DoZ (P36) Design Problems and Solutions

[EchoWars]

After fighting my battles with the transformer hummmmm on my DoZ, I thought I had the bugs worked out. However, there was one problem that really bugged me...on power-up, the bias current is very very high...my meter, which is slow, shows about 5A while there is still 35+ volts on Q3 (see schematic). I suspect that, if I could see instantaneous values, I'd see 6 amps or more current through the output stage while nearly the whole supply voltage is still on Q3. It falls 'fairly' quickly, but I was worried about the instantaneous values that I could not really see and suspected that the SOA was being violated. I wrote Rod and he replied:

Start-up current is usually high, but if you decrease the value of C6 you can speed up the 1/2 voltage settling time. The value selected was based on a compromise between start-up current and speaker thump. You could try a lower value, and noise shouldn't be a problem if you have the zener stabilisation in place.

I liked the silent start-up, so I left C6 alone.

After about 3 weeks, one channel blew. It was not overheated, as this amp is fan cooled. All it could be is SOA violation at power up hammering away on the output transistors, eventually killing 'em.

I have a buddy, ilimzn, who frequents this forum, as well as Audiokarma.org. I asked him to look into this huge current spike at power up, and perhaps suggest a way (besides using more than one pair of transistors) to make this a relaible amp. Here's his take:

There are a couple of things I do not like about the execution of the current source in the DOZ. it's almost as if Rod could not decide whether to use a voltage or current driven current source, and this I think is the root of several problems.

The reason you have a turn on current spike is because the time constant formed by VR1 and C6 is longer than that formed by the VR2/R6/R7 divider and C4 - a LOT longer. Whie C6 is charging, the output of the amp is pulled close to ground level which allows C4 to charge to a far higher voltage than under normal operation. The current into the base of Q2 is determined by the voltage drop through VR2 and this will be several times higher at startup - normally, with a 40V power supply, VR2 only sees about 33 - 20 - 2*(2(Vbe)) ~~ 10.6V, while at startup it will see about 3 times as much. Couple that with the gain of the transistor increasing right through that range of currents, and you get well over 3x the current compared to normal operation out of that current source.

Let me back up here a moment. The absence of a noticeable turn-on thump is assured by C6 charging much slower than the time constant made out of the output cap and the load, assuming 8 ohms, this would be ~~ 37.6ms, which BTW may be a bit low, yielding a -3dB point at 26.6Hz. VR1 and C6 need to form a time constant about 10x that to get a maximum 1/10th of quiescent voltage on the speaker during turn on )I'm doing a lot of simplifying here). Since the time constant is also influenced by the input stage of the amp, some testing and correction will be necessary.

Now, to avoid the turn on increase of current source output current, (VR2||(R6+R7))*C4 needs to be at least equal and if possible much higher than whatever you choose for the VR1-C6 time constant. This way, the input drive of the current source will be clamped down by the charging of C4, preventing the current spike. That should be enough to solve the immediate problem.

There are a couple more issues, which all stem from the unusual way the current source is designed. The good part is the xistence of C4, this is a bootstrap capacitor which enables the current source to work right to the point of saturation of Q2. The bad part is that the current source is actually a current amplifier, so it's output current equals gain vs input current. The only reason why it does not go into thermal runaway is that transistors are not ideal, and I do not like that. higher temperature will invariably mean a higher surrent for this source, and that in turn may mean a yet higher temperature, etc. It seems like the intention was to use the voltage on C4 as the reference value for setting the current. If so, the circuit is missing a crucial element, an emitter resistor for Q3. Your question about current sharing is on the right track. Ordinairly, you want at least 50-60mV on it, which will be a negligible reduction in maximum output. The smallest practical resistor will however be 0.1 ohm, so you can either live with 150-170mV of headroom loss or you can use two in parallel to reduce it to half that value.The current source will perform much more conjsistently since now we have a local feedback loop - of particular note should be better temperature stability. Given a minimal emitter resistor, the current source could probably be made completely thermally stable for nearly any temperature by replacing R7 with a resistor + diode near heatsink string (the resistor would be about 1.5k) but this could also be overkill. A 0.1ohm resistor should tame the circuit just fine.

Q5 does not need an emitter resistor because it is within the feedback loop and therefore already stabilized.

Well done!!

I replaced C4, the 100µf 63V cap with a 2200µf 10V cap, and I reduced C6 from 470µf 25V to 330µf 25V. C4 does not need to be rated at anywhere near 63V...at power up, the voltage on it rises to a max of about 5.2V, and then falls to about 4.1V as the bias stablizes. I also got a .05 ohm 3W resistor designed for current monitoring and stuck it on the emitter of Q3.

NOW when I fire up the amp, the current rises slowly to a maximum of about 2.5A instead of 5+ amps, and then falls to about 1.5A where it slowly waits to rise as the transistors begin to heat up. Also, because the current rise at power up is slow, C6 has a chance to charge, and instead of a large portion of the supply voltage being on Q3 when the current peaks as it used to, the voltage is almost centered when the current peaks and therefore is well within the SOA of the output devices. I've also noticed that the addition of the .05 ohm resistor has made the bias current much less dependant on the temperature of the output devices...

I would add that if parallel output devices are used and current sharing is working, the amp may very well deal with the turn-on current spike just fine. My amp and others, using only one pair of devices per channel, just can't handle that huge current.

I just wanted to share the info for others, as the DoZ seems to be a popular project. I'd bet Rod would have tackled these problems had he actually built the amp and used it for some length of time, but he admits that he has not. But I have read numerous posts about the DoZ here that go something like "It sounded pretty good until it blew up for the 5th time". Perhaps with these modifications the DoZ fans here might be able to enjoy a bit of reliability to their amps.

[

[EchoWars]

Interesting to see that, even with all the DoZ aficionados here, that no one has a comment.

So be it...just trying to help. I know that several here could not keep theirs running long and I thought they might have been interested in a solution.

[ilimzn]

Well, I could comment that I am ashamed of my own typos in that quote above
There was a thread on the DOZ, maybe this should have been posted there, or at least referenced from there. Or maybe people are more interested in gold leaf papyrus in virgin olive oil capacitors wound by peruvian virgins during a fool moon (now movin into my flame proof bunker )

[ultrachrome]

I'm just lurking. I won't be playing with my DoZ until I have more efficient speakers. Although I am curious to see what others make of this.

I am interested in the evolution of this design.

[sith]

good work Echo... and ilimzn

I'm workin on my first DoZ, and its nice to see this thread...
Planning to use parallel outputs, still I think I'll make these mods

[EchoWars]

Well, I could comment that I am ashamed of my own typos in that quote above

Lol!! I fixed a few, but the concepts are there, so we don't care about stinkin' typos...

I can't thank you enough for your help...I consider myself quite fortunate to know you.

sith, the changes have made a definate improvement in reliability. I would recommend them to anyone.

Here's something Rod doesn't give you:



Note that I show the 1K resistor for the current limit for the zener as well as the cut trace, and the zener is on the lower right. I also do not trust the pot at the input to be the only DC return, so I added the R702, the 464K resistor. I also show the trace I cut to add the .05 ohm resistor...I drilled holes in the board and put the resistor on the foil side to keep it all neat. Lastly, no filtering on the zener bugged me, and I added the 47µf 50V cap.

The capacitor, C6, could probably be reduced even further to 220µf and that would probably reduce the current spike at start up even more, but the values shown work fine, and I'm not worried about 2.5A @20V on the output devices for as short a time as it lasts.

[Greg Erskine]

Hi EchoWars,

Thanks for posting your findings on the DoZ. I hadn't replied as my DoZ test amp has been out of action for last week waiting for the temporary wiring to fixed up. Your post has prompted me to complete the wiring. I'm using the pre-zener circuit and so far, nothing has failed after a month. I hadn't noticed the surge current issue you mention until I turned it on tonight. Initially my multimeter showed a overload condition and by the time I changed scales the current was at 2+ amps. It soon dropped to normal. I haven't had a chance to let everything cool to try again.

I realise you have been concerned with reliablility but have any of your mods had an effect on the sonics of the amp.

BTW your link worked the first time I looked at it.

[still4given]

Hi guys,

I have been seeing a lot of posts about the DoZ.

May I ask what the reason is that so many folks build this amp? I have never heard one. Is there something special about the way they sound as say opposed to Rod Elliot’s P101 or P3A?

I am very new to this hobby so this may seem like a dumb question to some of you, but I have read many posts about this amp and have yet to really understand the attraction for this design.

Thanks, Terry

[ultrachrome]

The DoZ, JLH, Zen variants are all class-A amps. P3A and P101 are class-AB.

I think the idea is to eliminate crossover distortion and produce "happy" distortion somewhat like tube amps. Check out Rod's Class A Amplifiers - A Brief Explanation. Last updated 5/02/2005.

Also look start looking at Nelson Pass' Zen articles at http://passdiy.com

As to why DoZ v. JLH v. Zen? I chose the DoZ a couple of years ago as my first DIY amp. I knew nothing at the time (I know next to nothing now). With a whole forum dedicated to Pass and his throngs of admirers, I chose to root for the underdog. Largely an emotional decision. Also, the DoZ seemed more economical to build at the time, but I cannot tell you if that is actually true.

I'd like to go through the Zen series soon and start to really understand how all this stuff works.

[still4given]

Hi ultrachrome,

Thanks for the reply. So, in your estimation, does the DoZ actually sound "happier" than a P101? I have built the P101 and it sounds really good.

In almost every post I see about the DoZ, the guys are asking what they can do to make it sound good. They also are asking what they can do to get rid of some this heat.

In my mind, that heat = wasted electricity. Do these class A amps use more energy than AB amps of the same wattage?

Thanks, Terry