Welcome to Solid State Guitar Amp Forum | DIY Guitar Amplifiers. Please login or sign up.

March 28, 2024, 03:44:15 PM

Login with username, password and session length

Recent Posts

 

My Attempt at a Hybrid Design

Started by Bakeacake08, December 20, 2013, 06:41:39 PM

Previous topic - Next topic

Roly

Quote from: Bakeacake08I'm still not sure what makes you think I'd get away with using the MPF102.

I'm an optimist?  :-[   Okay, I'm trusting to a certain amount of luck, pushing the envelope a bit, but if it's bothering you then start scouring your local supplier catalogues for a FET that has a higher voltage rating.


If you have a volt of signal at the FET Gate and a tonestack with a 20dB insertion loss then the input to the tonestack will be 20dB higher;

dBV = 20 * log10(Av)

where Av is voltage gain, Vout/Vin (negative in this case)

Which resolves to log10(1) or 10 times (loss, i.e. 1/10th).

So there will be 10 volts of signal coming out of the last preamp valve for 1V at the FET.

Now the gotcha here is that with full treble or bass boost the tonestack loss is closer to zero (because the passive network can only get "boost" by having less loss).

Why I think you will get away with the lower voltage FET is because, as you say, with a volt in, the output will be fully driven, that is at the onset of clipping, and any more than this will start to give you overload shred (and such clipping generally sounds better done by diodes in the preamp, e.g. anti-parallel diodes across the signal path after the tonestack).  Considering the FET is only about a buck I'd risk it, but you're the designer here, it's your call.

How much gain you are going to get out of each of the valve stages on reduced HT is anybody's guess, but under normal conditions it would be somewhere between x30 and x50 per section.

Clipping your DMM to your guitar output and hammering it should show you something, but a CRO is normally a better method.  My normal rule of thumb is that quiet strumming gives about 100mV and Death Metal pounding a maximum of about a volt, so somewhere in that range.


Output current.

Quote from: Rolyyou only have a single 35VDC supply.  35Vp-p = 17.5Vpk --> 17.5 / root(2) = 12.4Vrms.

So the output voltage is about 12VRMS.

Assuming a nominal 8 ohm speaker then it's simply;

I = E / R  = 12.4/8 = 1.55amps

So your output coupling cap has to cope with a maximum ripple current of around 1.5 amps.  Ripple current ratings are often not quoted in catalogues, so you have to use those you can find as a guide, go for a low ESR cap, and use the physically largest one you can find.  Looking through a local catalogue that does quote, 470uF/35V is 1400mA.  Ripple current goes up with capacity and down with voltage, i.e. 1000uF/35V are about double while 470uF/63V are about half.  For the same value and voltage a physically larger cap will generally have a higher ripple current rating.

Note that current goes around the circuit, so the main PSU filter cap has to deal with this as well.

Theoretically you can get away with a voltage rating half the supply voltage, but if you have an amp failure that shorts high then the whole supply will be across your output cap.  If this then fails the whole supply will be across your speaker; so my practice is to use a cap rated for the supply voltage.  It's a bit of a vague point about soft v. catastrophic failure - just the chip, or chip then cap then speaker.

Quote from: Bakeacake08I just need to explain to my wife...

Ah ... that's not an electronic problem, and most of us have to deal with that as best we can.  Collect some data on similar commercial gear and show her the prices, then point out how cheap your bits are.  If she knows anything about dressmaking she should get the point; you are building Dior Haute Couture for a sub-Target price.


There comes a point with a design when it is mostly resolved that more head scratching and calculation runs into the Law of Diminishing Returns, and that an actual prototype build will throw up a lot more practical questions to be resolved.  It can be a hard point to pick, and the temptation is to go around the paper design just one more time, but if you want to keep the design moving forwards then a build is in order, and it's very rare that this first build is a total waste, often only needing a few mods to get you there.
If you say theory and practice don't agree you haven't applied enough theory.

Roly

{dwngs+title block, version control/sortable date}

See appended; one 12AX7 section 36V supply

32dB gain

dB = 20 * log10(Av)

32/20 = log10(Av)

32/20 = 1.6

10^1.6 = 39.8 or about x40

This is more than I thought and we will see how that triode model matches reality.

I'd reduce the cathode resistor to centre the anode voltage on the supply, ~18V, that way you will get some clean headroom, but I suspect that even the first stage will overload easily.  It that's what you want, fine, but if you want some clean too, that could take some tweeking.

The value of the grid resistor, 1M - 2.7M, has a small influence on the Q-point, the steady state bias point.


We still need an updated draft of the cct with the power supply at the output end, and some bypassing in the supply rail between the power amp and preamp (lest it start howling in it's own ear).
If you say theory and practice don't agree you haven't applied enough theory.

Bakeacake08

Quote from: Bakeacake08However, I'm still not sure what makes you think I'd get away with using the MPF102.
I hope this didn't come across as too confrontational and/or abrasive. My son was waking up from a nap, so I didn't take the time to proof read. What I was trying to say was really that I don't really understand how JFET parameters work, or how they break down. I think I get it with BJTs, but I've never used FETs before, so they're still new to me.

Quote from: Rolybut a CRO is normally a better method.
Yeah, I definitely have one of those on my wish list.  :) I will play around with my DMM until I can find one.


I've attached an updated schematic. I referenced one side of the heater circuit to ground, which I think is what you were explaining I needed to do. I can also get a 6.3V transformer instead and ground the center, if that's better (I remember reading about that somewhere on the internet awhile back). I moved the connection to the power supply to the power amp end and put some filter caps in front of the pre-amp stage. What is the benefit of putting the power supply here? My gut says it has something to do with preventing interference, but I'm not sure. Also, would there be any drawbacks to wiring each section in parallel? That is to say, what if I ran a line from each section straight to the power supply output?


I really appreciate your help with this design. It might be awhile before I actually get to build it as we're looking for a bigger place right now (happily gave up my tinker area for the new baby), but hopefully it gets done sooner than later and I can get on to the next project.  :)

Roly

Quote from: Bakeacake08I hope this didn't come across as too confrontational and/or abrasive.

No;
a) it's a very fair question when I am taking a risk with your posterior;

b) as I would tell my trainees, electronics (design or repair) isn't about being right, it's about getting it right.  Often these can become quite heated differences of opinion.  Two effective ways of resolving such differences are mathematics/modeling, and just bloody build it and measure it.

I thought that 12AX7's on 30-odd volts wouldn't have a lot of gain - they certainly won't have a lot of headroom*, but LTSpice tells me I was wrong about the gain (but are its models still valid at such low anode voltages?).  Okay, now for a build with a real 12AX7 to truly settle the question.

(* and in a mains powered amp it would be almost trivial to run them from 2-300V; flick the FET and stick in a second 12AX7 cathode-follower, and some other possibilities open...)

CRO
Far from ideal, but better than nothing and free - there are a few freeware "CRO's" that will turn a laptop into better than nothing, provided that you are well aware of the hardware connection problems (protection) and the limitation of most of these "CRO" programmes.

Visual Analyser v8
http://www.sillanumsoft.org/download.htm


Heater and CT
Quote from: Bakeacake08I referenced one side of the heater circuit to ground
...
I can also get a 6.3V transformer instead and ground the center, if that's better

The object of a center-tapped heater supply is to end up with a virtual earth point in the middle of the heater inside the cathode, so the signals into the capacitive coupling from each end of the heater to the (effectively) grounded) cathode are in opposite phases and thus cancel out.

This can be done with an actual CT on the winding, a phantom CT "hum-dinger" consisting of two 100 ohm resistors across the winding with their mid-point to ground, or a 100 ohm w.w. pot with its wiper grounded, giving an adjustable "hum-dinger" (normally only found in old up-market valve amps like Leak's.

This can only make a difference where the two twin-triode heaters are wired in parallel for 6.3VAC.

As a sidebar, taking the heater "ground" point to a +ve voltage can also have a beneficial effect on hum and noise (in your case simply to B+ instead of ground).


Quote from: Bakeacake08I moved the connection to the power supply to the power amp end ... What is the benefit of putting the power supply here?

The power supply is really part of the output stage - what it does for the preamp is almost incidental.

By-passing/De-coupling
When a signal goes through any stage it causes changes in the stage current, and these currents come together in the ground and supply rails.  If a circuit, such as a Class-A preamp, happens to be signal sensitive to changes in the supply rail (and op-amps aside, they mostly are), these changes will get injected into the signal path.

But these changes themselves are due to what is coming down the signal path into the output stage, so we have a potential signal loop.  If these signals are out of phase then they will tend to cancel and reduce available gain.  If they are in-phase they will increase gain and introduce the dreaded "instability" - oscillation, normally at very low frequencies, but can be anywhere.

If you turn on an amp and it starts puttering like an old motorboat then you go looking at the electrolytic decoupling caps on the supply rails because it's a fair bet that one or more have "dried out" lost capacitance, and no longer "bypass" low frequency signals effectively from the supply rail to ground.  They are supposed to tie the supply rail(s) to ground for AC, and they have stopped doing that.

There is something still missing from your bypassing - the series resistor.  A cap can only drop against the source impedance, and the raw supply is a very low impedance, meaning that you would need a very large cap to effectively form a low pass network (DC, but no AC getting past) with it.

So we inject a smallish resistance in series with the preamp supply so the preamp bypass cap has something to "work against", a highish source impedance, and a much lower value of cap will still have a very low reactance at infra-sonic frequencies - compared to the series resistor.

Since the preamp doesn't draw much current we can afford to make this resistor quite large before the preamp supply starts to drop too much.

So we need a resistor in the HT feed to the two preamp triodes.

I'll leave it up to you, but I'll grab a value out of the air to illustrate how it goes.

2K2 resistor, preamp bypass cap reduced to 100uF.

Preamp current, less than 1mA, so the drop across the 2K2 will be less than two volts, supply 35V, preamp 33V, which is fine.

What is the effective AC resistance of a 100uF cap at, say 10Hz?

Xc = 1/2 Pi f C

1/(2 * Pi * 10 * 100*10^-6) = 159 ohms.

Supply attenuation is then;

159/(2200+159) = 0.06740144, or a reduction to about about 5%, which is pretty good.

Loop with different values/assumptions to taste.

The resistor can go just about anywhere it will fit, but the cap should be as local to the stage(s) it bypasses as possible (thus the loop formed by the cap for the signals in that stage is mall and they get the shortest path "home").


{Now you understand why Nicola Tesla never got married...



NICOLI!  TURN THAT DAMN THING OFF AND COME TO BED!   :duh
}

If you say theory and practice don't agree you haven't applied enough theory.

Bakeacake08

Design time has been in short supply lately, but I have a couple hours to kill before picking up a rental truck (moving today--hooray!), so I've been playing around with Eagle. I wasn't planning on getting into PCB construction, but now that I am, it seems like it's going to be pretty fun. I wanted to separate the tone section so that I can switch it out easily if I want to, but here is my design for the power amp section. The "INPUT" connection on the left will come from the output of the tone stack. VCC and GND will connect to the power supply board. Q1 is based on the MPF102 layout. From top to bottom the pins are: Drain - Source - Gate. I used the part numbers from the TDA2050 datasheet to make it easier to route everything. I'll update them when I finalize my schematic.


I'll have a nice-sized garage in our new place, which means plenty of tinkering room. This amp might just get built yet!

teemuk

#20
Think for a while about currents that those traces will be carrying... You seriously need to consider thickening up some of those traces.

I like how the ground currents properly return so that higher currents do not have to return through stages with lower currents and much more sensitive inputs. Audio and power ground returns also seem to have fairly good separation. Local filtering caps are also nicely close to terminals of the IC but their layout and "noding" might need a few tweaks.

1st, that (probably 100nF or close) filtering cap (C5 I presume) should be as close to VCC terminal of that IC as possible. It's job is to act as local filter cap that counteracts the tiny tiny inductance of the power supply wiring. Placed too far away from IC's terminals its effect to do this becomes negligible. It must be placed close so it can do its purpose properly. The higher capacitance filtering cap can be further away but minimising length of power supply wiring (including traces on circuit board) is always a good practice. Now let me remind you about that adequate trace thickness. Several amperes of current will flow in some of those traces!
2nd, it's not really good "noding" to "T" capacitors. If you think about it, you want charging currents to flow to the terminals without much interference with currents pulled by the load. If you "T" the traces going to the capacitor's terminal both currents are forced to share their paths. If you just wire everything directly to the terminal, without that little "sideroad" to terminals, the current loops will have much less paths of interference. The same logic naturally applies to both terminals, positive and negative. Remember: current flows are loops.

Overall, I wouldn't etch that PC board just yet. Give it a few days of further thinking. You'll have much better success in building a flawlessly operating amplifier.

Bakeacake08

Thank you for your reply. I am brand new to PCB design, so I really appreciate your help. I understand what you're saying about "T-ing" capacitors. Now that I notice it, it intuitively makes sense; I just didn't know to look for it. (And to be sure, I think what you're talking about is how I connected the positive node on C3 to VCC. I should have just gone straight through it.)

As far as trace widths, I did thicken them up from how they originally routed. Not on any scientific basis, more just because they looked like they should be thicker. Looking at it now, I see that I could easily double--or maybe even tripple--the width of all the traces, with a little bit of component movement, with the exception of that spot where the VCC trace passes under the IC, which much be smaller to pass between the pads. How big of a problem will it be to make it thinner at this spot? In other words, would it make sense/be worth the effort to add in a jumper wire or something of that nature?

Also, a tutorial I followed said I should fill in the blank space with a pad connected to GND to help avoid interference (of some sort--RF maybe?). I believe the ground pin on the IC is connected to the heat sink, so it makes sense it would help with that too. Is there any sort of theory behind this that I should be thinking about when adding ground planes (or power planes)?

Don't worry that I will rush into production. I knew this was my first draft, so I'm mostly just pleased that the first comments made about it seem so minor (as opposed to, "Why on earth would anyone do that?").

teemuk

#22
Quotespot where the VCC trace passes under the IC, which much be smaller to pass between the pads. How big of a problem will it be to make it thinner at this spot?

Think of the currents involved.
The VCC trace that passes under the IC needs to withstand only the current draw of Q1 and its biasing circuit. That is few milliamperes at best.
The VCC traces that supply the power amp IC, and traces that run from TDA2050 output to speaker (return path of speaker's "ground" current included) on the other hand must carry currents in excess of 3 or 4 amperes, maybe more! Remember that internal limit of that chip is as high as 5 amperes.

You don't need to thicken all traces, just those ones that count. Basically the higher current the trace carries the thicker it must be. Thin trace = resistance, you don't want to mix that with high currents because the way that resistance dissipates its waste power is by heating. I'm sure a few minutes of google searching will find equations to calculate proper wire diameters / trace widths when currents are known. You probably don't have to care about that sort of stuff with the low current buffer in front, but in power amps that start to handle very, very high currents this becomes a very important design aspect. For same reason also proper routing of currents becomes important: you don't want amperes of transient current spikes return through the reference point of small signal stages.

Anyway, thickening all traces usually just introduces more drawbacks. Low current stages work well with thinner traces and thinner traces allow keeping more distance to traces that you don't want to "crosstalk". Basically every trace or wire running in parallel with another forms a tiny, tiny capacitor. Especially with high impedance circuitry ( >100K ) even very low capacitances start to mean a great deal. Picture a trace carrying the signal running a long way in parallel to ground trace, the capacitance might cause attenuation of higher frequencies. You can use this effect for your benefit too: Power supply traces benefit from running as close to their corresponding ground return traces.

Also, every trace and wire has inductive characteristics as well as a little magnetic field around it. Proper trace distance also helps eliminate interference from such in delicate circuits where tiny details like this begin to matter in utmost importance. For example, when frequencies involved rise above several hundreds of kilohertz's the series and parallel resistances, inductances and capacitances of the board design must be taken into consideration as real electronic components. Thank goodness guitar amp design doesn't focus much above 10 kHz.

J M Fahey

Agree and add:
besides current capability, given that will be a home made PCB try to keep tracks at no less than 30 mils wide, "necking"down to 20 mil to pass between relatively close pads, purely on ease of fabrication grounds.

Also increase pad size (which I bet is preset 62 mils) to at least 80 mils and wherever possible 100 mils.

Otherwise they will look good when printed but after the hole is made, you will only hace a thin copper ring around it.

To boot, home drilled boards are usually not that exact, in practice a few holes will be off center.
A tiny amount, but enough to leave little or no copper on one side.

That said, congratulations on your good job :)

You got the most difficult part, "connectivity".

Now you can tweak it, increase track and pad thickness where possible, put tracks parallel to some edge when possible for visual neatness, maybe reorient or move some part, etc.

:dbtu: :dbtu: :dbtu:

Bakeacake08

Still slow going on the design process, but here is my latest update to my power amp board. I increased most the track sizes and I figured out how to increase pad sizes. I also moved the 100nF cap closer to the IC.


In other news, we just moved into a house from our old apartment. It had some pretty fun electrical properties when we moved in. For instance, one of the light switch cover plates was loose, so I pushed on the screw thinking, "I should tighten this," when my finger started buzzing. Pulled out my trusty voltmeter and sure enough, there was 120V hooked up directly to the screw. The guy who swapped everything out said that the live and neutral wires were backwards on an outlet in the baby's room. Fun times.


Anyway, eventually I will have my workbench cleared off and the fun begins.

J M Fahey

This one looks much better.
Having made many, this one is "printable", "etchable", etc.

Bakeacake08

Fantastic! What do you recommend for drilling the holes? I'll be working with a hand drill (no ETA yet on when I'll be able to procure a drill press), and the smallest bit I have is 1/16". I've read that smaller than that might tend to break if not used in a press. I've also seen suggestions of using small dremel bit. Given your experience, should I worry about finding something smaller, or is 1/16" (which, if I've done the math correctly, works out to 62.5 mils) plenty big?

teemuk

inch = 25.4mm

Typical component lead diameters are 0.63 mm, 0.7 mm and 0.8 mm (for ordinary components, 0.7 mm being most common), 1 mm (high power resistors and transistors) and 1.5 mm (trimmers, potentiometers etc.). Usually the biggest component leads are rectangular instead of round.

So basically you need a set of different sizes of bits. Use the one matching closest to component lead diameter.

And yes, they break very easily without a drill press. In fact, they break so easily that I wouldn't even dream of drilling the PC board without one.