Here is another project inspired by Danny: it is based on the same principles as the BoosterPhone, low supply requirements, rail-to-rail output on a low impedance, in this case 600 ohm and decent audio quality without too much complications.
It is designed to match older sources, delivering 150~200mV to modern equipments requiring 700mV or more. It thus mimics a passive step-up transformer, but without the impedance penalty: a 1:4 step-up transformer increases the impedance by a factor of 16, which can be problematic. Instead, the output impedance is much lower than the input.
Although the X-FoBooster requires a supply, unlike a copper & iron one, it has no power switch and lasts forever on a single charge: the supply current is comprised between 55 and 85µA.
First, the simplest version optimized for 3.7V:
This is the graphical version, based on the real prototype. The sim uses the native transistor types of the LTspice.
Now the 5V version:
Here is the non-inverting version: a bit more complex, but it has a self-centering quiescent output bias, and a THD of 0.03% against 0.1% (at 0dBm/600ohm, actually measured)
The values shown as comments are the ones used in the physical circuit. The others are suitable for the sim.
All the transistors are also 2SA1015/2SC1815, except the lower output transistor Q6, which is a BC547C.
The output is referenced to the + supply, and it requires a DC link to work properly. If it used on a cap-coupled input, it is necessary to include a dummy load of 1 to 2K.
I didn't build the various versions as complete projects, because I have no use for them (for the moment), but I breadboarded them, and the measurements are based on the breadboarded circuits.
The 3.7V version can use the same battery management as the Boosterphone, and it will last even longer
It is designed to match older sources, delivering 150~200mV to modern equipments requiring 700mV or more. It thus mimics a passive step-up transformer, but without the impedance penalty: a 1:4 step-up transformer increases the impedance by a factor of 16, which can be problematic. Instead, the output impedance is much lower than the input.
Although the X-FoBooster requires a supply, unlike a copper & iron one, it has no power switch and lasts forever on a single charge: the supply current is comprised between 55 and 85µA.
First, the simplest version optimized for 3.7V:
This is the graphical version, based on the real prototype. The sim uses the native transistor types of the LTspice.
Now the 5V version:
Here is the non-inverting version: a bit more complex, but it has a self-centering quiescent output bias, and a THD of 0.03% against 0.1% (at 0dBm/600ohm, actually measured)
The values shown as comments are the ones used in the physical circuit. The others are suitable for the sim.
All the transistors are also 2SA1015/2SC1815, except the lower output transistor Q6, which is a BC547C.
The output is referenced to the + supply, and it requires a DC link to work properly. If it used on a cap-coupled input, it is necessary to include a dummy load of 1 to 2K.
I didn't build the various versions as complete projects, because I have no use for them (for the moment), but I breadboarded them, and the measurements are based on the breadboarded circuits.
The 3.7V version can use the same battery management as the Boosterphone, and it will last even longer
Attachments
Here are the clipping behaviours for 5V supply and a 1KHz triangle wave.
First, the simpler inverting version (on a 600 ohm load):
Then, the non-inverting variant. Slightly less clean, with a bit of rail-sticking visible on the positive side, but still quite acceptable:
First, the simpler inverting version (on a 600 ohm load):
Then, the non-inverting variant. Slightly less clean, with a bit of rail-sticking visible on the positive side, but still quite acceptable:
Many variations around the same theme are possible with this topology: take this circuit for instance
R4 is required because without it, the current through Q3 would be almost perfectly constant, due to the total bootstrapping of R2. This means that the current through Q3 would practically equal the current sunk by the bias control transistor Q7.
But this also means that Q6 has no way to get the additional base current it requires to drive the negative swing through the load.
R4 is just sufficient to provide the small supplement, without affecting too much the bootstrapping effect.
The bootstrapping keeps Q3 in a quasi-stasis condition, allowing for an excellent THD. To reconcile the minimum deviation from the ideal bootstrap, and the needs of Q6, R4 is made as large as possible and Q6 is chosen to have the highest possible Hfe, hence the -C selection.
Lower gain transistors could be used, but R4 would need to be reduced, increasing the THD.
It is however possible to combine a good bootstrap and a sufficient base drive for Q6: the solution is to use a non-linear element, like a LED, to provide the additional base current only when necessary :
Here, R4 and D3 provide the additional base current only when it is necessary, ie. during the negative swings. This results in a THD almost as low as the resistor version (0.04% instead of 0.03%), with the ability to use any transistor type.
The diode D1 is an antisaturation diode: it prevents the complete collapse of the base voltage of Q6 when it becomes inactive (during the positive swings).
C5 contributes to soften the transition, to iron-out the slightest remaining artefacts.
Increasing C5 can contribute to lower the THD if its value is increased, because the circuit becomes "adaptative", and shifts toward class A operation during large-signal conditions. This increases the current consumption, but the quiescent current remains unchanged.
The THD can also be reduced by increasing the total quiescent current beyond 85µA.
Having the input and output grounds on the same side of the supply would be nice, and it is feasible:
Unfortunately, this version has a significantly higher THD than the others, both in sim and reality
R4 is required because without it, the current through Q3 would be almost perfectly constant, due to the total bootstrapping of R2. This means that the current through Q3 would practically equal the current sunk by the bias control transistor Q7.
But this also means that Q6 has no way to get the additional base current it requires to drive the negative swing through the load.
R4 is just sufficient to provide the small supplement, without affecting too much the bootstrapping effect.
The bootstrapping keeps Q3 in a quasi-stasis condition, allowing for an excellent THD. To reconcile the minimum deviation from the ideal bootstrap, and the needs of Q6, R4 is made as large as possible and Q6 is chosen to have the highest possible Hfe, hence the -C selection.
Lower gain transistors could be used, but R4 would need to be reduced, increasing the THD.
It is however possible to combine a good bootstrap and a sufficient base drive for Q6: the solution is to use a non-linear element, like a LED, to provide the additional base current only when necessary :
Here, R4 and D3 provide the additional base current only when it is necessary, ie. during the negative swings. This results in a THD almost as low as the resistor version (0.04% instead of 0.03%), with the ability to use any transistor type.
The diode D1 is an antisaturation diode: it prevents the complete collapse of the base voltage of Q6 when it becomes inactive (during the positive swings).
C5 contributes to soften the transition, to iron-out the slightest remaining artefacts.
Increasing C5 can contribute to lower the THD if its value is increased, because the circuit becomes "adaptative", and shifts toward class A operation during large-signal conditions. This increases the current consumption, but the quiescent current remains unchanged.
The THD can also be reduced by increasing the total quiescent current beyond 85µA.
Having the input and output grounds on the same side of the supply would be nice, and it is feasible:
Unfortunately, this version has a significantly higher THD than the others, both in sim and reality
Member
Joined 2009
Paid Member
very good. I recognize the input stage bootstrap, something I used for my TGM8 amplifier and which I thought I was the inventor of 🙂 but likely I wasn't - anyhow I am a big fan of this architecture!
https://www.diyaudio.com/community/threads/tgm8-an-amplifier-based-on-rod-elliot-p3a.245619/
https://www.diyaudio.com/community/threads/tgm8-an-amplifier-based-on-rod-elliot-p3a.245619/
Last edited: