulk of modular supplies have 7812’s I thought everyone was working towards 11.5 anyway it was just a round up. Even the deeper supplies are around that amount.

[fredrik]

I suspect flimsy wiring also contributes to voltage drops (10 milliohm here, 20 milliohm there, …) which is why you can find corners of the synth internet where people insist on massive copper bus bars, etc.

(but while I think modern manufacturing, normal distribution, etc. saves you from worst case specs most of the time, I’m quite happy that I’m using a power supply design that can be populated with either 7812s or 317s – the latter is not just a better regulator, it also forces you to actually look at the voltages when you trim it…)

“fill in the blanks on rather patchy know how”

As long as you don’t forget the “don’t be scared to try it” bit

(I have a lot of respect for the “this seems to work” approach, especially when it consistently produces results; I just occasionally run into things where I don’t quite understand *how* they can work, which sometimes means that my understanding needs refinement, but sometimes means they actually don’t work, or at least don’t do what the designer thought they were doing )

[Maxhirez]

Oh snap I just saw this. So the Prophet 600 breaks down a lot? That sux! It is at least half again your own age though-you can’t be too hard on us relics!

Can you give us an elevator pitch for what you have in mind?

[lookmumnocomputer]

right so I have been doing a bit of playing sending to do as much as it can functionally, I am getting a current draw very small, its between 0.02 and 0.03mA (nothing else with more resolution) I thought maybe thats a bit small but, checked its all in order and thats what it is. thats a measurement between the 317 to the MC14504’s.

So to be honest the best solution that I think im going to give a go when they arrive is the suggestedTL431 , I guess the question is with the above reference voltage consumption readings, first of all does that sound at all realistic? its got me thinking its a bit low, however thinking of its function within the chip that its just using that reference into some pretty high impedance environments, maybe not??? also everything is buffered… so possibly??? but I digress, if this is indeed correct what sort of load resistor would I be looking at .

and as for the motto don’t be scared to try it indeed! it is indeed a productive way to learn and stumble upon interesting things… but yeah getting this reference voltage right is necessary! definitely a “this seems to work” motto with most things! but this module would be nice if it was in tune haha.

[lookmumnocomputer]

hold up im being a plonker! I knew it wasn’t right, got a new meter last week, different layout to my old one. no matter what is functioning it keeps a stable 1.3mA

[fredrik]

@lookmumnocomputer

also everything is buffered… so possibly???

Ok, I’d try 680 ohm 270 ohm in series to get a small margin, and a 5k1 resistor + 5k trimmer in the divider, to put the right position near the center of the trimmer.

(270 ohm gives you at least (11.5−10.6)/270 = 3.3 mA which is 1 mA for the TL431, 1.3 mA for the circuit, and 1 mA to spare. 330 ohm could work but then it’s getting a bit tight at 11.5 V)

EDIT: 220 ohm is probably better, since current is flowing through 5k+5.1k as well, see here.

[lookmumnocomputer]

haha max yes of course! the prophet dies all the time and im just after a simple poly to take on tour. with shared voice modular inputs, which is totally possible. its gunna be midi in for the notes and main env generators. VDO’s which are proving pretty stable without complicated autotune circuits. 1 per voice with a sub, so itll be like having a analog/digital wavetable hybrid. then just analog filters based on polys of old. most likely a 4 voice to keep DIY building to a minimum (mayyyybe 6), but it will be chain able for 8 voices. only thing in my way is finessing a filter for it, my attempts at the gameboy megamachine one is too complicated wanna keep part count as absolutely low as possible.