A headphone Transconductance Amp for a change!

[Antoinel]

The schematics in the the past threads showed the use of TWO Power supplies to drive the amp. No doubt a turn off to the DIYer. Do we need two power supplies?. Absolutely not. The past shematics also showed using a regulated power supply. Another turn-off! I had it at hand and it was convenient to use. Fortunately we do not need a regulated power supply to drive the amp. We only need one unregulated [dual] power supply per channel in the range of +/-10 VDC to +/-20 VDC. Figure 1 in the attached PDF Headphone 006 shows the [even] simpler schematic of the transconductance amp. It is mostly like that shown in the previous thread less the now-obsolete +/- 6 VDC supply. The schematic also shows the use of one dual +/- 15VDC power supply. The schematic of this common power supply is shown in Figure 2. It is comprised of a power transformer (12Vac secondary plus a center tap), a bridge rectifier and 2 filter capacitors (4700 microfarad each). Note its the battery model which I used in Figure 1. The CENTER TAP of the transformer's secondary joins the common junction of the filter capacitors, and most importantly it is the OUTPUT PORT Vo and not the customary ground.

In operation, the variation in the output voltage Vo is a trivial disturbance to the working of the JFET constant current sources (CCSs). The +/- 200 mV peak to peak level of Vo across the headphone [for my comfortable listening] is less than 1% voltage variation across the CCSs; i.e. (0.2V signal divided by 30 V power supply times 100 = 0.7%). Also, the value of the current passing through the JFETs is quite stable over a wide range of power supplies. Thus the JFETs resist the voltage variations in Vo and present themselves to the power supply as impedances of high value.

Food for thought.


Suppose I use two resistors (5.6 K each instead of the JFET contant current sources. One resistor is connected to the +15 V and the other to -15 V. The current passing through them is 30 V divided by 11.2 K = ~2.7 mA which also flows through the diodes to bias the output transistors. In operation the situation is quite different from the one I described above.

• The 200 mV peak to peak variation in Vo is still a minor disturbance to the current flowing through the resistors (5.6 K).
• There is now a substantial AC feedback and DC servo feedback from the output Vo to the input port. Maybe beneficial!
• The input impedance of the amplifier is now 5.6 K divided by 2 = 2.8 K. It is acceptable to sources of low output impedance.

[Antoinel]

I have an omission in Figure 2 of the previous thread. Figure 2 shows the schematic of a common power supply. The secondary of the power transformer equal to 12.6 Vac plus a center tap should give ~+/-18.5 VDC instead of the quoted +/-15 VDC. I used a variac or autotransformer on the primary of the power supply transformer to tweak the output voltage to +/-15 VDC.

[Antoinel]

What value does this headphone transconductance amplifier or [attenuator; see below for more discussion] bring to the table? How does it compare with other voltage source amplifiers on several issues?

• The output port of this amp can be shorted to ground indefinitely without damage. No one dares to do this with any of the voltage source headphone amplifiers which currently populate the general heading of Headphone Systems. It is contraindicated!
• A tool to measure the Impedance versus Frequency for a loudspeaker, its individual drivers and headphones. The attached Word document ImpedanceGrado80i shows such a graph. The two headphones were connected in parallel for the measurements. The Y axis is 20 X Log of the ratio of the [voltage at any frequency] to the [voltage at 1 KHz]. The X-axis is the Log of the frequency spanning 30Hz to 20 KHz. The graph is a typical Impedance graph of an electrodynamic driver; after the sharp edges (due to measurement inaccuracies) are smoothed out. The trends are there. The driver has a peak at resonance at low frequency and a rising impedance with frequency due to the inductance of the drivers' coils. This graph may remind you of an output of a "tone control circuit". The hidden bonus in this graph is a boost in the output power by the headphone for bass and treble frequencies. By contrast; a voltage source headphone amplifier is NOT a tool to graph Impedance versus Frequency and will not claim this auto boost of bass and treble frequencies; at least for Grado 80i. The bass boost is beneficial for bass-shy headphones and the treble boost is a plus for those of us who have lost hearing in the high end due to older age. A voltage source headphone will give the user the exact opposite effect which is a simultaneous muting of the bass and treble frequencies. The tonal balance of the same headphone is different from to a voltage source and a current source devices.
• Use the circuit which I described in Headphoone001 to graph the impedance versus frequency of loudspeakers or individual loudspeaker drivers. Bypass the 220K hearing protector at its input and drive the unity gain amp with a 1-4Vpp signal. DIYers usually find the need to graph the impedance-frequency profile of woofers. This is straight forward with a sine function generator and an AC voltmeter so as to cover the important range of 20Hz to 150 Hz.

As I continue to refine the circuit, it became apparent to me that the design is a Headphone Attenuator rather than an Amplifier. From now on I will call it a Headphone Device. Here's why for my case. My music source is a compact disc player; which may also be or could be yours. It has an output as seen on an oscilloscope of up to 6 Volts peak to peak. I am actual forced to attenuate this generous input level to a mere 200 millivolts peak to peak at the output port. This is a comfortable level to listen to the Grado 80i ($99) or to a lower cost pair of headphones (~$10) from Radio Shack or to the in-the-ear-canal ones commonly used on iPODs etc. The most important quality and specification to gloat about in a Headphone Device is it MUST Protect my hearing. What am I after anyway; pain and hearing loss or pleasure?

[Antoinel]

As I made the device simpler [in design], hum from the power supply crept in the output. Hum became noticeable and/or objectionable. So, power to the device is now enabled by using rechargeable batteries; which rendered it portable if that is an important attribute to the DIYer. The attached PDF Headphone 007 shows two schematics. The first is for the active circuitry and the second details the circuit which can be used to recharge the power supply batteries when the device is idle. Let us focus on the first schematic. Note the following:

• The TIP 120 and 125 transistors were replaced by MJE15028 (NPN) and 15029. The TIP transistors are recommended for low speed switching and may not be suitable for this application.
• The JFET constant current sources in PDF Headphone006 were replaced by resistors (4.7K). They accomplished 3 functions. The first is to supply current to the red LEDs so as to bias the output stage in Class A. the second is to provide corrective AC feedback. And thirdly is to enable a DC servo action so as to minimize DC offset at the output port Vo. The output DC offset will not rise above 100 millivolts after unplugging the headphones; this is highly protective of the device!
• Note the use of the rechargeable 9V Nickel-Metal Hydride batteries. The DIYer may wish to consider using rechargeable 12 V Lead-acid batteries instead; the gel type or other.
• Blocking capacitors in the input circuit are not recommended or needed.
• For a minimum listening level, the input impedance of the device is equal to 25 K (vol control) + 10 K + 4.7K/2 = 37K. The 10 K resistor is for my [and your] hearing protection.
• For a maximum listening level, the input impedance is ~12 K.
• My listening input impedance is variable and centered around ~[12K (vol control) + 10K + 4.7 K/2 = ~27 K]. I am using Grado SR 80i headphones. The oscilloscope trace at the output port was ~200 milliVolts peak to peak for my comfort zone. Note that the input signal Vi from the CD player resigstered upwards of 6 Volts peak to peak.

The useful listening time was a minimum 90 minutes when using the standard-sized 9 V NiMH rechargeable batteries. This battery has an acceptable amount of in-use energy for its small size! This is a "long time" for me in one sitting. So, I took a short ergonomic break. I turned off the device, and replaced the used batteries with freshly-charged ones. Thereafter I resumed the listening activity for another 90 minutes. This indicated to me an "easy' feasibility for using rechargeable batteries to operate the device. In a separate non-listening experiment, I monitored the quiescent operating conditions of the device at 5 minutes after turn on and just before turn off at 90 minutes. The attached WORD document HeadphoneDevice tabulates the salient voltage measurements. The key conclusion was the device operated in needed Class A during the period of 90 minutes. The current draw from each battery was 40 mAh; well below the maximum specified of 170 mAh. The terminal voltage of each used battery recovered after 12 hours to 8.8 V [open circuit] before the next cycle of recharge. I did not want to abuse these expensive batteries ($14 ea.) by "deeply" discharging them. I believe NiMH batteries dislike it unlike NiCd ones; which can also be used here instead.

The battery-recharging circuit is straight forward. It will be useful for small 12 V Pb-acid batteries. Two commercial or home-made rechargers are needed for both channels with one recharger for each power supply rail. A [single pole triple throw] switch is used for each channel. One position recharges the batteries for both power supply rails and both audio channels. The opposite position of the switch disables both rechargers completely and simultaneously turns on power to both channels of the headphone device. It will be a useful tool to have and or build so to use instead of the classical power line supply.

The sound of the device (one channel driving both headphones), was detailed and tonally balanced. Understandably, there was no 60 Hz hum in the headphones, but a mere 2 milliVolt peak to peak broad band noise at the ouptut port as seen on the scope. I am inclined to "box up" two channels. Pictures of my progress will be forthcoming.

[Antoinel]

The two attachments did not upload in the previous thread. Here they are.

[xnor]

Above you wrote that you smoothed out the sharp edges of the impedance plot due to measurement inaccuracies. Are you sure those are measurement inaccuracies? Afaik, the impedance is not as smooth and with a transconductance amp this will cause similar "inaccuracies / sharp edges" in the frequency response.
Dunno if you have the equipment needed for this but I'd love to see a sweep that you recorded from your Grados powered by this amp.

[Antoinel]

I agree with your analysis. Your recommendation will be the ideal experiment. I have a function generator with a digital readout and an oscilloscope. I dialed the desired frequency and then measured the attendant output voltage on the scope's 0.1 V/division scale. I utilized 4 divisions each of which contains 4 subdivisions. The measurement inaccuracy was in my reading the scope's subdivisions. Was it 3.6 or 3.5 subdivisions? Please go to the website www.firstwatt.com. Mr Pass has 2 articles which discuss transconductance (current source) amps and their use to drive speakers. He showed graphs of your proposed experiment; except on high end loudspeaker drivers. Regards