Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the THAT1646 As A Power Transistor Pre-Driver

Post by mediatechnology »

I've been meaning to get back to this project and incorporate a Vbe multiplier instead of an LED or diode bias string. A diode bias string required too much reference current to put the output into 75 mA Class-A. An LED reference, divided down by a divider, worked quite well as a voltage reference but lacked proper temperature tracking. A Vbe multiplier was an obvious choice.

Edited 12:22 PM after bench testing.

The following schematic was tested for DC performance. I will do distortion measurements soon. Using the previous bias arrangements the THAT1646 and output stage have proven themselves to be a very good performer at -90 dB 2nd and 3rd despite operating open loop. I originally estimated approximately 850 mV to set a 75 mA output current; the actual Vbe for 75 mA turned out to be quite close at 871 mV. I also added reverse Vbe protection diodes for the BD139/BD140. Edit: I used the "-16" version of the transistors selected for hFE 100-250 min/max (@ Ic=150 mA) rather than the lower gain "-10" (hFE 63-160) ones. I don't think it's that critical but why not.

Image

Some of you may be interested or curious as to why I chose a BD139 for the Vbe multiplier. The short answer is because it has a hole in it making heatsink mounting far easier. With its' 75 mA Iq this is not a "green" low-cost design. Anyone wanting to use a 2N3904 here is certainly welcome to. I couldn't find one with a hole in it....

Distortion Measurements
Added 14:48 11/08/07

Image
Distortion, 68R load, 4V P-P, Iq=75 mA, No build-out R

Image
Distortion, 33R load, 4V P-P, Iq=75 mA, No build-out R

Image
Generator Residual

Updated 11/09:

I put Visual analyzer into "B-A" mode to subtract the generator residual from the amp. This is a 200 sample average in order to reduce display noise and allow the actual distortion added by the amp to be more visible. You can see that the dominate added products are even-order.

This is with a 68R load under the previous test conditions:
Image
B-A Display mode showing the dominate even-order distortion added by the amplifier.

This is with a 33R load under the previous test conditions:
Image
B-A Display mode showing the dominate even-order distortion added by the amplifier and the slight rise in odd-order when heavily loaded.

Since the amplifier will have an internal 33R build-out resistor, distortion performance with 32R headphone loads will be similar to the 68R test results. Those tests did not use a build-out R.

The high-level performance is outstanding: Loaded in 600 ohms and driving 20V P-P the differences in generator and amplifier distortion can't really be displayed well. The numeric difference is quite low approximately 0.0003%.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the THAT1646 As A Transistor Pre-Driver/Headphone

Post by mediatechnology »

Hello everyone - I wanted to post a couple of updates.

I think I would like my build of this project to include a polarity switch and I've posted one using the THAT1240 here in this forum. I discussed at Headwize the use of the AD823 FET as a buffer for the pot/stepped attenuator.

The question arose in the Headwize forum about driving the 1646 directly from a level pot and I discouraged it for a couple of reasons. One, the input loading is fairly low, 10K, and this would undoubtedly bow the taper of a precision attenuator. I suppose that in some apps this wouldn't be an issue. The second reason I gave was bias current that might appear in the pot and the need for AC coupling as well as any Ibias-related Vos that might develop in the input resistor. I estimated based on other THAT parts an Ibias of around 1 uA...

To help everyone follow here's an image of the 1646 block diagram:
Image

Here's a copy of that post:

I checked with the designer of the THAT1646, Gary Hebert, and he gave me some specs for the inputs that are not in the datasheet.

The input resistance at the Cin+ input is quite high and swamped by the two 20Ks in parallel to ground at the pins 2 and 7 sense inputs, resulting in a 10K net resistance, and the 10 pF shunt C following the 20Ks. The actual 1646 Cin+ input itself is >50 M ohm times the available loop gain and for practical purposes can be ignored. Thus the 1646 will load a level pot in approximately 10K.

EDIT: It occurred to me to point out that if one were to leave one of the sense inputs open the resulting gain would be less than unity, approx -2.5dB, and the input resistance would rise slightly to 13K (20K||40K). Grounding one sense pin, while driving the other, would produce an internal 6 dB voltage divider and the input resistance would be 10K (20K||20K). So by driving both, floating one or grounding one sense pin we can have gains of 0 dB, "-3dB" or -6dB. For an exact 3dB attenuation an external resistor could be connected to the "open" sense line.

The input bias current for the Din+ input is 650 nA typically so my 1 uA guess wasn't that far off. But the Cin+ input, the one we use here, is another story being roughly 40 nA typically with a maximum of around 100 nA.

Other than the 10K load I don't see any reason why the 1646 could not be driven directly from the wiper of a 10K pot if accurate control law, i.e. taper, isn't important.

Adding the polarity stage is another matter: It needs to be driven by a low impedance source to prevent level shifts from source loading variances. So a buffer between the input jack and the 1240 remains a good idea. It would appear however that the pot can either be at the actual input or behind the 1240 between it and the 1646.

I'm still inclined to put the pot first and load it in a very high impedance buffer to preserve taper - particularly with a precision stepped attenuator.

I did measure a few of my BD139/140 pairs and at the test current of the Fluke 8050A I observed Vbe differences of approximately 15 mV between the NPNs and PNPs. The measured output offset is significantly lower than this.

Not apparent in my original schematic, but visible in the 1646 diagram are the 25 ohm internal buildout resistors in series with the BD139/140 bases. These 25R resistors do contribute to DC stability and deserve mention for those inclined to analyze it. Re, at 3R3 ohms, is significantly greater than Rb/beta. With a minimum hFE of 100 (at Ic=150mA) this ratio is greater than 10:1.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Post by mediatechnology »

I finally had a chance to finish the Vectorboard and run a few quick tests with RMAA. My power supply connections and load returns could be a little better (clip leads) but here's the test results I got:

http://www.ka-electronics.com/Headphone ... ut_30R.htm

Here's the summary updated 12/20:

Image
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Post by mediatechnology »

Some pictures:

Image

The heatsinks make nice hand warmers with dissipation of a little over 2W each at idle. With an efficiency of perhaps 2.5% I don't think I'll go for an "Energy Star" rating. Not shown on my original schematic are the 22 uF bulk bypass in the foreground.

Image

Here we have the Vbe multiplier using a second BD139 bolted to the rear of the heatsink. I used a 27R build out rather than the 33R on the original schematic. I had 27Rs on hand and it's probably a better choice anyway. All recent tests BTW were with the 27R. Note the carbon comp resistors from the last century. On my final build I'll use good metal films and a vertical trimmer.

On levels: With my 63 ohm MDR-7506s I have plenty of drive voltage and still use 10-12 dB of attenuation with typical-level "-10" sources. For my application I will dispense with an input gain buffer to keep the signal path short.

As it is now, the whole thing is DC-coupled. The low-end totally rocks. The output Vos is about 3.5 mV (at the Re junction) on the left channel and 6.5 mV on the right. The Vbe mismatch on both pairs (NPN to PNP) is 15 mV.

I'll continue to do the buffer for those that need it: I may end up using one anyway to add polarity inversion.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Post by mediatechnology »

Happy New Years everyone!

I decided to do some higher level measurements into various loads. The following measurements do not take into account the external generator distortion which has been posted earlier.

At 7.6 Vrms into 270 ohms, 213 mW, 1 KHz THD reads 0.0021%.
At 1.4 Vrms into 30 ohms, 65 mW, 1 KHz THD reads 0.0039%.
At 2.0 Vrms into 30 ohms, 133 mW, THD reads 0.0093%.

The above voltages are into the actual load, at the right-hand side of the 27R build out resistor and take into account its IR drop.

A pair of 300 ohm or 600 ohm phones can thus be driven to over 20V P-P. I checked the THD up to 6 KHz fundamental and there appears to be no rise. (My measurement beyond 3rd harmonic is limited by the 20 KHz measurement bandwidth of the soundcard.)

It should be noted that voltage gain upfront will be required to produce high level drive but with low impedance phones unity gain operation of the basic circuit is reasonable.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Post by mediatechnology »

Thanks Roger. Don't have a finished schematic yet but the one posted is really close. The only changes are:

1) Added 22/25V bypass caps, one set.
2) Changed the output R to 27 ohms from 33.
3) Added a dual 10K audio taped pot with the wiper tied to the input.

I do think it needs an input gain stage for higher drive requirement phones.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Post by mediatechnology »

While going through my office I found the following article describing a simple null tester to measure low-level THD: http://electronicdesign.com/Articles/Ar ... /8188.html

This allows the generator THD, to a significant degree, to be nulled while amplifying the device under test's THD to more measureable levels. Since the article cited above has broken links to the figures I've reproduced a copy:

Image
Image Courtesy of Electronic Design, June 21, 2004 authored by Jerry Riddick of Texas Instruments.

While not perfect, this technique allows the generator THD and fundamental to be nulled significantly to perform lower-level THD measurements. If the device under test has low phase shift and non-inverting operation, null testing with this device becomes a useful tool. The distortion and fundamental of the generator are nulled by the common mode rejection of the amplifier while the distortion of the device under test is amplified by 40 dB making measurement of it easier.

Since the THAT1646-based headphone amp has very low distortion, "almost" unity gain and very little phase shift it was a good candidate to measure. I did introduce slight attenuation, not shown, in the input leg of the instrumentation amp, a THAT1510, because the THAT1646 headphone amp does not have infinitely low output impedance and its' level is reduced by loading. This allows the CMR to be adjusted for maximum nulling. The gain of the 1510 was set to be 40 dB.

The output sample was taken at the junction of the amplifier's emitter resistors. The load was connected after the 27R build out and power measurements are with respect to the load on the right-hand side of the build out. This allowed measurements to be taken with smaller adjustments in CMR and without resistive attenuation by the build-out.

Please note that, to avoid confusion, the right hand scale of the spectral plots was replaced in an image editor with values reflecting the gain of the 1510. Thus, the -100 dB line, representing the actual THD level prior to amplification by the instrumentation amp, was in fact -60 dB with regard to the analyzer '0'. A 200 sample average was used to provide display clarity. The analyzer was operating in 16 bit 48 KHz. Please note that the 1 KHz is input that has not been fully canceled. The level of this, if nulled reasonably well, does not affect the harmonic measurements of the DUT given that the generator second and third are so far in the mud as to not matter.

The following is without load:

Image
No output load
The fundamental has been attenuated by ~70 dB. Note the complete cancellation of generator THD which typically ranges from -90 to -100 dB.

This is with a 30 ohm load driven to 25 mW:

Image
30 ohm load, 25 mW
Note the rise in second and third harmonic to approximately -95 dB each.

This is with a 30 ohm load driven to 65 mW:

Image
30 ohm load 65 mW
The third harmonic has increased to -83 dB which is 0.007% of the actual fundamental. With headphones having a sensitivity of 106 dB/mW, 65 mW would be, to say the least, quite loud at approximately 124 dB SPL.

This is with a 330 ohm load driven to 150 mW representing the higher end of modern headphone impedance (not withstanding vintage AKG240s). The drive level is 20 V P-P.

Image
330 ohm load, 150 mW, 20V P-P
Here we have an increase in second harmonic and may begin to be limited in our high level tests by the common mode range of the instrumentation amp.

Conclusion: When you don't have an Audio Precision System Two or a device that's too clean to measure with what you have sometimes you just have to make do.
radiance
Posts: 46
Joined: Thu Jun 26, 2008 4:50 am

Re: Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Post by radiance »

raf wrote:Wayne

Good stuff! You must have a "finished" schematic for this, no?
Deserves a good layout, with ground plane and all that!
Any news on the "finished" schematic yet?
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Post by mediatechnology »

This is the one:

Image
A headphone amplifier made from the THAT1646 and BD139/BD140 NPN/PNP pair.
radiance
Posts: 46
Joined: Thu Jun 26, 2008 4:50 am

Re: Using the THAT1646 As A Transistor Pre-Driver/Headphone Amp

Post by radiance »

Thanks!
Post Reply