Hi Guys
I'm rebuilding an amp I built ages ago into a Two Channel Clean/Distortion amp! I've put together a design and I just want to know if anything looks like it won't work to you guys. I've attached a screenshot of the LTSpice Sim that I've made and the Sim itself. Both channels in the sim give out near enough the same output, although I can obviously measure and adjust this when I build the amp. The chip amp is a TDA2050.
I've made the distortion side near enough like the Marshall JCM800 2203 Pre Amp, mainly because you'll get distortion with it at all levels on the gain knob. The clean is just a two stage pre amp with a fender tonestack. I've biased the FETs using a guide from runoffgroove.com to make them sound more valve like. Thats why they aren't bypassed at all and they don't need to be anyway as there is plenty of distortion.
I should mention that the input will always be connected to both channels, the switching will be done just before the master volume.
Ryan
Not ignoring you; looks interesting, just a bit preoccupied ATM, keen to see how it turns out. :dbtu:
No problem Roly. The only part I'm unsure of is the distortion pre amp. I know I could easily do this with valves but with transistors would I get blocking distortion at all doing this?
No blocking distortion.
It will roughly distort the way tubes do.
A couple details:
1) simulations are fine, but real world Fets are spread all over the map, so in a protoboard setup a 25V supply, 1K and 33K resistors source and drain, gate grounded, and plug there your Fets before actual soldering in the PCB, so you preselect the ones which bias reasonably close.
And also use 1K/33K at the tone stack driver, there's no reason for that 820 ohms resistor there.
Buy 2X what you'll actually use and select.
Or if you think you'll use them later, buy 50 or 100 (no kidding) , measure and separate them in groups.
ROG "avoids" this by using a trimmer at each drain, which is horrible ... and expensive.
2) you need gain 20X or greater at the TDA20xx or it becomes unstable.
If too sensitive you can always attenuate input a little.
Thanks for your reply J M
1) The JFETs that I will be using are J201s, except for the stage before the source follower which is a J202 so thats why it is biased differently. Reason for using a J202 here as it shouldn't distort as quick as a J201. The source follower is a 2N3819 as I have a few of these and thought it'd be a good chance to use one.
I think I've only about 10 at the moment, but I think J201s and J202s are dirt cheap so I'll just order a load off ebay.
I definitely won't use trimmers, not faffing about with that, I'll just experiment like you said. One thing I need to ask though, how do I test the JFETs? I read about people testing them to find out their properties as I know they are very inconsistent but how do they do it? Do they just build a circuit like I've got and test them that way?
2) That could be an issue as currently my gain is about 3......So I've made a few adjustments as per your recommendation. I have made a voltage divider just before the TDA2050 and brought the gain back up to 18. Will this be ok?
I've attached a revised schematic and LTSpice file.
I should add that I may try and use current feedback on the TDA2050 but am struggling to simulate this in LTSpice, so it will have to wait until I start to build the amp.
QuoteDo they just build a circuit like I've got and test them that way?
That's exactly what I suggested, using a free and of a protoboard just for testing, afterwards you disassemble it:
Quoteso in a protoboard setup a 25V supply, 1K and 33K resistors source and drain, gate grounded, and plug there your Fets before actual soldering in the PCB, so you preselect the ones which bias reasonably close.
Then you separate them in little labelled boxes or simply tape them to a piece of cereal box cardboard and write the actual "plate" voltage you got, which with a 25V supply
should be between 10/15V .
If most are either too high or too low, you must choose another value of drain resistor.
One problem with simulators is that they use an idealized transistor model.
The actual part you buy
may be far from it, specially in Fets..
Ah ok I see now. Well my plan of action was to solder the power amp onto the stripboard, then breadboard each pre amp to ensure they work correctly, then I am going to use little 3 pin sockets for the transistors so I can swap them easily in the final test.
One thing I found with the first solid state amp I built is that the attack was VERY harsh. The only cause that I could think of is that I'm using a voltage regulator for the Power Supply and as a result there is no sag. As a result the initial attack of the note being played isn't softened at all and is thus very fast and harsh. Does this sound right?
Transistors are very fast, hence "harsh".
The PSU regulation will be noticed when clipping, not a "clean" levels, so in practice it won't be noticed.
By the way all SS guitar amps have standard unregulated supplies.
What speaker you were using?
Check the Randall RG100 schematic, their "Presence" is actually a "cut" type control, Pres. max is actually flat and min. is cutting highs.
Adapt it to your circuit, rising pot value (they use 2K) and lowering the cap by the same factor.
I guess something 10X higher and lower (say 10K or 22/25K pot and corresponding 5/10X smaller cap) will work well to tame shrillness but be able to restore bite at higher power or in some horrible room.
Or put a rotary switch with various cap values to ground after Tone Control, same thing.
Just played through that amp again and it seems alright compared to before. I think the harsh treble was possibly due to me having it too loud in a small room. The cut control I can't implement as it loads down the tonestack, although I could put a source follower just before the TDA2050 which would help seperate the tonestack from the input of the chip and I could then put the cut control in. Will have a think.
Have to say the amp as it is sounds quite nice clean tone wise, hasn't got a tonestack so has a nice Fender Champ sound. Might remove the tonestack from the clean channel when I rebuild it but we'll see.
Just need to draw out the stripboard and once I've done that I can start ordering parts :dbtu:
QuoteThe cut control I can't implement as it loads down the tonestack,
You can add just 1 humble FET like Randall (RG100ES) does.
In fact, easier and cheaper, you can use almost any NPN small transistor there, same biasing and emitter/source resistor.
I use BC547C (Hfe~500) , search for any NPN with Hfe >200 and you'll be fine.
I'd use 10X higher impedances, so 1K + 2K pot (hard to find) + .1 ceramic would become 10K + 20/25K pot + .01 ceramic .
That Randall RG100 is interesting, nice to see JFETs being used instead of Op Amps.
I'll see what I have for the source follower as I have a collection of JFETs and some BJTs.
Had a quick go on the simulation and it seems the master volume control messes with the cut control. So I'll move the master to before the source follower and it should be fine then.
I also thought of a good idea for squeezing the 6 controls onto the front of the amp. As the tone control for both channels is just a treble I was thinking I could stick them both onto a stereo pot.
Quote from: J M FaheyThe actual part you buy may be far from it, specially in Fets.
While this is certainly true my experience with MPF102's of late is that I can just bang one in to the simulated circuit values and they work fine, the spread in practice today seems to be much closer than the 10:1 on the datasheet.
I remember my early encounters with MPF102/2N3919's where I would by a bag of 10 to get the one I needed; did that a while back and got 10/10 usable. {tighter process control? Or maybe I was just lucky?}
Quote from: LittlewyanI definitely won't use trimmers
Nuthing wrong with a little trim between friends where something might change during the course of the amplifiers' life, such as output bias or offset if the output transistors have to be replaced.
A trimmer can be valid to determine the correct value during construction, then replace with a fixed resistor. I think the main objection to the ROG mess is that the adjustments have been placed in the Drain rather than the Source. To me this is a loud warning bell that the "designer" didn't actually know what they were doing. {"You never find just
one roach in a kitchen"}
Unfortunately I don't have any trim pots for this, however I will be breadboarding the Pre Amps so won't be too difficult to switch out resistor values and I do have a fairly good selection............he says.
Anyway here is the updated schematic and LTSpice file. The cut control works and the master volume works (in the simulation anyway). I chose a 5K Pot for the Cut Control as going up to 10K didn't seem to make any difference but we'll see when I build it for real. What do you guys think?
Edit: Roly its funny you should mention your JFETs being used with simulation values as last time I used JFETs (2N3819 I believe) the source resistor was 500ohms off simulation values. I mean that as in 5.6K in the simulation gave the same result as 4.7K in the actual amp. Maybe things have gotten better.
You forgot the resistor after C24 . (check the Randall control an my earlier post)
Either 1K + 2K pot (5K may work) +.1uF cap OR 10K + 10/20K pot + .01 cap
You missed the R in the RC network.
It ytill works, sort of , (and you needed the larger value cap) just because the Fet follower does have some internal resistance ... but it's a random not well defined value, which will change amp to amp.
Better use predictable parts.
I've just added that resistor in and it made a huge difference. Tried the values you mentioned but they seemed to cut a lot of high end even with the pot on 10. So I've played around and for the moment settled on 1K + 10K Pot + .022uF Cap. But obviously the real test will be once I've built it and hearing how the control affects the amp. Its all fine and dandy looking at simulation results or doing equations but at the end of the day it all comes down to the sound. Thanks for pointing me in the right direction though :). I think the design is pretty much complete, just need to check it all over again and have a think to ensure I haven't missed anything.
I was looking at the Source Follower just before the TDA2050 and do you think I should have it connected to the PSU before or after the 1K Resistor? Reason being if its after the 1K resistor I'm thinking I might run into issues when I use the Distortion Channel as that has it's own 1K Resistor and we might get issues with the two channels interfering with each other.
Edit: Also do you think I'd have issues with interference if I crossed the Power Track with a resistor? So basically the signal path crosses the supply path.
What size transformer would I need for this? I read on another forum that someone recommended a 100VA Transformer with a 20V Secondary. Thats 5 Amps! Sounds a bit extreme to me. I was thinking more 20VA. What do you guys think?
The Quiescent drain current of the TDA2050 is 90mA Max with 25+/- supply. I'm using a 25V DC Single Rail Supply so I imagine it'll be a bit lower than that.
Quote from: LittlewyanSource Follower
...goes into the non-inverting input of the chipamp, so I'd put in another leg of decoupling, a 1k between R5 and J7, and another 100uF to ground.
Similarly I notice that you have three cascaded stage inversions on the same supply rail in the overdrive section. Again I'd be inclined to add another decoupling leg of 1k between R12 and R23, with another 100uF to deck.
Quote from: LittlewyanAlso do you think I'd have issues with interference if I crossed the Power Track with a resistor? So basically the signal path crosses the supply path.
No problem. Your supplies appear to be well bypassed so there shouldn't be a lot of AC on the supply rails, just don't make them too cosy.
Quote from: LittlewyanWhat size transformer would I need for this?
The output stage is 50% efficient, so you need a tranny rated for double the required power output (generally given in Volt-Amps, VA, read "watts"). So a 50 watt amp requires a 100VA transformer stack.
A 25V single supply rail with an 8 ohm load will give;
25Vpeak-peak
12.5Vpk
12.5/1.414 = 8.8Vrms
P = E^2/R
(8.8^2)/8 = 9.7 watts out, requires a 20VA stack.
Voltage
25/1.4=17.9
25/1.1=22.7
So an 18 to 22VAC winding @ about 1 amp.
HTH
Ok I've done as you said and did some more decoupling in the PSU. So basically transistor amps have the same rules as Valve Amps, decouple the supply every 2 gain stages. I didn't bother with much decoupling originally as I saw other manufacturers didn't but then again they were using Op Amps and they were probably skimping on parts.
For the Power Transformer I've bought a 2x18VAC 30VA Toroidal. Next size up from that was a 2x25VAC 100VA Toroidal but I didn't think the very small gain in volume and large gain in heat was worth it.
Managed to find some Transistor sockets on ebay as well, 10 for £2.50. The amp is coming together :D. Just need to draw out the layout of the PCB and then order the PCB making kit.
In an amp using transistors or valves each stage generally inverts the signal. This means that with two cascaded stages any unwanted feedback from the anode/collector of the second stage to the first is negative, so the worst outcome is less overall gain than expected. But when you add a third stage the stage current is now back in phase with the first stage, and there is gain in between.
Positive feedback and gain means what?
So as the decoupling cap on this rail ages and its impedance (ESR) rises the positive feedback will increase until something obvious happens, such as low frequency instability, putt-putt-putt "motorboating". Before that you may have transient or parasitic instability, damped oscillations triggered by signal edges, or depending on certain signal conditions to excite them (making them easy to miss).
You can get lax about bypassing with op-amps because they are designed with a very high power supply rejection, typically about 120dB, which single stages aren't (and really can't be while they are single-ended rather than balanced). In contrast our typical single-ended gain stage will have a supply rejection of only about 6dB.
This can lead to the seriously mistaken idea that op-amp supplies don't need bypassing. They may not need as much bypassing at audio frequencies, but having monumental open loop gain and wide bandwidth they sure as hell need high frequency by-passing, and right across the chip supply pins if you can, but close because we are dealing with very high gains and frequencies in the supersonic to RF that make stray capacitance circuit couplings significant.
Ah ok I see. I always knew you should decouple every two stages but didn't know why :D.
In the notes for the TDA2050 it mentions having separate grounds for the two inputs and the speaker output in case of high current. I'm guessing it is basically telling you to avoid a ground loop and to keep the grounds as far apart as poss?
Well we need to be a little more specific about a couple of things.
It's not just two stages, but two inverting stages. It's when you get the third inverting stage that the troubles start.
In your circuit you have a Source-follower J6 coupled to inverting gain stage J5. This follower does not count as a "stage" because a) its current is already in phase with J5, and b) as a follower it has, by definition, less than unity voltage gain, so J5 and J6 count as a single stage.
The idea behind this split grounding is that the high currents in the output/speaker circuit do not share common traces with grounds in the input side - we are trying to avoid ground resistances in common that would cause voltage drops due to output currents being injected back into the input.
The entire path from the transformer, rectifier, filter caps, output stage, speaker, and back to the tranny via the rectifier again, should be treated as a high current path with thick low resistance wiring (and the power supply loop from tranny through reccy, caps and back to the tranny, in particular; very high impulse charging currents flow around this loop)
Forgot these attaches;
Yahoo Groups LTSpice group is simply stuffed with all sorts of useful models (you just have to figure out the user interface to find them ::) ).
O yes forgot about the potentiometer symbols!
Right I see, so basically make sure the input grounding doesn't get mixed up in the high current grounding by giving it a separate trace back to the star ground?
Yup. :dbtu:
I now have all of the components so I can finally design my board layout. I'm planning on having separate boards for the PSU and Amp, however I'm unsure about the Pre Amp filtering. Do you think the Pre Amp will be ok if I keep all of the filter caps on the PSU Board and just have jumper wires going over to the Amp board? I mean I've done this kind of thing before in a valve amp but unsure about solid state amps.
Also what would be the easiest way to wire up my pots? Solder wire permanently into the board or perhaps have some sort of plug/socket system to make it easy to take pots out?
Bypass caps should really go right across the stage(s) they are bypassing, while the series resistors can go anywhere (generally back in the PSU). The HT and ground lines should radiate from the PSU, not daisy-chain if you want to avoid supply paths in common and unwanted coupling.
Solder the pots in. Unless you are particularly ham-fisted they should only require replacement every 10-50 years, and over that time span any sliding connector will also have a significant failure rate of its own. In general, avoid if possible; these are to facilitate mass production.
Try to fit bypass caps right at the preamp (at least for the first stages) , mainly to avoid having long ground paths which late will drive you crazy.
Thanks guys. I still haven't got round to this yet. Been tinkering with my valve amps lately as I've been using them a lot for band practices and just trying to get the tone right. Hoping to do this project over xmas.