How do I do it? I'm stumped (because I don't understand analogue signals at all).

I have a 0V to 3.3V analogue signal that I would like to convert to an audio line level -1V to 1V signal.
This thing above. It's just a 600Hz saw tooth generated by feeding an 8bit counter into a MAX5102 DAC. Let's say yellow is the right channel. And pink is the left channel outputting the inverse. Green is 0V.

I have an op-amp and an inverting charge pump and I think I should be able to use them level shift the 3.3V signal to at least -1.5V/1.5V. Optionally to also scale it down to two thirds.

Is an op-amp the way to go? I can run it at +ve VDD -ve GND using the charge pump but all I've succeeded in doing is clipping the signal rather than offsetting it into the negative. I've searched the internet for op-amp tutorials and calculators but none have taken a -ve virtual ground into account. And I really do not understand op-amps.

How do I do this? Any help would be most appreciated thanks!

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

He does actually have sawtooth waveforms there on channels 1 and 2 (yellow and purple). One is just the reverse of the other.


Here is sort of what you need:

_________________
Bill

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

They are there, just difficult to se. The channel that is active in the control group is the purple one, so it's going to be "on top". Unfortunately it's much harder to see than the yellow.

Edit: It would show a lot better if he did a proper screen capture too.


Good idea.


Of course! But I did say "sort of"

_________________
Bill

That's perfect thanks! In the end I used a 0.1μF MLCC cap (because it's the smallest I have but I suspect smaller would be fine [EDIT] - a 0.01μF MLCC cap was not fine). And I also needed a 1MΩ bleed resistor, anything smaller caused the signal to 'warp'.

I changed the DACs reference voltage from 3.3V to 2.2V and that gave me a suitably scaled analogue output signal.

However plugging the output into my amplifiers Line In dragged it's voltage to zero. I tried a bigger DC-blocking cap and another MΩ resistor in series after it but that just gave me a rounded signal that still went to zero when plugged into the amp.

In the end I brought the op-amp back again and used it to buffer my signal giving me a final schematic:
and a signal (into the speaker amplifier) of:The blue and green traces are the level shifted outputs and seem to remain pretty stably centered around 0V.

Thanks BillO! In the end I just poked roughly valued components in until I got a decent signal
(Oh, the oscilloscope is not drawing the vertical saw tooth line nicely because I left the scope in 'dotty persistent' mode rather than line mode.)

And no post is complete without the messy breadboard, the MAX5102 is hidden behind the mass of jumpers on the right

Andrew, your 'signal dragged down to zero' suggests a significant impedance mismatch. Generally audio signals are fed into a circuit with a defined but fairly low impedance - 600 ohms is standard in broadcasting and probably throughout the recording industry - and so need to be delivered from an output with a very low impedance. An op-amp delivers that low-impedance output that you need.

Just to be different, 'DIN' connectors on domestic audio are (were?) usually defined as much higher input impedances; tens of k ohms. These are less sensitive to the output impedance of the driving signal.

Neil

I'd guess so? I know very little about impedance. I don't know what the output impedance is to match it to the 600Ω input impedance. The datasheet mentions a 10KΩ in parallel with a 100pF load but I think that's just to work out a standard DAC settling time.

With that said my contribution to this is a link to an

ONLINE SMITH CHART TOOL

that can be run offline by following instructions at the

EEVblog Electronics Community Forum

For now I've got no tools, or even variable capacitors or inductors, that would help me troubleshoot this so it shall have to remain an op-amp.

Impedance matching is definitely something I'm going to have to learn so I'm glad you've let me know.

Cheers,
Andrew!

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

Thank you muchly. I think this is it. I want to drive any standard audio line-in device.

I've made the changes you suggested and if using an op-amp is not a bad way to deal with the impedance matching issue then I think I'm done. Both the op-amp and the charge pump are so cheap that, even if overkill, I don't mind using them. If not right then any more suggestions would be most appreciated.

At a slightly higher level what I'm trying to do is create a generally useful digital sound device. Something that could be easily interfaced with anything that can output an 8bit signal.

To do that it's almost as simple as four 74HCT40105s sitting on the data bus*. Their outputs are ticked at a constant 16KHz (for 16Khz 8bit stereo sound) and fed into the MAX5102. In my case I have a 2KHz NMI derived from the same 32KHz clock that ticks the '40105s which I use as a general watchdog / OS timer thing. As long as I can feed two times 8 bytes into the '40105s every NMI then... sound.
I've marked the changes I made in red but I forget I also added a bypass cap between the MAX5102 REF and GND (suggested by the datasheet).

That's intentionally a VHC flip-flop. At KHz speeds it should handle 3.3V VCC just fine - and I have a few knocking around that I picked up for something like a US cent each.

*Address decoding sold separately

[EDIT - I just realised the 10μF cap between -5V and ground is backwards. Ground is relatively positive here.]

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

A difference between UK/Euro and US pro audio standards: we usually defined 0dBm as 775mV into 600 ohms as zero level. A broadcast PPM (signal level meter) had zero level at an indicated 4 on the meter; other marks were 4dB apart. Most audio amplifiers in the BBC used 24v supplies.

US and Euro systems wire XLR audio jacks differently, too...

Neil

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

Euro/UK XLR audio wiring is pin 1 shield, pin 2 in phase, pin 3 antiphase (balanced signal of course); US wiring swaps pin 1 and 2 IIRC. It all sorts of works until someone puts phantom power on it

Our output impedances were usually well below 10 ohms but always 600 ohms in.

It's only in recent years that Euro domestic equipment moved away from din-3 and din-5 connectors to phono jacks (RCA) - largely the influence of Far Eastern audio equipment. Of course, if you look at an old Quad II amp, you'll find a Jones socket on the side...

It's been a few years now but I spent over thirty years as a BBC engineer in various capacities.

Neil

Things I've learned. Audio is hard.

Like, seriously.

But back to the topic of just getting a signal into a load and not having the load distort the signal. I pulled the charge pump and op-amp out. The 0.1μF DC blocking cap was way to small. I replaced it with a 47μF cap (and then a 100μF then a 330μF) and then plugged the Line In on the amp directly into it. The signal stayed unchanged.

Yay! Except not.

Audio is hard.

I tried a few different waveforms. And got a high pitch hiss / squeak out of a what I thought would be a nice low frequency triangle wave. The low frequency triangle wave was predominant but it definitely wasn't right. I inserted some 220Ω resistors in series after the caps and that seemed to help. I also tried 1KΩ and 10MΩ just to see hear what would happen. I think for a chosen update frequency of 16KHz it should be possible to smooth out the very sharp steps from the 8bit DAC but I don't know if that's the way to go.

I tried sampling the output sound using my phone and it looked sort of okay. Maybe. There were significant variations between the waveforms' peak-to-peak for that 500Hz triangle wave I was testing with. But the amplified speakers I'm using are an ancient pair of Logitech Desktop Computer speakers. Their sound reproduction could be quite poor. And the mic on my phone probably isn't great either.

Anyway I tried the same triangle waveform on Windows desktop PC using a tone generator and it looked consistent and properly triangular on my phone. And sounded 'purer'. So I guess that answers that, phone mic is good enough.

Except it was played back in 16bit 44Khz. And then the power went out so I had to stop testing.

I think my next step is to move my circuit to a PCB because there seem to be noisy glitches every time the DAC updates and I'm hoping that's a breadboard issue. But I must do it in a way that allows me to test different 'Line Out' boards against my one 'DAC board'. I also need to test with sound that isn't just a tone. That's going to take some proper programming time.

A question on the Nyquist limit. If the inital waveform was sampled at 32KHz but played back at 16KHz would that still be an effective 8KHz or is it 16KHz?

Thanks for all the help!
Andrew