Hey John, I played around with that Glass Blower circuit and have a modified version of it if you are interested in building another one.
When I was attending college in the early 80’s, I experimented with discrete differential pair circuits that bootstrapped the output back to the supply to a have greater output. I remember that those circuits used more current when compared to an equivalent circuit (without bootstrapping) running with higher voltage rails. So, I breadboarded this Glass Blower circuit a couple of summers ago when I discovered it on the web… because Valve Wizard implies on his website that it is a more efficient method than using a charge-pump circuit for headroom. I was very curious about the efficiency claims.
Valve Wizard’s statement:
“Most other 9V booster pedals run out of headroom around 6Vp-p, or even less, but the GlassBlower can deliver up to 27dB of boost and 12Vp-p of clean output signal! There are some pedals on the market that can do this (for example, the VisualSound Truetone), but they use voltage multipliers to increase the supply voltage internally. Such multipliers are, at best, 80% efficient, so your 9V battery only lasts 80% as long, which is a bit of a waste, considering you don't actually need all that headroom most of the time. The GlassBlower is different. I acheives its extra headroom with a devilishly-simple-yet-sophisticated rail boostrapping technique that forces the supply rails to follow the signal. In this way the extra voltage is there ONLY when it is needed. This is effectively what advertisers call Class G (I disapprove of the term though).”
My test results:
When I built the circuit I did not include the bypass circuitry (Q3, D2, R12, R13 and the BP switch) because I was just interested in the meat of the circuit. I also used 2N4401 and 2N4403 transistors for Q1 and Q2.
Opamps: Some opamps that were tried in this circuit wouldn’t work properly. I had two brands of 072’s (TI and STmicro) and one of them wouldn’t come close to working properly in this circuit… I don’t recall which brand wouldn’t work in this circuit. The 072 that somewhat worked, had nasty transient distortions for large signals and with steady state large signals there were small wave shape distortions on the positive half cycles. The TLC2272 seemed to work best for me in this circuit without obvious distortions and it has a lower idle current than the 072.
The circuit also had small glitches on the output signal where the bootstrap transistors were turning on/off. These glitches may just have been from my breadboard layout and a properly designed PCB may not have these issues.
Since this is implied to be more efficient than an equivalent circuit with a charge-pump, I measured the source current for various output signal levels with a 100Hz signal. The current waveform was measured across a 1ohm source resistor with an o-scope set to differential mode. Peak current waveforms were recorded.
Output Vp-p vs Source current (peak)
No Signal 2.4mA
1.0Vp-p 2.4mA
2.0Vp-p 3.2mA
3.0Vp-p 5.0mA
4.0Vp-p 7.0mA
5.0Vp-p 9.0mA
6.0Vp-p 11.0mA
7.0Vp-p 13.0mA
8.0Vp-p 15.0mA
9.0Vp-p 17.0mA
10.0Vp-p 19.0mA
11.0Vp-p 20.0mA
12.0Vp-p 21.0mA
Then I disabled the bootstrap circuit, installed a charge-pump for a negative supply and made the same current measurements. The negative voltage rail was connected to the grounded side of R4 on the bias divider and to the negative power pin of the opamp… everything else remained the same. A MAX1044 charge-pump was used with its osc freq boost enabled, which lowers the pumps conversion efficiency a bit. The supply voltage had to be lowered to 8Vdc because of the TLC2272 limitation of +/-8V.
Output Vp-p vs Source current (peak)
No Signal 4.0mA
1.0Vp-p 4.0mA
2.0Vp-p 4.0mA
3.0Vp-p 4.0mA
4.0Vp-p 4.0mA
5.0Vp-p 4.0mA
6.0Vp-p 4.0mA
7.0Vp-p 4.0mA
8.0Vp-p 4.0mA
9.0Vp-p 4.0mA
10.0Vp-p 4.0mA
11.0Vp-p 4.0mA
12.0Vp-p 4.0mA
The current measurements would be greater for both circuits if the LED indicator was installed. But since the LED is only powered from the positive power rail, both circuits would increase by the same amount and the charge-pump conversion efficiency would not change.
So, the only time the Glass Blower circuit will be efficient is when the signal output is below the Vbe of the bootstrap driver transistors or approx 1.2Vp-p… otherwise it is dumping a lot of battery current across the power rails every time Q1 and Q2 are conducting. Some will argue that your signal will not be driven over the 1.2Vp-p zone very often, so you will still have a fairly efficient boost amplifier. I believe that the whole premise of this boost pedal is to have large amplitude signals drive the preamp tubes into submission for beautiful harmonic distortion. So I think it is safe to assume that you will most definitely be running a lot of your signal greater than the 1.2Vp-p efficient zone. I’m pretty sure all of my solid-body bass guitars output signals greater than 1.2Vp-p.
The statement that Valve Wizard made about charge pumps being only 80% efficient and wasting battery current is very misleading. Granted that there is always some conversion loss with charge pump circuits, but you need to compare charge pump efficiency against the efficiency of his bootstrapped circuit… apples to apples. If you look at the two source current measurements for the 12Vp-p output signal, one can make the argument that the bootstrapped circuit has a 500% lower efficiency than the equivalent circuit with the charge-pump. And at a reasonable 4.5Vp-p output signal, the bootstrapped circuit has half the efficiency of the equivalent charge-pump circuit. Apples to apples.
The original Glass Blower:
Just for kicks, I made a different version of this circuit that uses the second opamp as a bootstrap driver for Q1 and Q2. I wasn’t trying to improve battery efficiency, or to make a design for production, but I wanted to eliminate the signal glitches when the bootstrap transistors turned on and off. It was just a bench exercise to improve upon what I had in front of me. I named the file Glass Crusher for my own amusement. The benefits of this modified circuit are that the Vbe barriers of the bootstrap driver transistors are essentially eliminated by opamp gain and feedback control to allow for an electrically quieter bootstrap drive and the audio signal only passes thru one opamp. It also buffers the bootstrap drivers from the audio signal. The elimination of the on/off switching of the bootstrap transistors at Vbe stopped the signal glitches from transistor switching current spikes (although a well laid out PCB of the original circuit may not have these signal glitches at all). All of the inefficient power aspects of the original circuit remain in this circuit as well… actually this modified circuit is a bit less efficient than the original with output signal levels less than the transistors Vbe drops of 1.2Vp-p. Because the bootstrap transistors in this modified version are continually active, the measured source current below 1.2Vp-p increased to approx 4.5mA with a 1Vp-p 100Hz signal… approx 2mA more than the original circuit. If you desire to have the same Vbe efficiency zone as the original, then you can move the feedback connection from the second opamp’s negative input from the Q1 and Q2 driver output, directly to the opamp output (pin 7). But with all of the current wasted for larger signals, I prefer the quieter drive. There are also some component value changes from the original...
A modified version:
This is a fun and novel circuit, but IMO, a properly designed gain boost circuit with a proper power supply for headroom is a much better method for a large signal boost. Think about how all opamp manufacturers specify clean and stable DC supplies for use with their opamps. They don’t recommend that you put AC signals on the supply pins.
-Frank
