Pre Emp Circuit

[OldskoolPirate]

Hi Necks, has anyone got a nice 50us pre-emp circuit that doesnt bring the audio volume down ? i tried alberts 10k and 2x 103 caps but the audio is way to low after. Maybe i need a stage after to bring it back up ? I dont really want to use active circuits if i can help it.

Ta!

[jvok]

Not really possible to have a passive circuit without bringing the level down because passive circuits can only cut, not boost. So an active circuit would give you a proper treble boost, a passive circuit fakes it by cutting the bass instead.

Edit: Figure 3 on this page has an active circuit that will do preemphasis without any loss: https://sound-au.com/project54.htm

[Albert H]

Simple, cheap ACTIVE pre-emphasis circuit:

Mono-pre-emphasis.png

This is about as simple as you can get. As long as you use a reasonably low noise transistor, it won't add hiss. If you want exactly 50µs, use two 100n capacitors in series (to give 50n).

[OldskoolPirate]

Thanks Albert, that looks nice. I’ll try that.

[Shedbuilt]

Simple, cheap ACTIVE pre-emphasis circuit:

Mono-pre-emphasis.png

This is about as simple as you can get. As long as you use a reasonably low noise transistor, it won't add hiss. If you want exactly 50µs, use two 100n capacitors in series (to give 50n).

That won't give an exact 50us curve (or even particularly close), because the capacitive element of the 50us time constant, has a 100R resistor in series with it. As we know, this is there to flatten the high frequency gain (which is a good thing), but it will push the turnover point higher, and reduce the gradient beyond the turnover point. The capacitor needs to be made larger than 50us / 1k, to more closely approximate the 50us curve.

[jvok]

Simple, cheap ACTIVE pre-emphasis circuit:

Mono-pre-emphasis.png

This is about as simple as you can get. As long as you use a reasonably low noise transistor, it won't add hiss. If you want exactly 50µs, use two 100n capacitors in series (to give 50n).

That won't give an exact 50us curve (or even particularly close), because the capacitive element of the 50us time constant, has a 100R resistor in series with it. As we know, this is there to flatten the high frequency gain (which is a good thing), but it will push the turnover point higher, and reduce the gradient beyond the turnover point. The capacitor needs to be made larger than 50us / 1k, to more closely approximate the 50us curve.

If you do the maths it turns out you want almost exactly 56n which is a standard value

[Shedbuilt]

Simple, cheap ACTIVE pre-emphasis circuit:

Mono-pre-emphasis.png

This is about as simple as you can get. As long as you use a reasonably low noise transistor, it won't add hiss. If you want exactly 50µs, use two 100n capacitors in series (to give 50n).

That won't give an exact 50us curve (or even particularly close), because the capacitive element of the 50us time constant, has a 100R resistor in series with it. As we know, this is there to flatten the high frequency gain (which is a good thing), but it will push the turnover point higher, and reduce the gradient beyond the turnover point. The capacitor needs to be made larger than 50us / 1k, to more closely approximate the 50us curve.

If you do the maths it turns out you want almost exactly 56n which is a standard value

I can see how you came to that, but it's not quite as simple as that. At 56nF, the sum of the resistance, and the reactance of the capacitor, will equal 1k at 3183Hz (as per 50us TC), but the curve is not as steep (as a straight, parallel RC), because the 100R series resistance stays constant with increasing frequency.

[jvok]

Ah you're right - I made Xc+100=1000 ohms when it should be sqrt(Xc^2 + 100^2)=1000 ohms. When you do the maths right it works out to 50.3n so a 50n cap is fine.

That said imo theres no point trying to be too accurate with preemphasis circuits. Component tolerances will give you bigger errors than that and you won't hear the difference anyway. 47n would be fine.

[Shedbuilt]

The 100R in series with the cap, curtails the caps ability to bypass the 1k resistor. This becomes more significant with increasing frequency. So if you make it accurate at 3183Hz, it will be a long way below the 50us curve at say 10khz or 12khz. To make it (overall) closer to a 50us curve, the cap would need to be significantly larger. This means the effect starts at a lower frequency, but because it rolls in more gently, it’s all a compromise. I agree there’s not much point in getting too hung up on ultimate accuracy - for most purposes.

[Shedbuilt]

Back to eat humble pie. I had a quick look ally this again yesterday, with fresh eyes. I this looks a lot closer to the text book curve, than I initially thought it would. The HF roll off is at 5us (ie around Fc 31830Hz), which is good for limiting out of band gain. I’d spotted that initially, but still thought the added resistance would affect the in band plot, more than it does / should.

[Albert H]

You're right. It's reasonably close to the right shape (and about as right as you'll get with a simple one transistor circuit!). This is (as is common with my designs) a "close enough" compromise.

Designing "proper" pre-emphasis circuits and 15kHz lowpass filters is a major undertaking. My latest commercial design relies of "gyrator"-based filters (emulating inductors) and lots of tight tolerance resistors and capacitors. Each channel uses a total of seventeen low noise op-amps!

After much experimentation, I went back to the "good old" NE5532 dual op-amps. If properly decoupled right at the power pins, these are just as good as many of the expensive, esoteric "audio" op-amps: They can drive into low impedances without distortion, don't suffer from latch-up if over-driven, and have vanishingly low noise. Not bad for a dual op-amp that costs €0.35 in quantity!

[XXL]

Simple, cheap ACTIVE pre-emphasis circuit:

Mono-pre-emphasis.png

This is about as simple as you can get. As long as you use a reasonably low noise transistor, it won't add hiss. If you want exactly 50µs, use two 100n capacitors in series (to give 50n).

I just tried this circuit. It sounds brilliant but it also amplifiers up all kinds of electronic noises from the sound card so its no good. Its a real shame because it sounds so good. Probably needs more circuitry to filter it maybe ?

[Albert H]

Nope - you have a problem with your sound card!

Listen - carefully - to the output of the sound card without the pre-emphasis in place, and you'll probably hear the same noises. If the output of the sound card on its own is clean, then you might have an impedance mismatch. The input impedance of the given circuit above is (roughly) 10 kΩ. Your sound card may want to see a lower impedance. You can scale the impedance the quick and dirty way by reducing the value of the level pot, or you can try putting a resistor across the input to the pot to load the output of the sound card. If you try (for example) 1 kΩ, you'll have to turn the level up a bit, but it might reduce the noises by providing more load to the sound card.....

[XXL]

It does it if I plug my phone in aswell, but only when it’s on charge. If I bypass the pre emp, there’s no noise. I’m not saying it’s not there without pre emp, but the transistor is definitely amping up the noise a lot. It’s obviously very low without it. Or like you say, some kind of mismatch causing it to make noise.

[Albert H]

In the case of your phone, it'll be crud coming from the charger.

[Albert H]

OK. Let's reduce the gain and lower the input impedance....
If you change the 10k collector resistor for 3k3, the upper base bias resistor from 470k to 47k and the lower bias resistor from 47k to 12k, you'll reduce the overall gain, whilst retaining the pre-emphasis shape. You'll also have a better match from the 10k level pot to the input of the transistor......

[reverend]

The input impedance of this circuit will vary slightly with frequency as the input impedance is, give or take, the hfe (gain) of the transistor multiplied by the emitter resistance. Assuming a gain of 100, at 100Hz where only the 1K resistor is at the emitter to any real extent, the input impedance would be 100K in parallel with the 470K and 47K resistors. The input impedance would roughly halve at 3kHz and would continue to decrease as the frequency increases.

Changing the resistors as suggested by Albert above would swamp the transistor input impedance with the resistors and lead to a flatter input impedance across all frequencies.

[Albert H]

Thanks Rev - you're exactly right!

[reverend]

The following will give you the correct 50 uS pre-emphasis and at the same time provide a little bit of low pass filtering. It's only about 0.5 dB down at 15 kHz and no better than 6dB down at 19 kHz, but that's often sufficient to stop the pilot light from flashing on and off with high treble content. It has roughly unity gain and an input impedance of about 4K Ohms.

50us-preemphasis.jpg

Any power supply from 9 to 15 Volts works fine. The transistors should be BC548B or 2N2222 types (or any NPN with low input capacitance and an hfe/Beta of about 200). Other transistors will work fine, but the frequency response may not be as flat.

[XXL]

OK. Let's reduce the gain and lower the input impedance....
If you change the 10k collector resistor for 3k3, the upper base bias resistor from 470k to 47k and the lower bias resistor from 47k to 12k, you'll reduce the overall gain, whilst retaining the pre-emphasis shape. You'll also have a better match from the 10k level pot to the input of the transistor......

I tried this and the results were much better. The transistor wasn’t swamped so much with RF. Before it was pulling in RF and the pll was going crazy. Now it’s fine but if these values aren’t correct, is there any way of reducing the gain correctly on this circuit ?