Author Topic: Designing and Building a VU Meter Buffer (Read 1962 times)

APZX · « **on:** December 10, 2019, 10:12:10 PM »

So, a while ago I felt that I really should buffer my monitor controller’s “Meter Output” to my VU Meters. This is one of things that is quite “in theory” territory, but is most certainly “good practice” territory as well.

First lets start with what is inside of a VU meter. Originally, they were essentially 200uA DC ammeters fed from a full wave copper-oxide rectifier (before silicon diodes were a thing, this is one way they converted AC to DC

). Now, I don’t know if that is necessarily the case with my Sifam meters, but the operation is likely quite similar even if the materials are different. Due to the inherent design of a VU meter the response is quite slow. It takes roughly 300ms for the needle to rise, this is called its “integration time”. At any rate, the real concern for me is that on my monitor controller the meter output is tapped right after the input buffer. This means that the VU Meter can effectively load my input buffer to the point of audibility. So, I set out to fix that.

My first attempt at sorting out this problem was successful.

VU Meter Buffer mk1

And the schematic

What is going here is the following. Looking at J1, the signal is AC coupled via C3 into the non-inverting input of U1.1. Additionally, the non-inverting input of U1.1 is also biased at approximately half of VCC via R5 & R4. U1.1 is then acting as a unity gain buffer. The signal is then AC coupled via C7 to the output on J3. And since I’m dealing with balanced signals I needed to do this for both the Hot (+) and Cold (-). The resistor after each 10uF capacitor is a discharge & pull down resistor. Their purpose is to keep that side of the capacitor at ground. The circuit is also a single supply. In this case I designed it to operate at 28VDC. So, the non-inverting inputs of the opamps are held at approximately +14VDC. Basically, the desired AC signal is offset by +14VDC and the input AC coupling capacitor superimposes the desired audio signal onto that DC offset. The output AC coupling capacitor then removes that DC offset. This AC coupling is fundamentally no different than what you see in a guitar pedal for example, the big difference is the increased voltages at play. That is basically the buffer part of the circuit. Pretty simple stuff. But what about that mess on the right?

All right, so this is something called a Capacitance Multipler. Without getting too crazy into the theory of things, it is basically amplifying the effect of the two stage passive RC filter created by R1, R2, C1, & C2. I’m doing this because I want a super clean DC power rail for the circuit, and this will remove all the junk coming in off the 28VDC AC-DC adapter I selected. This costs a bit of overall voltage, but that is not a problem really. Of note as well is the arrangement of Q1 & Q2. If you look up the part numbers you will find that the D45H11 is a power transistor in a TO-220 package, whereas the BC546C a small signal TO-92 transistor. So, what gives? Well that particular arrangement of transistors is something called a Sziklai Pair or Complementary Feedback Pair. This is a compound transistor in the same vein as a Darlington Pair. The big difference though is that a Sziklai pair uses two differing types of transistors. The advantage here is the voltage drop across the pair vs a Darlington arrangement. Instead of the roughly 1.2V you’d expect from a Darlington pair you have about half that or roughly 600mV. There are other differences as well, but they are not all that important in this situation. The other thing to note is that the Sziklai pair behaves as the same type of transistor as the driver transistor. So, in this case I’m using a NPN driver transistor and a PNP pass transistor. Thus, the Sziklai pair behaves as a NPN transistor. Break out your analysis skills if you don't believe me

And that more or less covers the first iteration of my VU Meter buffer. It is a simple design that works quite well, except that it is lacking in one critical area, gain. I want gain so that I can properly meter lower level average signals that I commonly work at, and this is where I shall introduce my VU Meter Buffer mk2. Which I'll begin talking about in the very near future

-APZX

APZX · « **Reply #1 on:** December 11, 2019, 05:37:09 PM »

Now, onto the VU Meter Buffer mk2 in all its glory. I feel at this point I should mention that when you design something you need to set yourself goals and limits. For example on my mk1 iteration I basically went out looking at what was readily available and while my circuit likely has some small differences the fundamental operation is probably nearly identical. You feed it power, give it a signal, and it spits it back out at unity. That is all well and good, but I wanted a more comprehensive design. Thus, I set out to improve my VU Meter Buffer mk1 by doing two things mainly. Goal 1, give the circuit gain. Goal 2, single DC input of 12VDC. Let us begin that design journey.

To begin I shall take you on a journey of enlightenment and discovery! So, I knew I wanted to keep the circuit operating on a single supply rail, but I wanted a higher operating voltage so that I have more headroom (marginal increase but that is okay). This also served as a means for me to better understand switching converters. Really a win win win, you know? To begin most standard opamps operate up to about +/-18VDC or about 36VDC if you single supply them. However, locating a 36VDC AC-DC power supply is near impossible. 12VDC, 24VDC, and 48VDC are pretty easy to find. So, I decided on rolling my own main power rail as it were. In this case I’m taking 12VDC in and stepping it up to just over 34VDC (chosen kind of arbitrarily honestly).

The first logical question to ask would be how do I take 12VDC and make it 34VDC without converting it to AC? Well, this is where we enter into the wonderful world of switching converters and in particular something called a boost converter. To get a general idea of how these work you can go watch YouTube videos on them, but they basically use the fact that when an inductor is quickly turned on and off with a capacitor an increase in voltage potential is observed. By rearranging the components it is also possible to step the voltage down in something called a buck converter. And for a 48VDC AC-DC power supply this would have also worked plenty well. However, I chose to step up the voltage instead because of Goal 2.

To that end I went chip surfing to find a suitable switching converter and selected a part from Texas Instruments’ “Simple Switcher” line. These parts are designed to be well simple to use, design, and layout. The particular part is the LM2577-ADJ or even more pedantic because I used the TO-263 version the LM2577SX-ADJ. It is a surface mount part, but hand soldering a TO-263 is not too bad. It also has a nice advantage because I do not need an external heatsink, I can just use a big ground pour to act as the heatsink

So, the datasheet is quite good and the typical application circuit shown on page two of the datasheet is fairly similar to what I ended up using. I’ll spare the formulas and calculations, but there are eight values that are selected and a total of twenty-one numbers associated with design of this chip. Everything is fairly loosey goosey with how the numbers end up. There are some gotchas here such as the ESR of output capacitor along with its rated ripple current. For example, I selected a 470uF output capacitor not because I need 470uF of output capacitance, but to meet the ESR & ripple current requirements I needed a physically larger capacitor which meant its value went up. This then changed two other values that needed to be shifted around which then meant I needed to double check two other values. In the end it is not too bad, but can be a bit trial and error based to get everything to work out.

Now, since I settled on a switching converter that means ripple that I do not want so I returned back to the capacitance multiplier. However, upon closer examination of how the circuit would behave on startup caused me some serious concern. Namely, in the power dissipated by the PNP pass transistor. She would be getting up to about 11W of dissipation during start up. I did not like that. I then spent an inordinate amount of time trying to think my way through a good avenue to make the circuit start up in a more controlled manner. I was not too concerned about the LM2577, but the poor D45H11 got my attention a bit. I then thought to limit the current charging up the circuit initially. However, trying to sink 3A of current is not a particularly easy task. Then it occurred to me use a constant current source. Now, these do exactly as they say. They provide a constant current. However, if you go looking for one you will find that they are almost always “low side” that is to say they are between the load and ground. I wanted it to be between the LM2577 and the rest of the circuit or a "high side" constant current source. So, I went to the wonderful world of the interconnected tubes and found a rough and ready PNP based constant current source. It uses two diodes, two resistors, and a PNP transistor. Basically, the current is governed primarily by the voltage drop of one of the diodes and the emitter resistor. It is not too bad to analyze if you think about it for a few minutes (try to think about it as a voltage loop ignoring the base resistor, that makes things make sense)

However, as I said it is a bit rough and ready. It suffers from thermal runaway, but since it isn’t running for too long that isn’t too much of a concern. Ahhh yes, that transitions nicely to the next challenge I had to overcome.

So, with there now being a constant current source to aid in softly starting up the capacitance multiplier I had to come up with a way to remove it out of the circuit when it was not necessary anymore. Basically, I needed a way to automatically switch the constant current source out of the circuit and also prevent the constant current source from powering the actual VU Meter Buffer. The best way I could think to do this was with relays. However, I also needed way to control when the relays would turn on. What I really needed was a soft start circuit. Now, the type of soft start circuit I ended up using is normally used to charge up the primary side of a big transformer, but there is no reason why it will not work here. Additionally, since I am powering this circuit with 12VDC then I have a known 12VDC voltage to work with as well. Though, I am still left with the problem of when do I turn on the relays? The most sensible way would be to turn them on when a certain threshold voltage or some such is reached on the capacitance multiplier. Almost like I need to compare a couple voltages… Ahhh yes a comparator!! Now, I could have easily just gone with most any old comparator IC and been fine, but since I have a thing for doing things the hard way I opted for a discrete comparator. Now, I did not design this comparator myself. I ended up using a design by Rod Elliott. It offers the side effect of actually being able to provide power. Most comparator ICs have something called an open collector output and they require pull up or pull down resistors to provide the functionality I needed. It really would not have been that big of a deal. At any rate, the inverting input of the comparator is tapped off the +12VDC provided by the AC-DC converter and holds the inverting input at about +4.9VDC. This causes the comparator to be shut off hard. The non-inverting input samples the output of the capacitance multiplier and uses a voltage divider such that when there is approximately +20VDC on there the comparator turns on. At this point two 10uF capacitors begin charging via resistors which begins to turn on a couple of 2N7000 MOSFETs. Additionally, when the relay coil reaches approximately 600mV of voltage drop across it with respect to the +12VDC rail a PNP transistor turns on HARD and very quickly charges up the 10uF capacitor causing the MOSFET to very quickly enter saturation. There is also a delay for when each relay turns on to prevent the constant current source from ever being connected to the VU Meter Buffer itself.

The completed schematic is below. I’ve covered the theory of operation basically by explaining my thought process on how I designed it and dealt with the problems it presented. Have a gander and see if you can figure out what parts are doing what.

And the completed PSU itself

In the next installment of this I will talk about the actual VU Meter Buffer itself. As I stated prior I wanted to add gain and that means overall the buffer itself is more complicated, but only in the sense of what I parts I can obtain.

-APZX

APZX · « **Reply #2 on:** January 22, 2020, 10:06:44 PM »

All righty. I have not abandoned this project! In fact when I started this I had all the boards built and tested. What I did not have was the rack case. I just got that today I did about 80% of the final assembly. I still have to solder the 2x6 rotary switch to the VU buffer board. So, I figure now is a good of time as any to talk about that.

First, lets start with why things had to get more complicated. My current VU Meters have a built in attenuator. It has the following steps: 0dB, -3dB, -6dB, -9dB, -12dB, and -15dB. The reason for the attenuator is so that the range of the meters is expanded. In the say -6dB position 0VU is actually -6dB from the nominal. Very handy for observing very hot signals

However, I typically do not work at those kinds of levels so I thought it would be really great to be able to amplify the incoming signal to increase the overall range even more! And thus I set out on my quest of adding gain!

The very first thing I had to do when I did this was choose gain settings. I figured I would do the opposite of the attenuator settings. So, 0dB, +3dB, +6dB, +9dB, +12dB, and +15dB. Now, this is actually a pretty simple straightforward design. The easiest way is to use some sort of multi-position switch and simply switch in a different feedback resistor or feedback resistors on a non-inverting opamp. With a stereo unit that means I need a 2 pole 6 throw switch. A rotary switch is quite well suited to the task, and 2x6 rotary switches are quite common and not too terribly expensive. Fantastic! Unfortunately, the real world is ever that nice to you. See, I am working with balanced signals here. That means for a stereo unit I would not actually be amplifying two signals, but four signals. A 4x6 rotary switch is not nearly as easy nor cheap as a 2x6. They do exist, but you go from say a $3-6 switch to a $40-100 switch. I mean I do not mind spending money, but sometimes you have to weigh your options lol.

But that did not deter me. Some other options would be to use a bunch of relays. Other than a bunch of board space, a lot of extra diodes, and about $2-3 per relay those are actually a very sensible choice. I could have even gotten quite clever with switching various relays and reduced resistor count more than likely. But I decided at the end of the day I wanted to use a simple 2x6 rotary switch because the actual parts are easier to obtain, and while the overall circuit is a bit more complex, the actual building blocks are still very simple.

Okay, now I have to rethink my initial design a bit. I am no longer going to be able to get away with two opamps. That just ain't gonna happen. The first thing that I needed to do was take the balanced signal and make it single ended. While just using a single opamp for the purpose of a differential amplifier is not the "best" choice it will work plenty fine for this purpose. If you really need precision in your particular circuit then you need something called the Instrumentation Amplifier. It is built from three opamps, and has much better performance than a single opamp. Furthermore, you can buy dedicated instrumentation amplifier chips such as the INA134 or THAT124x. Technically, they're just differential amplifiers, but the opamp and resistor matching on the chips are basically tailor made to the job at least for audio work

Now, in this instance resistor matching is very critical. The "absolute" tolerance is not "per se" all that important. What truly matters is the ratio. In my case I'm using 10k and 20k resistors, and all 8 of these resistors are matched as closely as I could manage. This is utterly overkill for the application, but there is no harm other than time

The input stage is then an OPA2134. Now, I chose the OPA2134 because it is a very well behaved JFET input opamp. It is slower than the old TL072 series, but it does have lower overall noise, and like the TL072 has very low input bias currents. This is important because I am running this off of a single supply rail meaning that I'll have to inject a "virtual ground" somewhere.

Speaking of the virtual ground. Unlike in my mkI version of this I am not individually biasing the inputs. Namely, because it is a little more complicated. Instead I opted to use an opamp to generate the virtual ground and inject that where necessary. Since I was going slightly overkill in the design in the first place, I decided to opt for a precision opamp in this role. In this case I opted for the jelly bean precision opamp the OP07. The big difference here is that this opamp has basically bugger all DC offset. So, if I say feed it +12VDC at its non-inverting input and put it into a unity gain configuration it will spit out +12VDC rather than say +12.1VDC that say a TL072 might. Additionally, since I'm only biasing this single opamp I can select lower value rail splitting resistors meaning that I'll get a more accurate split because the voltage divider can deliver more current before the voltage drifts too far away from the desired value. Another thing that you'll notice with the schematic is a 10k resistor in the negative feedback path. Now, why would there be a resistor there if it isn't adding gain or some such? Well, while an ideal opamp would have zero current flowing into its input pins that simply is not the case. In the datasheets for opamps on top of something called Input Offset Voltage there is something called Input Offset Current. Without getting too nitty gritty into the details here this is basically means that opamp's output voltage will also shift, and quite often, way out of the spec you would expect from its Input Offset Voltage. This is because in the differential input pair of the opamp since they are getting different currents into their bases or gates (in the case of a JFET) changes the output voltage ever so slightly. The easiest way to deal with this is to use a resistor in the feedback path. This resistor more or less equalizes the offset current for you. Not trying to get "too nitty gritty" here. This is exceptionally deep rabbit hole if you want to go down it lol.

Moving on. With the signal converted to a single ended signal via the OPA2134 input this signal then flows into the venerable NE5532. All this NE5532 is doing is applying that sweet sweet gain. Really I could have used any number of opamps here. The LM4562, NJM2068, LM833, an OPA2134, etc... However, what I wanted was a low noise, reasonably fast, dual opamp here. The LM4562 is an excellent candidate, very low noise, very high speed, etc... The NJM2068 is actually a very nice little opamp for not a change, in quite a few aspects it actually rivals the LM4562 for far less. The LM833 is really National Semiconductor's original competitor to the NE5532, and in general is quite similar. The OPA2134 is a very good little opamp, but it is more $$$$ than the NE5532 and has worse noise. Really, the NE5532 is just a hard opamp to beat for 99% of audio rate applications lol. How, the gain works is that I'm simply switching in a different resistor that increases the gain of the NE5532. There is no fancy nothing going on here. While the resistors are a nominal 1% tolerance they're all matched to within 0.25%.

Finally, onto the output. Here there are a total of four opamps in the form of a pair of NJM4556s. Now, these are interesting little devices. A quick inspection of their datasheet shows that these are not particularly impressive parts. GBW of 8MHz, Slew Rate of 3V/us, and a Noise figure of 10nV/sqrt Hz. Like honestly, at that point a TL072 is a better pick. More GBW, better slew rate, but slightly higher noise. However, the NJM4556 has a rather interesting "banner spec". It can deliver A LOT of current. It is specified for up +/-70mA, and in practice they'll sink & source double that without a problem. The venerable NE5532 kind of stops around 30mA. The LM4562 is spec'd for about 25mA. Without going to a discrete opamp getting that kind of current is not possible without using external transistors on most opamps. So, these suckers are taking a single ended signal and converting it to a balanced signal again, and they're operating at unity gain and are AC coupled on their output.

And here is the schematic!

Tomorrow I should hopefully have everything finished up and I show off the glory pictures of the unit, but as of right now it is not quite done again about 80%. So, almost there.

-APZX

APZX · « **Reply #3 on:** January 23, 2020, 08:13:47 PM »

And finally the conclusion!

Front

Guts

-APZX

bleakapathy · « **Reply #4 on:** January 25, 2020, 09:15:53 PM »

Clean build man, looks great! One potentially silly question though... This was a VU meter.. where is the metering? I didn't see any led or needle for visual feedback.

APZX · « **Reply #5 on:** January 26, 2020, 07:52:39 AM »

Quote from: bleakapathy on January 25, 2020, 09:15:53 PM

Clean build man, looks great! One potentially silly question though... This was a VU meter.. where is the metering? I didn't see any led or needle for visual feedback.

Not silly. This is a buffer. So, it sits between my monitor controller and my VU Meters.

That is the back of the MC624. My actual VU meters are from Crookwood. So, this buffer sits between the monitor controller and the VU meters and isolates them. This is important because a VU Meter can be thought of as a nominal 600 ohm load (it is very likely to change mind you). And while the OPAx134 series opamps are rated for +/-40ma of short circuit current having them drive a VU meter AND the attenuator in the monitor controller would be extremely taxing and may cause the output waveform of the otherwise excellent OPAx134 not be so excellent (increased THD, noise, etc...).

In other words this entire thing's purpose is to isolate the two devices and add gain

-APZX

bleakapathy · « **Reply #6 on:** January 26, 2020, 08:57:42 AM »

Oooh yep, probably should have gleaned that from the title of the post, that certainly makes sense now. I do love your threads like this pal, so cool to see things come together into a custom piece.

zanepepper · « **Reply #7 on:** January 26, 2020, 04:13:19 PM »

That is so rad! Really impressive.

APZX · « **Reply #8 on:** January 28, 2020, 08:18:23 AM »

For a lot of utility items in a studio things aren't too bad. Like distribution amps or devices such as this. Where things start to go off the deep end is with complicated processing. At some point in the future I want to build a full on console, but I am a long long long long way off from that haha.

-APZX

Author Topic: Designing and Building a VU Meter Buffer (Read 1962 times)

APZX

Designing and Building a VU Meter Buffer

APZX

Re: Designing and Building a VU Meter Buffer

APZX

Re: Designing and Building a VU Meter Buffer

APZX

Re: Designing and Building a VU Meter Buffer

bleakapathy

Re: Designing and Building a VU Meter Buffer

APZX

Re: Designing and Building a VU Meter Buffer

bleakapathy

Re: Designing and Building a VU Meter Buffer

zanepepper

Re: Designing and Building a VU Meter Buffer

APZX

Re: Designing and Building a VU Meter Buffer