Low noise mosfet amplifier

Yes, mosfets are very noisy. I don't know that a "low-noise, low-voltage mosfet" exists, or would help a lot.

Cascoding isn't a bad idea, but use an opamp or a bipolar transistor for lower noise. Possibly you can do the trick where the opamp V+ supply pin drives the mosfet source.

The other thing to do is close a feedback loop on the output, through an opamp. That's a little trickier, as you have loop stability issues to deal with.

How much output current do you need into what sort of load? At what speed? Maybe my HV optocoupler amplifier circuit would be useful.

John

I wonder how a higher voltage BJT would behave in terms of noise. The BUX85 comes to mind. One has to watch out whether the demise of CRT television won't do them in. Not much demand for H-flyback anymore. At least this one was transferred to RoHS but Digikey only has the non-RoHS in stock. Under 50c in qties.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

deja vu .....already.

What the hell does low noise have to do with power amplifiers?

It's the low level sections where this is a relevant issue.

Noise is present in the absence of the signal, and is usually the product of power supply rejection ratio in any power stage I'm familiar with.

RL

He who hath done full duplex RF links knoweth what evil things power amplifier hash can do.

Then there's lasers, high end audio and so on.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

But if the PA really is the source of an 'unnacceptable level of audio noise', can you imagine what his small-signal input stage must be doing?

If it really is audible, isn't it more likely to be the result of gross instability?

RL

No idea since I can neither read his original post nor his responses because he is using that Google domain. Depends on what his application is. I have worked with laser diodes where the audio noise figure and in particular the 1/f knee of the drive power stage really mattered. Problem was that manufacturers don't generally publish such data for larger devices because they don't have that data.

That could very well be. Only a fast scope will tell. Sometimes audio amps can sing the blues at several hundred MHz. I've seen prototypes where an FM radio went mute after turning on the big power supply.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Ted,

Does the output stage have gain, or is it source follower?

It's a bad thing. Apex used to build some hv opamps (actually, Lucent fabbed them for Apex) that used an all-mosfet fab process, and it was very noisy. They published an appnote that added a low-noise bipolar opamp to make it into a composite amp that was much better for both noise and drift.

There's nothing a low-noise front-end can go to fix a noisy power stage, unless you close a loop on the whole mess. And that can't fix hf noise out of the loop bandwidth. So the cascode is a good idea.

Not if the output stage adds lots of noise. In some situations, like RF, it's not feasible to close a loop around the whole system.

Mosfets are noisy without much help.

John

Hi John

I want to avoid BJTs for reasons I'll not go into at the moment. I've observed the output on a decent scope and seen no evidence at all of any instability, but I've isolated the output stage and its driver and that's where a significant amount of the overall noise originates.

Peak output requirement in thie prototype is about 150mA into the primary of a push-pull output transformer, though subsequent versions could be several times that, but still well within the MOSFET's current capability. Frequency response of 20kHz ideally, although could live with less for the application. Not sure how an optocoupler would help remove the noise if it originates in the output devices themselves.

Thanks for the input.

Ted Wilson

Not a source follower - common-source with feedback from output back to phase-splitter input.

Regards

Ted Wilson

, t

Whatever noise may be being contributed by the earlier stages can be sorted fairly easily, as there are a lot more low-noise device options available at the lower supply voltage the early stages are required to run at.

Whether the output noise is particularly bad because I'm using a relatively high voltage device, or simply because it's a power MOSFET, I'm not entirely sure, but a significant amount of the output noise persists when I disconnect the early stages.

Re the sugestion of instability - I've monitored the output on a decent quality scope and seen no hint of instability.

Regards

Ted Wilson

Hi Joerg

As I've said in my reply to John Larkin, I don't want to use BJTs - this defeats the whole object of the excercise. There is a closed loop around the output stage and its phase splitter, although it doesn't include an op-amp - at least, not in the signal path.

Thanks for your input.

Ted Wilson

Actualy, now I think about it, the cascode tail device could be a JFET and there are certainly low-noise versions of these available. That being so, the issue then boils down to whether or not the tail device controls output noise and I'm inclined to think it would, since that's the component controling output current.

Ted

d,

ut

Hi again Joerg

I'm using Google as it's the only newsgroup reader I can find at the moment that works with my isp - TalkTalk. I've tried setting up Outlook Express to use that, but can't find a newsgroup server that it works with. I used to use Free Agent, but that's gone now - much preferred the subject display that gave.

Ted

Pity, I'd be interested in knowing why you can't use bipolars.

I've

The optocouplers *are* the output devices! But not at your power level.

John

Ted, it seems you're seeing more noise in the MOSFET stage than theory predicts, and more than measurements of MOSFET performance predict. Despite your measurments, it's likely this is due to RF oscillation. Sadly, high-voltage MOSFETs love to make self-contained oscillators at high RF frequencies (15 to 40MHz) whenever they are operated in the linear mode with Vds greater than say 25 volts, and Id above 0.25mA.

It's not always easy to see this RF oscillation externally, unless you look at the right place with the right tools. But it's often easy to fix, by placing small ferrite beads on the leads. It's possible the exact type of bead may matter, but once I had good results with a certain bead, I bought a pile of them and stuck with it, so I can't advise you there. My beads have about 20 ohms of loss resistance above 20MHz.

The disappearance of your excess noise will be your best evidence that the MOSFET was in fact oscillating.all along. Let us know what you learn.

One has to be careful if it's for production. The number of HV bipolars has dwindled quite a bit. Just like the days of $1/gal gas are over :-(

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

[snip]

We whine and we cry, buying into the leftist weenie BS, then we see fact...

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: "skypeanalog" | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

| 1962 | America: Land of the Free, Because of the Brave

He doesn't need a HV bipolar for the botttom of a cascode.

John