A Difficult Circuit !

[Robert Reed]

I need a Variable Gain Amplifier That will also act as a variable attenuator of 0 - 20 Db. I actually need to boost a 400 mv signal up to 1600 mv and then be able to attenuate this signal from 0-20Db. The circuit must have a flat frequency response (+/- 1Db) across a range of 0.3 Mhz to 70 Mhz. Control linearity is not important as long as it has relatively flat response at any setting.

Heres what I need:
Gain of 4
Attenuation factor of 10 (20 Db)
Output capable of a 2500 mv signal (p-p)
Flat response 0.3 to 70Mhz (+/- 1 Db)
Less than 5% distortion (this is a sinewave signal)
Thru hole components with 2 sq. in. of board real
estate

Heres what I have Tried:
Various FET configurations as passive devices in the
in the triode state.
Various VGA chips.
Differential ampl. fed from an adj. constant current
source.
Pin diodes (which is what I normally use above 10 Mhz)
and I pcked one with the longest carrier lifetime I could
Find, but still introduced distortion below 3 Mhz.

All of these circuits have Basically had fair results but have fallen short in one way or another. In most cases stray capacitance is the killer in regards to flat response. The analog control (pot) for this circuit must be located about 6" from the electronics. Believe it or not, an in path 100 ohm potentiometer has worked as well as anything to date. However the connecting coax and mechanics of the pot set up broad low `Q' parasitic tuned circuits that cannot be overcome.
Obviously, with the huge collection of FETs and VGAs on the market ,I can not try them all so I may have probably missed some that would be better suited for the job at hand.
If anyone has built up circuits of this sort that would meet my parameters or even come close to it, I would certainly appreciate hearing about it.

[Chris Smith]

You may need to split up the task?

A volage controlled attenuator [Fet to 70DB] and the Variable Gain [op amp] Amplifier?

Tying them together you may work out?

[Externet]

Hi Robert.
Would an operational amplifier capable of that frequency response (which there is many, many of them) with a set gain of 4 and its input signal on a potentiometer work for you ?
At maximum potentiometer, applying full signal will amplify 4X, and at minimum potentiometer will attenuate as much as you trim it to divide the input signal.
Miguel

[Robert Reed]

Exter
Welcome to the world of high frequency and good test equipment standards.The key word here is flatness of response over a wide frequency range. The stand alone amplification is no problem and I have accomplished this with a single transistor stage. However attenuation is another matter. As mentioned in my previous post, a potentiometer has certain drawbacks in this application. Once you get into and espescially beyond 10 Mhz territory,components start to exibit different behaviors than what they were initially intended for. For example a standard panel potentiometer has inductive and capacitive properties as well as resistance. These are totally insignificant at low frquency (below 100 Khz), but will bcome very significant at high frequency. And to add to this its mechanical position will vary these properties at any given position. I had mentioned a combo amp & attn. as one package to keep part counts low, but so far have been unsuccessful in accomplishing that. For wideband flat response, low resistance values are mandatory, so that unwanted circuit capacitances can be quickly and fully charged and discharged to allow proper operation. Even the fet attenuator schemes I have tried thus far have failed to make the grade due to their inter element capacitance Consider an fet with an off resistance of > a few hundred Kohms and shunt internal capacity 5 pf. You now have for all accounts an open circuit at 1 Mhz, but only an impedance of a couple hundred ohms at 70 Mhz. Stray and parasitic capacitance - therein lies my problem. I need a circuit that can get around this. The pin diodes are wonderful attenuators, but do not work well below several Mhz.

[MrAl]

Hi Robert,

You realize that is quite an order right?

Did you check the Analog Devices site for a variable gain amp chip?
With a variable gain amp chip you would supply a control voltage
or current that changes the gain of the amp. The control signal is
dc so you can do it with a pot.

[rshayes]

Hello Robert,

Analog Devices used to make a Gilbert Cell type of multiplier that was capable of about 200 MHz bandwidth. If I remember correctly, it was in an 8 pin DIP and gave differential current outputs. A wideband transformer could be used to convert the output to a single ended, ground referenced signal.

The old MC1495 multiplier could give about 50 MHz bandwidth when operating into 50 ohm load impedances. This is a little short of what you need, but the part is over 35 years old.

The original application for this type of multiplier was as a variable gain amplifier in wideband oscilloscopes. Barrie Gilbert published a pair of papers on this type of circuit in one of the IEEE journals (probably the Solid State Circuits Journal) about 1969.

[Robert Reed]

RS Hayes,MrAl
Yes I have experimented with some VGAs by Motorola and others. Most of these were chips I already had on hand. These were for the most part diff. input & diff. output although thy can all be operated single ended OK with some reduction of gain.The problem here is that they are restricted in maximum input levels - usually less than 50 mv p-p. Although they have enogh gain to get you back up to where you want to be, I hate to attenuate first and then have to use higher amplification to achieve my purpose as this requires more stringent circuit layout in terms of noise and stability. My latest hope is a differential long tailed pair configuration, but I am having balancing and matching problems of the transistors involved. This circuit would involve only three PN junctions and is controlled by a constant current source via DC potentiometer control. I believe the more PN junctions involved, the more difficult it would be to stay within my set parameters- hence the long tailed pair.I see a lot of VGAs advertised in surface mount, but I need thru hole for this application, as it is an upgrade to an existing peice of test equipment. I do not want to get into SMD adapters here as in this application circuit lay out is very critical (short point to point wiring; minimal parasitic capacitances,etc.) Actually if I could come up with a good variable 20 Db attenuator, the rest (gain block) is duck soup. As I had mentioned. my attempts with FETs operated in their triode region didn't quite make it. The problem here is they need to have an Rds on of <30 ohms, Low Rds ons require larger gate areas and this increases inter element capacity. These capacities are killers for reasons previously stated. Lower capacity FETs go hand in hand with higher Rds ons - makes it difficult for flat wideband attenuation. I will research the Gilbert cell ICs and the suggestion of Scope vertical amplifiers sounds very interesting. I had not given thoght to that before, but it seems like it would be the circuit I am looking for as long as its not to complex(limited board real estate here).

[Chris Smith]

My source book from 1968 shows a double FET with A volage controlled attenuator to 70 DBs with a heavy load.

Great place to start?

I have over 800 pages of these oldes but goodies.

Sourcebook of Electronic Circuits

[rshayes]

Hello Robert,

Look at the Analog Devices AD834 (also ap notes AN-212 and AN-216). The input range is plus or minus 1 volt, which should be adequate for your purpose.

The MC1495 data sheet has a fairly extensive section describing the operation of this type of multiplier. The MC 1495 itself is just a little slow for your application, and it is a little bulky since it is in a 14 pin DIP package.

[Robert Reed]

RS Hayes
I misread your prior reply and dug out a MC1445/1545 from stock. This is another wideband VGA. I breadboarded a quick circuit. Comes close to what I want. Flat response to 12 MHz; - 3Db down at 30 Mhz - and thats on a breadboard which greatly inhibits hi frequency operation! This looks worthy to build up a test circuit on a ground plane circuit board. On your last post, I saw my error- MC1495. I dug thru my spare parts and actually came up with one (MC1595L). The specs look worthy of a try (80 Mhz / transconductance).Also downloaded AN834 data sht. and related ANs. If it looks good on paper, I hope I can still get it in the DIP package. Well it looks like I have a lot to chew on for a while

[Chris Smith]

Did you try a Mosfet or the Cascade mosfet set up?

[k7elp60]

Robert you might want to consider a MC1490 or it's sister the MC1350 RF/IF amplifier. The characteristics for gain looks close, and it has a DC voltage AGC control. If you think it might work for you, I have a number of each in my stock of parts if you would like to try them.
Ned

[Robert Reed]

K7
Thank you very much for your kind offer, but yes I have already run through some initial designs with the MC1350 and MC1590 (mil spec of the 1490) chips. They fell short of my expectations in that they require a very low level input to maintain low distortion at the higher attenuation levels. It seems that most these style chips run their output levels very close to Vcc thereby limiting large output swings (~2vp-p). I guess their primary purpose is for low level RF and IF circuitry in which case you would probably never approach these kind of levels. At any rate hang onto these chips as I saw some on the web priced near $50 apeice.
I can't beleive that in all my recent research and going through older circuit manuals I have not been able to come up with a circuit like I am looking for. In the past, I have designed many high frequency amplifiers upwards of 200 Mhz to translate Lo level signals to digital logic levels. But these were bang the signal rail to rail with no concern of fidelity or depth of collector saturation. Now I must maintain purity of sine wave and a constant flat output across the band. As the post reads - very difficult!