Ampeg SVT-5 Pro distortion

[Grant_More]

No worries Barry! It's difficult to disguise the Scottishness!

Cheers,
Grant

[Mars Amp Repair]

Say great video!
I'm wondering if the fan might be a bit worn causing it to stall when it is supposed to start up, hence causing the extra current draw as well as the lack of heating. I have seen some of these fans do that.
I like the idea of extra heatsink on the regulator anyway. glen

[Steve Conner]

Excellent work Grant! Sounds like you've discovered a design flaw in the thing. The regulator won't get cooled until the fan starts, but by the time the fan wants to start, the regulator has got so hot that it can't deliver the current needed to start the fan. We can forgive you for taking brown sauce on your chips instead of vinegar.

Increasing the heatsink seems like a sensible idea to me too, or maybe extend the leads and mount the regulator on the chassis.

Or you could fix it Glasgow style, just short out the fan control circuit so it runs all the time.

[Grant_More]

Hi Barry,

I've got another thread running on this, but thought I would cross-post my thinking here incase it's useful for you.

I think the Ampeg design team have calculated the heatsink requirement without taking into account a couple of important factors and have hence fitted a heatsink which has too high a thermal resistance (i.e. is not good enough at getting the heat out of IC4, causing it to overheat in still air (when the fan’s not running)).

Let me explain.

We know that IC4 is getting too hot with the stock heatsink in still air (no fan) so we need to decrease what's called the thermal resistance between the silicon chip die (the working electronic bits) inside IC4 (which we call the "junction") and the free air inside the amp case (which we call the "ambient”).

We can calculate the thermal resistance of heatsink we need in still air (no fan) based on the power which is being drawn through IC4 and the maximum junction temperature of IC4.

The maximum operating junction temperature of IC4 is 125degC as per the manufacturer datasheet. At some point above a junction temperature of 125degC, IC4 will sense an overheat and shut down. If we keep the junction temperature of IC4 at 125degC or less, it will never shut down and our problem is solved. Great!

Let’s assume the maximum ambient temperature (Ta(max)) inside the amp case is 40degC when we turn our amp on. At this heatsink temperature the fan will most probably be running, but to be safe let’s assume this is the turn-on temperature of the fan and it’s possible it won’t run until this heatsink temperature is reached.

This means that the maximum allowable temperature (Tja(max))difference between Tj(max) and Ta(max) is 85degC in still air (no fan). So, with maximum power dissipated in IC4, we’ve got to rate our heatsink to ensure that we can get rid of enough heat to ensure this temperature difference is never exceeded.

How much power is dissipated in IC4? Well, we can consider IC4 to have a maximum rated current of 500mA and the voltage difference of 22-15V, so we’ve got 500mA running through IC4 with a 7V voltage difference. That’s 3.5W of power dissipated in IC4.

We’ve got 3.5W dissipated in IC4 and a maximum Tja(max) of 85degC, hence our thermal resistance between junction and ambient (this is formally called Theta JA, but we’ll call it Rja here to keep it simple) can be a maximum of 85/3.5 = 24.3degC/W.

Now, there are several thermally resistive interfaces between the junction and the ambient and we must consider the thermal resistance of all of them when calculating the heatsink required. The interfaces here are:

1. Interface of junction to case (Rjc)
2. Interface of case to heatsink (Rch)
3. Interface of heatsink to ambient (the air) (Rha)

From the manufacturer’s datasheet for IC4, the interface of junction to case has a thermal resistance (Rjc) of 5degC/W.

The next one (case to heatsink) is debatable. Ampeg have been cheap here and haven’t applied any thermal transfer compound between IC4 and the heatsink and hence I think this is a non-trivial value. I’d guess 5degC/W without the compound. In any case, we’re going to apply heat transfer compount to the new heatsink, so let’s be conservative and assume a 1decC/W thermal resistance here with the compound applied.

Finally, we have the thermal resistance of the heatsink, which is what we’re trying to calculate.

So, Rja = Rjc + Rch + Rha. We want to know Rha, the thermal resistance of our heatsink to ambient.

Rearranging gives us:

Rha = Rja – Rjc – Rch.

Filling in the numbers gives us:

Rha = 24.3 – 5 – 1
Rha = 18.3degC/W

This is the magic number we need for our heatsink thermal performance, assuming we use thermal transfer compound.

Interestingly, if we don’t use thermal transfer compound, using my “5degC/W without the compound” assumption above, the heatsink would need to have a Rha thermal resistance of

Rha = 24.3 – 5 – 5
Rha = 14.3degC/W

Guess what, the heatsink which Ampeg have used has a thermal resistance (Rha) of 21decC/W (see Farnell part number 171-0625). Too high a thermal resistance – IC4 may overheat under the conditions described! ;0)

I think they’ve forgotten to take into account the thermal resistance of the IC4 case and the interface between the case and the heatsink. Hey, everyone makes mistakes – it’s such a great sounding amp, I don’t mind doing their heatsink maths for them! =0)

The great news is that Farnell make a heatsink which has an identical footprint to the stock heatsink, but is longer and has a thermal resistance of 16degC/w. See Farnell part number 171-0628.

I think we can pop the old heatsink off (it just slides off once the glue is released, cut it with a knife) and pop the new heatsink back on, of course with a light smear of thermal transfer compound (it’s white stuff – from any electronics store or from Farnell part number 317-950. A very light smear is all you need, don’t glug it on or it will act as an insulator and increase the thermal resistance.

So, if I’m right here – we have a sub £5 ($10) fix which will take ten minutes and hopefully completely fix our amps.

I’ve got a couple of these heatsinks on order from Farnell due Tuesday, I’ll let you know how I get on.

Cheers,
Grant

[Grant_More]

Hi Steve,

It's brown sauce all the way mate!

Great suggestions there - could just have the fan running full tilt all the time, bit noisy, but you're right it would fix the problem!

It bamboozles me how Ampeg managed to miss this - it's such a subtle, but annoying problem. Maybe it's why they stopped selling these units!

There are some nice big chunks of alloy begging to be drilled and tapped for a TO-220, but it would probably mean taking the main PCB out. I'm thinking if this is design flaw (which it looks like it is), everyone with one of these will have this problem and I guess that most folks won't want/have the know-how to take the guts out. If we can come up with an easy (and non-technical) fix, everyone's a winner!

Glen, thanks very much for your suggestion. I actually just replaced the fan with a brand new one of near idential specification - same issue. Might be an idea to characterise the load current versus supply voltage for the fan, perhaps there's a peak at low voltage causing the high current (certainly looks that way from experimentation). It's 250mA rated and I think the tube heater is around the same - a lot of load for a 500mA rated regulator!

Farnell parts arrive Tuesday - I'll keep you guys posted!

Thanks again!

Cheers,
Grant

[Mars Amp Repair]

Well, it all sounds easy enough. Adding a bigger heatsink is a very intuitive solution sans all the formulae. I will say tho that I have repaired many of these amps & have never had anyone mention this an issue with cycling power.

did anyone ask if this particular amp was cycling as a result of the fan stalling?

I wonder if this is more of an anomally than a chronic issue. As mentioned before, it may be possible that the whole issue revolves around a worn fan that is dragging or stalling & hence increasing the current through the regulator that cannot be cooled sufficiently as the fan has stalled or slowed.

If the issue results from the fan stalling or dragging, wouldn't it be prudent to replace the fan as well as add the heatsink to preclude premature thermal shutdown of the regulator as a precaution?

Either way I believe adding a larger heatsink to the regulator is certainly a viable, reasonable, & inexpensive solution esp if these fans have a propensity to wear & stalling. I am glad to have been privy to this information. glen

[MarkusBass]

Grant,

Can you measure the current drawn by the fun when it's working correctly and when you stop it with your finger?
It seems to me that doing so you just increase the current, which causes all other problems. In this case the Q1 transistor is still conducting. Normally, when the fun stops the transistor is not conducting. Then the current is much smaller and this problem does not occur.
Does this problem happens also without stopping the fan? How old is the fan? Maybe it has problems with start.

Mark

[Grant_More]

Hi Glen, Mark,

Great suggestions and awesome source of debate as to why we're seeing this issue - thanks!

Glen, apologies, I did reply to your suggestion of a dragging fan, just in the body of an earlier reply! See below:

"Glen, thanks very much for your suggestion. I actually just replaced the fan with a brand new one of near identical specification (for an unrelated issue – plastic mounting frame was cracked) - same issue. Might be an idea to characterise the load current versus supply voltage for the fan, perhaps there's a peak at low voltage causing the high current (certainly looks that way from experimentation). It's 250mA rated and I think the tube heater is around the same - a lot of load for a 500mA rated regulator!"

Mark, I think this reflects your suggestion and I think you're right - the fan draws more current when it's stalled/starting and hence causes extra regulator heating due to the extra regulator load. However, the fan will need the stall current momentarily to start and that seems to be when the issue manifests – on fan startup. When the fan is free to spin, I still occasionally have the power on/off problem at fan startup (and the massive pops associated with it!).

The interesting thing as you point out Glen is that some amps seem to do it, some don't. You've not heard a report, but I know of at least five amps which suffer from this problem. Question is: why some and not others?

Is this something to do with the loading of regulator? For example, replacement of the pre-amp valve or degradation of the heater within? Does the valve heater current go up as valves age? Were different valves fitted to different regions/amps? Is there drift in the fan control circuit which is turning on the fan at a higher temperature as the amps age? Is the heatsink compound drying out and stopping the temperature sensor from accurately sensing the heatsink temperature? I can’t believe Ampeg would have let this out of the factory like this!

My point is, to make this system 100% robust, the regulator needs to operate in still air (no fan) to guarantee that the issue won't occur and the engineering calcs show this isn’t necessarily the case – feel free to tear my calcs to bits, I’m just applying a little bit of know-how and I know that can be pretty dangerous!

The fan is switched from a sensor on the main heatsink for the power FETs and takes no account of the heat of the regulator heatsink. Thinking back to my Control Theory classes many years ago, that’s an open control loop – poor system design unless both parts of the system are designed to operate in isolation. If the fan is intended to keep the regulator heatsink cool, it better be running when the heatsink regulator needs it and that isn’t the case as we’ve found.

What do you think? I’d love to get to the bottom of this!

Cheers,
Grant

[Steve Conner]

My (maybe not totally serious) opinion is that anyone who would let an amp design out of the door with vertical MOSFETs in the power stage is committing a major design flaw already. So he's quite capable of screwing up an auxiliary supply too.

It seems like a classic design gotcha: maybe when Ampeg made prototypes, they had a strong batch of regulators that could take the overload, but now they've hit some weaker ones.

Or maybe the unregulated voltage is higher when you use the amp in the UK, than it was in Ampeg's lab in St. Louis. The UK is at the high end of the European "230V", as we're still 240 on average, and I've measured 250+ in the lab at my old work. That means more volt drop across the regulator and more heating. If you have access to a variac, try the amp on that. Does it still play up at 230? 220? etc.

A 12AX7 valve/tube draws 150mA of heater current with the heaters wired for 12.6V, and this shouldn't change as the valve ages.

I have a home-made amp with a fan, that runs off a 7809 regulator, that also runs the panel lights. Every time the fan starts, the lights dim briefly and then come back up. But I don't know if this is because of the regulator limiting, or simply dropping out because the unregulated supply sags too badly. (I've never had to care because I didn't run anything vital off the +9.)

Another cheap and nasty idea: Resistor in series with the fan to limit the surge at startup?

[Grant_More]

Hi Steve,

Thanks for your suggestions there - they actually triggered an interesting thought process and I did some checks which revealled something interesting.

On the schematic, the AC output from the transformer is shown as +/-20V relative to ground (I guess it's RMS). That's around 28V peak. I measured it and I'm getting around +/-21Vrms - a little more than I expected, but my mains is 244Vrms here! Good suggestion to check

The DC recified level on the input to IC4 is shown as 22V DC. Hold on a minute - surely the DC rectified level is something like the peak level on the AC input - therefore it should be around 28V with good smoothing. A quick measure and it's around 27.5V - not 22V as shown on schematic.

So.... I wonder if the Ampeg guys have rated the heatsink for the bugged voltage shown in the schematic rather than the voltage which appears on this node in reality.

If that's right, that would be not (22-15)V*500mA = 3.5W, but 27-15V*500mA = 6W dissipated in the regulator. Ouch! I already calculated that the heatsink would need to be 18.8degC/W at 3.5W power dissipation!

So, using the same assumptions:

Required Rja = 85/6 = 14.17degC/W.

Rha = Rja – Rjc – Rch.

Filling in the numbers gives us:

Rha = 14.17 – 5 – 1
Rha = 8.17degC/W

I think your idea of mounting it on the case is looking more and more attractive now...

Maybe we don't need to rate it for the full 500mA current draw, I've not seen the 78M15 current limiting (dropping voltage) - it's only thermally overloading.

We'd need to characterise the current draw of the fan, but you think the tube is 300mA at 15V, right? I guess the pre-amp is tens of mA.

Let's say 350mA constant current draw in a typical startup case until the fan kicks in.

27-15V*350mA = 4.2W dissipated in the regulator.

Required Rja = 85/4.2 = 20.23degC/W.

Rha = Rja – Rjc – Rch.

Filling in the numbers gives us:

Rha = 20.23 – 5 – 1
Rha = 14.23degC/W

We might just get away with the new heatsink at 16degC/W!

Thanks again - great pointers for further investigation!

What's the issue with vertical FETs BTW? I'm green on mahoosive power amps, stuff I'm used to is all sub 2W!

Cheers,
Grant

[Enzo]

What if we just fiddled with the voltage divider resistors in the comparator reference to set the fan on at low speed to start with, then the sensor would run up the speed from there? Seems like we could set a minimum speed.

[MarkusBass]

When the fan is free to spin, I still occasionally have the power on/off problem at fan startup (and the massive pops associated with it!).

Grant, I'm rather a practical guy. Instead of all these nice calculations I would rather make some measurements. I don't have currently SVT5Pro but I looked at SVT3Pro which I was fixing yesterday. The fan is rated 24V and 140mA. I can believe that the current consumption is slightly higher on startup but I think this is only for few seconds. I would measure voltage on the input (you say it is 27V), voltage on the output (we can assume it is 15V) and actual current consumption. Did you measure that this is 500mA? You can also look at the resistor that is in series with the fan. It is 10 Ohms and only 0.5W. Have you checked that it has correct value in your amp? And what if you just increase it to 15 Ohms? Based on these values (10 Ohms, 0.5W) you can also make some nice calcutations.
If you say that the the voltage on the input of 7815 is too high, you can put a small resistor there. Instead of heating the 7815, the resistor will get some heat. This is another way to solve the problem. Let me know whether it helped.

Mark

[Steve Conner]

So.... I wonder if the Ampeg guys have rated the heatsink for the bugged voltage shown in the schematic rather than the voltage which appears on this node in reality.

That voltage isn't necessarily bugged. It could be the voltage that really appears there when the amp is running off its native 117V line voltage with the transformer configured for a 120V line. I'm stressing this point because if you want to develop a fix that'll suit everyone, and promote it on the net, then you need to take worldwide line voltage variations into account. Markus' suggestion of putting a resistor ahead of the regulator may hurt the low end of the voltage range.

Of course, maybe it's only people in high line voltage areas who ever see the problem, so the fix doesn't need to address low line voltages! Unless the owner moves to Brownout, North Dakota and has to take the fix out again, and then one day he gets a gig in a bar right next to a major substation, etc, etc...

Does the amp have a speaker protection relay, and is that operated off the +15? There's something else to add into your power budget.

The issue with vertical MOSFETs is explained on Rod Elliott's site: Using HEXFETs in High Fidelity Audio

You can tame them, but they were never really meant for audio service, and need careful design and usually selection. If your solid-state SVT blows its output stage, you can't just throw in any old IRFP-whatever, you're supposed to order a matched set of MOSFETs from Loud.

Unfortunately the lateral ones are less efficient and more expensive. They're almost obsolete and, fascinating fact, most of the ones you can get nowadays are made in Glenrothes. :-)

[MarkusBass]

Grant,

There is only 1 valve in this amp and the heater works in 12,6V mode, so the current consumption is only 150 mA and not 500 mA as you used in you calculations. I think it would be much easier if you just put a small resistor in front of 7815. Or correct the calculations.

Mark

[Mars Amp Repair]

Well as fate would have it...we just got our first amp with this issue! the universe works in funny ways.
I'll let you know what seems to be the cause/effect on this one when we get into it. I'm always for mounting underrated regulator/ic's on the main chassis or heatsink if they get too hot.
I recall years ago when VCR's were the big thing one NEC VCR with a transistor as a series regulator circuit where that transistor just got smoking hot. We replaced the transistor with a stonger variety (2SC1265) & merely extended the wires from the PCB & mounted it on the chassis. Voila, problem solved.

Hopefully this regulator is of the plastic case variety so we don't have to monkey around with mica insulators & such. I haven't looked into it yet. If it weren't, I'd just replace it with the plastic variety anyway. They're pretty cheap.

I'm practical, too. I'll just do it & see what the result is. thanx, glen