Valve Car Amplifier

The completed amplifier with remote control.

See the amplifier operating at

The Problem with Radio.
This project came about because I was fed up with listening to the radio while driving. Because of my location and the places I drive to, most trips are at least an hour return. And I like my music, so until now I have used the car radio to provide entertainment. The problem is the very conservative formats of Australian radio. For the kind of music I like, it is only provided by commercial stations. However, playlists are extremely limited and thus quite repetitive. There is also a lot played that I don't like. Next, we get to the advertising. You can actually get a good idea of what time it is because of how the ads are played. Generally, there will be a block of ads 40 minutes past the hour. Then for about 5-6 minutes before the end of the hour is another lengthy block. Somewhat depressingly, for the music format I like, such radio stations are geared to those on death's door. This means lots of funeral ads, ads for graveyard holding pens, and ads about getting sick in various forms. Having finally got through those is the next annoyance - 10 minutes of news. The same news we heard last hour, and the hour before that...and what we'll hear again the next hour. Furthermore, in the morning and afternoon, we get news twice an hour! Driving 100km, it doesn't take long for the next lot of news to come round. They've lost me as a listener as I drive.

What to do? Well, CD's are one option. Indeed, I used to travel with a box of CD's, but I seemed to be changing them at a surprisingly high rate. Also, home burnt CD's don't work reliably in my car CD player. Because of being burned to a lesser depth than commercially made CD's, output is lower, and the laser servo struggles to provide output with high road vibration. There was also the problem of predictability after a while, with the tracks in a fixed order.

Then I found an iPod on the side of the road whilst out riding the Trek Y5. With the awful iTunes bloatware/spyware removed, and Anapod Explorer installed instead, loaded up with about 860 of my favourite tracks, this was the final answer. On shuffle mode, it's just like listening to the perfect radio station.
Now, the iPod is of course for headphone use only, so how to use it in a car? My first attempt was to use an FM transmitter and receive it on the car radio. The problem here was that while it worked, in some locations the signal would be swamped by another transmitter. It was too difficult to find a clear part of the band for all the areas I drive in.
Then I tried a MW transmitter running from an inverter. This was better in that regard, but again, and particularly at night, an annoying heterodyne could be heard in some locations. Also, because I hadn't regulated the power supply, the frequency drift was quite bad as the car electrical system voltage changed.
Really, the proper way to do it is to simply amplify the iPod output and drive the speakers directly. This is what this project is all about.

If it seems like I'm contradicting myself with the amount of radio receiver projects on this site, it needs to be pointed out that listening to the radio in a car is very different to at home. In a car, one is captive to the program, but at home there are other distractions so one can mentally tune out when the ads and news come on.

Concept Stage.
The amplifier would obviously use valves, which of course required a vibrator power supply. Given my car has stereo speakers, the amplifier might as well be stereo too.
I thought about installing extra speakers in the doors, but then in some HRSA junk was a nice little Clarion speaker relay box. It was intended to use a car cassette deck with the existing car radio speakers. When the cassette deck was in use, a 12V output would go high, actuating a 4PDT relay to switch the speakers over from the car radio to the cassette deck. This neatly solved my speaker problem, for I could save myself a lot of work and just use what was already there. In my application I would use the "power antenna" output of the car radio to actuate the relay. Thus, the speakers would be fed via the N.O. contacts when the radio was on, and from the amplifier via the N.C. contacts when it was off.

Clarion Speaker Relay Box makes it convenient to use the existing car radio speakers. It's mounted near the car radio.

There aren't a lot of places to put the amplifier in my car, so the obvious choice was under the seat - as had been done with a two way radio used by the previous owner. Because of the location, it would mean a remote volume control and power switch would be required. I wanted a quick release mounting system too - I like good service access, and designing this myself meant it was easy to include it. The iPod would simply connect by a flying lead with 3.5mm plug.

A car is not a hi-fi environment. Road noise, speakers mounted in a plastic dashboard with no tuned ports etc, and the position of the driver not being situated halfway between the speakers make it scarcely worthwile to have an amplifier of above average quality. Additionally, the sound quality from mp3 files is not hi-fi anyway.

For the valves, 6BM8's would be a good choice as I have hundreds (maybe thousands?) of them. I've used them in many projects before, and the power output is sufficient for this application. Importantly, with the limited space, the fact that a triode is included in the same envelope as the pentode is useful. A single triode pentode isn't quite enough to obtain the gain required from the line level iPod output, if full power output is to be obtained, so a preamplifier stage would also be required. Something like a 12AU7 or 12AX7 would do, with one triode for each channel. The remote volume control could be conventional, or it could be a DC controlled amplifier which would simplify the wiring. I had used such a scheme with the Model T Ford radio so knew it was practical. The speaker transformers would be 100V P.A. line transformers because I have a few, and the size is right.
With my latest bulk acquisition of vibrators from the U.S., I thought it was time to break with tradition and use one of them, rather than reach for yet another locally made Oak. The transformer would have to be a toroid given the low profile case. Using a commercially made mains transformer in reverse would unfortunately preclude synchronous rectification, or even a valve rectifier, so a solid state bridge would have to be used.

Design Begins.
First thing was to design the 6BM8 circuits. This simply entailed using Philips data to suit a 5K load impedance, for this is standard for most 5W P.A. line transformers.
The data can be found here. Similarly, the triode operating conditions were also according to Philips data.
Initially, I proposed a 12AU7 for the pre-amplifier stage as this gave a more appropriate gain than a 12AX7. However, doing this would require a multi-way shielded cable from the amplifier to a dual gang potentiometer on the dash for the volume control. While this is of course practical, a much more neater scheme is two unshielded wires to a single potentiometer, and to use DC control. This potentiometer can be connected so one side is earthed, then requiring only one wire to connect it.
I had used this method with a 6CS6 valve for my Model T radio, which like this amplifier, uses remote control. By controlling the bias on the valve, the gain can be altered over a wide range. Stereo operation would require two 6CS6's. However, I had seen a circuit operating on the same principles using a single valve; a 6BU8 or 6HS8. This seemed to be a very elegant way to control the volume, and so a test circuit was built on a breadboard. It had ample gain so would take place of the 12AU7 pre-amp, and the volume control was excellent.
As a digression, despite having a large quantity of 6BU8/6SH8's, the first one I used had a short between the cathode and first grid! It's very rare to find a faulty valve, but it can happen.
At this point, the subchassis for the valves was designed to accomodate three nine pin sockets. The enclosure for the whole thing was designed, cut and and bent up. The mounting in the car was to be two brackets attached to the floor, and the amplifier would slot into this and be secured with 6mm wing nuts.

Major parts laid out and secured. The Delco vibrator shown was not used in the final construction.

The major parts were laid out and secured to the enclosure. Another subchassis was made for the vibrator.
Wiring according to the theoretical circuit was commenced, with the exception of the power supply. To start with, I used a BWD215A valve power supply to power and test unit. The vibrator supply would then be designed and constructed once the amplifier part was satisfactory.

First tests after the amplifier stages had been wired. The power supply was yet to be built. Here, the resistors are being optimised for the DC volume control.

At this stage, all seemed to be working well, and it was time to proceed with the power supply. The details of this will be discussed in greater detail further on.

Completed amplifier.

The amplifier was completed and installed. First thing was to mount the brackets on a piece of plywood and screw that to the car floor with 8 gauge stainless steel screws (so they don't rust where they protrude under the floor).

Mounting screwed to the floor. The speaker plugs can be seen to the right.

Amplifier mounted and secured with wing nuts.

Seat reinstalled.

To complete the car installation, the next thing was to make up the remote control. With the DC volume control, this is a very simple affair, with just a switch and 10K pot. It connects to the amplifier with a four pin plug and socket at the amplifier end to make removal easy. The 12V supply is fed from the remote control end. I bent up a piece of black Marvi-plate for the control panel.

The 12V supply was fed from the remote control end. Here, the easiest place to get 12V was from the cigarette lighter. I made up a connection using the existing bullet connectors.

Cigarette lighter provides 12V to amplifier.

Completed control panel installed. It was kept plain so as not to attract attention. The switch looks just like an alarm warning light.

The Maiden Voyage Test.
Everything seemed in order by the time the installation was finished on Sunday evening. Monday was a drive into work which takes about two hours, and would be the perfect first test.
Two things did become evident, but despite that, the amplifier worked well.
Firstly, the sound seemed harsh and trebly. I suspected this might happen because of the P.A. line transformers. As described here, they can be problematic when used as valve output transformers in single ended mode. This is because the iron core is not designed with steady DC flowing through the windings. In short, the core saturates to a lesser or greater degree and this reduces inductance as well as undistorted power output. As the measured frequecy response was quite good, I can only put the peculiar sound to some kind of harmonic distortion.
Some P.A. line transformers, such as the obsolete M-1100 type, are quite good as is, but others work poorly. The transformers I used looked remarkably like the M-1100 having a very similar appearance and the same tappings. I suspected at first they may have been M-1100 clones. Indeed, a quick check on the frequency response with a CRO seemed to indicate all was well. The bench speaker I used for testing is a 4" type in a plastic cabinet so there was no way the amplifier's quality could be judged from that.

The way around the transformer problem is to operate the output stages in partial triode (ultra-linear) mode, or full triode mode. As far as I can tell, the lowered inductance of the primary winding is more suited to triode operation. Subsequently, I rewired the output stages in ultra-linear mode by feeding the 6BM8 screens from the 2.5K tapping on the output transformers. This improved things and I drove around for a week listening to the amplifier like this. However, it still wasn't quite good enough.

Next was to try full triode operation. This time, I used decent bookshelf speakers to test the amplifier, as I was aware that the car speakers have poor low frequency response anyway.
Now the amplifier sounded as it should.

The other problem encountered on the first drive was rather severe overheating. I could smell something warm about an hour into the trip. Reaching under the seat revealed the amplifier case too hot to touch. Once at work, I took the cover off so I could continue using it without it overheating. Evidently, my passive flow through ventilation idea wasn't good enough. Fan forced ventilation would have to be installed. For this, I found a processor fan seemed to provide adequate air flow.

Computer fan is required to force air flow. Dissipation inside the case is about 36W.

With a 12V input, the inside of the case reached 55 degrees C in a 15 degree room. With the fan installed, this was kept down to 23 degrees. To reduce the fan noise, I experimented with reducing the voltage to it. At eight volts, the temperature was 24C which seemed satisfactory, and the fan was much quieter. However, a few months after building the amplifier with the onset of summer, it was found the cooling was inadequate. This was fixed by running the fan at its full 12V rating. Given the 14V supply, I found 8.2R to be a suitable dropping resistor.

The Circuit.


The initial circuit was identical except that the 6BM8 screens were supplied from the 200V rail via 470R stopper resistors, as per Philips data. In this pentode configuration, output power was 1.8W into the voice coil for both channels. Input sensitivity at the 6BM8 triode grid was 212mV rms for full output.
When partial triode mode was then tried, the 470R resistors were removed, and the screen grids connected to the 2.5K tapping on the transformer primaries (not shown). Power then dropped to 900mW per channel. Frequency response did not change between modes. It was 55c/s to 12Kc/s at 3dB points.
For full triode mode, the tradeoff for lesser distortion was a much lower output, now 500mW per channel. This demonstrates just how inefficient triodes are compared to pentodes.
Neverthless, in order to use the P.A. line transformers, it was necessary to do it this way. Output power is still sufficient, and it is well to remember there are two such channels, effectively doubling the power. The .0047uF plate bypass capacitors are probably not needed, now the pentodes are wired as triodes, but they do no harm and it was easier to leave them in situ.

Rearranging the transformer laminations and providing an air gap has been suggested in the past, but I found that doing so was worse. The problem is the primary winding is wound for a certain inductance. When you convert to an E-I format with air gap, the inductance with the same winding is reduced. It would entail rewinding the primary with more turns to get this to work. Having said that, some 5W P.A. line transformers do have tappings up to 30K, and it may just be possible to compensate this way.

Aside from the pentode operating conditions, the circuit around the 6BM8's is conventional. Pentode bias is obtained with the usual cathode resistor and bypass capacitor. I used a lower value of grid resistor than is usual, to reduce the effects of grid emission which is exacerbated with high cathode temperature, which will result when the car electrical system is operating at around 14V.
Plate load for the triodes are 100K. Contact bias is used. This works by virtue of electrons being emitted from the cathode building up on the grid. If the grid resistor is the usual 1M kind of value, the electrons drain away rapidly. However, if the resistor is made much higher, the electrons don't drain away so quickly, and a useful negative voltage is obtained, sufficient in this situation to provide suitable bias. Typically grid resistors are 10M to 22M for contact bias. Note that this scheme is not suited to output valves and such because the voltage developed is insufficient, and the high grid resistor required would be problematic with regards to grid emission and thermal runaway. Therefore, it is suited best to high mu triodes handling a low signal voltage input.

Negative feedback is applied by means of the 4.7K and 150R voltage divider fed from the 4 ohm secondary. The feedback is applied to the 6BM8 triode cathode. Note that the 150R is not a bias resistor. Its low value at the low cathode current does not make any useful contribution to the bias.
Across the 4.7K is a phase correcting capacitor. Because of the phase shift in the transformer as the frequency rises, it is possible for positive feedback to actually occur. The .0033uF compensates for this. This capacitor is easily selected by feeding a square wave into the amplfier, because of its infinite harmonics, and selecting the capacitor for the minimum value that removes the overshoot on the waveform.
The feedback circuit was part of the original design as it is essential with pentodes. I questioned the necessity now with triode operation, but a quick distortion test revelead it was best to keep it. With the feedback, distortion is 2.6%, Without it, it was 3.5%. For those unaccustomed to valve circuits, these figures are very typical of single ended output stages of average quality. Without feedback, a pentode output stage has about 10% distortion. Yes folks, your average valve radio with no feedback produces 10% distortion, at least.

The DC Volume Control.
The 6BU8 or 6HS8 valve is something unique to the TV world. Here, it is designed to operate as a noise gated sync separator and gated AGC amplifier. The common control grid feeds the same video signal into both sections, but the plate and third grid circuits work entirely separately.
In this circuit, the left and right audio inputs feed the separate third grids, via the usual isolating capacitor (.068uF) and grid resistor (560K), and the amplified or attenuated output emerges from their separate plates. Gain is controlled by varying the bias of the first grid which is common to both sections.
To avoid the complications of a negative supply, the grid is earthed and the cathode (also common to both sections), is taken positive, acheiving the same thing. The 1.2K sets the intial and minimum bias, and the maximum gain required. The 10K pot then allows the cathode voltage to increase further to allow the gain to be at its minimum. A bleed current from the 100K ensures that sufficient voltage is developed across the 10K pot when the 6BU8 plate current falls as the bias increases. In practice, the 100K is set so the volume just becomes minimum with the pot at full resistance.
The 10uF reduces crosstalk between channels due to both sharing the one cathode.
The supply for the 6BU8 is stabilised with a 62V zener diode. This is to prevent changes in the supply voltage affecting the volume. An unforseen characteristic of 6BU8's was revealed during the building of the amplifier. I was rather surprised to find a large variation in gain between both channels. Trying a number of 6BU8's showed a lesser or greater variation. This seemed strange, for the valve has a common cathode and should therefore mean both sections are identical. However, with further thought, I could see what was happening. The valves I had chosen were all second hand. It would appear that the section used for sync separation would pass a different amount of plate current to that of the AGC stage. Therefore, the cathode surface associated with each section will wear at a different rate. The golden rule is don't expect second hand valves to work properly in this circuit.
I solved the problem by using a new valve.

Power Supply.

The power supply is conventional using a shunt driven non synchronous vibrator, transformer, and solid state bridge rectifier. It provides 200V at 75mA, most of which is drawn by the two 6BM8's.

Valve Heaters.
The 6BM8 heaters are connected in series for 12V operation. At this point, something needs to be said about connecting valve heaters in series like this.
The 6BM8 is not intended for series heater use. Proper valve data books will tell you if a valve heater is intended for parallel operation only, or if it's suitable for series heater circuits also. The reason for this is that parallel heater valves do not necessarily draw exactly the same current as each other, even if the same type number. Furthermore, the warm up time is not controlled. What this means is that parallel heater valves used in a series circuit may be exposed to excessive voltage during warm up, and may receive too much or too little voltage even when the valves have finished warming up.
With this amplifier, I did test a number of 6BM8's to find a pair with the same heater characteristics. There was quite a variation amongst randomly selected valves. For example, one would have 6V on the heater, and the other 7V, with a 13V supply. As the voltage rose to 14V, it was worse. Had I not been able to select from a sample, I would have included resistors across each heater to balance the current. Something also to keep in mind is that tungsten filaments gradually increase in resistance as they are used over time. This is due to the filament evaporating. It's something to consider when second hand valves are being used.
The 6BU8 requires a dropping resistor as there is no 12.6V heater equivalent.

The Transformer.
As previously mentioned, a Jaycar 20VA toroid was used because the current rating is just right, and the physical dimensions are ideal for the case. More importantly, toroids make very good vibrator transformers because of their high efficiency. For a B+ of 200V, allowing for drop across a filter resistor, the required AC output would need to be around 240V.
This requires a low voltage winding of 9-0-9V. See here for an explanation of vibrator transformer turns ratios.
Rather than buy a new 9-0-9V transformer, it so happened I had a few 15-0-15V types which would be ideal for rewinding. The transformer was found to have 163 turns on each 15V winding, resulting in a turns per volt ratio of about 9.6. Rewound with 91 turns, the voltage was 9.5V which was just right.

The Vibrator.
I decided to use one of the 3 pin Delco based vibrators I had. This type of vibrator was mostly, as the name suggests, used in radios made for GM cars. I have never seen a 6 volt type with this base, so assume it was part of the 12V changeover in the mid 1950's. This three pin base was never used in Australia, so the chances of getting a socket to fit would be nil.
The way around that was to use a capacitor clamp to mount the vibrator, and automotive bullet connectors on flying leads to attach to the pins. The outer two pins fitted perfectly, but for the centre pin the bullet connector had to be modified to make it suit a smaller diameter.
The vibrator I chose is a Teletronics 12V-3P. A rather meaningful part number, unlike most. I also tried a couple of Delco types and a repaired Cornell Dubilier, but thought I'd go with the Teletronics for no particular reason. I recognised the particular Delcos as being Mallory rebadges.

Circuit Design.
The timing capacitor was chosen to obtain the correct waveform. For the Teletronics vibrator, this was .01uF. The CDE required .005uF, and the Delcos required .02uF.
So you can hopefully see from this,  that just because a vibrator has the same voltage and pin configuration, it is unwise to change vibrators without checking the timing capacitor is correct.
For example, if I was to install a Delco vibrator in this amplfier, the timing capacitance would be inadequate, leading to a damaged vibrator.
There is no such thing as a one-size-fits-all timing capacitor. Its value is highly dependent on transformer characteristics, vibrator frequency, and duty cycle.
It is attention to detail and understanding how a vibrator power supply really works that results in complete reliability. The simple "buzzer switching a transformer" concept will lead to disappointment...
330R resistors are included across each of the contacts to dampen spikes when the contacts open.

Correct waveform and operating conditions makes vibrator life a non issue. This is the primary waveform of the vibrator transformer as fed with the CDE vibrator.

As a further check, the CDE vibrator was operated all day into the rewound transformer, with a 15W lamp load, and no sparking was visible. This particular vibrator had been abused in its previous life, but with care I was able to restore reliable operation.

Rectification is by a bridge rectifier consisting of 1N4007 diodes. Because what is now the secondary winding is not centre-tapped, it is not possible to use the contacts of a synchronous vibrator, nor is it possible to use a single ordinary rectifier valve. Half wave rectification is undesirable with vibrator power supplies because of the uneven loading of the vibrator contacts, as well as the DC component that will exist in the windings.
Filtering is by two 47uF capacitors and a 680R 10W resistor. Actually, the filtering is more than need be because of the high duty cycle and flat topped vibrator waveform. While the 680R only dissipates about 3W, a 10W type is used to spread the heat. It's also clamped to the chassis to get more of its heat outside.

12V Supply.
Current consumption is about 3A at 12V. A 6.3A fuse protects against short circuits. The switch mounted on the remote control includes a lamp, and for long life it is run at reduced voltage by means of a 120R resistor. It was easiest to pick the 12V source from the cigarette lighter as this was nearby, and has the current capability.
In reality, the supply voltage is more than 12V. The battery terminal voltage with the car running varies from about 13.8 to 15V. With the thin wiring in the car, by the time it gets to the amplifier, it's about one volt less.

Final Result.
This turned out to be one of those projects I wish I had done a long time ago. Now, driving is a totally different and enjoyable experience. Instead of constantly going through radio channels each time news, ads, or an annoying song comes on, I just relax with continuous good music that I like. I feel completely relaxed at the end of my drive, even if it was in peak hour traffic.
Postscript: After 14 months the amplifier is still performing perfectly. It is in use several days per week, sometimes for over four hours at a time.
Postscript-2: This amplifier went into service in May 2017. It's now September 2020, and the amplifier is still working well.
Postscript-3: It is now December 2022. It is now clear that:- 1) Good quality shunt drive vibrators are reliable. 2) Jaycar toroidal transformers make excellent vibrator transformers. 3) What Mallory says about selecting timing capacitors is correct.