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

March 28, 2024, 06:14:27 PM

Login with username, password and session length

Recent Posts

 

Single supply HiZ preamp to 1 Vrms out, HELP!!

Started by mickmad, March 30, 2013, 09:32:06 AM

Previous topic - Next topic

mickmad

#15
I'm not doing a cab simulation, I'm doing some simulations of a guitar preamp to 1 Vrms, a line preamp (reducer) from 2 Vrms to 1 Vrms, and a line driver with low pass and high pass filter. Actually I'm working on the last one, I'm finetuning the filter to achieve an flat response filter like a C-weight filter, or something like that, cutting frequencies below 10 Hz and above 22KH, with a 2x gain in the passband region; I'm testing different solutions right now, as you can see in the posts :)

edit: after hours of twiddling, studying, changing filter topology, smoothing Q, and banging my head on a wall, I finally got it! I won't be touching this output filter anymore: first, I'm connecting the DAC output to an input buffer, which biases the signal and cuts some low frequencies; the biased hi-passed signal is fed into a unity gain Sallen-Key lo-pass, its output fed to a passive lo-pass; then another Sallen-Key lo-pass, identical to the first one, gets the first stage out and smooths it again, while boosting up the signal with a 2X gain, its output fed to a passive hi-pass; with this topology, I got a really nice bandpass, with a steep cutoff below 10 Hz and over ~22KHz. Again, I'm uploading schematic, .asc file, and Bode plot of input,first stage output ,and second stage output, which clearly shows the super steep cutoff. NEAT!

phatt

Well I looked up Weighted filters and I understand now what that means ,,,,,,
But what you just posted has gone way way past those ideas???

So what you have created is in fact the response curve of some Famous Amplifier. xP 8|

True Marshall Amp output including cab response is a 50~60 Db cut from ~100 Hz down to 10 Hz. You only show a 20 Db drop.  :o  So you still have a long way to go.

The top end roll off looks like it will work,, if that is the aim?
Sorry but I'm still confused as to what it is you are actually trying to achieve?
Phil.

mickmad

#17
@Phil: if it resembles a Marshall Amp output, then I'm on a good road! :D

By the way, the output of that filter is going to be fed in a standard Line receiver, like that of a mixer input, or of an active monitor speaker.
But since the input signal of this filter/driver is an analog conversion of a digital signal, it could encounter some really harsh noise due to D/A conversion around the sampling frequency of the D/A converter, which ranges from 44KHz to 192KHz.
That's why I'm cutting a lot of high frequencies (I got -80dB/decade after 10KHz); I preferred to cut not too much around the low frequencies (I got 5 Hz at around -20dB) because I want to leave more room to the basses; if a potential user wants less basses, then he could cut them away.

In short terms, I want to achieve a somewhat powerful bass response, flat response over audible range, and total removal of very high frequencies components due to digital noise in the circuit. Hope that helps to understand better ;)

edit: I added a final gain reducer stage to the line driver; this way I can control the output volume with a potentiometer ;) I will post the revisioned schematic when I finish the preamp section, so to make a single post with all the nice stuff :D

Kaz Kylheku

Hi MickMad,

Have a look at the JFET input stage of the ADA MP-1. It's a neat circuit which places an N-channel JFET with a PNP transistor in a kind of hybrid complementary pair, which is then used as a source follower (unity gain: it's not a gain stage!)  The JFET provides the high input impedance, and the PNP delivers the current, allowing for a fuller voltage swing. The circuit runs on dual 15/15 rails.

I'm mentioning it because of the LTSpice appearance in this thread, and that circuit happens to be something I've been simulating in LTSpice recently. (The transistor part values are not the real ones.)

And, hey, use good parts in this USB interface! You will never beat the M-Audio or Behringer and whatever *s!!t* if you do what they do: use *s!!t* parts. 

I found JRC4558's in the preamp section when I opened a Tascam US-122L audio interface. Ugh!

Also read this document: http://www.rane.com/note151.html    And google "pin 1 problem".


   
   
ADA MP-1 Mailing ListMusic DIY Mailing List
http://www.kylheku.com/mp1http://www.kylheku.com/diy

mickmad

#19
WARNING: NECESSARY WALL OF TEXT AHEAD!

Thanks Kaz for your help, the preamp section is something that it's still driving me nuts, along with the supply problem... first of all, I'd like to clarify something more technical about this audio interface.

I'm designing this whole thing around two Wolfson chips, namely the WM8786 ADC and the WM8740 DAC, which are 16-24 bit @ 192KHz max. converters. They both have amazing specs, with 111 dB of SNR and -102 dB THD for the ADC, and 120 dB of SNR and -104 dB of THD for the DAC; also, both have differential input/output! I will interface both chips to the I2S bus of a microcontroller, the Freescale Kinetis K20 ARM Cortex-M4.

At this point of development, though, I decided to lay down a prototype board which will have a PSU, a pre section, a post section, an ADC section and a DAC section. The pre and post section are going to be, just for the sake of prototyping, the basic one that are suggested on the datasheets of the chips.

I need to filter both input and output signals, and this is done via active filtering with ultra low distortion opamps. Wolfson reccomends the MC33078 opamp for the filter section, with corner frequencies of 12.5 Hz and ~1MHz, roughly a 1MHz bandpass on both in and out. Other signal filtering is done digitally inside the chips, so that's nice.

The output filter section (the post) also has an AD797 as a differential to single-ended converter, for ultra-quality unbalanced line output.

I've seen the page you linked, and that's the subject of my sleepless nights during the whole last week! Mixed signals circuits are a pain in the a** to layout properly. And the fact that I may want to selectively test the ADC or the DAC looks like a problem; I'm currently aiming the prototype design to be as modular as possible. I want separation between ADC, DAC, preamp, postamp, and psu sections, so that I can debug any part of the whole thing alone. But this means that I have to split the pcb in multiple pcbs, which means multiple, separate, ground planes; which means possible EF antennas could create if I bridge some parts of the circuit over separate ground planes.... UGH!! My brain hurts!

In fact, I still got to finish the whole design as I want a nice, all-around, preamp, with selectable input impedance, and selectable, input-impedance dependent, gain control; I thought of something like a simple inverting buffer, with a switch to select between 1M input impedance and x20 gain, or 10K input impedance and x1 gain; this way I could plug a guitar and give it some good gain, or plug in a simple line output with no gain over it. I also want an attenuator stage; so my idea was in reality a fixed, selectable gain stage, with an attenuation stage. I also want to keep the gain-attenuation controls separate for each channel. The fact that I'm working with a balanced input ADC doesn't help either, as I might as well need to duplicate this circuit to make a balanced, fixed gain stage and attenuator... but this means that I will be using stereo pot for each channel, and big switches. This is more pain in my brain. And remember that I need to filter the input signal, so this gain-attenuation stage will and must be followed by the specified active filter. Considering the ideal setup, with balanced or unbalanced input, gain stage, attenuation, and filter, I should need 2 opamps for the gain-attenuation stage, and 2 opamps for the filter stage, for each channel; using the MC33078, which is a dual opamp on a single IC, I should need 4 ICs just for the input section. Add a pair of stereo potentiometers and a pair of something like QPDT switches (quad pole to dual throw, I don't even think they exist!), the preamp section easily became power demanding and expensive.

The output stage is somewhat simpler,; altough there is no analog gain control over it, I might use the DAC internal 256 levels attenuation circuit; I thought something like: read value with microcontroller from a random pot, map it to 8 bit and then pass it into the DAC. Simpler than it looks.  As I said, the post has already been designed as a balanced-unbalanced output, with an AD797 as a differential receiver to single ended for the unbalanced out. As for the preamp, a channel of the post section needs a dual opamp for the active balanced filtering, and a single opamp for the unbalanced output. So that's another 4 ICs.

Total: 8 opamp ICs for pre and post, two of which cost more than 7 € each (the AD797). Luckily, the MC33078 costs only ~0.5 € each, and for its specs that's way cheap, but then comes the cost of the switch and the stereo pots, which will easily outstand the cost of the opamps to which they would be connected to. Moreover, the supply I've designed, which is based around a Murata switching DC-DC converter module, with 12 V @ 0.5 A input -> +/-9V output @ +/-111mA, could not bear the whole circuit current requirements, as I have to get 5 V from it to power the microcontroller and the converters; considering that the microncontroller will consume ~30mA in idle mode, max 100mA when everything on it is used, and adding ~55 mA to power both converters, and adding the current drop from the 5V regulator and the 3.3 V regulator that is aside the microcontroller and that will drive the digital section of the converters... well it looks like I'm well out of power.

But let's forget price and power, for a moment. The main, real, important, and difficult problem is : NOISE. There's going to be huge amounts of noise into the converters, even if I suddenly become the master designer of mixed signals; I've made my math this night, and from the specs given by Wolfson, at least for the ADC, the SNR is 111 dB, with a 0dB signal of 2 Vrms. It means that the ratio of the 0dB signal amplitude in Vrms and the noise amplitude is around 10^(111/20) = 345813.389etc ; so the actual input noise to the ADC that is reckognized as zero signal is 2 V rms / 345813 = 5.63676 uVrms (u = micro). BUT, the thermal noise of a 1MOhm resistor, at 25 Celsius (which is the same temperature condition of the datasheet specs), over a bandwidth of 96Khz (assuming that the converter is working at 192KHz gives us a maximum input frequency of Fs/2 = 96K) is something like 12.5 uVrms; around 2.5 times higher than the zero input signal. Given that the usable range in Volts for the converter is ~ 5.65 Volts, the converter has an error of 1 bit if there's a difference in Volts of 5.65 / (2^24 - 1) = 0.337 uVolts. Given also that the peak to peak value of the 1MOhm noise is 35.52 uVolts, and removing the 5.63 uVolts of unreckognizable signal, it gives us around 30uVolts of error; dividing this value by the 1 bit error in volts gives us 89, which is the number that will be represented in bits when the noise of the resistor is converted. That's more than 7 bits of error! How can I possibly have a 1MOhm resistor in this circuit at all?!

I'm going so nuts about this thing, I haven't slept for more than 4 hours a day for a week straight, and then when I think that my design could be somewhat good, it is totally not! I really don't see that much need of using 24 bits converters in a circuit that easily obscures ~8 bits with white noise! I wanna use almost every bit out of it! I want it to perform! That's why I'm going to write an email to my electronics teacher and see if he can clarify something. I was thinking that I could plug in an instrumentation amplifier, as seen in all those electronic books, and use the uber high input impedance of the opamp alone. Don't know much, though. I might have been saying loads of bulls*** all the time.

The fact is: I'm 2 exams due to my graduation, the exams are in the next few weeks, my graduation admission deadline is September 24th, I still got to study for both exams (Microeconomics and Complex Calculus, btw), and I still got no design ready to send in production. Consider that I'm in a tight economic situation, so I need to use cheap chinese PCB fabbers, and the min. estimated order-to-delivery time is around 2 weeks (paying extra bucks to UPS); also, consider that I'm designing this thing in SMD technology, and I alone will be soldering it; give in that I still have to exercise on SMD soldering techinques (though those are overlymystified)... I think this is going to be a nightmare.


Link to the datasheets:
http://www.wolfsonmicro.com/documents/uploads/data_sheets/en/WM8786.pdf
http://www.wolfsonmicro.com/documents/uploads/data_sheets/en/WM8740.pdf

Btw: sorry for the wall o' text, but these questions are really driving me crazy and I could not explain my point without a deep insight of it.

Roly

re: grounding - actually with mixed mode circuits it's normal to have two grounds, an analogue and a digital, but the main thing you need to avoid is impressing digital switching current signals on the analogue ground, the digital ground will take care of itself as long as the supply is well bypassed.

re: setting the DAC internal 256 levels attenuation circuit - the problem with doing this from a pot is the possibility of dither between two values, and it would be better to use a two-phase roto-pulser or even up/down buttons.

You do not need a bandwidth to 96kHz.
  Choose something much more reasonable like 20-25kHz and recalculate your noise.

Yes, you have considerable excess S/N and conversion depth, and you are discovering why mic channels are nominally 600 ohms (1k4 in actuality).  Have a look at bootstrapping; this allows you to have a very high effective input resistance while using a quite low actual resistance, and hence low thermal noise.

"You can have it good, cheap, or quick, but you can't have it all three."

You are wedged between a tight timeline, tight economics, and a 24bit converter with, what, 111dB of dynamic range?  Frankly 24 bits is galloping overkill when CD's don't even use the full 14 bit resolution available - the bottom two bits are basically crud.

60dB is good.  80dB is very good.  100dB is really outstanding.  110dB is fugging MYTHIC - 1 part in 300,000; this is the span from the threshold of hearing to the threshold of pain, a feather landing to a jackhammer at 1 metre.  It's a great ideal but well equipped labs with lots of resources and time have difficulty with this range, and you need to get real about your situation.

TIME is your real enemy here and you have to get serious about what you can realistically deliver in the available time - money you can borrow, specs you can compromise, but you have to get something with development potential running by deadline.

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

mickmad

Roly, you made the point. TIME is the real "threat" here; the thing that worries me most is the USB audio class firmware implementation; luckily, I've had some good news from my electronics teacher: he said that the prototype construction, from pcb to actual population of it, should be taken care by the university, along with the costs associated; so I can try to achieve a very good quality product, with the advice of another teacher of my university., who is involved in mixed signals circuits :) also, it will leave me more time to actually debug the hardware and the software, since the device will be made just outside Rome I will avoid huge delays caused by ordering the pcb in China and then having to populate it by myself :) I will keep you up with this project ;)

Roly

That you can get some skilled help is a real bonus, BUT, given the tight time a full spec result is a big ask and you need to think of what part of the spec you can relax to make sure you get something over the line.

I've seen projects like this before where there was a refusal to compromise any part for a result, and the result was no result at all, no part of the project going properly, and a fail.

Think "proof of concept", and that you can debug the digital noise floor out of your 24 bit analogue for a real world product after you demonstrate it basically working, albeit not to full spec.
If you say theory and practice don't agree you haven't applied enough theory.

mickmad

I surely will; first of all, I'm giving up on designing an over-the-top power supply, and will stick with using an ATX power supply, which already has +/- 12 V for the amps, and +5/+3.3 V for the rest of the board. Second, I will stick with the datasheets' input and output filter circuits, which are obviously tested to be within the converters' specs. I know that sticking with those filters will limit me to balanced only line input, and balanced/unbalanced line output, with no analog "volume" control for either part. And of course the input will not accept a guitar; but still, if the whole thing works as it should, then the problem reduces to adding a guitar/unbalanced frontend for the input filter, and an attenuation frontend to the output filter.