Car Amp Design

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi Steve

Thanks for taking the time to reply.

I have prototyped that circuit with the 15/16 opamp configured as a unity gain buffer and it works very well infact, and it softstarts slowly when it comes put of protection. What exactly is the problem with it?

I have increased the resistance after the remote line very slighty while still maintaining 4 or 5ma or so through the zener and at least 1V across the base of Q8. This I feel should not stress the headunit.

As for my poor FET, i like using them :clown:
I agree in terms of cost engineering it isnt as good as two BJT's, but it works extremely well and I am not trying to get this amp through production - merely making it for my own pleasure. I have a whole bag of these FETs too :)
The 10uF is to keep it on for half a second or so and protects it from turning off under mild brownout conditions. The zener is to protect the FET from gate-source overvoltage. Not really necessary I guess cos it's 20V max and thats unlikely to be found in this circuit. Force of habit i guess - Poof! It's gone.
Again the S3G diode was there to provide a small drop to help keep the ICL7667 well within its SOA - but poof! its gone too


The 15V Zener is in fact a Transorb, or transient voltage suppressor and protects the circuitry from voltage spikes much like a zener would but far far quicker. It can absorb huge amounts of power for a very short period of time. It was initially there to protect my ICL7667 which is CMOS. I agree it is not completely necessary anymore, but it will not hurt and will protect the amp from reverse voltage connection until it or the fuse blows.
I work for a company that develops fleet manager on board computers for Siemens/VDO and we use just such a device on all our products. I have often come across a faulty unit where the transorb has blown but invariably it has blown into a dead short and in doing so protected the rest of the circuitry in front of it.

I have moved the diodes on the 7x15's as per your advice, and decreased the values of my primary- side snubber caps.

In this later revision of the design, there are no pots wired to the inputs of any 5532, only to the input of the amplifiers themselves.

As for the secondary inductors, I am a little hazy on this. Can I use a ferrite core like my main core or must it be iron? By common mode, do you mean both inductors wound in opposite directions on the same core? What value inductance would you recommend? I will be winding this(these) and my transformer myself on toroids.

As for isolating the grounds, should I just go the opto-isolated route with zeners? This works well and uses both rails for feedback. I thought isolating the grounds would not be necessary. I could also improvise a virtual earth with a voltage divider and an opamp, and then connect this as my signal ground and then connect it to power ground through a 100R or so to keep it near true ground potential.. what do you think?
I have implemented optocoupling with separate grounds in the latest revision.

Again, do you think the gain structure will be sufficient as is? I was worried about too much attenuation through the LPFs. I have a gain of two for the signals being fed through the LPFs, but no additional gain on the full range signal. The amp module has a pretty high gain configured as is tho.

Sorry to send u such a barrage of questions mate, perhaps a msn or skype chat would be quicker..

Anyways, thanks again.

Gareth
 

Attachments

  • main schematic rev3.pdf
    54.3 KB · Views: 262
If the Transorb is anything like the transient absorbers I found, it won't work as you want it to. The reason is that it has too high of a forward voltage (~0.8v). The intrinsic diodes in the Z44s have a lower forward voltage and will pass virtually all of the current. If you want to help protect the amp from reverse polarity, use another MUR820 across the power supply.

For the optocoupler circuit, you may want to use a low pass filter on the LED side. Also, there's essentially no reason to go from rail to rail since there is no way to control the positive rail and negative rail outputs independently. If you were building a mono amp, then there may be some argument for it but a bridgeable stereo amp will essentially load both rails evenly.

The instability of the PWM control circuit (if there is any) may not be noticed until you do some critical low level listening on some good speakers in a quiet environment.

For the totem pole drivers, you could tie the emitters together and use a single resistor. The resistor is actually not needed but can serve as a sacrificial component when/if the FETs fail. A 1/8 watt 10 ohm resistor would be enough to do the job. This will help to protect the drivers and the gate resistors in the event of a failure.

In my opinion, the overvoltage lockout isn't really important. Even if your alternator's regulator failed and caused it to produce 17+ volts, the amp would likely be OK. If you freed-up that op-amp, you could use it for the soft-start. Then you could tie the protection circuit to pin 4 instead of pin 16. This would allow you to use feedback between pins 2 and 3 for stability. Pin 16 would be tied to ground. Pin 15 could be connected to 5v.
 
Car amplifier design

Dear Gareth,

Here are some pointers.

TL494 - should not be operated with pin 3 joined to pin 15. You must have loop compensation between pins 2 and 3 - a series RC is fine. Simply join pin 15 to 5v ref.

Use a simple form of single ended feedback. Simply put pin 1 to a pair of resistors, one to ground and the other to +35v. The gnd end use 2K4 and the +35v end use a 33K series a 5K trimmer to set final HT.

Put a cap from pin 12 to gnd - 10-47mfd OK

Take out those 10 ohm resistors from your totems they only hurt performance.

If you wind your transformer with interleaved windings the leakage inductance will be low and then you can use lower gate resistors which will improve the switching of the Z44s

The common mode coil MUST be on a composite ferrite/powdered iron core of about the same size as the main xfr. The PI must be about 80-85% of the core and the ferrite (same material as your xfr) the balance.

Use #15 wire (or 2 x #18 in parallel) for each winding

Wind one winding around the complete circumference of the core with one layer. Start this winding at say 6 o'clock on the core.

Eind the second winding starting at 12 o'clock and place the same # of turns as the first winding.

From say your +ve rectifier come in on the first winding at one end (You choose this). Now the -ve rectifier connects to the other winding but on the end that is opposite to the first.

So if the first end came out of the top of the core, then use the other winding's end that comes from the bottom of the core.

If you mess this up you will find out quickly enough when you tuen on and look at you switching waveform on a scope. Always turn on through a series 2 to 4 ohm 50w resistor which is is series with your +12v supply and then bypass when power testing. We have done this forever and it sure saves amplifiers.

You MUST separate your power grounds (primary side of xfr) from your audio grounds (secondary side of xfr) To make the simple single ended feedback work put a 220 ohm 1 watt resistor between these gnd. if they are joined you WILL get alternator whine as there WILL be a ground loop once you connect your RCA inputs.

Pre-amplifier section:

Get rid of U! as it serves zero purpose. Your schem is wrong as U1 has no bias on the + inputs.

Change U5 to an inverting stage with a gain of say 2 (This will limit your maximum input to about 4.5 volts and I know of no head that puts out this kind of signal) I have tested too many heads with BS specs and NONE of them put out more than 1 to 1.5 volts at full volume. Configure U5 with say a 15K input resistor and then use a 30K feedback resistor with a 22pF in parallel.

Your LP section has 0dB gain in the pass band so there is no need for gain up. Simply split your signal from the inverting inout stage and feed the LP opamps together with the selector switch.

Now from the poles of your selector switch feed a variable gain inverting opamp stage which does two things. It brings the signal back to normal phase and it allows you to vary the gain over a wide range.

Let's assume we shall have a 200mV to 4v input sensitivity and that the main pwr amps can put out 22 volts.

Make the main amps have a gain of 11 x so they then need 2 volts to drive them to 22 volts.

The variable gain inverting stage is made from 2 resistors a dual pot and 33pF cap. it looks like this. Ground the + input of course.

Join the -ve input of the opamp to the slider of the pot. From the CCW end put a resistor to the output of the opamp and this is a 5K8 in value. Then from the CW end put a 1K6 which is fed from the poles of your selector switch.

Put the 33pF cap from the output to the -ve inputs. The second channel is identical of course.

So your full range signal path has only two opamps in it and your LP has the Linkwitz opamps as well.

This makes it simple and elegant.

Back to the power supply.

There is no need for over voltage protection as you have a regulated power supply.

Add more capacitance to your main 35v supplies, 1000mfd is not enough, go to at least 4700mfd and use 50v caps

There is no need for the 15v zener across the +12v supply

Put back diodes across your 7815/7915 regs to cure their latch up problems on turn on. 1N4002 are OK

Not knowing your power amp schems I cannot comment.

I am not a big fan of NE5532 there are many nicer opamps which sound way better.

By the way Cape Twon is my favourite city in the world. I lived in JHB all my life but went to the Cape many many times. I just love the scenic drives and Paarl was the best.

Is Chapman's Peak drive still open? That is a spectacular view which these guys in the USA can only dream about. Is the Hout Bay hotel still there? It used to have a restuarant called the Plaza Espania which served a Sunday evening buffet for R1.50 eat what you want --- crayfish tails till you burst and all the other good things we like to eat. This was back in the 60's and in the 70's when the price went up to R7.50 per head I thought we were being ripped off!! How times change.
Regards

Steve
 
Assuming that the inverting stage is left as is...

If U1 is removed, the amp (in full range) would not be able to be driven from a passive volume control (i.e. a pot inserted in the preamp line). It would have to be driven from an op-amp (used in virtually all commercially available head units and signal processors). The reason is that the resistance of the passive control changes the gain of the inverting section. This would cause the gain of the left and right channels to be different.
 
Amplifier design

Well NOBODY I know of runs their car amplifiers from a passive volume control. All car systems I have EVER seen have a head unit either CD/Cassette or Mini disc. To the best of my knowledge no head has a simple passive volume control. They all have electronics post the volume control and furthermore it would be suicide for a head manufacturer to supply a unit with a passive pot on it's output. The output impedance would be high and varibale over a wide range and any cable capacitance (5 metres or so) would hurt the high end.

Why one channel would unbalanced from the other (besides the natural inbalance between sections of ALL potentiometers) is beyond my knowledge of electronics.

"If U1 is removed, the amp (in full range) would not be able to be driven from a passive volume control (i.e. a pot inserted in the preamp line). WHO DOES THIS? It would have to be driven from an op-amp (used in virtually all commercially available head units and signal processors). The reason is that the resistance of the passive control changes the gain of the inverting section. YES WE KNOW THAT This would cause the gain of the left and right channels to be different." WHY?????

If the sliders of a dual passive potentiometer drive two identical invering opamp stages then the two channles WILL track equally within the tolerance of the potentiometer.So forgive me I do not understand the purpose of your post

Stephen Mantz

Zed Audio
 
I know lots of people who hate the digital volume controls whether they're rotary encoders or up/down switches. They often use a potentiometer mounted in their consoles to have a more convenient/user friendly control. Sometimes it's done for the signal feeding all of their amps and sometimes just for the bass. The Radio Shack 100K stereo volume control is the pot of choice in most cases.

The gain of the inverting op-amp is dependent on the input resistor. When the pot is near the middle of it's range, the resistance between the wiper and the signal source is added to the value of the input resistor. The higher input resistance lowers the gain in the middle of the potentiometer's range of travel. This will cause a gain imbalance between the left and right channels.

In his amplifier, if he removed the input buffer, in the full range setting the left channel signal drives a non-inverting input. On the right channel, the signal drives an inverting input. These are not two identical input stages (when you remove the buffer and run it in full range).

I noted that all commercially available units used op-amps on their outputs.
 
Car amplifier design

Dear Perry,

I did not ask Gareth to make one channel inverting and the other non inverting. I asked him to make BOTH channels inverting.

Good luck to those who want to run the outputs of their head into a 100K control and then run this fragile high impedance signal down an RCA cable to the amplifiers in a the nastiest noise situation an audio signal finds itslef - car vehicle.

For those who believe that one more digital pot will make a big difference - good luck. Does anyone know what the small micro volt signal wentr through to finally get put onto a CD or Vinyl?
If they did they would not worry too much about one more pesky control especially in a car!

Steve
 
I'm sure I misunderstood something in your post but... If he has 2 identical channels at U5, how is the amp then going to be bridgeable?

Wouldn't he have to have an inverting channel at some point? This was done at U5 (the rightmost U5B).

Unless I'm mistaken, the first (leftmost) U5 is supposed to be U2. That's where you wanted the 15k/30k resistors.

When he started this thread, I wonder if he REALLY wanted this much help. :)
 
Amp design

Simply make one of the power amp channels inverting or add a simple inverter opamp somewhere.

The help I offer is free and because Gareth is from my home country I have some kind of "kingsmanship" to him.

I have many years of designing car amps and he will have many problems with his design.

Steve
 
Hi Perry and Steve

Thanks for the feedback.

Perry - I fear you may be right about the FET's diodes having a lower drop than the transorb but I will check when I am home. The transorbs have a 0.55V forward drop.


Steve :-
I think I am going to can the over / under voltage protection. It serves no useful purpose as the transorb clamps the B+ line to 15V and if the remote line drops below 10/11V then there is not enough voltage on the base of Q8 to bias it on and hence turn on the FET Q7 and the rest of the amp.

I also think I will use your idea of latching the protection on. I have thought of using a two-transistor cct to do the latching or perhaps jsut getting the opamp to latch itself on with a diode? Otherwise there is always ttl :p

Using a resistor to separate the grounds is a clever idea that I wish I had thought of myself. Of course the feedback will work! Silly me, that allows me to negate the messy zener / optocoupler setup and the non-linearities associated with it- it is also more flexible.

I cant believe I forgot the ripple cap on pin 12. There is a 47uf there now.

I removed the 10E resistors and reduced the gate resistors to 22E.

I have had another look at the opamp stages. I really like the idea of a variable gain stage - Im sure it will help SNR at lower signal levels. Now I realise why car amp volume controls can seldom be turned right off - cos they are var gain and they cannot have a gain of 0! I will try get 4-ganged pots so that I may use var gain and passive ctrl together, that will give me the best of both worlds.

I have also added 15K bias resistors from the caps to signal ground at the input of the amps.

As for opamps, have you any better suggestions? I merely specced 5532s cos I thought them better than TL072s or 4588s but I am (as always) open to suggestions.

I will need to give the structure of the opamp stages some more thought. As Perry said, I have structured the design so that one channel inverts and the other does not, but I cannot have that mismatched stage as my input stage... hmm.

BTW, you said you cannot comment on the amplifiers cos you haven't seen them - they are posted in a separate .pdf here. They are based on the TDA7293 IC, almost straight out of the datasheet configured to drive low-impedances.
You said I should increase my rail caps, but there is only 1000uF on the output of the SMPS but a further 4000 per rail in each amp module (8000 per rail total). These will be placed right next to the IC's themselves.

I have moved diodes on the 7815 / 7915 so as to prevent lockup as suggested btw.

Thanks for the help re the choke, as this was a very hazy part for me. I had heard of common mode chokes but I didnt know how to implement one. I am trying to source my toroids at the moment.

As for Cape Town, ja I think it is the most beautiful place I have ever been (if a bit windy sometimes). I havent been to America before but I have been through Europe.

I stay in Bergvliet, next to Constantia in the southern suburbs but I work in Stellenbosch so I have a fantastic drive along the coast every day :)

Chapman's Peak was closed for many years after a rock fell on a motorist's car. It has since been reopened with a R20 toll to go from Noordhoek to Hout Bay or back but it is looking stunning. I will happily send some photos or a post card if you are interested - my girlfriend has been bugging me to take her round it for a while now :D
R7.50 for crayfish? You have been gone a while heheheh. Two litres of coke will cost you more than R10

Thanks again for taking the time to help me here, I really really appreciate it and I assure you I value your advice.
I am trying to learn as much about this particular field of electronics as possible and I feel my knowledge has increased a lot lately just through doing this design. Switchmode's have always fascinated me.

I will work the schematic this evening and repost it when I am done. I need to lay out the board soooon so I must finalise it quickly..

Kindest regards
Gareth
 
If you want to bring the signal level up abve the noise floor for both the low pass and full range sections, you could take the signal from U2 (pins 1 and 7) to feed the full range section. If you do that, U1 could probably be eliminated but I generally like to use a unity gain buffer at the input of anything I build.

The 5532 is a good op-amp but you should also look at the MC33078 and the OPA2227. I've used both and they are very quiet. If used throughout the preamp section, the noise produced by the amplifier section will be significantly more than that produced by the op-amps.

If you decide to use the variable gain section, look for Vimex or Piher potentiometers. Both have pots with hex holes that allow stacking.

If you're working with a prototype that's easily modified, you could try this for the input circuit... Omit U1. Use U2 as the input but instead of tying R7 and R17 to ground, connect them to the input shield ground. Change R7, 8, 17 and 18 to 20k each. Connect the positive inputs of U2 to the center of a voltage divider using 10K resistors (one divider per channel). The top of the voltage divider would be driven with the input signal. This gives you a balanced input with the same impedance on both the signal and the shield inputs. I've used this a few times and it works well. It works even better if you buffer all 4 inputs with unity gain amps. For this type of input, the secondary ground could be directly tied to the power ground. All inputs can be driven through 22uf bi-polar caps. The resistors for the input circuit should be 1% for best rejection of engine noise.

If you use the buffers for the above circuit, you can drive signal into the negative input of the right channel (the signal from the right shield would go to the positive input of U2b). If you do this the right channel would be inverted and you could use the variable gain that Steve described.
 
Hi Perry - I will experiment with the balanced inputs as it sounds like a nice way to get rid of noise altho I dont think I will go that route for this V1 of the design.

Steve - Have a look at the latest rev, has quite a few changes including a discrete latch for the TL494 protection and variable gain stages.

I must agree that I like having a unity gain buffer at the input too.

Please forgive the awful annotation. I will correct it when I am sure the circuit is finalised so that I can generate my BOM.

Thanks and regards

Gareth
 

Attachments

  • loudlester v1 rev4.pdf
    53.8 KB · Views: 213
If you use a diode in place of the 1m resistor in the thermal protection circuit, the op-amp will latch. If the op-amp tries to latch on power-up, add a small cap from the positive input to ground to delay the positive input's rising for a fraction of a second. If you get the op-amp to latch, you can lose the transistor latch.

Why not use the pot as the reference (on the negative input of the op-amp) for the thermal protection? A fixed value resistor would go from the thermistor to ground. If you want to get rid of the pot, look up the curve for the thermistor. Find it's value at the desired cutoff temperature. If the value is, for example, 8k and you use a 10k to ground, you would use an 8k and a 10k as a voltage divider feeding the negative input of the op-amp.

If you don't need a 4ch op-amp, an LM358 will work for your application.

One thing I think you should do is tie the wiper of the gain pot to one end (pin3) of the pot. This will still give you variable gain but if the pot ever gets 'dirty' and the wiper loses contact, the op-amp won't be allowed to go to open loop gain. The gain values will change but you can recalculate the values to give you the gain you want.
 
Hmmm, a diode eh? I thought that might work. Oh well, so much for my poor little transistor latch :( I'll give it a try.

I think ur right about the pots, I guess I must recalculate the gain then. *sigh* this is so much more work than I thought it would be.

I agree I dont need a 4-channel opamp anymore, but I have a tube of the little bastards and they are SO-14s so I'm unlikely to use them elsewhere. Also if I change opamps now then I must reorder SMD LM358s cos I only have leaded ones. Also it leaves room to use them in later revisions (if I have any), I will just tie their inputs low for now.

I have selected my cores:

Ferroxcube TN36/15 for my trafo with 3C85 core material good for upto 200KHz. Its a little small (36.6D x 22.3d x 15.6h mm) so I will step up my switcher to 100KHz or so - I will prob need to increase deadtime then.

Micrometals ST150-267 composite ferrite / powdered iron core for my common mode choke, 200uH at 6A and 1.5mH at idle (200mA). Dimensions 38.6D x 21.2d x 15.9h mm

Will work the schematic more tonight when I have some peace :rolleyes:

Thanks for the help / advice

Gareth
 
OK! Cores are on order at last.

I have ordered the ferroxcube 3C90 core for my trafo, the micrometals 26mix P-I toroid for my input filter.

I also ordered 2 ETD44's and 2 ETD49's to experiment with. I had no idea they were so cheap. The ETDs run around R10 each without bobbins (call it $1.50) and the toroids were about double that for some strange reason.

Unfortunately I could not get the composite core I wanted for my output filter so I guess I will use another PI core for that. Pity.

Otherwise parts procurement is going very well. I could only get through-hole 5532s :mad: but at least now I can mount them in sockets and swap them for better opamps when the opportunity arises.

I will take some photos of the parts and the housings soon, I am having them anodized next week. Faceplates will be made of 5mm perspex which I will have laser-cut :) No drilling for me!

Take care, and thanks to those who posted.
Herewith latest schematic (hopefully final).

Last question - do you think I should increase dead-time if I am going to switch at 100KHz?

Thanks!
Gareth
 

Attachments

  • main schematic.pdf
    51.3 KB · Views: 201
If you can properly control the FETs, you probably don't need to increase the DT. I looked up the core material you're using and you'll get 4x more core loss at 100kHz than at 25kHZ.

Since you're already using the dead time input for soft-start and the comp pin is being used for stabilizing the reg, there are a couple of other ways to vary the dead time. You can manipulate the Ct and Rt values to vary the dead time slightly. You could insert a low value resistor (~4.7-10 ohms) in series with the emitter of the NPN transistor in your totem pole driver. This can also be used to control switching transient problems. Having the resistor only in series with the NPN doesn't slow down the turn off time of the PNP transistor.

You'll likely have to lower R52 and R60 to ~330 ohms. I doubt that the 1k is going to be enough at 100kHz.

When you initially power up the supply, you'll probably be at less than full duty cycle due to the over-winding of the secondary. Turn the power supply (the DC power supply) down until the chip goes to full duty cycle. Watch the current flow. If the current flow increases at lower voltage, you may need to increase the dead time. I've seen many instances where a power supply seemed fine when the duty cycle was cut back even though there were problems. When it's at full duty cycle (B+ in reduced), use a heat gun to heat the core to at least 80C. Some cores change value enough with temperature to cause problems at higher temps. If you're running at 100kHz, it's not likely to be a problem with 4+4 turns on the primary but it's something you should check.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.