Hardy 990s run a little warm ?

[Samuel Groner]

I was under the impression that these old circuits performed best when terminated with 600 ohms, i.e. if the 2503 output transformer is connected to a 10k input, then the transformer should ideally have 600 ohm strapped across it's secondary. If outputing into a 600 ohm input, then the termination is taken care of. Is I wrong, or is I right?

It's impossible to give a general answer to the question whether to terminate vintage gear with 600 ohm or not--in many cases, high frequency response will be smoother with 600 ohm termination, but distortion and low-frequency response may suffer. Look at the Great River MP-2NV user guide, IIRC there is a section on this.

A compromise might be a zobel network load, e.g. 3.3 nF in series with 600 ohm--this should keep the HF response clean but decreases distortion in the audio range. Some experimentation with the optimum capacity value is needed though.

Not that the above consideration only apply to transformer balanced outputs. For other output stages less loading is better in about any circumstance.

Samuel

 

[peter purpose]

Interesting... thanks Samuel.. :thumb:

From the GR pdf.......

Loading
This button controls the output termination relay. When in the out position, the output
transformer is unloaded. When pushed in, a 600 ohm resistor is connected across the secondary.
Great River Tips:
The output transformer has an inherent resonant peak that affects the frequency response at
high frequencies. This peak is at approximately 50kHz, and is about 6db up from flat and is
cable and load dependant. The amplifier is designed to be correctly loaded by 600 ohms for best
flatness in both frequency and phase response, but some interesting effects happen when
unterminated.
A gentle high frequency lift, like an “air band” EQ and a bit grittier midrange are the effects you
should hear when the loading button is out.

 

[John Hardy]

From Peter:

The 2503 has a 75 ohm primary. The 990 is not designed for that kind of load.

From cj:

Anyway, the 2503 is indeed 75 ohms.
DCR is about 6 or 7.

From API 512C specs:

Output Impedance: Less than 75 Ohms, Transformer Balanced

Perhaps there is a misunderstanding of terms here: It is the output impedance of the API 512C card that is "Less than 75 ohms". This is not the load impedance that the 2520 or 990 sees --- it is the output impedance of the overall combination of the op-amp and output transformer. The op-amp itself typically has an output impedance of less than 1 ohm. It is essentially the same as taking the output of any op-amp and putting a 75 ohm resistor in series with it on the way to the output connector. If there were a 10,000 ohm load at the output connector, the total load seen by the op-amp would be 10,075 ohms because the 75 ohm resistor would be in series with the load. It is NOT a 75 ohm resistor going to ground. Therefore, the 2503 output transformer adds "less than 75 ohms" in series with the output of the 2520 or 990.

So I will repeat what I said earlier: "Basically, the AC impedance that the 990 sees is determined by the load that the secondary of the transformer sees. If the load on the transformer secondary (the next piece of equipment in the signal path) is a high impedance load, the 990 will see a high impedance load, not 75 ohms."

John Hardy
The John Hardy Co.
www.johnhardyco.com

 

[Samuel Groner]

Perhaps there is a misunderstanding of terms here: It is the output impedance of the API 512C card that is "Less than 75 ohms". This is not the load impedance that the 2520 or 990 sees--it is the output impedance of the overall combination of the op-amp and output transformer. The op-amp itself typically has an output impedance of less than 1 ohm. It is essentially the same as taking the output of any op-amp and putting a 75 ohm resistor in series with it on the way to the output connector. If there were a 10,000 ohm load at the output connector, the total load seen by the op-amp would be 10,075 ohms because the 75 ohm resistor would be in series with the load. It is NOT a 75 ohm resistor going to ground. Therefore, the 2503 output transformer adds "less than 75 ohms" in series with the output of the 2520 or 990.

The API 512C got a 1:2 step-up output transformer, IIRC. So if you load that output with 600 ohm, the opamp will indeed see a 75 ohm load!

Samuel

 

[rodabod]

[quote author="Samuel Groner"]
The API 512C got a 1:2 step-up output transformer, IIRC. So if you load that output with 600 ohm, the opamp will indeed see a 75 ohm load!
[/quote]

Do you not mean 150 Ohms? (600/4).

The 2503 is classed by Sowter as a 75:300+75 transformer which makes sense to me, unless I have missed something obvious.

http://www.groupdiy.com/specs/pdfs/Sowter-9825-Data.pdf

 

[Samuel Groner]

Oh, sure! Need to fresh up my mental arithmetic skills. :green:

Not sure how most people set up their output transformers but if used 1:3 the reflected load would be even lower than 75 ohm.

Samuel

 

[John Hardy]

From Samuel:

The API 512C got a 1:2 step-up output transformer, IIRC. So if you load that output with 600 ohm, the opamp will indeed see a 75 ohm load!

There is always a way to make an op-amp see a 75 ohm load if you are determined! Actually, you may need to SHORT the output of the 2503 to get the op-amp to see (approximately) a 75 ohm load, since there is ("less than") a 75 ohm output impedance introduced by the transformer. The transformer looks like approximately a 75 ohm resistor in series with the output of the 990 or 2520, so a 600 ohm load on the secondary would look like a 675 ohm load to the 990 or 2520 driving the transformer.

But is a 2503 supposed to be used with a 600 ohm load? Of course, I only use the Jensen JT-11-BMQ (essentially the same as the JT-11-BMCF) and I will be using the JT-11-DM for my upcoming Lunchbox-style cards (one size smaller transformer than the JT-11-BMQ, and more appropriate for the lower supply voltages in the Lunchbox format), and they don't need a 600 ohm load, or whatever. They just work great. They don't have any frequency response bumps that need to be dealt with by adding a terminating resistor (600 ohms or otherwise). They don't have the color and distortion of the 2503 (and the 990 doesn't have the color and distortion of the 2520), but that's another story.

Besides, that is not the original issue: an apparent misunderstanding of output impedance and load impedance.

John Hardy
The John Hardy Co.
www.johnhardyco.com

 

[[silent:arts]]

John,
your 990 24V work fine in my API clones.
they get warm, but not hot (doesn't matter if I terminate the Pre out with 600R, 10K or 100k)
and: it is the best looking OpAmp I have ever seen :grin:

the headroom is killer (32 dBu at >1% THD - will do some more tests with an AP when I'm back at work)

 

[Samuel Groner]

The transformer looks like approximately a 75 ohm resistor in series with the output of the 990 or 2520, so a 600 ohm load on the secondary would look like a 675 ohm load to the 990 or 2520 driving the transformer.

No, because this is a step-up transformer. If used as 1:2, a 600 ohm load looks like 600/2^2 = 150 to the driving opamp (ignoring DC resistances). Very likely most of the DC resistance will be in the secondary, so this will reflect as a lower impedance to the primary, causing a short to be less than 75 ohm at the primary.

But anyway, I think we agree that the opamp will not see 75 ohm in any reasonable real world use and that the 990 is doing fine here. :thumb:

Samuel

 

[John Hardy]

From Samuel:

No, because this is a step-up transformer.

I've strictly worked with the 1:1 Jensen output transformers (JT-11-BMCF, JT-11-BMQ and JT-11-DM). A 1:1 ratio makes more sense to me since I would think it would be more linear than a 1:2 or 1:3 ratio. I don't see much need for adding voltage gain at that point, just the isolation that the transformer provides, and any particular character the transformer might add. On the other hand, the voltage gain probably comes in handy in some instances. Perhaps a mic preamp with a ribbon mic where the preamp itself doesn't have quite enough gain, the step-up transformer can help out. If an op-amp is working at +/-15 VDC supplies (much of the API equipment), the voltage gain from the output transformer might be helpful in achieving higher output levels than would otherwise be possible, necessary for certain high output level situations. Etc.

I suppose the step-up ratio is part of the API sound as well.

John Hardy

 

[Flatpicker]

Good discussion, guys. I've studied this "how best to terminate a 2503" thing a lot myself, but it keeps boiling down to "how I want it to sound". Of course we need limits so as not to harm the op-amp or transformer, but otherwise I think it's all about taste.

John - The see-thru 990 looks awesome! Are you making them like that now?

 

[John Hardy]

From Flatpicker:

John - The see-thru 990 looks awesome! Are you making them like that now?

Thanks for the compliments.

I haven't made the 990 in the clear epoxy package since 1990 or so. It was a lot of fun designing the package, but a lot of work to make them that way. It was only the original version of the 990 that was available that way.

I always wanted a better package from a thermal standpoint, so the aluminum shell with a soft silicone encapsulant was the result. I designed the aluminum shell package when the "A" and "C" versions of the 990 were developed at Jensen.

In the aluminum package, the two output transistors, which create most of the heat, are along the edge of the p.c. board so their metal back plates are in direct contact with the wall of the shell and bonded to it with a high thermal conductivity epoxy for maximum heat sinking. The aluminum shell dissipates heat much better than clear epoxy, and it keeps a much more uniform temperature throughout the module. The soft silicone encapsulant puts much less stress on the components as the modules heat up and cool down, and expand and contract. The silicone actually expands and contracts at a greater rate than the clear epoxy (and they both expand and contract at a greater rate than the components), but the silicone is so much softer than the epoxy that there is much less stress on the components.

The aluminum shell is not as fun to look at, but it is more reliable. I usually bring an unpotted 990C to the AES conventions so people can see the (hopefully) high quality of design and components.

John Hardy
The John Hardy Co.
www.johnhardyco.com

 

[Flatpicker]

[quote author="John Hardy"]In the aluminum package, the two output transistors, which create most of the heat, are along the edge of the p.c. board so their metal back plates are in direct contact with the wall of the shell and bonded to it with a high thermal conductivity epoxy for maximum heat sinking. [/quote]
Which might explain why it's warmer to the touch - it's dissipating the heat more efficiently.

I remember some discussion about the first 990s being clear epoxy, but I thought maybe you were making them like that again. Of course, the aluminum shell and soft silicone makes more sense, though.

Thanks for all the good info.

 

[FredForssell]

John,

IMO the potting on that clear module was a work of art. I still have one that I bought back in the mid-1970s around cause I like to look at it.

Cheers

 

[John Hardy]

From FredForssell:

IMO the potting on that clear module was a work of art. I still have one that I bought back in the mid-1970s around cause I like to look at it.

Thank you very much. Just to put the time line straight, I started making them in the 2nd half of 1979.

One of my favorite stories was when I showed a clear 990 to a friend for the first time. He had just smoked a joint. He stared at the 990 from every possible angle. Then he shook it, and said "It would be cool if it snowed inside."

If the Star Trek concept of "transparent aluminum" ever comes true (Star Trek IV - The Voyage Home), I'll consider another clear package.

John Hardy
The John Hardy Co.
www.johnhardyco.com

 

[hejsan]

[quote author="John Hardy"]
One of my favorite stories was when I showed a clear 990 to a friend for the first time. He had just smoked a joint. He stared at the 990 from every possible angle. Then he shook it, and said "It would be cool if it snowed inside."[/quote]
BVAHAHAHAHAHHA!!

Seriously, you should make like a special christmas edition for your best customers or something! That's to good an idea!

 

[Flatpicker]

[quote author="John Hardy"]..."It would be cool if it snowed inside."[/quote]
That’s hilarious!

Wow, Fred and John posting here. Great to hear from both of you!

 

[FredForssell]

Thank you very much. Just to put the time line straight, I started making them in the 2nd half of 1979.

Ah, yes. Having come of age in the late 1960's my memory of that period isn't quite what is should be. I think we actually purchased those from you in the early 1980's when the 16 channels of API 550 EQs we had on the road started to die. We never had that problem with your modules.

Anyway, nice potting job. Right up there with Paul Buff's Transamp modules.

Cheers,

 

[John Hardy]

[quote author="FredForssell"]Anyway, nice potting job. Right up there with Paul Buff's Transamp modules.[/quote]
It was the TransAmp that inspired me to do the clear 990. A friend of mine was working for Paul Buff in the late 1970s, so I visited Valley People. The front panels for the Kepex and Gain Brain (actually Allison Research, a predecessor to Valley People) were made of molded epoxy. They had a fader that was part of an automation system, and its panel assembly was potted too. A lot can be done with silicone molds, but they have limited life. Many of the clear 990s had to be buffed to get the surface finish optimized.

Since Deane Jensen had published the schematic of the 990, I had nothing to hide by using the usual black epoxy, so the clear epoxy approach seemed like a cool thing to do.

I've been through the 1960's too. At least, I think I have.

John

 

[jspartz]

Wow,

Great conversation going on in this thread! I have 990 versions from both John and Fred so to see them here talking about the 990 design is wonderful. I love this forum. I can learn so much from so many people.

Thank you!

Jason