Cheap amplifier for DDS / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Cheap amplifier for DDS

previous 1 2 3 next

Move Thread

LAN_403

Hazmatt_(The Underdog)

Sat Feb 17 2018, 06:29AM

Registered Member #135 Joined: Sat Feb 11 2006, 12:06AM
Location: Anywhere is fine
Posts: 1735

This amp goes to 10MHz, but the board for it uses ground planes to help minimize capacitance and all the wierdness that affects gain of amps at higher frequencies. Also, the power transistors need heat sinks, don't attempt to run this amp without heat sinks, the transistors do get quite hot.

1518848980 135 FT181443 4017 Output Stage

Dr. Slack

Sat Feb 17 2018, 06:53AM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

tarakan2 wrote ...

Dear Dr. Slack, what does "a volt or so from 50 ohms" mean?

Dr Slack is referring to an output type and a power level invariably used on professional signal generators.

Let's say you have a source designed to drive 1v rms (2.828v peak to peak) into a 50 ohm load. That's 20mA rms into the load, so a power of 20mW, also written as +13dBm (10log10(power_versus_1mW_reference)).

There are several ways this could be done, including having a voltage source delivering 1v rms. The way it's invariably done in professional equipment is to have a 50ohm resistor in series with a voltage source delivering twice the voltage required at the output.

When I first went into engineering, I thought 'what a waste of power and voltage headroom.' However, for general purpose equipment, the driving concern is being able to use it generally. This means connect it to anything else while keeping its specified accuracy, use it to drive anything, and replace it with any other manufacturer's piece of kit, and still have it work the same way.

It's usual to have coaxial connections on sources and receivers, and make them 50 ohm impedance. That way, everything can be connected with 50 ohm cables and everything stays matched, which amongst other things means it's easy to compute what the voltages and current are at any point. In addition, you can load an output with any load impedance, including open circuit and short circuit, and the output continues to behave, still delivering a calculable signal.

This is why I made a distinction between measurement, and power. For power, you need what you need, 100v, 1000v? For measurement, you only need a reasonable level signal. This means enough to be well out of thermal noise, enough to drive an oscilloscope, and enough to drive a power amplifier to get to whatever level you actually want at your test system.

For most bits of kit, 'reasonable' means something between +10dBm and +20dBm, so between 10mW and 100mW into 50 ohms, or between 707mV rms and 2.23v rms into 50 ohms. That latter voltage level of 2.23v rms will need +/- 6.3v peak into the internal 50 ohm resistor, which is about the limit for +/-15v rail op-amps, which is why +20dBm is often regarded as the upper end of 'reasonable'.

So you see that 'a volt or so from 50 ohms' is something in the +10 to +20dBm range.

Just like on cryptography forums, when they say 'don't write your own encryption, use a tried and tested library', I would be inclined to say 'if you're making a general purpose source, then don't invent a new output method, do use an internal 50 ohm resistor in series with a voltage source'. It will mean then you can load the output with any impedance, drive it into coaxial cables without capacitive loading effects, and when you use or buy other pieces of kit, it will all work exactly the same way. If you drive a high input impedance amplifier with it, you'll just see double the calibrated voltage.

Once you've accepted the cost of the series 50ohm resistor, in terms of wasted power and voltage headroom, you can then appreciate the other advantages. For instance, the output amplifier always sees a minimum load, regardless of instrument loading. Any capacitive loading is isolated by the resistor, op-amps are notorious for not liking to drive capacitance.

If you do need a different output impedance to do some tests, then build a custom amplifier to do just that, don't build it into your general purpose source. Borrowing from the software guys again, it's separation of concerns.

Hazmatt's amplifier LM6181 looks chunky enough by itself to do +20dBm without the transistor followers. It would be somewhat easier to build without them, but that's quite a lively amplifier anyway, and would need care in layout to keep stable, which is why I said if you can get an evaluation board for a suitable amplifier, then all that's taken care of.

Conundrum

Sat Feb 17 2018, 10:10AM

Registered Member #96 Joined: Thu Feb 09 2006, 05:37PM
Location: CI, Earth
Posts: 4059

Yup, second that.

Hazmatt_(The Underdog)

Sat Feb 17 2018, 05:47PM

Registered Member #135 Joined: Sat Feb 11 2006, 12:06AM
Location: Anywhere is fine
Posts: 1735

The amp I posted comes from the BK 4017 signal generator.

I've been researching medium power wide band amplifiers for a while, and the wide band mosfet amplifler that
always seems to present itself, although I am not real happy with some of the design, I have not quite settled
on a scheme that I like that preserves the symmetry of the sine wave. You have to run the amp hot to preserve
symmetry.

Here's a gratuitous shot of a higher power driver up to about 1MHz, mosfet, note the 47 ohm input resistor.

Steve is right though, you really should specify desired power and impedance.

The power amp has nearly zero output impedance, and the wide band amp is 50 ohm, what are you looking for?

If you're looking for a medium power wide band amplifier, I would buy one from Amp Line, you can get the power block for about $500. I built this, wasting a tremendous amount of time, instead of just buying one of their amplifiers. I should have bought one years ago.

Amp line and ITECO basically use the same mos design, but don't tell them that.. its a secret ;p.

That would be my advice, you want a wide band higher power amp, just save up and buy one.

tarakan2

Sat Feb 17 2018, 08:57PM

Registered Member #3859 Joined: Sun May 01 2011, 03:47PM
Location:
Posts: 179

It is very likely that I will use a push-pull circuit with a gate driving transformer.

I am still thinking about what amp I would like to use since I am not sure what DDS arrangement I would stick to.

Perhaps I would need to start a different thread to describe what I would like to do in my experiment, what signals I am trying to obtain.
I have many questions about DDS itself and about my options of obtaining the signals that I would need.

Sulaiman

Mon Feb 19 2018, 09:24AM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

I bought one of the eBay ready made pcb dds signal generators some time ago,
AD 9850 or AD9851, I forget which, with lcd display.
The kit used single-ended output via an anti-aliasing (low pass) filter taken staight from the application notes ... for a 200 R load.
Even after adding a 1-0-1 : 2 rf transformer to the IC output, changing the bias,
(push-pull gives twice the current and adding a diode from xfmr. ct. to Ov doubles voltage range)
and adding a second transformer for 200 : 50 Ohms after the filter, the output is still only
a 'signal', requiring amplification for most uses.

IF I wanted a sig-gen for TC frequencies I'd consider something like this

You would also need a USB power source - old 'phone charger, usb battery power pack etc.
(you could save a little cash by buying a pcb without case - but you would regret it ;)

If the specifications are correct then for up to 500 kHz it should be more than adequate.

+/- 10 Vpk from 50 Ohm source is equivalent to;
+/- 20 mApk into a short circuit, or diode, base-emitter, led etc.
7 Vrms with no load
3.5 Vrms into a 50 Ohm load
... which is not too far from the resistive part of the impedance of a tesla coil secondary at resonance.
You may be able to illuminate a neon lamp from the secondary topload to earth ... or your body.

OR, for wider frequency coverage and data over usb for analysis/display

or similar

P.S. I use one of these

... probably older than many members here but still giving good service.

tarakan2

Mon Feb 19 2018, 06:24PM

Registered Member #3859 Joined: Sun May 01 2011, 03:47PM
Location:
Posts: 179

I am trying to avoid programming of Arduinos or FPGA. I don't enjoy programming and I would rather buy a turnkey product for a greater sum of money.

Is it possible to de-soldier oscillator crystals from DDS boards and somehow wire several of them to work from a single clock source?
I have seen this option on more expensive DDS generators, yet I am not sure what they propose for me to use as a clock source.
I would need 5 DDS generators with interface wired together for an experiment. They all need to work off the same clock.
If I needed 4, then I would have used the AD9959. However, I am not certain that the interface that they sell on Aliexpress would allow me to do what I want and I am not interested in creating my own. If I would team up with a programmer to create something of my own, it would be based on FPGA and there is a person that is an expert in this area that I am friends with. However, to do a preliminary experiment, I would like to wire five DDS generator boards to a single clock source.

This may not be so simple since I never soldered things that work ant tens of MHz.

Do they sell anything that would make such retrofitting convenient?

Dr. Slack

Mon Feb 19 2018, 08:29PM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

You could buy five of these things

for Â£75

or five of these

for Â£15

I'd imagine you could take the crystals off the first ones, which have the advantage to you of having manual front panel controls. I know you could take them off the second ones, but these would need a bit of software to control, but not a lot, they are SPI, and at least you know what the device is and can get data sheets, which might be difficult for the first ones.

It's fairly straightforward to move clock signals around at 10s of MHz. It needs a bit of care, you don't want long wires and haphazard paralleling of loads. A single hex HC buffer from the master clock source, each output driving a single source terminated transmission line to each driven device should be pretty bomb-proof.

Sulaiman

Mon Feb 19 2018, 09:37PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

Most DDS have external clock pins so can be operated by the same clock,
for synchronised operation

loads of other app notes available there.

Conundrum

Tue Feb 20 2018, 09:53AM

Registered Member #96 Joined: Thu Feb 09 2006, 05:37PM
Location: CI, Earth
Posts: 4059

How high will a DDS go, could it be modified to work with a satellite LNB?
Thinking DIY radar here (0.05mW)

previous 1 2 3 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob